Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Defect control in vacuum bag only processing of composite prepregs
(USC Thesis Other)
Defect control in vacuum bag only processing of composite prepregs
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
DEFECT CONTROL IN VACUUM BAG ONLY PROCESSING OF COMPOSITE PREPREGS by Lessa Kay Grunenfelder A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (MATERIALS SCIENCE) December 2012 Copyright 2012 Lessa Kay Grunenfelder ii Epigraph “ And r e memb e r, a lso… t ha t m a n y pl a c e s y ou wou ld l ike to se e are just off the map and many things you want to know are just out of sight or a little beyond y o ur re a c h. Bu t som e da y y ou’ll r e a c h th e m all, for wha t y ou l e a rn toda y , fo r no re a son a t a ll , will he lp y o u discove r a ll the w ond e r ful se c re ts of tom or row .” Norton Juster, The Phantom Tollbooth iii Dedication This manuscript is dedicated, with all my love, to Joshua Legg. iv Acknowledgements I am grateful to have had the support of my colleagues, friends, and family throughout the completion of this work. So many people have contributed to the realization of this accomplishment, and I owe all of them my most heartfelt thanks. I would first like to acknowledge my advisor, Professor Steven Nutt. I have been extremely fortunate to study under the guidance of Professor Nutt. His expertise, generosity, advice and encouragement have been invaluable. With his help and feedback I have become a stronger researcher, writer, and presenter. I am grateful to have spent the past five years working with Professor Nutt. My colleagues at the USC composites center have not only brightened my days, but have provided me with valuable feedback, assistance, and scientific discussion. I would particularly like to thank Nikil Kar, Ehsan Barjasteh, Christina Naify, Yuzheng Zhang, Shad Thomas, Byungmin Ahn, Yinghui Hu, Rohan Panikar and the late Warren Haby. I have also had the pleasure to work with several amazing undergraduate students, whose enthusiasm and work ethic have been inspiring. Thank you to Mike Asfaw, Chris Fisher, Sidi Huang, Stephanie Klimczak, and Christina Cabble. For funding and technical support I would like to acknowledge several companies whose personnel have contributed to my work. Technicians and scientists at Airbus, Northrop Grumman, Boeing, Cytec Engineered materials, TA Instruments, Airtech International, ThermoFisher Scientific, Mettler Toledo, Wyoming Test Fixtures, and Instron were always an email or phone call away to help me solve equipment problems v and come up with creative solutions. I also received material donations from many of the above listed companies. While considerable time was spent in the lab throughout my career in graduate school, I was also fortunate to have wonderful friends to take my mind off of work. All of my friends have made this process more enjoyable. Specifically, thank you to Melissa McMeekin, for your abundance of love and support. You are a true friend. And thank you Adrienne White, for sharing in both my frustrations and excitements, and always being up for an adventure. Last, but certainly not least, I would like to thank my family. My parents, Gregg and Catherine Grunenfelder, have been a constant source of encouragement and unconditional love. I grew up believing I could do anything. Not only did my parents encourage this belief, they also provided me with the resources and confidence necessary to follow my dreams. My sister, Rachel Dayka, is my greatest inspiration. Throughout my life she has provided me with a model for what a strong woman can accomplish if she puts her mind to it. Finally, I would like to thank my fiancé, Joshua Legg, for standing by me through every hurdle, and helping me to always see the bright side. Without him, this would not have been possible. vi Table of Contents Epigraph .............................................................................................................................. ii Dedication .......................................................................................................................... iii Acknowledgements ............................................................................................................ iv List of Tables ..................................................................................................................... ix List of Figures ..................................................................................................................... x Abstract ............................................................................................................................ xiv CHAPTER 1. Introduction........................................................................................... 1 1.1 Motivation ............................................................................................................ 1 1.2 Voids in composites ............................................................................................. 2 1.3 Vacuum bag only prepreg .................................................................................... 5 1.3.1 Sources of voids in VBO processed parts ..................................................... 9 1.4 Scope of dissertation .......................................................................................... 15 CHAPTER 2. Background .......................................................................................... 18 2.1 Experimental methods ........................................................................................ 18 2.1.1 Rheology ..................................................................................................... 18 2.1.2 Coulometric fischer titration ....................................................................... 20 2.1.3 Thermogravimetric analysis (TGA) ............................................................ 22 2.1.4 Differential scanning calorimetry (DSC) .................................................... 22 2.1.5 Fourier transform infrared spectroscopy (FTIR) ........................................ 23 CHAPTER 3. Demonstration of property equivalence .............................................. 26 3.1 Introduction ........................................................................................................ 26 3.2 Experimental ...................................................................................................... 27 3.2.1 Material and cure conditions ....................................................................... 27 3.2.2 Mechanical testing ...................................................................................... 28 3.3 Results and discussion ........................................................................................ 33 3.3.1 Interlaminar shear strength ......................................................................... 33 3.3.2 In-plane shear strength ................................................................................ 34 3.3.3 Compression after impact ........................................................................... 35 3.3.4 Open-hole compression .............................................................................. 36 3.3.5 Mechanical testing summary ...................................................................... 38 3.4 Conclusions ........................................................................................................ 39 vii CHAPTER 4. Cure cycle optimization ....................................................................... 41 4.1 Introduction ........................................................................................................ 41 4.2 Experimental ...................................................................................................... 42 4.2.1 Material ....................................................................................................... 42 4.2.2 Cure cycle variations................................................................................... 43 4.2.3 Rheology ..................................................................................................... 44 4.2.4 Laminate characterization ........................................................................... 44 4.3 Results and discussion ........................................................................................ 44 4.3.1 Airbus cure cycle ........................................................................................ 44 4.3.2 Laminate characterization ........................................................................... 46 4.3.3 Optimized cure cycle .................................................................................. 50 4.4 Conclusions ........................................................................................................ 51 CHAPTER 5. Impregnation and compaction ............................................................. 52 5.1 Introduction ........................................................................................................ 52 5.2 Experimental ...................................................................................................... 53 5.2.1 Material ....................................................................................................... 53 5.2.2 Prepreg initial condition .............................................................................. 54 5.2.3 Test samples ................................................................................................ 55 5.3 Results ................................................................................................................ 56 5.3.1 Prepreg initial condition .............................................................................. 56 5.3.2 Impregnation during cure ............................................................................ 57 5.3.3 Compaction ................................................................................................. 61 5.3.4 Isothermal holds .......................................................................................... 62 5.3.5 Activation energy for impregnation ............................................................ 63 5.4 Discussion .......................................................................................................... 65 5.5 Conclusions ........................................................................................................ 69 CHAPTER 6. Moisture and pressure effects on void formation ................................ 72 6.1 Introduction ........................................................................................................ 72 6.2 Experimental procedure ..................................................................................... 73 6.3 Model framework ............................................................................................... 76 6.4 Determination of model parameters ................................................................... 78 6.4.1 TGA and titration ........................................................................................ 78 6.4.2 Resin solubility ........................................................................................... 79 6.4.3 Gelation ....................................................................................................... 80 6.5 Results ................................................................................................................ 82 6.5.1 Model prediction ......................................................................................... 82 6.5.2 Void content ................................................................................................ 84 6.5.3 Measured vs. predicted ............................................................................... 86 6.6 Discussion .......................................................................................................... 87 6.7 Conclusions ........................................................................................................ 92 viii CHAPTER 7. Out time effects on VBO prepreg and laminate properties ................. 95 7.1 Introduction ........................................................................................................ 95 7.2 Experimentation ................................................................................................. 96 7.2.1 Prepreg characterization .............................................................................. 96 7.2.2 Laminate fabrication and testing ................................................................. 98 7.3 Results and discussion ........................................................................................ 99 7.3.1 Prepreg characterization .............................................................................. 99 7.3.2 Laminate characterization ......................................................................... 106 7.4 Age determination and quality prediction ........................................................ 110 7.5 Conclusions ...................................................................................................... 111 CHAPTER 8. Prepreg age monitoring ..................................................................... 113 8.1 Introduction ...................................................................................................... 113 8.2 Experimental .................................................................................................... 114 8.2.1 Materials ................................................................................................... 114 8.2.2 DSC ........................................................................................................... 115 8.3 Results and discussion ...................................................................................... 116 8.3.1 Validity of the method .............................................................................. 120 8.3.2 Mechanism of ambient aging .................................................................... 121 8.3.3 Degree of cure ........................................................................................... 126 8.4 Conclusions ...................................................................................................... 129 CHAPTER 9. Adhesive out-gassing ......................................................................... 132 9.1 Introduction ...................................................................................................... 132 9.2 Experimental .................................................................................................... 134 9.2.1 Materials ................................................................................................... 134 9.2.2 Methods..................................................................................................... 134 9.3 Results and discussion ...................................................................................... 136 9.3.1 Adhesive foaming ..................................................................................... 136 9.3.2 TGA-FTIR data ......................................................................................... 137 9.3.3 Adhesive pre-treatment ............................................................................. 144 9.4 Conclusions ...................................................................................................... 146 CHAPTER 10. Conclusions and future work ........................................................... 148 10.1 Conclusions ...................................................................................................... 148 10.2 Recommendations for future work ................................................................... 150 10.3 Broader implications ........................................................................................ 151 Bibliography ................................................................................................................... 152 Bibliography (Alphabetical) ........................................................................................... 160 ix List of Tables Table 2-1. Rheological properties [36] ............................................................................. 20 Table 3-1. Mechanical test matrix (images courtesy Wyoming Test Fixtures) ................ 33 Table 3-2. Values for E 1mm ............................................................................................... 35 Table 3-3. Summary of mechanical test data .................................................................... 39 Table 4-1. First Dwell Variations ..................................................................................... 43 Table 5-1. Resin impregnation rates ................................................................................. 63 Table 6-1. Values for total weight loss in sample wt% during TGA and wt% moisture in the sample determined by Fischer Titration ................................................................. 79 Table 6-2. Void content (vol %) ....................................................................................... 85 Table 8-1. Minimum, maximum, and average difference in predicted age and actual age (in units of days) for the three prepreg systems examined ....................................... 120 Table 9-1. Experiment parameters .................................................................................. 136 Table 9-2. TGA weight loss data .................................................................................... 138 Table 9-3. Weight loss during temperature ramp and hold ............................................ 144 x List of Figures Figure 1-1. Voids in a composite laminate ......................................................................... 3 Figure 1-2. Void content as a function of processing pressure [8] ..................................... 4 Figure 1-3. Decrease in interlaminar shear strength as a function of increased void content [8] ........................................................................................................................... 5 Figure 1-4. Function of TLP (top) and VPP (bottom) prepreg during autoclave cure [16] 6 Figure 1-5. Design of EVaCs in VBO prepreg ................................................................... 7 Figure 1-6. Composite panels cured under vacuum pressure only with fully impregnated VPP (left) and partially impregnated TLP (right) [5] .................................... 8 Figure 1-7. Void content as a function of vacuum gauge pressure [17] ........................... 11 Figure 1-8. Vacuum bagging scheme for VBO cure [23] ................................................. 12 Figure 1-9. Sandwich structure consisting of 4-ply composite face sheets and a 3 pcf nomex/phenolic core with 1/8 in cell size. Left: panel cured in autoclave, showing core-crush. Right: panel with same specifications cured under vacuum .......................... 14 Figure 2-1. Schematic showing phase angle [36] ............................................................. 19 Figure 2-2. Generator electrode without diaphragm [37] ................................................. 21 Figure 2-3. Schematic of FTIR spectrometer [39] ............................................................ 24 Figure 2-4. FTIR instrumental process [39] ..................................................................... 25 Figure 3-1. Airbus recommended cure cycle .................................................................... 28 Figure 3-2. Interlaminar shear strength data for (A) room temperature dry testing, and (B) hot/wet testing............................................................................................................. 34 Figure 3-3. In-plane shear strength data for (A) room temperature dry testing, and (B) hot/wet testing ................................................................................................................... 35 Figure 3-4. Compression after impact data for (A) room temperature dry testing, and (B) hot/wet testing............................................................................................................. 36 Figure 3-5. Initial open-hole compression strength data, and (B) image showing delamination caused by water jet cutting .......................................................................... 37 xi Figure 3-6. Open-hole compression data for carbide drilled samples .............................. 38 Figure 4-1. Resin viscosity during Airbus cure cycle ....................................................... 45 Figure 4-2. Thickness measurements ................................................................................ 47 Figure 4-3. Void volume fractions .................................................................................... 48 Figure 4-4. Surface dryness for samples cured at 110 °C [51] ......................................... 49 Figure 4-5. Surface dryness for samples cured at 130 °C [51] ......................................... 49 Figure 4-6. Resin viscosity during optimized cure cycle .................................................. 50 Figure 5-1. Recommended cure cycle .............................................................................. 54 Figure 5-2. Prepreg initial condition showing engineered vacuum channels in (a) unidirectional and (b) 5-harness satin (5HS) material ...................................................... 56 Figure 5-3. Impregnation (void removal) as a function of cure time and temperature ..... 58 Figure 5-4. Unidirectional test samples showing prepreg impregnation during cure ....... 58 Figure 5-5. 5HS test samples showing prepreg impregnation during cure ....................... 59 Figure 5-6. Magnified intra-tow region of 5HS material, showing absence of fiber tow voids after 5730 s of processing ........................................................................................ 61 Figure 5-7. Thickness (compaction) in unidirectional and 5HS samples. Inset micrographs show consolidation of fibers ........................................................................ 62 Figure 5-8. Impregnation during isothermal holds for (a) unidirectional and (b) 5HS prepreg .............................................................................................................................. 63 Figure 5-9. Impregnation rates as a function of inverse temperature. Slopes correspond to activation energies for impregnation ............................................................................ 65 Figure 6-1. Vacuum bagging assembly ............................................................................ 74 Figure 6-2. Cure schedule ................................................................................................. 75 Figure 6-3. Moisture content as a function of relative humidity exposure with parabolic solubility fit ....................................................................................................................... 79 Figure 6-4. Gel time as a function of relative humidity exposure .................................... 81 xii Figure 6-5. Rheological data (G'-storage modulus, and G''-loss modulus) displaying resin gelation for a humidity exposure of 70% ................................................................. 81 Figure 6-6. Predicted void growth at 70% relative humidity............................................ 82 Figure 6-7. Predicted void diameter as a function of relative humidity ........................... 84 Figure 6-8. Micrographs of cured laminates. Scale bars are 1mm ................................... 85 Figure 6-9. Predicted and measured void volume fraction data ....................................... 87 Figure 6-10. Void pressure as a function of initial relative humidity exposure ............... 90 Figure 7-1. Schematic of characteristic tack test data showing the tack calculation method............................................................................................................................... 98 Figure 7-2. Temperature cure profile ................................................................................ 99 Figure 7-3. (A) First heat ramp showing the glass transition of the uncured system and the exothermic reaction peak, (B) Second heat ramp showing the glass transition temperature of the cured composite ................................................................................ 100 Figure 7-4. Changes in (A) glass transition temperature of the uncured system and (B) exothermic reaction peak, as out-time is increased ........................................................ 101 Figure 7-5. Changes in glass transition temperature as out-time increases .................... 102 Figure 7-6. Changes in DSC data as a function of out-time. (A) Enthalpy of reaction, (B) glass transition of the uncured system, (C) glass transition of the cured composite 103 Figure 7-7. Energy of separation per unit volume as a function of aging time .............. 104 Figure 7-8. Continuation of the tack stress-strain curve showing detachment of adjacent prepreg plies. Images show (1) compacted plies, (2) adhesion of plies in tension, (3) ply separation ............................................................................................... 105 Figure 7-9. Energy of separation calculated for the tensile portion of the tack test until full separation of prepreg plies ....................................................................................... 106 Figure 7-10. Ultrasound C-scans of cured laminates ...................................................... 107 Figure 7-11. Laminate thickness measurements ............................................................. 108 Figure 7-12. Imaged cross-sections showing laminate compaction and void content .... 109 Figure 7-13. ILSS values ................................................................................................ 110 xiii Figure 8-1. Linear trends between B-stage T g and prepreg out-time for (A) Prepreg 1, (B) Prepreg 2, and (C) Prepreg 3 .................................................................................... 117 Figure 8-2. Age predicted using linear fit equations plotted against actual known age for (A) Prepreg 1, (B) Prepreg 2, and (C) Prepreg 3....................................................... 119 Figure 8-3. Reversing heat flow signals, showing B-stage glass transition temperature results for three samples of fresh prepreg (Prepreg 1), used to validate the accuracy of the modulated differential scanning calorimetry (MDSC) method ................................ 121 Figure 8-4. Degree of cure as a function of ambient aging time for (A) Prepreg 1, (B) Prepreg 2, and (C) Prepreg 3 ........................................................................................... 127 Figure 8-5. Degree of cure as a function of B-stage T g for (A) Prepreg 1, (B) Prepreg 2, and (C) Prepreg 3 ........................................................................................................ 128 Figure 9-1. Composite sandwich panel. Exploded view (left) and completed panel (right) .............................................................................................................................. 132 Figure 9-2. Adhesive samples before (left) and after (right) vacuum cure ..................... 137 Figure 9-3. TGA-FTIR data from adhesive A. (a) Intensity of Gram-Schmidt profile, weight loss, and temperature vs. time. (b) 3D plot of FTIR spectra during testing ........ 139 Figure 9-4. TGA-FTIR data from adhesive B. (a) Intensity of Gram-Schmidt profile, weight loss, and temperature vs. time. (b) 3D plot of FTIR spectra during testing ........ 140 Figure 9-5. FTIR spectra of off-gassing components ..................................................... 142 Figure 9-6. Example of the polymerization process for a polyimide [102] .................... 143 Figure 9-7. TGA weight loss curves during isothermal pre-treatment at 110° C for (a) adhesive A and (b) adhesive B........................................................................................ 145 Figure 9-8. Viscosity during cure for (a) adhesive A and (b) adhesive B. Photographs show foaming .................................................................................................................. 146 xiv Abstract Composite parts for commercial aircraft are traditionally manufactured using high-pressure autoclave processing of prepregs (carbon fiber pre-impregnated with epoxy resin). In recent decades, however, the use of composite parts for aircraft has increased, and aircraft markets have grown, creating pressure to increase production rates. To meet the growing demand for composite aircraft parts and to allow for the production of large composite components (i.e. wings and fuselage) alternative processing methods will be required. There are several drawbacks to autoclave processing, including a large capital investment, long cycle time, high cost of the nitrogen gas used to pressurize the vessel, size limitations, and poor energy efficiency. New out-of-autoclave processing methods have been developed to address these issues. One such method is vacuum-bag-only (VBO) processing of prepregs, a technique which uses atmospheric pressure alone to consolidate parts. VBO processing presents a potential solution for the manufacture of larger parts at faster rates using conventional layup and placement tools. However, before VBO methods can be used on primary structure, the quality of VBO processed parts must be shown to be equivalent to that of autoclave cured parts. The elimination of high external pressures during the cure cycle removes safeguards in the manufacturing process, resulting in the need for strict protocols in the layup and cure of VBO parts. To assess the feasibility of VBO processing for aerospace components a systematic study of the effect of process parameters on the quality of VBO parts is essential. Specifically, the mechanisms of void formation and growth in prepreg-processed carbon fiber composites xv are not well understood. The purpose of this work is to examine the potential causes of voids, to develop a complete understanding of the mechanisms of void formation. This knowledge will aid in the production of higher quality parts, and help to determine the feasibility of low-pressure VBO processing for large-scale structural components. As a starting point, carbon fiber/epoxy test laminates were manufactured using vacuum bag only methods as well as traditional autoclave cure cycles. Cured laminates were tested using aerospace qualification standards. Tests were performed on dry laminates as well as laminates that had been hot/wet conditioned. Mechanical properties were shown to be equivalent in vacuum bag only and autoclave processed laminates, and values for all test panels and test conditions exceeded the required level for structural aerospace applications. Cure cycle optimization was carried out to further improve the properties of out- of-autoclave processed parts. Variations in hold time and temperature were investigated for the first temperature dwell of the cure cycle. Several test panels were fabricated with a range of processing times and temperatures. Cured laminates were characterized for compaction (thickness), void content, and surface finish. Resin rheological properties were also examined. Based on experimental results, an optimized cure cycle was developed for the material system studied. The mechanisms of impregnation and compaction in vacuum-bag-only prepreg materials were investigated to determine the influence of fiber architecture on flow and void removal in out-of-autoclave processing. Material microstructure and laminate thickness were tracked as a function of cure time and temperature for a unidirectional xvi material and a 5-harness satin woven prepreg featuring the same resin system. Isothermal resin flow was analyzed to determine the activation energy for resin impregnation in each prepreg, a quantitative measure of the influence of fabric type on flow behavior. Impregnation in the unidirectional material occurred early in the cure process, followed by additional ply compaction. In contrast, impregnation and compaction occurred on a longer time scale for the woven fabric. The impregnation, compaction, and air removal mechanisms observed for unidirectional and woven fabrics were different, indicating that cure cycles must be tailored to fiber architectures in prepregs, and possibly to part geometry. Void formation as a function of resin moisture content was studied to better understand and control process defects in composite parts made from prepreg. Uncured prepreg was conditioned at 70, 80 and 90% relative humidity and at 35°C. Conditioned prepreg was laid up into 16-ply laminates and cured using vacuum bag only processing, as well as partial vacuum and autoclave processing. Moisture uptake in the resin was measured using coulometric Fischer titration. Void content was measured by image analysis of polished sections of cured laminates. Void content increased substantially with increasing moisture content in vacuum bag only processed samples, and a strong pressure dependence was noted. Under autoclave cure conditions, void-free parts were produced even at high moisture levels. Experimental results were compared with trends predicted using a diffusion-based analytical model. Changes in vacuum bag only prepreg properties were tracked as a function of room temperature aging time (out-time). A modulated differential scanning calorimetry xvii method was used to characterize the prepreg, and changes in prepreg tack levels were examined using an energy of separation technique. Laminates were cured from prepreg at various levels of aging using both traditional autoclave processing and low-pressure VBO techniques. Cured laminates were examined using ultrasound scanning, mechanical properties were tested, and laminates were sectioned for investigation of microstructure. Laminate quality as a function of out-time was examined in terms of prepreg properties and manufacturing technique. The results of the preceding study indicated that a need exists for an accurate and convenient method to monitor the extent of prepreg aging as a function of out-time. For this reason, a method to track prepreg age was developed, involving measurement of changes in glass transition temperature as a function of room-temperature aging time. Samples from three out-of-autoclave prepreg systems were aged in ambient conditions and tested periodically using modulated differential scanning calorimetry. A linear increase in glass transition temperature with prepreg age was noted. In addition to monolithic parts and flat laminates, prepregs are used for face sheets of sandwich structures, where they are bonded to low-density cores of honeycomb or foam. To reduce processing time and cost, co-cure of composite face sheets and sandwich structure adhesives is desirable. VBO processing is attractive for this application, as it allows for the use of lighter and less costly core materials, eliminating the risk of core crush that can result from high autoclave pressures. Under vacuum, however, sandwich structure adhesives often foam as gas species evolve from solution. The first step in eliminating adhesive foaming is to identify the volatiles evolving from xviii adhesives during cure. An on-line coupled thermogravimetric analyzer-Fourier transform infrared spectrometer (TGA-FTIR) technique was employed to identify volatile components evolving during the cure of two polyimide film adhesives. Overall, the work presented here provides an analysis of the influence of various process parameters on voids, leading to an improved understanding of the mechanisms of void formation in VBO processed parts. The results of this investigation shed light on the cure protocols required for the production of high quality composite parts in the absence of autoclave pressures. 1 CHAPTER 1. Introduction 1.1 Motivation A paradigm shift is currently taking place in the aerospace industry, in which traditional metallic structures are being replaced by composite designs. Composites offer advantages in terms of weight savings, fuel efficiency, and resistance to corrosion and wear. Additionally, the demand for air travel is increasing rapidly, with projected demands for aircraft expected to quadruple in the coming decade. Composite parts, however, can be costly and time consuming to manufacture. Meeting the growing production demands with composite structures will require new and innovative manufacturing methods. Composite parts for aerospace applications are typically produced by lay-up using automated methods, followed by compaction and cure in a pressure vessel called an autoclave. Autoclave manufacturing, however, is inefficient and expensive. Drawbacks to autoclaves include the high capital investment required to obtain the equipment, the high cost of the nitrogen gas used to pressurize the vessel, poor energy efficiency, and long cycle times [1-5]. Autoclaves are also size-limiting, restricting the dimensions of the part that can be produced to the inner diameter of the vessel [4, 5]. It will be impossible to meet the growing demand for composite aircraft structures with autoclave cure methods. For this reason, the industry is looking to new out-of-autoclave (OOA) processing options. To reach the production rates necessary to replace metallic parts with composite components, OOA processing methods will be required. Moving processing out of the 2 autoclave will allow for the production of larger parts at faster rates. Eliminating the need for autoclaves will also reduce processing cost. However, while traditional autoclave processing involves high pressures that suppress formation of porosity, OOA methods supply a maximum pressure differential of 1 atm. Thus, implementation of low pressure processing methods requires a clear scientific understanding of the effects of process parameters on porosity and defects. Before aerospace structures can be produced out-of-autoclave, the ability to manufacture high quality parts at a large scale by OOA methods must be assessed. Critical to this analysis will be reaching an improved understanding of void formation mechanisms. Voids in composites are known to be highly detrimental to mechanical performance. With autoclave processing, high compaction pressures are used to hold voids in solution and facilitate compaction, but with OOA curing, this safeguard is eliminated. Thus, before OOA methods can be used reliably, the influence of low pressure processing on void formation and overall part quality must be examined. 1.2 Voids in composites The critical factor restricting composite part acceptance for aerospace applications is void content. Voids, or porosity, are open spaces within a composite part, which act as strength-limiting defects. A micrograph of a composite sample containing voids is presented in Figure 1-1. Typically a porosity level of less than 1-2% is required for aerospace structures. Parts with void content exceeding this limit are rejected, as voids adversely affect the mechanical performance of composite components. 3 Figure 1-1. Voids in a composite laminate Several studies have examined the influence of porosity on part quality for autoclave processed composites [6-13]. These studies have shown that reduced cure pressures lead to increased void content [6-9]. Additionally, voids have been shown to affect mechanical properties [6-9, 14, 15]. Interlaminar shear strength, a matrix driven property, is particularly sensitive to void content [7, 8]. In addition to static mechanical properties, voids negatively influence fatigue behavior, moisture absorption, and material toughness [7, 8, 15]. Olivier et al. investigated the influence of autoclave cure pressure on composite void content [8]. Their findings, including inset details regarding void distribution and size, are presented in Figure 1-2. These results indicate that autoclave cure pressures exceeding 0.7 MPa will result in void free parts, while lower cure pressures lead to the occurrence of ellipsoidal voids, similar to those shown in Figure 1-1. 4 Figure 1-2. Void content as a function of processing pressure [8] By implementing reduced cure pressures, Oliver et al. produced composite panels with varying void content. The mechanical properties of these samples were tested. Results for interlaminar shear strength (ILSS) for two different composites materials (denoted Composite A and Composite B) are presented in Figure 1-3. Also included in the graph in Figure 1-3 are additional data sets from the literature, showing a similar trend in reduced ILSS as a function of sample void content for other composite material systems. These results highlight the detrimental effect of voids on part quality. 5 Figure 1-3. Decrease in interlaminar shear strength as a function of increased void content [8] While it is clear that voids reduce the mechanical properties of composite parts, the main source of voids is unknown and debated. Historically, high autoclave pressures have been relied upon to suppress voids and facilitate compaction. With out-of-autoclave processing, however, the safeguards to void formation supplied by high cure pressures are removed. Thus, an understanding of the mechanisms of void formation is critical to successful defect reduction in low-pressure processed parts. 1.3 Vacuum bag only prepreg The key to production of void-free parts in the absence of autoclave pressures is the design of the prepreg materials used in out-of-autoclave processing. A new generation of prepregs has been introduced, specifically designed to facilitate air removal. Originally developed for the autoclave cure of thick parts, these materials contain engineered vacuum channels (EVaCs) [16]. These channels consist of areas of dry fiber created by partially impregnating fiber tows or sheets, thus providing open pathways through the 6 material. The porous nature of the prepreg provides pathways for entrapped air and other volatiles to escape upon application of vacuum. When vacuum is pulled, the porosity network in the prepreg becomes evacuated. As the resin is heated and begins to flow, it is pulled into the evacuated channels, resulting in a void free part [16]. The differences in the void r e moval pr oc e ss f or f ull y im pr e g na ted m a ter ials (r e fe rr e d to a s “v oid produc ing pr e pr e g” or V P P ), a nd p a rtiall y im pr e g n a ted ma ter ials c ontaining EV a C s (“ thi c k laminate pr e pr e g” or T L P ), du ring a standa rd a utocla ve c ur e are presented in Figure 1-4. Figure 1-4. Function of TLP (top) and VPP (bottom) prepreg during autoclave cure [16] Due to the high cost and energy consumption of autoclave processing, a method for composite cure under vacuum pressure only was developed based on TLP technology [5]. This technique is referred to as vacuum bag only, or VBO, processing. Prepregs formulated for VBO processing are porous TLP materials, which contain EVaCs to facilitate air removal under vacuum. VBO prepregs are primarily permeable in the in- 7 plane direction, resulting in the need for breathable edge dams to allow for air evacuation during cure [5]. Typically, VBO prepregs are partially impregnated from the top and bottom surfaces, leaving a dry fiber channel at the center of each prepreg ply. A schematic representation of the design of partially impregnated VBO material is presented in Figure 1-5, along with a scanning electron microscope image of an actual EVaC in a unidirectional prepreg ply. Figure 1-5. Design of EVaCs in VBO prepreg Past studies have shown that the presence of EVaCs allows for the cure of void- free parts out-of-autoclave, while fully impregnated prepregs lead to high void contents in the absence of autoclave pressures [5, 17]. Figure 1-6 shows two composite panels, cured using the same vacuum bag only cure procedure, with the same carbon fiber fabric and resin system [5]. The only difference in the two parts was the initial impregnation level of the prepreg. The panel on the left was cured with traditional, fully impregnated VPP material. The void content in this panel is greater than 5% [5]. The panel on the 8 right was cured with partially impregnated TLP material. For this panel, void content is less than 1% [5]. Figure 1-6. Composite panels cured under vacuum pressure only with fully impregnated VPP (left) and partially impregnated TLP (right) [5] In addition to the design of VBO prepregs, the prepregging technique itself is beneficial to void mitigation. Though there are several techniques available for prepreg processing, the two most pre va lent a re the “ solve nt- dip” a nd “ hot m e lt ” met hods [4]. Fully impregnated prepregs are typically produced using the solvent-dip method. In this technique, a low viscosity resin is achieved by addition of solvent. This solvent and resin mixture is then coated onto the fabric, fully wetting the fibers. To complete the process the solvent is removed by passing the prepreg through heated solvent recovery towers. Recovery is never fully efficient however, so some degree of solvent remains in solvent- dip processed prepregs [4]. Unrecovered solvent is likely to evolve during the cure process, leading to void formation in the absence of autoclave pressures. 9 To mitigate void formation due to residual solvent content, VBO prepregs are manufactured using the hot melt method. The hot melt process, in addition to being faster and less expensive, provides benefits in terms of void reduction. In hot melt processing the resin is mixed at high temperature to yield a low viscosity product, eliminating the need for added solvents [18]. The hot melt method is most effective in producing prepregs with a lower degree of impregnation, allowing for full consolidation during the cure process [4]. The resin systems implemented for VBO processing are addition-cured, meaning no gasses are evolved during the cure process [19]. Controlling the degree of impregnation and avoiding the addition of solvents and evolution of gasses leads to a prepreg well suited for out-of-autoclave consolidation and cure. The design features and manufacturing methods implemented in the production of VBO prepregs assist in the elimination of voids. However, while there have been many studies on void formation and mitigation for autoclave processed parts [9-12, 20], little is known about the translation of void formation mechanisms to VBO processing. 1.3.1 Sources of voids in VBO processed parts The mechanisms of void formation and growth in composite parts made from prepreg are not well understood. Historically, high autoclave pressures have been used to eliminate voids regardless of their source. High pressures hold moisture and volatiles in solution and facilitate compaction. Unfortunately the safeguards to void formation supplied by high pressures are removed in VBO processing [2, 4, 5, 16, 21]. Therefore, a major concern regarding VBO processing is that the reduced cure pressures will lead to parts with unacceptable porosity levels. 10 1.3.1.1 Monolithic parts There are generally thought to be three leading sources of porosity in composite parts. These sources include entrapped air, evolved gasses and volatiles, and insufficient resin flow. The composite manufacturing process involves the lay-up of multiple plies of material, followed by cure under applied temperature and pressure. Heat is required for resin flow and crosslinking, and pressure is necessary to consolidate plies and reduce void content [9]. Without proper control of process parameters during this lay-up and cure process, voids can result. Air is often entrapped between plies during composite layup. If this air is not removed prior to, or during cure, voids will remain in the cured composite. For VBO processed parts, common reasons for void formation from entrapped air are improper temperature cure cycles and insufficient vacuum level. Composite cure cycles typically involve two temperature holds. The first is at an intermediate temperature, and designed to allow for resin flow, compaction, and void removal. If the time or temperature of this hold is not carefully chosen, cured parts may contain defects. Additionally, for high quality VBO processed parts, vacuum level should exceed 50 mm Hg of absolute pressure [5, 22]. The influence of vacuum level on void content in a VBO prepreg system is presented in Figure 1-7. It is important to note that the data in Figure 1-7 reports gauge pressure (the difference in pressure from ambient barometric pressure) rather than absolute pressure (the difference in pressure from prefect vacuum) [5]. 11 Figure 1-7. Void content as a function of vacuum gauge pressure [17] For large or complex parts, room temperature debulk cycles (vacuum holds) may be required for air removal. Debulk cycles aid in porosity reduction, as vacuum holds allow additional time for removal of trapped air [5]. Sufficient edge breathing, and proper bagging, are also critical for air removal and porosity reduction. A schematic showing the proper vacuum bagging process, and detailing all required consumables for VBO cure, is presented in Figure 1-8. 12 Figure 1-8. Vacuum bagging scheme for VBO cure [23] A second source of voids is the evolution of volatiles during cure. VBO prepregs are formulated to have low volatile content. However, epoxy resin in prepreg will absorb moisture from the air [24, 25]. If this moisture is not removed prior to cure, or held in solution during processing, it will evolve, resulting in voids. Increased temperature during the cure cycle leads to a reduction in the solubility of dissolved species (increased likelihood of volatilization), which can result in void nucleation and growth [22]. Increased pressures increase solubility, but with VBO methods the maximum applied pressure is 1 atm, making void formation due to volatilization more difficult to control. Void formation can be reduced in VBO processing by the implementation of low temperature cure cycles, or the removal of dissolved species prior to cure. 13 The last main source of void formation in monolithic parts is insufficient resin flow. This is unlikely to be a problem with fresh prepreg materials. However, carbon fiber epoxy prepregs undergo chemical aging at room temperature, which influences resin viscosity [26-32]. To prevent this aging process, prepregs are stored frozen until used [26-32]. In the manufacture of large structures, however, lay-up can take several days. During this time, prepreg is exposed to ambient conditions, and flow properties are adversely affected. In over-aged prepregs, resin viscosity is high, resulting in insufficient resin flow and dry areas (voids). Thus, material age must be carefully monitored to ensure void formation does not occur as a result of reduced resin flow. 1.3.1.2 Sandwich Structures In addition to monolithic parts, composites are often utilized in sandwich structures. Sandwich panels consist of face sheets bonded with adhesive to a core material such as foam or honeycomb. Sandwich structures are attractive for weight- critical applications as they provide strength and stiffness, as well as weight reduction. To reduce processing time and cost, composite face sheets can be co-cured with film adhesives. Current co-cured sandwich panels, however, are overdesigned to withstand high autoclave pressures, as light-weight core materials are crushed by autoclave pressures. VBO processing is attractive for co-cure of sandwich panels, as light weight core materials can be utilized with vacuum cure. Figure 1-9 displays two sandwich structures laid up with the same composite prepreg and honeycomb core materials. The panel on the left was cured in an autoclave, and exhibits core-crush. The panels on the right were VBO cured. 14 Figure 1-9. Sandwich structure consisting of 4-ply composite face sheets and a 3 pcf nomex/phenolic core with 1/8 in cell size. Left: panel cured in autoclave, showing core-crush. Right: panel with same specifications cured under vacuum To eliminate core-crush and allow for optimization of weight-reduction with light-weight core materials, VBO processing of sandwich structures is desirable. Sandwich panels, however, are susceptible to the same void issues as monolithic parts, but also present additional challenges. The cellular core material in honeycomb sandwich panels results in a redistribution of applied pressure. This redistribution of forces can lead to pressure differences in different areas of the structure, influencing resin flow and void formation [22]. Not only must entrapped air be removed from composite face sheets of sandwich panels, the large quantities of air inside the core material must also escape the part. Additionally, out-gassing of film adhesives during co-cure can lead to adhesive foaming and voids in the bond area. Sandwich panels are a commonly used aerospace material, so understanding void formation in these structures is critical. Other researchers have focused on air removal, permeability, and pressure distribution during VBO cure of sandwich structures [1, 3, 33- 35]. In this work, the evolution of volatiles from film adhesives will be examined to 15 further improve the understanding of void formation in low-pressure processing of sandwich panels. 1.4 Scope of dissertation Moving composite processing out of the autoclave eliminates the need for high- pressure, low-efficiency equipment, resulting in faster production with reduced energy demands. Eliminating high autoclave pressures from the manufacturing process, however, introduces issues with regard to void formation and porosity. Voids in composites are detrimental to mechanical performance. With traditional processing methods, high autoclave pressures are sufficient to suppress void formation. Low pressure out-of-autoclave methods, however, do not have this safeguard, resulting in the need for a scientific understanding of the process parameters and manufacturing methods required to control and reduce defects in composite parts. The goal of this work is to develop an improved understanding of void formation and growth in low-pressure processed parts by systematically investigating various mechanisms of void formation. As a starting point for a void formation study, an initial qualification of VBO prepregs was performed, showing that under ideal processing conditions parts with autoclave-equivalent quality can be produced with vacuum bag only methods. Details of the qualification testing process, and mechanical property data, are presented in Chapter 3. In Chapter 4, cure cycle optimization is discussed as a means to further improve part quality. This study goes beyond mechanical property testing, focusing on resin flow characteristics and laminate microstructure. An improved understanding of resin 16 impregnation, material compaction, and air removal mechanisms in VBO prepregs is developed in Chapter 5. Subsequent chapters go on to discuss common void formation mechanisms, and examine the influence of various process parameters on void content in VBO processed parts. In Chapter 6, a study of the effect of dissolved moisture content is presented, along with a discussion of the influence of vacuum pressure on void content and morphology. The influence of out-time (room temperature aging) of prepreg on part quality is discussed in Chapter 7, with an examination of out-time effects on prepreg properties and cured laminate characteristics. Because of the degradation in laminate properties that occurs as a function of aging time, a method to monitor prepreg age was developed. This age monitoring method is presented in Chapter 8. The majority of the work presented here relates to void formation in flat, monolithic parts. The scope of the work is expanded in Chapter 9 to include defect control in composite sandwich structures. Under vacuum, sandwich structure adhesives often undergo foaming, leading to voids in the bond line. Out-gassing of sandwich structure adhesives is examined using a method of coupled TGA/FTIR analysis. Volatile components evolving during cure are identified and thermal pre-treatments to reduce foaming and improve part quality are discussed. As the aerospace industry moves toward the use of more composite parts, and the demand for air travel increases, new manufacturing methods will be required. The current method of compaction and cure in a high-pressure autoclave cannot meet the imposed limits of time, cost, or energy consumption. However, moving processing out-of- 17 autoclave requires an improved understanding of the influence of process parameters on void content. The objective of this work is to investigate various mechanisms of void formation in low-pressure processed parts, with the goal of improving part quality and controlling manufacturing defects. Three leading sources of void formation (entrapped air, evolved volatiles, and insufficient resin flow), are examined in this work, providing an improved understanding of the mechanisms of void formation and growth in VBO processed parts. Conclusions of this investigation are presented in Chapter 10, along with suggestions for future work. 18 CHAPTER 2. Background 2.1 Experimental methods To reach an improved understanding of void formation mechanisms in composite parts made from prepreg, multiple polymer characterization methods were employed. Descriptions of the background principles behind these methods are presented in the following sections. 2.1.1 Rheology Rheology is the science of the deformation and flow of material [36]. A rheometer is used to determine the viscosity, viscoelastic properties, and transient response of a material. The rheometer used in this work applies a controlled stress to a material, and measures a strain response. The information obtained from the rheometer is thus a stress- deformation relationship [36]. Under deformation, most materials display viscoelastic behavior, acting partially as a solid and partially as a liquid. The levels of viscous and elastic response vary based on stress, time, and temperature. A stress-controlled rheometer works by applying an oscillatory torque to a sample. The deformation is sinusoidal with a user defined amplitude and frequency. A displacement sensor measures the strain response of the material. The difference in the stress stimulus wave and the strain response wave is called the phase angle (Figure 2-1). 19 Figure 2-1. Schematic showing phase angle [36] A purely elastic material (ideal solid) has a phase angle of 0°, while a purely viscous material (ideal liquid) has a phase angle of 90° [36]. Most materials display both viscous and elastic behavior. Using the complex stress (stress values in a dynamic experiment are referred to as complex stresses) and strain signals, and the measured phase angle, several properties of the material can be calculated. The complex modulus, which is separated into storage (elastic) modulus and loss (viscous) modulus, and damping behavior are all calculated from stress, strain and phase angle. Complex viscosit y is dete rmine d b y incor po ra ti n g the osc il l a ti on fr e que n c y , ω i n ra d /sec . An y of these properties can be examined as a function of applied oscillation rate, time, or applied temperature. Such values are very useful in the study of the flow and cure behavior of polymers. Table 2-1 shows common rheological properties, and how they are calculated. 20 Table 2-1. Rheological properties [36] Parameter Definition Calculation Complex Modulus, G* Measure of the materials overall resistance to deformation G*= stre ss*/ stra in=G ’+ iG ’’ Storage Modulus, G’ Measure of the elasticity of the material – ability to store energy G’ = (str e ss*/ stra in)c os δ L oss M odulus , G’ ’ The ability of the material to dissipate energy G’ ’= (stre ss*/ stra in)sin δ Tanδ Measure of material damping Tan δ = G ’’ /G’ Complex viscosity, η* Resistance of the material to stress deformation (flow characteristics) η *= G*/ω 2.1.2 Coulometric Fischer titration Karl Fischer titration is a method for determining the water content in a sample. Titration utilizes the Bunsen Reaction between water, iodine, and sulfur dioxide to determine the amount of water, by weight, in a sample. In an alcoholic solvent, such as the methanol-based reagent used in this work, the reaction between iodine and water occurs in a stoichiometric 1:1 ratio [37]. Titration reagents are typically composed of iodine, sulfur dioxide, and imidazole, dissolved in an alcoholic solvent (here methanol). There are two types of Fischer titration; volumetric and coulometric. Coulometric titration is used when sample water content is low (1ppm – 5%), as was the case for the prepreg samples examined in this work [38]. A coulometric titration cell contains an anode and a cathode. In classical titration units, anode and cathode compartments are separated by a diaphragm. The titrator used in this work, however, did not contain a diaphragm. A schematic of the coulometric titration cell without a diaphragm is presented in Figure 2-2. 21 Figure 2-2. Generator electrode without diaphragm [37] During coluometric titration, iodine is generated from iodide at the anode through electrochemical oxidation, during which process iodide ions (negative) release electrons and form iodine, which reacts with water [37]. The reaction at the cathode is the reduction of hydrogen ions (positive) to hydrogen [37]. In a titration cell without a diaphragm, the generation of hydrogen gas at the cathode forms gas bubbles, which prevent iodine from reaching the cathode and being reduced to iodide [37]. Iodine is generated at the anode by means of applied current pulses. The amount of moisture in a sample is determined through measurement of the electrical current (in Coulombs) used to generate iodine [37]. In the anode reaction, shown in Equation 2-1, two electrons are released. The resulting iodine reacts 1:1 with water. It requires 96,485 C/mol to produce one mole of a chemical substance requiring one electron [37]. Thus, 192,970 C/mol are required to produce one mole of water during the Karl Fischer 22 reaction. The molecular weight of water is 18.015 g/mol, meaning that 10.712 C of current correspond to 1 mg of water [37]. (2-1) To test solid samples, such as prepreg, a drying oven accessory is utilized. The oven heats the sample, releasing moisture, which is transferred to the titration vessel via applied gas flow [38]. 2.1.3 Thermogravimetric analysis (TGA) Thermogravimetry is a thermal analysis technique used to measure volatile content and degradation behavior of materials. In TGA testing, a sample is heated in a furnace (typically under nitrogen purge) using a controlled time-temperature program. During testing, sample weight loss (weight percentage) is recorded as a function of time and temperature. As temperature is increased, volatile components and absorbed moisture evolve from the sample, causing a decrease in sample weight. TGA testing to high temperatures provides information on decomposition, corrosion kinetics and oxidation behavior. 2.1.4 Differential scanning calorimetry (DSC) Differential scanning calorimentry is another useful thermal analysis technique for characterizing polymers. In DSC testing, small material samples (typically 8-15 mg) are enclosed in disposable pans. DSC measures the difference in the amount of heat required to change the temperature of the sample and a reference (empty pan). The principle behind DSC is that when a material undergoes a thermal transition, more or less heat will need to flow to the sample to hold it at the same temperature as the reference. 23 By tracking heat flow to the sample, thermal transitions can be identified. If a sample undergoes an endothermic transition, more heat will need to flow to the sample then to the reference to hold both at the same temperature. For an exothermic transition, less heat will need to flow to the sample. During DSC testing, a controlled temperature profile is applied, and changes in heat flow are tracked. DSC measures the amount of heat absorbed or released during a thermal transition. DSC is often used with polymer samples to identify melting points, exothermic cure reactions, and glass transition temperature. 2.1.5 Fourier transform infrared spectroscopy (FTIR) Fourier transform infrared spectroscopy (FTIR) is used to identify material samples based upon their molecular composition. In IR spectroscopy, infrared radiation is passed through a sample. Some radiation is absorbed, and some is transmitted, creating a unique spectrum [39]. The absorption peaks in an FTIR spectrum represent the vibration frequencies of atomic bonds in the material [39]. A schematic of a typical FTIR instrument is presented in Figure 2-3. Radiation generated by the IR source is directed through an aperture to an interferometer. In the interferometer, a beamsplitter divides the IR beam into two optical beams [39]. One beam is reflected off of a fixed mirror, and the other is reflected off of a mirror that is allowed to move a short distance away from the beamsplitter and back. The beams recombine again at the beamsplitter. One beam has traveled a fixed distance, while the other has traveled a constantly shifting distance. Thus, the signal that exits the interferometer, referred to as an interferogram, is a result of the two optical beams interfering with one 24 another [39]. The interferogram has all infrared frequencies produced by the source encoded into it [39]. Because of this, the measurement of an interferogram results in simultaneous measurement across all frequencies. Figure 2-3. Schematic of FTIR spectrometer [39] During testing, the interferogram signal passes through the sample of interest, and is directed to a detector. The interferogram data is digitized and transferred to a computer. To decode the interferogram signal into individual frequencies, to produce the frequency spectrum required for material identification, a mathematical Fourier transform is used [39]. The fast Fourier transform (FFT) calculations are performed by computer, and a 25 material spectrum is output for further analysis. A schematic of this process is shown in Figure 2-4. Figure 2-4. FTIR instrumental process [39] In this work, FTIR is used in conjunction with thermogravimetric analysis. In coupled TGA-FTIR testing, volatile components evolving during TGA testing are transmitted, through a heated transfer line, to an FTIR cell, where the spectra of the vapor is collected. The coupled technique allows for the identification of off-gassing components. Spectra are identified through comparison to pre-existing vapor phase spectral libraries. 26 CHAPTER 3. Demonstration of property equivalence 3.1 Introduction Traditionally, composite parts for aerospace applications are manufactured using high pressure autoclave processing. The pressure supplied during the cure cycle holds voids in solution and facilitates compaction [2, 5]. As the industry moves toward the use of more composite components in commercial aircraft, however, autoclave processing will not be feasible. The major issues related to autoclave curing are size limitations, cycle times and cost. As the demand for air travel continues to increase, production rates will rise to levels unreachable with current autoclave technology. Additionally, the cost associated with autoclave cure is prohibitively high. A potential solution to these limitations is a new generation of prepregs formulated for out-of-autoclave, vacuum bag only (VBO) cure. VBO processing eliminates the need for a pressure vessel, thus removing size constraints [4, 5]. Additionally, VBO prepregs are formulated for cure at lower temperatures, with shorter cycle times. Removing the need for an autoclave also eliminates the capital investment required to obtain a pressure vessel, and the high cost nitrogen gas used to pressurize the vessel [4]. While VBO processing is an attractive alternative to autoclave curing, there are issues which remain to be addressed. With VBO processing, the maximum compaction pressure applied is 1 atm, a value 3-6 × lower than typical autoclave pressures [1, 2]. Before aircraft components can be produced using low pressure processing methods, the material and technique must be qualified, and be shown to produce parts equivalent in quality to autoclave manufacturing. 27 Previous work has shown that parts produced out-of-autoclave can achieve properties comparable to autoclave cured samples [5, 40]. However, each new material formulation must undergo a qualification process. To qualify a material and manufacturing method for use in aerospace applications, a series of panels are made and mechanical tests are performed. In this study, a series of tests dictated by Airbus was carried out on laminates manufactured both with traditional autoclave methods, and vacuum bag only cure procedures. The laminates manufactured for this study were small (approximately 305 mm × 305 mm) and flat, an ideal geometry for producing high quality parts. Cured laminates were subjected to a series of mechanical tests, both in the dry condition and after hot/wet exposure. Mechanical property data for VBO samples was compared to data for autoclave cured samples, as well as industry requirements set by Airbus. Results presented here demonstrate that under ideal manufacturing conditions, vacuum bag only processed parts display quality equivalent to that of autoclave manufactured laminates, and meet the mechanical requirements set by the aerospace industry. 3.2 Experimental 3.2.1 Material and cure conditions The material used in this study was a prepreg system specifically formulated for autoclave or VBO processing, consisting of a 2 × 2 twill carbon fiber fabric and a toughened epoxy resin (MTM44-1/CF5804A, Advanced Composites Group, UK). To manufacture test samples, prepreg plies were laid up, and then bagged according to the manuf a c tur e r’ s re c omm e nde d method. One set of laminate s wa s cur e d in a n a utocla ve 28 under vacuum with 5 atm external pressure, and another set in an oven under vacuum pressure only. All laminates were cured according to the temperature profile specified by Airbus (Figure 3-1). Figure 3-1. Airbus recommended cure cycle 3.2.2 Mechanical testing Cured laminates were tested using four standard composite test methods utilized in the qualification of a material for structural aerospace applications. Sample dimensions and lay-up configurations were dictated by test standards. All tests were performed at room temperature using dry samples, and three of the tests were repeated for hot/wet conditioned samples. Samples for hot/wet testing were conditioned in a temperature humidity chamber for 1000 h at 70° C and 80% relative humidity prior to testing. 29 3.2.2.1 Interlaminar shear strength The out of plane shear properties of a composite are measured using a three-point- bend test, also referred to as a short-beam shear or interlaminar shear test. Test samples rest on two supports and are loaded in compression by a central force. As the load is applied the sample bends, creating a shear force that causes the individual plies to slide against one another. This sliding motion results in a failure between the layers of the laminate, testing the strength of the inter-ply bond. The area between composite layers is resin-rich, so interlaminar shear failure is matrix-dominated. Interlaminar shear strength was tested in accordance with the American Society for Testing and Materials (ASTM) standard D2733 [41]. Test laminates were manufactured using an 8-ply quasi-isotropic layup. Test samples were cut to dimensions of 10 mm × 20 mm. During testing a compressive load was applied with a controlled displacement of 1 mm/min. Samples were loaded to failure, and interlaminar shear strength was calculated using Equation 3-1. (3-1) He re , τ is t he sh e a r str e n g th (MP a ), P is the maximum load before failure (N), w is the width of the test sample (mm) and t is the thickness (mm). 3.2.2.2 In-plane shear strength In-plane shear strength testing measures the resistance of the laminate to sliding in the in-plane direction. For this test the laminate is composed of composite layers placed at a 45° angle to the vertical axis. The angled specimen is tested in tension, resulting in a 30 shear force in the laminate. For the material to slide the fibers must break. In-plane shear is thus a fiber-dominated failure. In-plane shear was tested in accordance with Airbus Industry Test Method (AITM) 1-0002 [42]. Test samples consisted of 8 plies of alternating ±45° ply angles. Before testing, samples were cut to coupon dimensions of 230 mm × 25 mm. End tabs measuring 50 mm in length were then applied to each end of the test samples, resulting in a final test span of 130 mm. Samples were tested in tension at a controlled displacement of 5mm/min. In-plane shear strength was calculated using Equation 3-2. (3-2) He re , τ is t he sh e a r str e n g th (MP a ), P max is the maximum load reached before failure (N), w is the width of the sample (mm), and t is the thickness of the sample (mm). 3.2.2.3 Compression strength after impact To determine the effective toughness of the material, a compression after impact test was performed. Toughness is an important material property, which measures resistance to failure in a sample after an initial defect is introduced. In compression after impact testing, a series of test laminates is struck at various impact energies, until a dent depth of 1 mm is achieved. Multiple samples with a 1 mm dent depth are then compressed to failure, and the compression strength is recorded. Compression after impact testing was carried out in accordance with AITM 1- 0010 [43]. Samples were 16-ply laminates laid up in a symmetric, quasi-isotropic orientation. Sample dimensions were 100 mm × 150 mm. 31 For each sample type and condition 12 specimens were prepared. The first 9 test samples were impacted at different impact energies, in an effort to determine the energy required to produce a dent depth of 1 mm. The first sample was impacted at 50 J, and the dent depth corresponding to that energy was measured using a depth gauge. If the dent depth for that sample was greater than 1 mm, subsequent samples were impacted at 10, 15, 20, 25, and 40 J. If the dent depth at 50 J was measured to be less than 1 mm, subsequent samples were tested at 20, 25, 40, 60, and 70 J. In both cases, 3 additional samples were tested at an impact energy of 30 J. Using the dent depth measurements for each energy level, the impact energy required to produce a dent of 1 mm (E 1mm ) was determined through linear interpolation. The final 3 samples were tested at E 1mm . Samples impacted at E 1mm were then tested in compression. The crosshead speed during testing was 0.5 mm/min. Samples were tested to failure, and the load at failure was recorded. Equation 3-3 was used to calculate compression strength after impact. ( ) (3-3) Here, σ r (E 1mm ) (MPa) is the compressive strength after impact at an impact energy of E 1mm , P r (N) is the load at failure, w (mm) is the width of the sample, and t (mm) is the thickness of the sample. 3.2.2.4 Open-hole compression strength Open-hole compression strength provides information on how mechanical properties change when holes are introduced into a part for joining or fastening. Open hole compression testing was performed in accordance with AITM 1-0008 [44]. Test samples for open-hole compression testing were 8-ply quasi-isotropic laminates. Samples 32 were cut to dimensions of 76 mm × 32 mm using a water jet cutting technique. Holes with a diameter of 6.35 mm were introduced into the test samples using two methods, first water jet cutting and second drilling with a carbide drill bit. Samples were compressed to failure at a constant displacement of 0.75 mm/min. The load at failure was recorded and open-hole compression strength was calculated using Equation 3-4. (3-4) Here, σ coh (MPa) is the open-hole compression strength, P u (N) is the maximum load before failure, w (mm) is the width of the test sample, and t (mm) is the thickness of the test sample. Due to time and material constraints, open-hole compression testing was not performed for hot/wet conditioned samples. 3.2.2.5 Mechanical testing summary A summary of all tests performed, including images of the test fixtures used, is presented in Table 3-1. 33 Table 3-1. Mechanical test matrix (images courtesy Wyoming Test Fixtures) Property Condition Test Standard Fixture Interlaminar Shear Strength RT Dry & 70° C Wet ASTM D2733 In-plane Shear Strength RT Dry & 70° C Wet AITM 1-0002 Tabbed Tensile Test Compression Strength After Impact RT Dry & 70° C Wet AITM 1-0010 Open-hole Compression Strength RT Dry AITM 1-0008 Results for each test were compared with the requirements set by Airbus for a 2 × 2 twill carbon fiber prepreg cured at 180° C. These requirements were obtained using the Airbus Material Specification document AIMS05-01-013 [45]. 3.3 Results and discussion 3.3.1 Interlaminar shear strength Results of interlaminar shear strength testing are presented in Figure 3-2. The values reported are averages over 5 samples. Data is shown for both room temperature dry testing (Figure 3-2a) and testing after conditioning at 70° C and 80% relative humidity for 1000 h (Figure 3-2b). Reported along with test data are the required values for the material system used, as dictated by Airbus. 34 Figure 3-2. Interlaminar shear strength data for (A) room temperature dry testing, and (B) hot/wet testing These results show that for both the dry condition and the hot/wet condition, laminates cured using VBO methods show properties equivalent to or exceeding the properties of autoclave processed laminates. Both VBO and autoclave cured test samples also display properties exceeding the required values. 3.3.2 In-plane shear strength Data for in-plane shear strength is presented in Figure 3-3. Again, values are reported for samples tested in the room temperature dry condition as well as samples that were hot/wet conditioned, and the requirement set by Airbus is displayed. Data presented is an average over 5 samples. 35 Figure 3-3. In-plane shear strength data for (A) room temperature dry testing, and (B) hot/wet testing In this case, values for VBO cured samples are similar to results for autoclave cured laminates, and both VBO and autoclave processed parts meet the Airbus requirements. 3.3.3 Compression after impact Initial impact testing on both vacuum bag and autoclave processed laminates was performed to determine the energy required to produce a dent depth of 1 mm. Values for E 1mm , determined through linear interpolation, are presented in Table 3-2. Table 3-2. Values for E 1mm Compression After Impact E 1mm (J) VBO 60 Auto 57 Three samples were impacted at E 1mm for each test condition. Those samples were then compressed to failure. Data for compression strength after impact is presented in 36 Figure 3-4. No required value was listed for compression after impact testing in the Airbus Material Specification, but values for high-performance toughened epoxy systems are typically in the range of 240-275 MPa in the dry condition [46, 47]. Figure 3-4. Compression after impact data for (A) room temperature dry testing, and (B) hot/wet testing Results for compression strength after impact show that the material system examined in this study has properties similar to other commercially available epoxy systems. 3.3.4 Open-hole compression Initial samples for open-hole compression testing were machined to size, and drilled, using water jet cutting techniques. Data for these samples is presented in Figure 3-5. Results plotted are an average over 6 test samples. 37 Figure 3-5. Initial open-hole compression strength data, and (B) image showing delamination caused by water jet cutting Open- hole compression strength for these samples is well below the value required by Airbus. Close inspection of test samples revealed that considerable delamination occurred during water jet drilling of the test specimens. This delamination can be seen in the image in Figure 3-5b. To avoid such delamination, a strength-limiting defect, a new set of samples was drilled using a carbide drill bit. Drilling eliminated the delamination caused by water jet cutting. Data for samples machined with a carbide drill bit is presented in Figure 3-6. 38 Figure 3-6. Open-hole compression data for carbide drilled samples These results show that the delamination was in fact the cause of reduced mechanical properties. Samples drilled with a carbide bit are shown to meet the requirements set by Airbus. This fact underscores the importance of proper sample preparation and machining. Due to the time and material constraints, open-hole compression testing was not performed in the hot/wet condition. 3.3.5 Mechanical testing summary A summary of all mechanical test data is presented in Table 3-3, along with the calculated percent difference between VBO results and autoclave results. Percent difference was calculated using Equation 3-5. ( ⁄ ) (3-5) 39 Table 3-3. Summary of mechanical test data Test VBO (MPa) Auto (MPa) % Diff Meets Requirement? Interlaminar Shear Strength Dry 66 66 0 Yes Wet 64 61 4.8 Yes In-Plane Shear Strength Dry 100 103 -3.0 Yes Wet 96 102 -6.1 Yes Compression Strength After Impact Dry 278 274 1.4 NA Wet 223 236 -5.7 NA Open Hole Compression Strength Dry 320 324 -1.2 Yes Wet -- -- -- -- Examination of Table 3-3 shows that for all tests and all conditions presented here, both VBO processed samples and autoclave cured parts meet the requirements set by Airbus for use in aerospace applications. Additionally, VBO cured samples show property equivalence to high-pressure autoclave cured laminates to within 6% difference for all mechanical properties tested. 3.4 Conclusions Mechanical testing of laminates was performed in an effort to demonstrate property equivalence between low-pressure, VBO processed parts and traditional autoclave cured samples. Four standard material qualification tests were performed. Data for all tests showed equivalence between VBO and autoclave processed parts to within 6% difference, and both manufacturing methods yielded properties acceptable by Airbus standards. 40 As mentioned, test samples manufactured for this study were small and flat, a geometry which results in high-quality, void-free parts. Low-pressure VBO processing has been shown here to result in mechanical properties acceptable for structural aerospace applications under these processing conditions. As the industry moves to implement VBO processed parts in aircraft, however, additional issues will arise. Real aircraft parts contain complex curvatures and features such as stiffeners and ply drop-offs. Additionally, parts for aerospace applications will be much larger than the lab-scale test panels examined here. Complex geometries and large part sizes will make void elimination more difficult, particularly with low processing pressures. Thus, before VBO methods can be implemented for structural parts, a better understanding of void formation mechanisms in low-pressure processing must be achieved. 41 CHAPTER 4. Cure cycle optimization 4.1 Introduction A critical component of the composite manufacturing process is the material cure cycle. Fabrication of composite structures from prepreg involves placing several plies of material onto a tool, bagging the layered assembly, and curing the part. This cure process requires application of temperature and pressure. Increasing the temperature initially lowers the viscosity of the resin, allowing resin to flow. Under pressure, compaction is facilitated and voids are removed. Continued application of temperature then leads to the cross-linking and cure of the resin. The choice of an appropriate cure cycle for a composite material is essential to the production of high quality parts. Inadequate cure conditions can lead to the presence of voids in composites. For autoclave processing, there are three parameters that can be altered to optimize a composite cure schedule: time, temperature, and pressure [48]. Typically, increased pressures are used to reduce voids and improve part quality [6-10]. The prepreg examined in this study, however, was formulated for processing out of autoclave using a vacuum bag only (VBO) method. The maximum cure pressure for VBO processing is 1 atm, and successful VBO manufacturing is dependent on achieving high vacuum levels during cure. Thus, for optimization of a VBO cure cycle, full vacuum is used, and the parameters that can be varied are cure time and temperature. Thermoset composite parts made from prepreg are generally cured using a two- step process [48, 49]. First, the part is heated at a steady ramp rate to a moderate temperature. The temperature is then held for a period of time to allow for resin flow, 42 void elimination, and compaction [48, 49]. This initial temperature dwell is followed by a second, high temperature, dwell, during which cross-linking takes place and the cure is completed [48, 49]. It is important to choose the dwell time and temperature for these steps carefully, to achieve high quality parts while keeping the cycle time low. As compaction and void removal occur during the first temperature dwell, this is the critical step to optimize to ensure the production of high quality VBO composite parts. The goal of this work was to determine an optimum temperature and hold time for the first dwell of a VBO cure cycle, to reduce porosity and improve part quality. 4.2 Experimental Optimization of composite cure cycles is typically carried out through experimental trial and error [49]. An existing cure cycle, developed by Airbus for the material used in this study, was taken as a starting point for cure optimization. This cure pr of il e will be re fe r re d t o a s the “ Airbu s c ur e c y c le.” Resin viscosity during the Airbus cure cycle was measured, and the results were used to determine possible variations in dwell time and temperature for improvement of part quality. Test laminates were manufactured for each variation, and cured laminates were characterized. Through this laminate characterization, an optimized cure cycle was developed. 4.2.1 Material The resin system examined in this study was a toughened epoxy formulated for out-of-autoclave processing (MTM44-1, Advanced Composites Group, UK). Both neat resin film and prepreg were studied. The prepreg was composed of a 2 × 2 twill carbon fiber fabric (CF5804A). 43 4.2.2 Cure cycle variations The Airbus cure cycle, used as a starting point for this study, consisted of a temperature ramp to 65 °C, followed by a hold at 65 °C for 0.5 h. After the initial dwell the temperature was ramped again to 180 °C (the cure temperature for the material), and held for 2 h (see Figure 3-1). Examination of this cure cycle showed that a hold at 65 °C was insufficient to allow for resin flow and compaction. It was thus determined that a higher dwell temperature and longer dwell time should be implemented. The increased tempe ra tur e s e x a mi ne d i n thi s s tud y w e re 110 ° C , a nd 130 °C (the re sin manuf a c ture r’ s recommended dwell temperature). Dwell times were studied in half-hour increments up to a dwell time of 2 h. All sample variations are presented in Table 4-1. Table 4-1. First Dwell Variations Sample # Dwell Temperature (°C) Dwell Time (s) 1 110 0 2 1800 3 3600 4 5400 5 7200 6 130 0 7 1800 8 3600 9 5400 10 7200 To allow for completion of cure and cross-linking, all samples were post-cured for 2 h at a temperature of 180 °C. 44 4.2.3 Rheology Rheology testing was carried out on neat resin to examine viscosity during the cure cycle. Tests were performed using a parallel plate rheometer (TA Instruments AR2000), at a controlled frequency of 6.28 rad/s. Test samples were prepared by stacking several layers of resin film. 4.2.4 Laminate characterization To examine part quality as a function of cure time and temperature, test panels were manufactured. Samples were 8-ply quasi-isotropic laminates, measuring 127 mm × 127 mm . P re pr e g pli e s w e re laid up a nd v a c uum ba gge d a c c or din g to the manuf a c tur e r’ s recommended method. Cure time was varied by venting the vacuum bag at designated points during the first temperature dwell (see Table 4-1). Vented samples were post-cured for 2 h at 180° C. Cured laminates were characterized to track part quality as a function of dwell time and temperature. Thickness was measured to investigate ply compaction. Density was measured and used to calculate void content, in accordance with the American Society for Testing and Materials (ASTM) standard 2734-94 [50]. An additional set of test panels was fabricated at the Korean Aerospace University (KAU). These samples were used to characterize surface dryness. 4.3 Results and discussion 4.3.1 Airbus cure cycle The cure cycle in use at Airbus was taken as a starting point for the cure optimization study. To determine possible changes to implement for improved part 45 quality, resin viscosity during the Airbus cure cycle was examined. Rheology results for neat resin are presented in Figure 4-1. Figure 4-1. Resin viscosity during Airbus cure cycle It is clear from the viscosity profile in Figure 4-1, that a dwell at 65 °C is ineffective. The purpose of the first temperature dwell is to reach, and maintain, a low resin viscosity to allow for flow, compaction, and void removal. The temperature hold at 65 °C r e sult s in a pla tea u a t fa irl y hi g h viscosit y ( ≈ 10 3 Pas), indicating insufficient resin flow during the dwell. Subsequent increasing of temperature to the second dwell at 180 °C leads to an initial decrease in viscosity, followed by a steep increase to a cure plateau (reached approximately 10000 s into the cure cycle). To improve part quality, the first dwell must be increased to a higher temperature. The goal of a higher dwell temperature is to achieve lower resin viscosity for a longer period of time, allowing for optimized void removal and compaction. A longer initial 46 dwell time may also be required to eliminate porosity and achieve high part quality prior to resin gelation and finalization of the cure process. To arrive at an optimized cure schedule, dwell times of 0, 1800, 3600, 5400, and 7200 s were examined at temperatures of 110 °C and 130 °C. 4.3.2 Laminate characterization To investigate the influence of higher dwell temperatures and different dwell times, ten test samples were fabricated (see Table 4-1). Thickness measurements and void content calculations were performed for each sample. An additional set of samples was manufactured at the Korean Aerospace University to test surface dryness. 4.3.2.1 Thickness Cured laminates were measured for thickness. Thickness measurements were made at the center of each panel. Thickness is taken as an indication of laminate compaction. Results are presented in Figure 4-2. 47 Figure 4-2. Thickness measurements Thickness data shows very little variation for the panels held at 110 °C, suggesting that compaction is not facilitated at this temperature. Panels held at 130 °C, however, show decreasing thickness as a function of dwell time. The decrease in thickness continues until a dwell time of 5400 s, after which the value stabilizes. 4.3.2.2 Void content Void volume fraction was determined for each test panel. Results are presented in Figure 4-3. 48 Figure 4-3. Void volume fractions Consistent with thickness measurements, little change in void content is observed for the 110 °C samples. Void content for the 130 °C is shown to decrease with dwell time, again leveling off after 5400 s. 4.3.2.3 Surface dryness Additional test panels, fabricated and tested at KAU, were used to examine the quality of laminate surface finish as a function of dwell time and temperature. Insufficient resin flow during the cure cycle can result in resin-starved areas on the surface of the panel. Results for panels held at 110 °C for 1800, 5400, and 7200 s are presented in Figure 4-4. Results for panels held at 130 °C for 1800, 5400, and 7200 s are presented in Figure 4-5. 49 Figure 4-4. Surface dryness for samples cured at 110 °C [51] Figure 4-5. Surface dryness for samples cured at 130 °C [51] After a dwell time of 1800 s at 110 °C, the panel surface is shown to contain a high percentage of dry (unimpregnated) area (38.6%). Panels held at 110 °C for longer dwell times show a reduction in surface dryness, but after 7200 s, 11.4% dry area remains. Samples held at 130 °C, on the other hand, show 9.7% surface dryness after 1800 s, with full impregnation (0% surface dryness) after 5400 s. 50 4.3.3 Optimized cure cycle Thickness, void content, and surface dryness data show that 110 °C is an insufficient temperature to result in resin flow and compaction in the material system studied. Increasing the initial dwell temperature to 130 °C, however, results in void removal and ply compaction. Thickness, void content, and surface finish are optimized after a dwell time of 5400 s (1.5 h) at 130 °C, with little to no improvement in properties observed when the dwell times is increased to 7200 s. As it is desirable to keep cure cycle times low, an optimized dwell time and temperature for the material system examined in this study was determined to be a 1.5 h hold at 130 °C. This initial dwell is then followed by a 2 h hold at 180 °C to allow for completion of the cure process. Resin viscosity was examined for this optimized cure schedule. Rheology results are presented in Figure 4-6. Figure 4-6. Resin viscosity during optimized cure cycle 51 The optimized cure cycle, with an initial dwell at 130 °C, allows for a minimum resin viscosity to be reached during the first temperature dwell, and a low viscosity to be maintained to facilitate resin flow and compaction. A subsequent ramp to 180 °C results in a cure plateau approximately 12500 s into the cure cycle. 4.4 Conclusions Examination of a composite cure cycle in use by Airbus showed insufficient resin flow during the initial temperature dwell. To improve part quality higher dwell temperatures (110 °C and 130 °C) and longer dwell times were investigated. It was found through laminate characterization that a dwell time of 5400 s (1.5 h) at 130 °C resulted in optimized part quality. Examination of resin viscosity profiles during the Airbus cure cycle and the optimized cure cycle verified this result. Cure optimization is critical for composite manufacturing out-of-autoclave. With autoclave processing, increased pressures can be used to eliminate voids and facilitate compaction. With vacuum bag only processing, however, the maximum cure pressure is fixed at 1 atm. Because of this constraint, it becomes important to optimize the time and temperature parameters of the cure cycle to ensure the production of high quality parts. VBO manufacturing removes the safeguards associated with autoclave pressures, but when the proper cure procedures are implemented, low pressure processing can be used to manufacture void free composites. 52 CHAPTER 5. Impregnation and compaction 5.1 Introduction Prepreg materials formulated for out-of-autoclave curing (VBO prepregs) are partially-impregnated by design, a feature that is critical to the manufacture of high quality parts under vacuum pressure. The VBO prepregs contain areas of unimpregnated, dry fibers typically sandwiched between impregnated surface layers. These unimpregnated areas create a network of engineered vacuum channels (EVaCs) through the prepreg [17, 24, 52-54]. When vacuum is applied during cure, this network is evacuated, and air is removed from the material through edge breathing mechanisms [5, 16]. During the remainder of the cure cycle, dry areas in the prepreg must be fully impregnated with resin to produce a void-free finished part. While the function of engineered vacuum channels is critical to VBO processing, as discussed in Chapter 4, another crucial factor to the manufacture of high quality composites is an optimal cure cycle [49, 55-58]. Traditionally, thermoset composite cure cycles have been developed based on resin chemistry [48]. The most common cure profile for epoxy prepreg materials is a two-temperature dwell cycle. The first dwell occurs at an intermediate temperature (below the gel temperature of the resin), and is intended to facilitate resin flow, void removal, and compaction [17, 48, 49]. After the first dwell, the material is ramped to a higher temperature to cross-link the resin, which affects the final glass transition temperature of the cured system [48, 49, 58]. Cure cycles recommended by material manufacturers are resin-specific, and the same cure cycle is recommended regardless of the fiber architecture. In this work, we investigate resin flow 53 during cure for two VBO prepregs with the same resin system and different reinforcements to determine the influence of fiber architecture on impregnation, compaction, and porosity removal. The mechanisms of resin flow during a VBO composite cure cycle are difficult to observe, as cure takes place in a sealed vacuum assembly. Past works have used interrupted cure cycle methods to observe the evolution of microstructure during the cure of single VBO prepreg systems [52, 53]. Here, we employ a similar method to investigate the differences in time scale for impregnation and compaction for the same resin system in two different carbon fiber reinforcements; unidirectional tape and 5-harness satin wove n fa b ric . I n a ddit ion t o obse rva ti ons of impre g na ti on dur in g the ma nu f a c ture r’ s recommended cure cycle, we present data from isothermal holds. Resin flow rates vary with viscosity (a temperature-dependent parameter) [53, 58-60], so isothermal data can be used to determine the influence of processing temperature on flow behavior, and to calculate an activation energy for resin impregnation. Results are analyzed in the context of the differing mechanisms of impregnation, compaction, and void removal during VBO processing of unidirectional and woven prepregs. 5.2 Experimental 5.2.1 Material Two carbon fiber epoxy prepreg materials were examined in this study. Both materials featured the same toughened epoxy resin (CYCOM 5320, Cytec Industries, USA). The manuf a c tur e r ’s r e c omm e nd e d c ur e c y c le f or thi s re sin s y stem i nc ludes two dwells, and is presented in Figure 5-1 [23]. 54 Figure 5-1. Recommended cure cycle One prepeg was comprised of unidirectional fiber reinforcement, while the other prepreg consisted of a 5-harness satin (5HS) woven fabric. The unidirectional tape was composed of Thornel T40-800B intermediate modulus carbon fiber. The 5HS fabric was composed of Thornel T650/35 high modulus PAN-based carbon fiber, with 6K fiber tows (6000 filaments per strand). 5.2.2 Prepreg initial condition The initial microstructure and impregnation level of each prepreg material was examined with a scanning electron microscope (SEM). To prepare material samples for imaging, stacks of prepreg plies were laid up in a quasi-isotropic orientation and cured at room temperature. Room temperature cure was used to prevent any resin flow or compaction of plies. After cure, samples were sectioned with a diamond saw to produce a smooth surface suitable for SEM imaging (JEOL JSM-6610-SEM). 55 5.2.3 Test samples To track impregnation during cure, and determine an activation energy for resin impregnation, multiple test samples were prepared. All test samples measured 152 mm × 152 mm (6 in × 6 in). Unidirectional samples were 16 plies in thickness, and 5-harness satin samples were 8 plies thick. All samples were laid up in a symmetric, quasi-isotropic orientation. Samples were bagged in a standard vacuum bagging assembly and cured in an air circulating oven. No room temperature vacuum hold (debulk) was performed prior to cure. One set of samples for each material system was cured according to the cure schedule in Figure 5-1 (the manufacture r ’s r e c om mende d c ur e c y c l e ). A dd it ional sa mpl e s were cured under isothermal conditions at a range of dwell temperatures. The resin impregnation process was halted at various points by cutting vacuum pressure and venting the vacuum bag. By eliminating the applied pressure, resin flow (impregnation) and further laminate compaction was halted. This interrupted cure method was utilized to obtain “ fr e e z e f r a me” im a ge s of ma te ria l m icr ostr uc ture a t va rious point s during the c ur e cycle. After removing vacuum from test samples, resin cure was completed with a 2 hr dwell at higher temperature (177 ºC). After curing, test panels were sectioned and polished. Sections were prepared from the center of each laminate by mounting in epoxy and polishing. Void content and impregnation level were determined through image analysis using a light microscope (Keyence VH-Z100R). Material compaction was quantified through thickness measurements of cured panels. 56 5.3 Results 5.3.1 Prepreg initial condition The initial impregnation condition of the materials studied was documented through SEM images of prepreg in the as-received condition, as shown in Figure 5-2. Areas of unimpregnated fibers are clear in these images. These dry regions constitute engineered vacuum channels that allow for air removal and void elimination during VBO processing and cure. Figure 5-2. Prepreg initial condition showing engineered vacuum channels in (a) unidirectional and (b) 5-harness satin (5HS) material 57 5.3.2 Impregnation during cure Resin impregnation and void removal during the cure cycle was measured using image analysis of polished sections. To obtain sections at various points in the cure cycle, vacuum pressure was disconnected at specific times to halt the flow and compaction process. The cure was then completed in the absence of applied pressure to allow for cutting and polishing of samples. Microstructure was tracked through the end of the first temperature dwell. Figure 5-3 shows the progression of void removal as a function of time and temperature during the standard material cure cycle. In the unidirectional material, void removal is complete prior to the first temperature dwell (after 3300 s of cure). For the 5HS material, on the other hand, the void removal process transpires during the entirety of the first ramp and initial temperature dwell (after 7560 s of cure). This difference in process kinetics can be understood through examination of material microstructure. 58 Figure 5-3. Impregnation (void removal) as a function of cure time and temperature Micrographs tracking the void removal process for unidirectional samples are shown in Figure 5-4. The unidirectional tape contains only intra-ply porosity in unimpregnated fiber regions at the center of each ply. Resin quickly infiltrates these regions, resulting in a void-free panel. Figure 5-4. Unidirectional test samples showing prepreg impregnation during cure 59 Micrographs of 5HS test samples are presented in Figure 5-5. In contrast to the unidirectional prepreg, the woven fabric contains both porosity within fiber tows, and larger porosity regions between the intersections of tows, and between plies. Interlaminar porosity occurs in fabric prepreg due to the uneven surface morphology of woven tows, which results in tow nesting in adjacent plies. Figure 5-5 shows that the intra-tow regions (hereafter referred to as fiber tow voids) are impregnated with resin in early stages (by 3300 s), while the inter-tow and interlaminar regions are not fully infiltrated until the end of the first temperature dwell. Figure 5-5. 5HS test samples showing prepreg impregnation during cure 60 The images in Figures 5-4 and 5-5 provide insight into resin impregnation and the mechanisms of void removal in VBO prepregs. In unidirectional materials, the engineered vacuum channels (EVaCs), which facilitate air removal, are composed of an interconnected network of dry fiber regions. The 5HS material, on the other hand, contains dry areas not only within fiber tows, but also in the regions between woven fiber tows and between plies. Fiber voids in both materials seal off after 3300 s of cure. For the unidirectional samples, where these dry fiber regions are the only pathway for void removal, parts are void-free at this point in the cure cycle (3300 s). For the woven material, however, voids remain in the regions between fiber tows well beyond 3300 s into the cure cycle. The mechanisms of void removal in the 5HS prepreg were more carefully examined in additional microscope images. The light micrographs in Figure 5-5 provide a wide field of view and a global perspective of the material microstructure. The resolution of these images, however, combined with abrasive polishing methods, does not reveal sufficient detail. Figure 5-6 shows the light microscope image of a sample after 5760 s of processing, along with an ion-polished SEM sample showing that the regions within fiber tows were indeed void-free in later stages of material processing. 61 Figure 5-6. Magnified intra-tow region of 5HS material, showing absence of fiber tow voids after 5730 s of processing 5.3.3 Compaction The first temperature dwell in a thermoset cure cycle is designed not only to remove trapped air, but to facilitate compaction of plies. Compaction can be determined through thickness measurements of test samples. As the degree of compaction and fiber volume fraction of a sample increase, the thickness of the sample will decrease. Void removal is completed in unidirectional test samples early in the cure cycle (before the first dwell). However, test samples at later times in the cure process reveal that during the first dwell, additional laminate compaction occurs. This increase in compaction with dwell time is documented in Figure 5-7. Unlike unidirectional samples, void removal occurs throughout the first dwell in 5HS samples. For this reason, compaction in the 5HS system follows a trend very close to that of void content. Compaction data for both materials is presented in Figure 5-7. 62 Figure 5-7. Thickness (compaction) in unidirectional and 5HS samples. Inset micrographs show consolidation of fibers 5.3.4 Isothermal holds To examine the temperature dependence of resin flow in prepreg materials, isothermal holds were performed, and impregnation was tracked as a function of hold time. As before, vacuum was disconnected at the desired analysis point for each panel. Cure was completed by a ramp to 177º C, followed by a 2 h hold. Percent impregnation was determined from examination of polished sections, and defined as (5-1) Impregnation rates for unidirectional and 5HS prepregs are presented in Figure 5- 8. Note that isothermal holds for the woven material were performed at higher temperatures than unidirectional holds. Initially, isothermal data was recorded for the 5HS prepreg at 75, 90 and 105º C, but resin flow was minimal at these temperatures, even 63 after extended dwell times (>1.5 h). Thus, to generate useable data, the temperature was increased for the woven fabric. Figure 5-8. Impregnation during isothermal holds for (a) unidirectional and (b) 5HS prepreg 5.3.5 Activation energy for impregnation To calculate the activation energy for impregnation, the slope of the linear region of each isothermal plot was determined (dashed lines in Figure 5-8). Each slope was taken as a temperature-dependent resin impregnation rate, k. Rate values are presented in Table 5-1. Table 5-1. Resin impregnation rates Unidirectional 5-harness satin Temp (ºC) k (s -1 ) Temp (ºC) k (s -1 ) 75 0.00436 120 0.0125 90 0.00912 125 0.0172 105 0.01183 130 0.0201 From this data, the activation energy for impregnation was calculated using the following Arrhenius equation, 64 ⁄ (5-2) where A is a constant, is the activation energy, R is the ideal gas constant, and T is absolute temperature. Taking the natural log of both sides of the equation results in the following expression ( ) ( ) (5-3) Plotting ln(k) vs. ⁄ yields a straight line, the slope of which can be used to calculate activation energy, as (5-4) A plot of ln(k) vs. ⁄ for unidirectional and 5-harness satin prepreg materials is presented in Figure 5-9, along with the calculated slope of each line. Using Equation 5-4, the activation energy for unidirectional material was determined to be 36.64 kJ/mol, while the activation energy for resin impregnation in the woven fabric was approximately 1.7 × greater, at 63.51 kJ/mol. 65 Figure 5-9. Impregnation rates as a function of inverse temperature. Slopes correspond to activation energies for impregnation 5.4 Discussion The results presented above reveal that impregnation mechanisms depend strongly on fiber architecture, even when a single resin system is examined. Fiber orientation and tow configurations directly influence resin permeability and material porosity. Impregnation of unidirectional VBO prepreg requires only flow of resin through the dry regions between aligned fibers. This flow occurs relatively quickly, prior to the first dwell of the cure cycle. The 5HS material, however, exhibits both flow into dry regions within fiber tows and flow into the spaces between woven tows and plies. The dual nature of resin flow in woven fabrics, and the larger flow distances required for impregnation, result in longer times for full resin saturation, with impregnation and void elimination occurring throughout the entirety of the first dwell. 66 Differences in resin impregnation and compaction behavior can be attributed to fiber architecture. Structural differences in unidirectional materials vs. woven fabrics result in different EVaC configurations that facilitate removal of entrapped air. For unidirectional materials, where aligned fibers are partially impregnated on the top and bottom surface, EVaCs constitute a connected network of fiber porosity. For woven fabric, on the other hand, unimpregnated areas consist of fiber tow voids plus porosity regions between layers of prepreg and between fiber tows. Additionally, the bulk dimensions of unidirectional materials and woven fabrics differ significantly. For reference, the thickness of a single uncured prepreg ply for the materials studied here was 0.17 mm for the unidirectional tape, while the 5HS fabric was 0.62 mm thick. The greater thickness of the woven material, combined with the larger inter-tow and inter-ply network of porosity, requires larger flow distances for full impregnation and void elimination. The total thickness change for the two materials during cure is also quite different, as shown in Figure 5-7. The greater compaction required for the woven fiber bed and the higher initial void content contribute jointly to the longer impregnation times for 5HS material. The difference in impregnation time scale can be quantified by calculating an activation energy for impregnation in each material system. The activation energies for impregnation calculated here are consistent, in order of magnitude, with the activation energy for viscosity reported for a similar epoxy resin system [61]. This indicates that, as expected, a critical factor in composite impregnation is resin viscosity. Resin flow rates, however, depend not only on viscosity (a matrix property), but also on the permeability 67 of the fiber network [57, 59, 62-66], as evidenced by the differences in activation energy reported here. For the same resin system, the activation energy for impregnation in a woven reinforcement is 1.7 × greater than the activation energy for impregnation in a unidirectional prepreg. Thus, barriers to resin flow are greater in more complex woven reinforcements. Note that in this work, no vacuum debulk was performed on samples prior to cure. I n fa c t, t he m a nufa c ture r ’ s re c omm e nde d c ur e c y c l e f or these m a ter ials doe s not include debulking when samples are flat, or lightly contoured [23]. The addition of a room- temperature vacuum hold, however, alters the porosity distribution in a woven fabric. Past studies have shown that extended hold time under vacuum eliminates and/or reduces large porous regions between plies and between tows [53, 54]. However, debulk cycles do not remove the dry areas within fiber-tows; eliminating such voids requires increased temperature (decreased resin viscosity) to facilitate flow into small spaces between fibers. Thus, when a room-temperature vacuum hold is performed prior to the cure cycle, EVaCs in woven fabrics will consist primarily of dry regions within fiber tows, and function much like the EVaCs observed in the unidirectional samples analyzed here. Impregnation during cure, therefore, will occur more quickly if an initial debulk is performed. For large or complex parts, trade-offs will be required, as vacuum debulks shorten the required cure time, but add additional time prior to the start of the temperature cycle. These factors dictate that material cure cycles must be tailored to reinforcement architecture, as well as to resin flow, cure characteristics, and part geometry. In particular, for a two-dwell cure schedule, the first dwell temperature and duration must be 68 specific to the fabric architecture. Ideally, the first dwell temperature should be specified to allow for flow through the fiber network (without exceeding the gel temperature of the resin), and sufficiently long to achieve full saturation. However, longer cure times increase manufacturing cost and may not improve final part quality for all material types [49]. Cure cycles recommended by resin formulators are not always suitable for every application [56, 58, 67]. For example, the unidirectional prepreg in the present study can be cured using a shorter cycle time and yield high-quality parts. The 5HS woven material, however, requires the full recommended cycle time to achieve low porosity, and may require additional cure time and/or debulk cycles, for large or contoured parts or complex geometries. While the different impregnation rates observed for unidirectional and woven material is not surprising because of inherent differences in the structure and permeability, different fabric architectures (weaves) are expected to display different impregnation rates as well. The permeability of a fabric depends on the tortuosity of the flow pathways and gap sizes, which vary substantially for common fabric types, such as plain weave, twill, satin weave, and non-crimp fabrics. Permeability differences in turn affect impregnation kinetics. For example, Saunders et al. studied the compaction and impregnation behavior of different glass fiber fabrics in a wet layup compression molding experiment [68], and showed that different fabric weaves displayed markedly different compaction and impregnation behavior. Differences were attributed to the varying tortuosity of flow paths in the fabrics studied [68]. In related work, Centea et al. examined compaction in plain-weave and 8HS out-of-autoclave prepregs, reporting 69 similar differences in compaction and impregnation behavior due to fabric areal weight and weave type [54]. These data and the results reported here highlight the need to optimize cure cycles for specific reinforcement architectures. 5.5 Conclusions We have investigated resin impregnation, void removal, and ply compaction in partially-impregnated VBO prepreg materials with unidirectional material and woven fabric. Test samples sectioned at different points during the cure cycle revealed a difference in the rate of impregnation for the two materials, as well as differences in compaction behavior and the mechanisms of void removal. Impregnation of unidirectional materials involves only resin flow between aligned fibers, while impregnation of woven fabric involves dual-scale flow. Dry fiber areas in unidirectional materials were impregnated early in the cure cycle, prior to the first dwell, resulting in void free samples. Impregnation and void removal in the 5HS woven fabric, however, required the entirety of the first temperature ramp and dwell. While fiber tow voids in the 5HS fabric were removed early in the cure cycle, interlaminar voids and void area between fiber tows remained. This observation suggests that the nature of air entrapment and void removal differs in unidirectional and woven materials. Measurement of changes in sample thickness during cure provided insight into the compaction behavior of the materials studied. Following void removal in unidirectional samples, slight additional compaction occurred. The total thickness change for the 16-ply unidirectional laminates throughout the monitored region of the cure process was 0.19 mm. Thickness changes for the 5HS material were much greater due to the woven nature 70 of the fabric. For the 8-ply samples examined in this work, the total thickness change was 1.58 mm. The large initial thickness and void content of the woven prepreg contribute to the long times required for impregnation and compaction in 5HS samples. To quantify the difference in time scale for impregnation in unidirectional and woven materials, isothermal holds were performed, and an activation energy for resin flow was calculated. The activation energy for impregnation via resin flow was 1.7 × greater for the woven fabric than for unidirectional material. Both materials contained the same resin system, and thus the findings highlight the influence of fiber architecture on flow behavior. The mechanisms of air removal (and entrapment) also differ in woven fabrics and unidirectional prepregs. In unidirectional material, air removal occurs through direct pathways between fibers, a process that is completed early in the cure cycle. For woven prepreg, in the absence of an intial debulk, void areas exist not only within fiber tows, but also in the regions between plies and between tows. Small-scale fiber tow porosity is removed early in the cure process, while porosity in larger interlaminar and inter-tow regions require longer times for impregnation. The results presented here indicate that cure cycles should be developed to reflect not only resin chemistry, but also fiber architecture. The goal of cure cycle optimization is to ensure high -quality parts (full impregnation, low porosity, high compaction) with a minimum cycle time for a given resin and reinforcement architecture. To optimize cure cycles efficiently, analytical tools are required to predict impregnation behavior, air 71 removal, and compaction in VBO prepregs. The work presented here provides insights and a foundation for development of such tools. 72 CHAPTER 6. Moisture and pressure effects on void formation 6.1 Introduction Voids in composite structures are known to severely degrade mechanical performance [6-9, 14, 15]. Multiple studies have addressed the causes of void formation and mitigation measures for parts produced by autoclave processing of prepregs [6-13]. However, a new generation of prepregs have been introduced that do not require autoclave processing, known as out-of-autoclave (OOA) or vacuum bag only (VBO) prepregs. These prepregs are formulated for processing without autoclaves at much lower pressures (3-6), a factor that has great appeal in the aerospace industry. However, the introduction of this new class of VBO prepregs raises an important question about the extent to which our understanding of void formation mechanisms and mitigation measures will translate to VBO processing. Furthermore, there is not universal agreement regarding the source of voids in prepreg processed parts, and part manufacturers typically rely on high autoclave pressures to suppress voids. A strong and growing need exists, particularly in the aerospace industry, to find lower cost alternatives to autoclave processing, such as vacuum assisted resin transfer molding (VARTM), pultrusion, and in the present case, vacuum bag only (VBO) processing of prepregs. The appeal of VBO processing over autoclave processing stems from the lower capital investment, elimination of the need for costly nitrogen gas, greater energy efficiency, and reduction of size constraints [1-5]. However, before VBO processing can be utilized on structural components, part quality must be equivalent to parts produced by autoclave processing. Because of the high pressures applied by the 73 autoclave during the prepreg cure cycle, species dissolved in resins generally remain in solution, and void-free parts can be produced consistently. However, a vacuum bag assembly involves a relatively small pressure difference, and thus production of void-free parts presents a greater challenge. The first step in addressing the challenge is to reach a deeper and improved understanding of the mechanisms by which voids form. The mechanisms of void formation and growth in laminates produced from prepregs are not well-understood, and opposing views are held regarding the primary cause(s) of voids. Some researchers contend that moisture dissolved in the resin is the primary source of voids [20], while others claim that entrapped air and volatiles are the leading causes [1, 4]. Still others suggest that all three factors play a role [6-8, 11-15], although there is no consensus regarding the relative impact of each factor. However, prepregs intended for OOA processing are designed and formulated to suppress void formation from volatiles and entrapped air. For example, VBO prepregs are manufactured by a hot-melt process [4], and thus contain negligible solvent content [11, 18]. Furthermore, VBO prepregs are designed with engineered vacuum channels to facilitate air removal [2-5, 16]. Thus, in the present work, the effect of moisture content on void formation was examined to begin to clarify the causes of voids in composite laminates produced from prepregs. 6.2 Experimental procedure The material used in this study was a carbon fiber/epoxy prepreg designed for VBO processing (MTM44-1/CF5804A, Advanced Composites Group, UK), featuring a woven 2 2 twill fabric (6k Tenax HTA fiber). Control materials were fabricated by 74 laying up 16-ply quasi-isotropic laminates 203 292 mm in the as-received condition, followed by curing in an autoclave and in atmosphere, using a standard vacuum bag assembly. To determine the effects of moisture on void formation, prepreg plies were humidity-conditioned and laid up and cured using identical processing conditions. Humidity conditioning consisted of exposure for 24 h at 70%, 80% and 90% relative humidity and at 35° C. Control panels were also fabricated from plies conditioned in an oven at 35 °C with no added humidity (RH = 30 ± 5%) to eliminate the possibility of heat effects as a source of voids. After conditioning, plies were laid up in a [0/±45/90] 2s orientation and vacuum bagged as shown in Figure 6-1. Figure 6-1. Vacuum bagging assembly The bagged assembly was then cured according to the temperature profile in Figure 6-2. One set of laminates was cured under vacuum with an applied autoclave pressure of 5 atm, and one set under vacuum only. Compaction pressures were defined as the total pressure difference between the pressure within the vacuum bag and the pressure in the surrounding environment. The pressure for the vacuum bag only case was taken to be 1 atm, while the pressure in an autoclave (pressurized to 5 atm) was assumed to be a total of 6 atm. 75 Figure 6-2. Cure schedule Image analysis was performed on polished sections of cured laminates to determine void volume fraction [6, 8, 9, 14]. Sections were prepared from the center of cured laminates (where void content is likely to be largest) and examined using a digital stereo microscope (Keyence VHX-600). Void volume fraction was determined by measuring the void area in the imaged cross section. The final value for volume percent of voids in each laminate was taken as an average from four samples, with each sample sectioned and polished four times to obtain a representative measurement. The thickness of the laminate at the center of each panel was also measured. Neat resin films were humidity-conditioned and subsequently measured for weight loss by thermogravimetric analysis (TGA Q5000, TA Instruments) using a ramp rate of 15º C/min. This weight loss was correlated with resin moisture content measured 76 by Fischer titration using a coulmetric titrator (Mettler Toledo C20 with D0308 drying oven). Experimental measurements of void content were compared to analytical predictions obtained using a diffusion-based void model, described in the following section. 6.3 Model framework An analytical model was developed and employed to predict void formation as a function of resin moisture content. The model was based upon previous analysis of diffusion-based void growth by Kardos et al. [20]. Void growth during a thermoset cure process is time dependent, making diffusion-based analysis a natural choice. Because of the nature of VBO prepregs moisture in the resin was examined as a leading cause of voids. The theoretical basis for the model stems from the assumption that voids grow via diffusion of water from the surrounding resin. The driving forces for this diffusion process are temperature and pressure, and diffusion can favor either void growth or dissolution, depending on the solubility of moisture in the resin and the concentration gradient [8, 12, 13]. Void growth will occur when the pressure within a void exceeds the hydrostatic pressure in the surrounding resin. A void containing air will collapse under applied pressures, but when a void contains water, the water vapor pressure will increase exponentially with temperature, causing the void to stabilize and grow [7, 10, 11, 20]. Additionally, the diffusion coefficient of moisture in the resin will increase exponentially with temperature, resulting in accelerated void growth as temperature increases. Higher 77 temperatures and lower applied pressures facilitate void growth, as both factors accelerate moisture diffusion through the resin. To simplify the analysis, void growth was assumed to occur in a pseudo- homogeneous medium [20]. Because of this assumption, the situation resembles that of mass diffusion bubble-growth, a phenomenon that is well-understood [12, 13, 69, 70]. Details of the assumptions of the analytical approach have been presented elsewhere [20]. The governing equations for the model define the void diameter d (mm) [7, 10, 20, 69] and the growth driving force [7, 10, 20]: √ (6-1) (6-2) In these equations, D (mm 2 /h) is the diffusion coefficient of water in the resin, t (h) is time, C bulk (g/mm 3 ) is the concentration of water in the bulk resin, C void (g/mm 3 ) is the concentration of water at the surface of the void, and g (g/mm 3 ) is the gas density (in this case water vapor density). Examination of Equation 6-2 shows that void growth will occur only if C void < C bulk ,. For this reason, the time and temperature for which C void = C bulk is used as a starting point for the analytical model [8, 20]. The temperature-dependent diffusion coefficient of water used in the model calculations is given below [20]. ( ) (6-3) 78 Although the specific diffusivity for the selected prepreg resin is unknown, the value above (also used by Kardos et al. for a comparable epoxy system) is similar in order of magnitude to the diffusivity of water for a range of uncured epoxy resins [1, 71]. The gas density within the void was taken to be the density of a pure water void, defined as [20] (6-4) where M H20 (g/mol) is the molecular weight of water, p (atm) is the total pressure in the resin, which is assumed to follow the applied pressure during the cure cycle, and R (mm 3 /molK) is the gas constant. The concentration of water at the surface of the void depends on temperature and pressure and is given by [10, 20] ( ) (6-5) The final input parameter, C bulk was determined experimentally. 6.4 Determination of model parameters 6.4.1 TGA and titration To verify the validity of a moisture-based model, results of thermogravimetric analysis were compared to resin moisture content measured by Fischer titration. Within the margin of error, the values for wt% moisture in the resin were equivalent to the values for total sample weight loss during TGA (Table 6-1). The equivalence supports the assumption that escaped volatiles did not contribute appreciably to void growth. 79 Table 6-1. Values for total weight loss in sample wt% during TGA and wt% moisture in the sample determined by Fischer Titration Total Wt% Loss Wt% Moisture As received 0.26 ± 0.04 0.24 ± 0.03 35°C 70%RH 0.61 ± 0.05 0.64 ± 0.03 35°C 80%RH 0.78 ± 0.01 0.77 ± 0.04 35°C 90%RH 0.95 ± 0.11 0.99 ± 0.05 6.4.2 Resin solubility The value of C bulk , the final input parameter required for the analytical model, was derived from resin solubility data. Moisture content in resin samples conditioned at selected relative humidity values was measured, and a parabolic solubility curve was fit to the data. Figure 6-3. Moisture content as a function of relative humidity exposure with parabolic solubility fit 80 The parameter S in Figure 6-3 is the coefficient of resin solubility obtained from the parabolic fit. Multiplying by the initial relative humidity 2 o RH yields the solubility of the resin (wt% moisture / unit wt% resin). The coefficient for resin moisture content is defined as [20] (6-6) where r (g/mm 3 ) is the density of uncured resin (1.25 10 -3 for the selected system). Substituting the relevant values yields the equation for C bulk ( ) (6-7) 6.4.3 Gelation The resin gel point was determined to define an end-point for the analytical model. Resin flow ceases after gelation, and any voids remaining in the matrix are thus trapped [4, 10, 12]. No void growth or dissolution occurs after the resin has gelled, making gelation an effective stopping point for a void formation model. To determine the gel point of the resin (MTM44-1), rheometry was performed on conditioned resin using a parallel plate rheometer (TA Instruments AR2000). Humidity conditioning altered the resin gel point, which showed a second order dependence on relative humidity. To increase the accuracy of the analytical model, this dependence was incorporated into the analysis. Figure 6-4 shows the decrease in gel time with relative humidity exposure, while Figure 6-5 shows characteristic rheology data for a humidity of 70%. 81 Figure 6-4. Gel time as a function of relative humidity exposure Figure 6-5. Rheological data (G'-storage modulus, and G''-loss modulus) displaying resin gelation for a humidity exposure of 70% As shown in Figure 6-5, the gel point was taken as the time when the curves of storage modulus (G´) and loss modulus (G ˝) intersect [72]. Using the second order fit to 82 the gel point-humidity plot (shown in Figure 6-4), the analytical model for void formation was truncated at the appropriate time for each humidity level to provide effective endpoints for void growth. 6.5 Results 6.5.1 Model prediction Using the parameters outlined above, model predictions for diffusion-based void growth were obtained for the designated cure schedule at applied pressures of 5 atm, 1 atm, and 0.84 atm. To illustrate the temperature and pressure dependence of the model, results for time dependent void growth at the three pressure levels examined are presented in Figure 6-6 for a relative humidity of 70%. Figure 6-6. Predicted void growth at 70% relative humidity 83 Keeping in mind the governing equations of the model, several interesting features can be noted on the plot in Figure 6-6. First, the starting point for void formation can be seen to shift earlier in time with decreasing pressure. This pressure dependence is a result of the fact that void growth does not being until C void <C bulk . This condition is met on a shorter time scale for lower applied pressures and never met under autoclave cure conditions (evidenced by the absence of void growth for the case of autoclave processing). As expected from Equations 6-1 and 6-2 an exponential growth is exhibited during the initial stage of the cure when pressure is held constant and temperature is increasing, while a square root dependence is predicted during the portion of the cure where both temperature and pressure are held constant [20]. Lastly, the final value for void diameter can be seen to increase with decreasing applied pressure for a constant humidity level. Model predictions for void diameter after exposure to various humidities are presented in Figure 6-7 for the three pressure levels examined. An exponential increase in void size with increasing moisture content is predicted for both VBO conditions. In contrast no void formation is predicted for the case of autoclave manufacturing, again due to the fact that the C void < C bulk condition is never met with autoclave pressures. 84 Figure 6-7. Predicted void diameter as a function of relative humidity 6.5.2 Void content To determine the accuracy of the model-based predictions, a series of experiments was performed. Panels were fabricated using prepreg conditioned at different relative humidities, then cured using VBO and autoclave techniques. Polished sections were then prepared and examined to determine void content. Characteristic images for each panel are presented in Figure 6-8. Numerical values for void volume fraction appear in Table 6- 2. 85 Figure 6-8. Micrographs of cured laminates. Scale bars are 1mm Table 6-2. Void content (vol %) Autoclave Full Vacuum 0.84 atm Heat Only <.1 <.1 .18 ± .10 As Received <.1 <.1 1.07 ± .22 70% RH <.1 .08 ± .08 4.10 ± .41 80% RH <.1 1.00 ±.29 4.37 ± .08 90% RH <.1 2.62 ± .48 4.23 ± .35 Samples cured under autoclave pressures can be seen to be void-free, as predicted in model calculations. VBO samples manufactured under full vacuum display an exponential increase in void content with increasing moisture levels, again as predicted. VBO samples cured under partial vacuum, however, display a much less straightforward response to increasing moisture levels. Additionally, the morphology of voids is 86 noticeably different for the two VBO cure conditions. With full vacuum, an increasing number of large voids (0.5-2mm in length) can be seen for increasing moisture levels. Void formation under partial vacuum, however, takes on a much different formation. L a r g e voids a re pr e se nt, but pre vious l y unse e n ‘p e ppe r’ voids a lso a ppe a r. This re sult suggests that a secondary void formation mechanism is at work when a partial vacuum is used to consolidate the laminate. 6.5.3 Measured vs. predicted To compare the model prediction to the experimentally determined void content, the predicted void diameters were converted to void volume fractions. By modifying Equation 6-1, the following expression for void volume fraction was derived [10] [ ( ) ] (6-8) where V m (mm 3 ) is the unit matrix volume, which is used to scale the model results. This unit matrix volume represents the volume of resin required to produce an observed void volume fraction with a single void of a given diameter [10]. The value of V m is expected to remain constant for a given resin system [10]. Using predicted void diameter values combined with measured values of void volume fraction (V v ), a characteristic V m value for the resin was determined. The plot in Figure 6-9 shows the measured void fractions plotted as a function of initial humidity values, as well as the dependence predicted from the diffusion model. 87 Figure 6-9. Predicted and measured void volume fraction data 6.6 Discussion A fundamental question motivating the present work concerned the origin of voids in composite laminates produced from prepregs. The nearly identical values of weight loss determined by TGA and Fischer titration (Table 6-1) support the conclusion that in the present study, the weight loss can be attributed almost entirely to moisture. Thus, any volatiles present in the resin system exist in negligible quantities, and the contribution of volatiles to void formation can be neglected. Although the moisture content in the prepreg appears relatively small when expressed as a weight percent, it represents a much larger mole percent and a significant potential volume fraction of water vapor, indicating that dissolved moisture is a potential cause of void formation [8, 11]. 88 The measured void fraction dependence on moisture content is consistent with the model predictions for autoclave and full vacuum VBO processing, as shown in Figure 6- 9. Because the model predictions have been scaled, the measurements constitute validation of the predicted dependence of void content on moisture content (as opposed to the specific values). Note that the model calculations predict an absence of voids in autoclave-processed parts, and the prediction was corroborated by examination of polished sections. This finding has also been reported in a previous study examining void content in autoclave cured prepreg materials [16]. Autoclave pressures effectively force moisture to remain in solution, thereby preventing the formation and expansion of voids [4, 12]. However, the safeguard supplied by high external pressures is absent in VBO processing, rendering laminates more susceptible to void formation from dissolved moisture. This pressure dependence is particularly noticeable with reduced vacuum VBO processing. Void volume fraction results for low pressure VBO cured samples are shown to deviate considerably from the predicted trend (Figure 6-9). This result can be understood in terms of the formulation and design of VBO prepregs. While the materials are manufactured for cure under either autoclave or VBO conditions, maintaining a full vacuum is critical for part quality. The deviation of the experimental results from the moisture based model is likely a result of a secondary, pressure related, void formation mechanism. VBO prepregs can be consolidated under vacuum pressures as a result of engineered vacuum channels present in the resin geometry [2-5, 16]. Upon application of 89 vacuum these channels become evacuated and consequently entrapped air is drawn out of the laminate via edge breathing [5, 16]. Engineered vacuum channels successfully eliminate void formation due to entrapped air between plies in small flat panels under full vacuum. Figure 6-8 shows that with full vacuum processing large isolated voids are formed. Agreement between void volume fractions and analytical predictions suggests that these large voids are formed via diffusion of moisture. Data obtained at a vacuum level of 0.84 atm, however, shows a much different void formulation. It is clear from inspection of polished sections that under partial vacuum compaction between plies is not fully obtained, resulting in long thin inter-ply voids and the previously mentioned small ‘pe ppe r voids ’. The se p e ppe r voids a re li ke l y c a us e d b y a ir tha t i s tra ppe d b e twe e n pli e s and unable to escape due to the reduced vacuum level. As the analytical model takes into account only moisture-based void formation, the presence of voids due to entrapped air provides an explanation for the deviation between experimental results and the predicted trend for the case of reduced vacuum processing. To further illustrate the issue of compaction pressure, void pressure as a function of moisture content was examined. The following equation was derived by equating Equation 6-5 and Equation 6-7 and solving for pressure [20] ( ) (6-9) In this expression, the left hand side represents the applied compaction pressure, while the right hand side represents the water vapor pressure in a void as a function of temperature and relative humidity exposure. As described previously, void growth will 90 take place only when C void < C bulk , so when the above inequality (Eq 6-9) is satisfied, void growth via moisture diffusion cannot occur [20] To illustrate the dependence of void pressure on temperature during the cure cycle, a void pressure map was created (Figure 6-10). Pressure dependence on temperature was plotted for two relative humidities, 100% and 45%. To prevent void formation and growth, the applied compaction pressure must exceed the water vapor pressure [8, 11, 20]. The 100% RH curve represents an upper bound on void pressure, while the 45% RH curve is included to display the maximum moisture content that can be successfully held in solution by full vacuum pressure. Figure 6-10. Void pressure as a function of initial relative humidity exposure This graph illustrates the effectiveness of autoclave pressures in suppressing void growth, as well as the sensitivity of VBO processing to resin moisture content. A relative humidity of 45%, using the solubility curve defined earlier, corresponds to a resin 91 moisture content of approximately 0.25 wt%. The resin in the as-received condition typically has a moisture content of 0.24 0.03, a level slightly greater than that which can be controlled by atmospheric pressure alone, and certainly too high to be controlled with reduced vacuum processing. After an out time of 24 h in an uncontrolled lab environment (RH = 50 ± 5%), this level increased to 0.30 ± 0.01. Manufacture of large parts often requires several days of out-time for cutting and lay-up. Consequently, control of the ambient humidity may be necessary to ensure part quality when manufacturing parts with VBO processing. Despite this concern, void-free laminates were consistently produced using VBO processing. Laminates that were laid up and cured using the as-received prepreg, under full vacuum, exhibited void contents and thicknesses identical to those cured in the autoclave. The findings demonstrate that void-free laminates can be achieved by VBO processing using appropriate protocols. Additionally, laminates fabricated under full vacuum from VBO prepregs exposed to a temperature of 35º C for 24 h were void-free, confirming that dissolved moisture, not temperature exposure, was the source of voids in humidity conditioned laminates. Note that although the diffusion-based model provides an accurate fit to the experimental data, there are limitations to the approach used here. The addition of moisture to epoxy resins affects material properties, often unavoidably, as in the case of gel time. Such properties include tack, resin reactivity, and resin viscosity [73, 74], and all are likely to influence void formation. However, the goal of this work was to test the hypothesis that resin moisture content affects void formation. Additionally, our 92 measurements were performed using small flat panels, and thus the contribution of entrapped air was neglected. With large and/or contoured parts, pathways for escape of entrapped air are generally longer and more tortuous. Factors such as sample geometry and part size are expected to affect gas removal and must be investigated to develop a more complete picture of void formation in VBO-processed composites. 6.7 Conclusions Void formation was examined as a function of resin moisture content and processing pressure. During autoclave processing, high pressures were sufficient to suppress void formation, while with full vacuum VBO processing void volume fraction increased exponentially as a function of moisture content. A diffusion-based model of moisture-driven void formation was consistent with experimental data for these processing conditions. VBO processing at a reduced vacuum of 0.84 atm resulted in a unique void morphology and values deviating substantially from analytical predictions. This deviation is a result of secondary, pressure-related, void formation mechanisms which are not accounted for in the moisture-based model. The findings demonstrated that the production of void-free parts is possible with VBO processing, but the absence of high external pressures requires careful control of the lay-up environment to prevent excessive moisture uptake in the resin. Both analysis and experimental measurements demonstrated that while autoclave pressures were sufficient to suppress void formation in the presence of dissolved moisture, low-pressure processing was more sensitive to moisture content, resulting in voids with high levels of dissolved moisture. Other manufacturing parameters are likely 93 to give rise to similar effects, and low-pressure processing will require special care and attention to suppress void formation. One potential issue that must be considered is the extent to which vacuum bag processing can be successfully translated to the manufacture of larger parts. With the application of autoclave pressures, large parts generally can be processed without voids, largely because the high pressures suppress the evolution of dissolved species. However, the pressure differences associated with VBO processing are much lower, and thus careful control of process parameters will be more critical, particularly for production of complex parts. Examples of such parameters include site- specific vacuum levels and breathe-out distances. With large or complex parts, maintaining full vacuum at all locations throughout the cure is often difficult. Additionally, for large part production, pathways for air and volatiles to escape must be maintained over longer distances. The longer the pathway to a vent (breathing edge), the more difficult it becomes for entrapped air to exit the part [16, 74]. The present investigation of moisture effects suggests that in controlled lab environments, moisture can be eliminated as a significant source of voids in flat panels. This result sheds light on the controversy surrounding the role of moisture content vs. entrapped air as a leading source of voids, suggesting that small flat laminates produced under full vacuum with prepreg stored in a controlled environment will not develop voids as a result of moisture content. While these findings contribute to the general understanding of void formation, the role of other factors remain unclear. Systematically addressing additional factors affecting void formation will lead to a better understanding 94 of void formation in prepreg processed parts, and clarify the limits of VBO processing for composites manufacturing. 95 CHAPTER 7. Out time effects on VBO prepreg and laminate properties 7.1 Introduction An issue relevant to any composite processing method is ambient aging of material. The resin in prepregs is unstable at room temperature, so to prevent advancement of cure, material is stored in freezers until used. However, for the manufacture of large parts, layup can take several days [25, 28]. It is thus important to understand how material characteristics will change with prolonged exposure to ambient conditions. Several out-time studies have been performed on a range of prepreg material systems [25, 26, 28, 30, 31, 74-79]. Most of these studies have focused on autoclave processing [25, 28, 30, 31, 74-79], though vacuum bag only processing has been examined as well [26]. What has yet to be explored is the difference in part quality for the same material cured using both autoclave and VBO techniques. Aging will impact processing and final properties, and it is possible that the quality of VBO laminates will have a different dependence on out-time than that of autoclave processed parts. This work presents an evaluation of prepreg properties and associated laminate quality as a function of not only aging time but also processing technique. Characterization of prepreg is aimed at determining a simple method to estimate prepreg age, and laminate testing provides a means of relating this age to anticipated part quality. 96 7.2 Experimentation 7.2.1 Prepreg characterization 7.2.1.1 Modulated differential scanning calorimetry A Modulated Differential Scanning Calorimetry (MDSC) technique was utilized to track changes in the resin system as a function of ambient aging time. MDSC is a powerful thermal analysis technique that results the separation of two distinct thermal transition types: reversing and nonreversing [80]. This separation is accomplished through a method of modulated heating, a sinusoidal temperature profile that is applied on top of the standard temperature ramp. Reversing reactions (i.e. glass transition) are heating rate dependent, and thus change based on the application of a sinusoidal heating and cooling cycle [80]. Nonreversing reactions, on the other hand, are kinetic events that are dependent only on absolute temperature, and cannot be altered by cyclic heating and cooling (i.e. crystallization, exothermic cure reactions) [80]. The separation of heat flow into these two components allows for the identification of subtle transitions. MDSC was performed on prepreg samples sealed in aluminum DSC pans using an instrument capable of modulation and cooling to subambient temperatures (TA Instruments Q2000). Precut samples were aged in open DSC pans, then sealed and tested. During testing, a temperature modulation of ±1º C every 60 seconds was applied to a temperature ramp from -90ºC to 280ºC at 3ºC/min. The same heating profile was then repeated to determine the glass transition temperature of the cured system. Tests were performed under a nitrogen purge. Subambient temperatures were included to track transitions in the uncured system. DSC testing was performed every 7 days. 97 7.2.1.2 Tack P re pr e g tac k, or “ sti c kin e ss,” is a pr ope rt y that h a s be e n shown to be se nsit ive t o aging time [73]. Tack is important for laminate quality as it affects the control of the layup of plies. When prepreg is too sticky, misalignments cannot be corrected, but without adequate tack plies will not remain positioned properly during layup. There are several qualitative methods for determining tack, but to obtain quantitative data an energy of separation technique was used. The tack test was performed on a stack of 5 plies of prepreg. The plies were first compressed, at a rate of 3.05 mm/min, to a maximum load of .0801 MPa. This maximum load was held for 30 s, after which the plies were pulled apart in tension at a rate of .26 MPa/min. A tabletop load frame (Instron 5567) was used to perform the test. Figure 7-1 shows a typical stress-strain curve for the tack test in the compressive region. Section A in Figure 7-1 represents the input energy during the compression step, while section B represents the output energy as the plies are pulled back to a zero position [81]. Energy of separation per unit volume was calculated as the area of section B on the stress-strain curve of the tack test. Changes in energy of separation were tracked as a function of aging time. Details of the test method and analysis technique can be found elsewhere [81]. Tack tests were performed every 3-4 days until the prepreg consistently exhibited a tack level of zero. 98 Figure 7-1. Schematic of characteristic tack test data showing the tack calculation method 7.2.2 Laminate fabrication and testing 7.2.2.1 Test panels The material used in this study was a Cytec prepreg composed of 5320 resin and 5-harness satin carbon fiber fabric. The out-life and tack-life of the material system are reported by the manufacturer to be 3 weeks and 2 weeks respectively. Prepreg plies trimmed to 210 × 210 mm were stored in an ambient lab environment. Every 7 days two 8 ply quasi-isotropic laminates were laid up. One laminate was cured with traditional autoclave processing (5 atm external pressure), and the other using a vacuum-bag-only, out-of- a utocla v e c ur e pr oc e ss. Th e manuf a c tur e r ’s re c omm e nde d l ay-up procedure and temperature cycle (Figure 7-2) were used for both cure methods. 99 Figure 7-2. Temperature cure profile 7.2.2.2 Laminate characterization To examine the quality of laminates, several characterization methods were used. Each laminate was scanned using water-coupled ultrasound, then cut and measured for thickness. Samples from the center of each panel were polished and imaged to examine microstructure. To track mechanical properties, interlaminar shear testing was done on 5 samples from each laminate (ASTM D2344). Results from various laminate tests were correlated with prepreg age to determine the useful out-life of the material. 7.3 Results and discussion 7.3.1 Prepreg characterization 7.3.1.1 MDSC Modulated DCS runs were performed for two heating cycles on each sample. Full data sets from fresh prepreg are shown in Figure 7-3. Modulation allows for the splitting of the heat flow signal into reversing and nonreversing components. Through this process 100 three important events can be identified. During the first heating cycle (Figure 7-3a) a glass transition in the uncured system can be seen in the reversing heat flow signal. The nonreversing signal shows a clear exothermic cure reaction peak. A second heat cycle was performed to determine the glass transition temperature of the cured system. This value can be seen clearly in the reversing heat flow signal for the second temperature ramp (Figure 7-3b). Changes in these three features were tracked as a function of aging time. Figure 7-3. (A) First heat ramp showing the glass transition of the uncured system and the exothermic reaction peak, (B) Second heat ramp showing the glass transition temperature of the cured composite Samples were tested every seven days as ambient aging progressed. Representative changes in the glass transition of the uncured system and the exothermic 101 cure peak are presented in Figure 7-4. The glass transition shifts to higher temperatures as the resin ages (Figure 7-4a), and the exothermic peak decreases in height (Figure 7-4b). These changes occur as a result of the progression of cure of the epoxy during ambient aging. Figure 7-4. Changes in (A) glass transition temperature of the uncured system and (B) exothermic reaction peak, as out-time is increased Similarly, as shown in Figure 7-5, the glass transition temperature of the cured composite shifts higher with aging time. This trend, which has been reported for other epoxy systems [30, 31], results from an increased level of crosslinking, and thus an increase in the degree of cure of the system [30]. 102 Figure 7-5. Changes in glass transition temperature as out-time increases Data points from each day of testing were plotted for the three points of interest. The value of the area under the exothermic peak (enthalpy of reaction) was calculated using a sigmoidal tangent method. All MDSC runs were performed using prepreg, and results were not altered to take into account resin content. Such changes would result only in a relative shift in the results, not a change in the trends observed. Data fitted with linear trend lines is presented in Figure 7-6. 103 Figure 7-6. Changes in DSC data as a function of out-time. (A) Enthalpy of reaction, (B) glass transition of the uncured system, (C) glass transition of the cured composite From the plots in Figure 7-6 it is clear that the best linear relationship (R 2 =0.94) is obtained for the glass transition of the uncured system as a function of out-time. As it is often the case that the exact out-time exposure of a roll of prepreg will be unknown, it is desirable to have a simple test to predict prepreg age. It is possible from the results presented in Figure 7-6b that the glass transition of the uncured epoxy could be used to accurately predict prepreg age for the 5320 system. Fit lines for enthalpy of reaction (R 2 =0.79) and glass transition temperature (R 2 =0.71) show a much higher degree of scatter, making them poor candidates for age prediction. 7.3.1.2 Tack Another property closely related to prepreg age is tack. To determine tack an energy of separation technique was used. Results are displayed below. 104 Figure 7-7. Energy of separation per unit volume as a function of aging time It is clear from the data presented in Figure 7-7 that a large degree of error was present in the calculation of energy of separation. This is likely due to the sensitivity and degree of control available with the load frame used for testing. No clear trend in tack level as a function of out-time can be determined, but the data does show that after 22 days in ambient conditions the prepreg consistently displays a tack level of zero. Because the energy of separation technique described in the test method used showed a large degree of scatter, a second method was developed to calculate a representative measure of tack. The test method described by Ahn and colleagues [81] uses only the area compressive region of the stress-strain plot to specify output energy. Actual tests were run, however, until the prepreg plies fully separated, which occurred significantly beyond the zero-point in fresh samples. This extension of the stress-strain 105 curve for an unaged sample is shown in Figure 7-8 with photographs showing the adhesion and separation of adjacent prepreg plies. Figure 7-8. Continuation of the tack stress-strain curve showing detachment of adjacent prepreg plies. Images show (1) compacted plies, (2) adhesion of plies in tension, (3) ply separation Calculating the area beyond the zero-strain point produced a plot of tack level vs. out-time showing an asymptotic decay in prepreg adhesion. The tack level can be seen to reach zero after 22 days, in agreement with the data in Figure 7-7. Both tack calculation methods show that the prepreg retains some stickiness after the 14 day tack-life specified by the manufacturer. Figure 7-9, however, provides additional insight, showing that with even very limited room temperature aging prepreg tack is dramatically reduced. 106 Figure 7-9. Energy of separation calculated for the tensile portion of the tack test until full separation of prepreg plies 7.3.2 Laminate characterization 7.3.2.1 Ultrasound scans Ultrasonic C-scan images were generated for each laminate. A 2.25 MHz transducer was used and scans were performed in the reflected mode in a water tank. The color scale on the images represents the reflected signal intensity. For the images in Figure 7-10 green represents a reflected signal of approximately 90%, yellow 50% and red 10%. All equipment settings were kept the same for all scans, so changes in color indicate changes in laminate quality (void content, compaction). 107 Figure 7-10. Ultrasound C-scans of cured laminates C-scan images show a decrease in reflected signal with increasing out-time, signifying a decrease in laminate quality. The green circles that appear in later scans are a result of air bubbles that formed at surface defects. The stated out-life of the material used in this study is 21 days, and it is clear from ultrasound scans that laminate quality begins to decrease after this period (at day 22 in autoclave samples and day 28 in VBO laminates). 7.3.2.2 Thickness measurements Thickness of laminates was measured at the center of each panel. The thickness of the sample was taken to be an indication of laminate compaction. Consistent with ultrasound scans, thickness measurements show that compaction level begins to decrease for autoclave samples after 22 days and for VBO samples after 28 days. VBO samples cured after long out-times (over 30 days) were found to be thicker than autoclave cured samples. This is likely a result of autoclave pressures forcing greater compaction then vacuum pressure only. 108 Figure 7-11. Laminate thickness measurements 7.3.2.3 Microstructure Samples were polished and imaged to examine changes in microstructure. Laminates cured with fresh prepreg show a high level of compaction, though voids were present in both autoclave and VBO cured samples (void volume fraction of 0.62 0.2 for autoclave panels and 0.85 0.2 for VBO panels). Void content was reduced after a week of aging (0.31 0.15 for autoclave samples and 0.22 0.2 for VBO samples). This trend of reduced void content with increased out-time has been noted by other researchers [26]. Reduction of void content for aged samples is a result of reduced tack levels. The prepreg used in this study is formulated for edge-breathing, and lower tack prepregs trap less air during lay-up and allow for more efficient air removal during cure. 109 Figure 7-12. Imaged cross-sections showing laminate compaction and void content Reduction in void content after 8 days of ambient aging can be seen in the micrographs in Figure 7-12, and is verified by ultrasound scans (Figure 7-10). While there was not a clear increase in void content with increasing aging time, laminate compaction was reduced. For samples cured after long out-times each prepreg layer is easily distinguishable due to a lack of compaction (Day 56, Figure 7-12). This is the result of reduced flow in aged resin. Reduction in compaction can also be seen in thickness measurements (Figure 7-11). 7.3.2.4 Mechanical properties Interlaminar shear strength was tested following ASTM D2344 [41]. Again, consistent with both ultrasound scans and thickness measurements, autoclave laminates begin to show reduced properties after 22 days of ambient aging, while VBO samples 110 show a reduction after 28 days. Reductions in ILSS are attributed to reduced laminate compaction. Figure 7-13. ILSS values 7.4 Age determination and quality prediction For the resin system examined, it was found that the glass transition of the uncured system followed a linear trend, providing a reliable way to predict prepreg age. Examination of cured laminates was then carried out to relate a given age to laminate properties for both autoclave and VBO processed panels. Laminate characterization showed that quality samples could be produced prior to the 21-day out life of the material with both autoclave and VBO processing. VBO laminates produced after 28 days of room temperature aging were also high quality, though autoclave samples cured after 28 days showed reduced properties. These findings confirm prepreg that has been out of the fr e e z e r for less than 21 da y s (the manuf a c tur e r’ s state d out -life) can be used to produce 111 quality parts. Current findings also suggest that prepreg cured using vacuum pressure only will result in laminates of equal or better quality then autoclave processing for out- times of less than 30 days. 7.5 Conclusions Prepreg and laminate characterization was carried out to examine ambient aging effects on carbon fiber/epoxy composites. To characterize prepreg, differential scanning calorimetry and tack tests were performed. Modulated differential scanning calorimetry showed an increase in the glass transition temperature of both the uncured resin and the cured system, and a decrease in the heat of reaction as a function of aging time. The glass transition of the uncured epoxy showed a strong linear relationship to out-time. It is thus possible that this transition could be used to predict prepreg age, though verification of the trend in other resin systems will be necessary to support this claim. Laminate characterization was performed to correlate prepreg age with part quality. It was found that quality began to decrease after the manufacturers stated out-life of 21 days. Ultrasound scans, thickness measurements and interlaminar shear strength measurements showed that VBO processed laminates began to degrade in quality after 28 days of prepreg aging, while autoclave parts show reduction in quality after 22 days. This result suggests that VBO processing can be used to produce parts with quality equivalent to, or better then, autoclave processing for ambient aging of approximately three weeks. This finding is promising for the application of VBO processing to large-scale aerospace parts. 112 Results reported suggest that the glass transition temperature of the uncured epoxy system can be used to determine room temperature out-life for the Cycom 5320 system. Laminate characterization shows that if age is determined to be less than 22 days, quality laminates can be produced with confidence. 113 CHAPTER 8. Prepreg age monitoring 8.1 Introduction The layup of prepreg is performed at ambient temperature. The epoxy resin in composite prepregs, however, undergoes aging at ambient temperature [26-32, 78, 79]. For this reason, prepregs are stored frozen to prevent advancement of cure, then thawed prior to layup [26-32]. Despite this precaution, ambient aging is inevitable for the production of large parts, as the lay-up of large-scale structures can take several days and e ve n we e ks, a pe riod kno wn a s ‘‘ out - ti me’ ’ [26-29]. Additionally, the chemical age of a given roll of prepreg is often unknown. With the push in the aerospace industry to manufacture large components such as airplane wings and fuselages with out-of- autoclave prepregs, an understanding of the out-life limits of these materials will be critical. Also of importance from a manufacturing standpoint is the development of a test method for monitoring prepreg age that is not only accurate but simple to carry out and suitable for production use [26, 27, 32]. The degree of cure of epoxy resin in prepregs advances with room temperature aging time. As shown in the previous study (Chapter 7), the chemical changes that occur during ambient aging influence the mechanical properties of cured parts, as well as the practical attributes of prepregs, such as tack and drape [25, 26, 32, 76, 78, 79, 82]. The ability to assess prepreg age is important for the manufacturing of high-quality parts. Studies on ambient aging of traditional autoclave prepregs have been performed for a variety of resin systems [25-28, 30, 31, 73, 74, 76, 78, 79, 83] These studies examined multiple methods for monitoring prepreg age, with the most success coming from 114 infrared (IR) spectroscopy [76, 78] and high-performance liquid chromatography (HPLC) techniques [83]. A more recent study demonstrated the use of photoacoustic spectroscopy to track prepreg age [26]. In this study, we examine a method for monitoring prepreg out- time via differential scanning calorimetry (DSC) methods and verify the validity of this method for out-of-autoclave prepreg systems. In the past studies of ambient aging, changes in IR spectra and in HPLC results were attributed to chemical aging effects. In the present study, these same effects resulted in a linear increase in the glass transition temperature (T g ) of the B-stage epoxy. Results from the past aging studies, as well as the general glass transition theory, provide insight into the mechanism behind this relationship. The findings are of interest because T g is readily and conveniently measured using conventional DSC. The results of this study, combined with past literature, suggest that DSC test methods could be implemented on the factory floor to monitor the age of a wide range of epoxy prepreg systems. 8.2 Experimental 8.2.1 Materials Three carbon fiber/epoxy prepregs were selected for this study, all formulated for vacuum-bag-only processing. The first (Prepreg 1) was a five harness satin (5HS) carbon fiber fabric with a toughened epoxy resin (CYCOM 5320, Cytec Engineered Materials, USA). The second (Prepreg 2) was an eight harness satin fabric (8HS) with a toughened epoxy resin formulated for long out-life (CYCOM 5320-1, Cytec Engineered Materials, USA). The third prepreg system, from a different manufacturer, was included in the study to assess chemical aging behavior for a prepreg formulation from a different source. This 115 prepreg system (Prepreg 3) was composed of a 2×2 twill carbon fiber fabric and a toughened epoxy matrix (MTM44-1/CF5804A, Advanced Composites Group, UK). The nominal resin content for all three prepregs was 35 –40% b y we i g ht. The m a nufa c tu re rs’ stated out-life for the three materials was 21, 30, and 21 days, respectively. The specific formulations of the epoxy resins examined in this study were proprietary. 8.2.2 DSC Modulated DSC (MDSC) was used to track changes in the resin systems as a function of room-temperature aging time. Material was stored in unsealed plastic bags in ambient laboratory conditions (20±2°C, 50±5% relative humidity) and tested periodically by MDSC as aging progressed. DSC is used to measure the difference in the amount of heat required to change the temperature of a pan containing a material sample and an empty reference pan as a function of temperature. In modulated DSC, a sinusoidal temperature modulation is applied over the standard linear temperature ramp. This temperature modulation allows for the splitting of the total heat flow into reversing and non-reversing components. Reversing heat flow is a function of the heat capacity of the sample as well as the rate of temperature change, while non-reversing heat flow is a kinetic component dependent only on time and absolute temperature (not influenced by a sinusoidal temperature modulation) [80, 84]. Glass transition temperature is a reversing (heating rate dependent) event and was thus calculated from examination of the reversing heat flow signal [80]. In contrast, exothermic cure reactions are kinetic events, and thus degree of cure was determined by examination of the non-reversing heat flow signal. In this study, temperature modulation was used during DSC experiments to increase the 116 accuracy of measurements. However, standard DSC methods (without temperature modulation) can also be used for the same purpose. For prepreg characterization, MDSC tests were performed periodically as aging progressed. A DSC with temperature modulation capability and sub-ambient cooling was used (TA Instruments Q2000). A temperature ramp from -90 to 280° C was performed on each sample. A temperature modulation of ±1°C every 60 s was applied with a ramp rate of 3°C/min. All tests were performed under a nitrogen purge. Note that in this study, glass transition measurements were performed on prepreg samples that were not fully cured. The low-temperature T g values measured in this study are characteristic of the B-stage resin, while the glass transition of the cured systems are much hig he r ( ≈200°C ). T o a void conf usion with t he T g of the cured resin, the glass transition temperatures reported in this study will be referred to as B-stage T g . 8.3 Results and discussion Past aging studies have shown that the degree of cure of thermosetting epoxy systems increases as a function of room temperature aging time [76, 78, 83]. Additionally, based on glass transition theory, T g increases linearly with degree of cure [85]. This suggests that the B-stage T g should depend linearly on ambient aging time. This trend has been validated for one autoclave resin system [30], but proof of applicability to a range of resin formulations and an analysis of the accuracy of the trend as a predictor for prepreg age has not yet been demonstrated or reported. We have performed MDSC testing on three out-of-autoclave prepregs. To establish the linear trend in B-stage T g as a function of out-time, two sets of samples 117 taken from different regions of the prepreg roll were tested for each material system. Data were plotted and linear fits were applied. Results are presented in Figure 8-1, with linear fit equations and calculated correlation coefficients. Figure 8-1. Linear trends between B-stage T g and prepreg out-time for (A) Prepreg 1, (B) Prepreg 2, and (C) Prepreg 3 118 One objective of this work was to investigate the reaction kinetics governing ambient aging and to develop a test method for monitoring prepreg age. To determine the accuracy of predicting age using glass transition temperature, a third set of samples was tested. This sample set, from still another area of the prepreg roll, was used as blind data to establish the accuracy of the linear fit. Using the B-stage glass transition temperatures obtained for each sample, a predicted prepreg age was calculated using the linear fit lines displayed in Figure 8-1. Predicted age was plotted against actual known age for all three sample sets. 119 Figure 8-2. Age predicted using linear fit equations plotted against actual known age for (A) Prepreg 1, (B) Prepreg 2, and (C) Prepreg 3 Figure 8-2 displays predicted age for the prepregs, calculated using measured B- stage T g values and the linear fit equations (shown in Figure 8-1), plotted against actual prepreg age. The solid lines in each plot are reference lines for which the actual and predicted age values are equal. The accuracy of using glass transition temperature as a 120 measure of prepreg age was evaluated by examining the difference between actual age and age predicted using MDSC results. Table 8-1 displays the minimum, maximum, and average difference (in units of days) for all three prepreg systems. The method was accurate, on average, to within 3 days (Table 8-1). Table 8-1. Minimum, maximum, and average difference in predicted age and actual age (in units of days) for the three prepreg systems examined Min Max Average Prepreg 1 0.36 6.26 3.08 Prepreg 2 0.04 6.71 2.33 Prepreg 3 0.11 4.28 1.87 8.3.1 Validity of the method To verify the precision of the experimental testing and to evaluate the error inherent in the MDSC test method, a validity test was performed. Three samples were prepared from adjacent regions of a single prepreg ply (Prepreg 1). One sample was tested each day for 3 consecutive days. All samples were stored frozen until tested. Because samples were taken from the same region of the prepreg roll, the resin content in each sample was assumed to be constant. Additionally, because the samples were stored frozen, the resin was not exposed to environmental factors such as moisture and temperature variations. The results of the validation experiment, presented in Figure 8-3, demonstrate that the error inherent in the test method is negligible. The standard deviation in glass transition temperatures measured was 0.03° C. Thus, the scatter in the measurements reported in Figure 8-1 is assumed to arise from differences in temperature and humidity 121 exposure during aging, and variations in resin content and specific thermal history in different regions of the prepreg roll. No variance is attributed to experimental test errors. Figure 8-3. Reversing heat flow signals, showing B-stage glass transition temperature results for three samples of fresh prepreg (Prepreg 1), used to validate the accuracy of the modulated differential scanning calorimetry (MDSC) method 8.3.2 Mechanism of ambient aging A clear linear dependence of B-stage T g on ambient aging time was observed for all three resin systems examined. This linear dependence is attributed to chemical reactions occurring in the resin during ambient aging. During room-temperature chemical aging, the number of unreacted epoxide groups in the resin decreases. These reactions are similar in nature to those reported in previous studies using IR spectroscopy [76, 78] and HPLC [83] and cause the gradual increase in B-stage glass transition temperature 122 reported here. Though we have examined out-of-autoclave resin systems in this work, the general aspects of epoxy aging are common to all prepreg systems. In previous ambient aging studies, IR spectroscopy results were used to define an ‘‘ e pox ide inde x ,’ ’ w hich w a s tra c k e d a s a func ti on of a g e . Th e e pox ide inde x , a mea sure of the number of unreacted epoxide groups, was calculated as the ratio of the area of the aromatic ring band of the IR spectra to that of the epoxide band [76, 78]. The reduction in epoxide index was reported to be a linear function of prepreg age [78]. A reduction in the primary epoxy component of prepreg resin during ambient aging has also been reported in HPLC results [83]. These studies showed that a reduction in free amine molecules (as measured by reverse phase liquid chromatography) occurs during the aging process [78]. The reduction in unreacted epoxide groups, resulting from an epoxy-amine reaction, is known as precuring. This precuring reaction shifts the glass transition temperature of the uncured system to a higher temperature. Glass transition marks a change in the resin system from a glassy state to a rubbery state. As the number of unreacted end groups is reduced during precuring, segment mobility decreases, leading to an increase in glass transition temperature. Such a phenomenon, though never examined as a means of monitoring ambient aging, has been discussed in terms of the general properties of the glass transition temperature of thermosetting resins [85]. The classical theory of glass transition in cross-linking polymers is complicated in the temperature region beyond the gel point of the resin, where molecular weight becomes undefined [86]. For the B-stage glass transition temperature examined in this study, however, the resin is well below the gel point, and 123 has a defined molecular weight. A simplified theory of glass transition temperature can thus be employed to describe the behavior observed. As reported in past IR studies of ambient aging, the concentration of unreacted epoxide groups decreases linearly with aging time [78]. This process can also be discussed in terms of the degree of cure (DoC) of the resin. In reference to an ideal backbone glass transition temperature ( ), glass transition as a function of degree-of- cure can be expressed as [85] ( ) (8-1) where p is the DoC of the system (or end group conversion) and K 1 is a constant for a given resin system. Here, DoC is defined as the ratio of the number of reacted sites to the total number of reactive sites available. The equation shows that a linear increase in glass transition temperature with end group concentration is expected at low conversions (small p) [85]. This simple relationship holds in the low-conversion region because the effect of cross-linking (the linking of polymer chains) is negligible. The observed change in g lass tra nsit ion t e mper a ture Δ T g is simply a function of the consumption of end groups (the reduction in the number of unreacted epoxide groups in the resin) that occurs during the epoxy-amine reaction as the cure progresses. Thermoset resins, such as the ones examined in this study, begin as low molecular weight oligomers, with a high initial concentration of end groups [85]. These end groups are consumed during the cure process, which influences the glass transition temperature of the material. Past investigations of the effect of DoC on glass transition temperature for thermoset resins have reported trends similar to the behavior predicted in Equation 8- 124 1 [85]. Characterization of these materials through a broad temperature range shows a linear relationship between glass transition temperature and degree-of-cure up to a conversion of approximately 0.3-0.4 (DoC of 30-40%) [85]. The relationship becomes non-linear only when branching and cross-linking reactions become dominant (far beyond the useful out-life of prepreg materials). These results are consistent with the trend observed here for B-stage T g as a function of aging time. Because the number of unreacted epoxy groups decreases in proportion to aging time, and because the general theory predicts that T g will increase in proportion to DoC, T g is also expected to increase linearly with ambient aging time. This trend, reported for a single resin system formulated for autoclave curing [30], is shown here to hold for a variety of out-of- autoclave prepregs as well. Although the B-stage T g exhibited a linear dependence on aging time for the range of out-times considered here, deviation from linearity is expected to occur at longer aging times, as predicted by Equation 8-1. In fact, the dependence of T g on out-time should become non-linear when the B-stage T g exceeds the cure temperature (here ≈22°C), because vitrification of a polymer network occurs when the glass transition temperature reaches the cure temperature [87]. Of the prepreg systems studied here, only Prepreg 1 reached T g values significantly greater than ambient temperature through the length of this study. Despite the fact that T g values exceeded 22° C for Prepreg 1, no significant deviation from linearity was observed. Similar behavior has been reported for an autoclave prepreg system, where measurements of B-stage T g as a function of out-time, showed little deviation from linearity until T g reached 45° C at an exposure time of 100 125 days [30]. For greatest accuracy in predictions of prepreg age, the linear relationship between B-stage T g and ambient aging time should be defined for B-stage T g values less than 22°C. This is a reasonable limit to impose for practical purposes, because the B- stage T g of an epoxy system will not exceed the cure temperature (vitrify) within the useful out-life of the material. While the chemical aging process described in this work is well understood, a means of implementing this understanding to monitor prepreg age in-service is new. Past studies have shown IR techniques and chromatography methods can be used to monitor prepreg age. However, these methods are not suitable for production floor use. IR measurements require dissolving resin from the prepreg with acetone to produce resin films, a labor-intensive sample preparation task [26, 78]. Other proposed age-monitoring methods (IR, HPLC, and photoacoustic spectroscopy) require post-processing and analysis that is time-consuming and skill intensive. DSC methods, on the other hand, require minimal sample preparation, simple analysis/interpretation of data, and can be applied to a wide variety of sample types [84]. Only small samples of prepreg (6-mm disks) are needed for DSC methods and are easily cut and no separation of resin from the carbon fibers is required. Lastly, data analysis is straightforward, and due to the linear nature of the trend observed in glass transition temperature, complicated statistical methods are not necessary. These factors highlight the potential for DSC testing to provide a quality control method for monitoring prepreg age suitable for production floor use. 126 8.3.3 Degree of cure As described in the preceding discussion, the degree of cure of the epoxy in the prepreg systems is expected to increase as a function of ambient aging time. To support this supposition, DoC was calculated and plotted as a function of both aging time and B- Stage T g . The degree of cure was calculated using the following equation: (8-2) Here, values for the number of reacted sites and total number of reactive sites were determined by integration of exothermic cure reaction peaks. Calculating the area of the exothermic cure reaction peak for unaged prepreg yielded the heat of reaction corresponding to full cure (DoC of 100%). As the cure progressed, the number of unreacted sites decreased, lowering the heat of reaction required to complete the cure. In other words, the exothermic cure peak decreased in intensity as a function of aging time. Degree of cure was thus determined by comparing the heat of reaction of the aged material (H r,aged ) to that of the unaged sample (H r,total ). This calculation is shown in Equation 8-3. ( ⁄ ) ( ⁄ ) ( ⁄ ) (8-3) Results for percent DoC as a function of aging time and B-stage T g are presented in Figures 8-4 and 8-5. 127 Figure 8-4. Degree of cure as a function of ambient aging time for (A) Prepreg 1, (B) Prepreg 2, and (C) Prepreg 3 128 Figure 8-5. Degree of cure as a function of B-stage T g for (A) Prepreg 1, (B) Prepreg 2, and (C) Prepreg 3 As expected, the DoC data showed a linear dependence on both ambient aging time and B-stage T g . The scatter in the data was attributed to differences in resin content in different regions of the prepreg. DoC measurements depend on sample mass and 129 consequently variations in wt% resin in each prepreg sample will cause variation in DoC values. 8.4 Conclusions In this study, we considered measurements of B-stage glass transition temperature as a means of tracking out-time and monitoring prepreg age. Measured values of B-stage T g showed a linear dependence on prepreg out-time for three out-of-autoclave prepreg systems from different manufacturers. This linear dependence resulted from the precuring reaction that occurs during room-temperature aging of epoxies. Analysis of degree of cure corroborated this mechanism of ambient aging. Comparisons with other aging studies, as well as consideration of general glass transition theory, supported the conclusion that precuring of epoxy prepregs causes a linear increase in B-stage T g with ambient aging time. The T g of uncured epoxy at low temperatures was measured using MDSC. The test method was precise, and relatively quick and simple, introducing minimal error into the results. Because of these attributes, measurement of T g by MDSC affords a convenient means for monitoring prepreg age. The method also is appealing because it involves virtually no sample preparation, requires only small sample sizes, and the data analysis is trivial. For these reasons, the method is well suited to a production floor setting, where it can be readily employed to accurately determine the remaining useful life of prepreg materials. The need for a method for accurately determining prepreg chemical age is growing, particularly as manufacturers undertake fabrication of larger and more complex 130 parts that require longer preparation times. Composite parts fabricated with prepreg that has exceeded the recommended out-life limits exhibit defects and degraded mechanical properties that are often unacceptable. Implementing a simple method to determine prepreg age would result in more efficient production of high-quality parts. Execution of the method presented here requires generating an initial calibration curve governing the chemical aging kinetics of the particular prepreg. The linear fit equations will be specific to each prepreg system because of differences in resin formulations. For example, although the specified out-lives for Prepregs 1 and 3 were the same, the more rapid increase in DoC for Prepreg 1 (and the higher values of B-stage T g measured) indicates that room-temperature chemical aging proceeds more rapidly in this material system. Prepreg 2, which was formulated for extended out-life, and Prepreg 3, show similar dependences of B-stage T g values on aging time. Thus, chemical aging occurs more gradually in these systems, indicating a longer useful out-life. Calibration curves are therefore required to characterize each material system, although these calibrations can be performed in a few weeks and require only grams of material. The method described here could be deployed on the production floor in straightforward fashion. An allowable limit for ambient aging time is first established by conventional means, such as mechanical testing of cured laminates, or simply by fol lowing the ma nu fa c tur e r’ s r e c omm e nde d out -life for a given material. By either method, manufacturers set a B-stage T g limit over which prepreg will be discarded. Prepreg age is then determined by measuring the B-stage T g and comparing to a pre- measured calibration curve specific to the system used. Note that the allowable out-time 131 will depend on storage conditions, including temperature and humidity. Similarly, the variation in age prediction reported here (Table 8-1) is assumed to arise due to variations in temperature and humidity during aging. Prepreg stored in a controlled environment, such as a clean room with controlled humidity and temperature, would display less variability, allowing for increased accuracy in age prediction. Nevertheless, the results reported here for out-of-autoclave prepregs, combined with past results for an autoclave curing system [30], indicate that the technique will be applicable to a wide range of composite prepregs. 132 CHAPTER 9. Adhesive out-gassing 9.1 Introduction Sandwich structures are commonly used in aerospace applications. The benefit of the sandwich structure design arises from a combination of increased stiffness accompanied by weight reduction [35, 88-91]. A sandwich structure consists of two face sheets bonded to a core (typically honeycomb or foam) using film adhesive [89]. When composite face sheets are used, processing time and cost can be reduced by curing the composite prepreg and the adhesive simultaneously, a process known as co-curing [90, 92, 93]. A schematic showing the construction of a composite sandwich panel is presented in Figure 9-1. Figure 9-1. Composite sandwich panel. Exploded view (left) and completed panel (right) Composite sandwich structures for aerospace applications have historically been manufactured using autoclaves. Co-cure processing with light-weight core materials, however, is problematic with autoclave processing because of the core-crush that occurs under high compaction pressures [35, 90]. Additionally, autoclave processing is 133 expensive, inefficient, and size-limiting [91, 94]. To combat these issues, out-of- autoclave, vacuum bag only (VBO) manufacturing methods have been introduced. VBO processing eliminates the high external pressure that can lead to core crushing, decreases processing cost, and increases manufacturing efficiency [95]. A major concern in the introduction of vacuum bag processed sandwich panels, however, is the adhesive foaming that occurs during co-cure under vacuum [94]. Foaming is attributed to volatiles in the adhesive that evolve during the cure cycle (adhesive degassing) [33]. Adhesive foaming can lead to unwanted porosity and a reduction in the structural integrity of the panel. The mechanical properties of sandwich structures are closely tied to the quality of the bond between the core and face sheets [91, 92, 95]. Past studies have linked voids caused by degassing in the adhesive layer of bonded structures to reduction in peel strength and bond durability [92, 95, 96]. It is thus critical for the production of high quality sandwich structures that foaming of adhesives under vacuum be understood and prevented. The goal of this work was to examine the evolution of volatiles during the vacuum cure of sandwich structure film adhesives and identify species evolving at specific temperatures. To accomplish this, an on-line coupled thermogravimetric analyzer-Fourier transform infrared spectrometer (TGA-FTIR) method was used with two commercial adhesives. The TGA-FTIR technique enables detailed analysis of the evolution of volatile species. By documenting the time and temperature at which various species evolve, optimized adhesive pre-treatments can be developed to eliminate (or reduce) foaming and improve part quality. 134 9.2 Experimental 9.2.1 Materials Two commercially available condensation polyimide adhesive films were examined in this study: adhesive A - FM ® 57 and adhesive B - FM ® 680-1 (Cytec Engineered Materials, USA). Both adhesives had a supported areal weight of 0.5 kg/m 3 , and both were formulated for cure at 177° C , f oll owe d b y a hig h e r te mper a tur e post - c ur e . P ol y im ide a dhe sives we re c h osen fo r this s tud y be c a use o f th e r e lev a nc e to hi g h se rvic e tempe ra tu re a ppli c a ti ons ( 300 ° C), and applicability to sandwich structure co-cure and composite bonding. 9.2.2 Methods 9.2.2.1 Vacuum oven cure To observe the foaming behavior of the polyimide adhesives under vacuum, samples were cured in a vacuum oven at 177° C. Adhesive samples were photographed before and after vacuum cure. 9.2.2.2 TGA-FTIR analysis Identification of adhesive out-gassing components was achieved using a method of coupled TGA-FTIR analysis. Coupled TGA-FTIR testing enables simultaneous measurement of sample mass loss, and information regarding the specific species evolved [97, 98]. TGA is used to measure weight loss of a sample as a function of temperature. For this study, gas vapors evolving during TGA testing were transferred to an FTIR gas cell via a heated transfer line, driven by IR-transparent nitrogen carrier gas [97]. The 135 transfer line and gas cell were heated to 220° C during testing to prevent condensation of vapors. Sample weight loss was tracked during a programmed TGA temperature profile, and FTIR spectra were recorded continuously throughout testing in the form of interferograms. A Gram-Schmidt orthogonalization procedure was used to process the data. The Gram-Schmidt profile represents the average of all FTIR peak intensities over the spectral range recorded (here 600 – 4000 cm -1 ) [97, 99]. Spectra were identified by comparison to a TGA vapor phase library to determine specific out-gassing components. Past studies have used TGA-FTIR methods to monitor the degradation behavior of polyimides through high temperature decomposition [98, 100]. This study, in contrast, was focused on identifying volatiles evolving during cure, rather than investigating degradation products. Thus, TGA samples were ramped to 177° C (the cure temperature of the adhesive) and held for the recommended cure time of each material (90 min for adhesive A, and 120 min for adhesive B). Off-gassing components were tracked as a function of time and temperature during cure. An optical spectrometer (a Nicolette 4700) equipped with a TGA interface accessory was used for FTIR measurements. Details of the experimental set-up and data collection are presented in Table 9-1. 136 Table 9-1. Experiment parameters FTIR Experiment Parameters Detector MCT/B Beam Splitter KBr Gain 1 No. of scans 8 Resolution 4 Optical Velocity 2.5317 A ramp rate of 10° C/min was chosen for TGA-FTIR testing to avoid temperature lag or inconsistency [98]. Additional TGA experiments were conducted at a ramp rate of 2° C/min to more closely approximate an actual adhesive cure cycle. Samples for TGA testing were prepared using a 6.25 mm diameter hole-punch to ensure consistent surface area and size. 9.3 Results and discussion 9.3.1 Adhesive foaming Adhesive samples were cured in a vacuum oven, to examine foaming behavior. Images of samples before and after vacuum cure are presented in Figure 9-2. 137 Figure 9-2. Adhesive samples before (left) and after (right) vacuum cure Foaming of the polyimide adhesive films under vacuum is clear in these images. Samples after cure in a vacuum oven (right hand side in Figure 9-2) exhibit a rough surface texture characterized by pitting and bubbling. This bubbling would lead to voids during vacuum bag only cure of sandwich panels. To reduce or eliminate foaming and void formation during cure, an improved understanding of the out-gassing behavior of film adhesives is required. 9.3.2 TGA-FTIR data Thermogravimentric analysis was performed to determine the quantities of volatiles evolving during the cure of the adhesives studied. Results are presented in Table 9-2. 138 Both adhesives displayed significant weight loss during the recommended cure cycles, with a greater percentage weight loss observed for adhesive A. Weight loss was comparable between samples ramped at 10°C/min and 2°C/min, indicating that FTIR data collected at a high ramp rate is an accurate representation of the gasses that evolve during the cure of the adhesives studied. Table 9-2. TGA weight loss data Adhesive A Adhesive B Sample # Ramp Rate (°C/min) % Wt Loss Sample # Ramp Rate (°C/min) % Wt Loss 1 10 27.38 1 10 20.68 2 10 27.52 2 10 20.61 3 2 25.57 3 2 20.75 As described in the previous section, FTIR interferograms were recorded continuously during TGA experiments. TGA-FTIR data is presented in Figure 9-3 for adhesive A and in Figure 9-4 for adhesive B. Part (a) of each figure displays the total FTIR spectral response as a function of time (Gram-Schmidt profile), along with the intensity of sample weight loss and temperature as a function of time, as recorded by the TGA. Part (b) of Figures 9-3 and 9-4 show a 3D display of the spectral response during testing. From this data, the specific off-gassing components of the adhesives tested were identified. 139 Figure 9-3. TGA-FTIR data from adhesive A. (a) Intensity of Gram-Schmidt profile, weight loss, and temperature vs. time. (b) 3D plot of FTIR spectra during testing 140 Figure 9-4. TGA-FTIR data from adhesive B. (a) Intensity of Gram-Schmidt profile, weight loss, and temperature vs. time. (b) 3D plot of FTIR spectra during testing Analysis of the FTIR spectra of both adhesives revealed two prominent off- gassing products. During early stages of the cure process, the evolution of water is apparent. The major volatile component at later stages of testing, however, was 141 determined to be the polar solvent 1-methyl-2-pyrrolidinone (NMP). These off-gassing components are indicated with arrows in part b of Figures 9-3 and 9-4. The spectrum obtained during testing (taken from 38.51 min into the cure of adhesive A) is presented in Figure 9-5a, along with the reference spectrum for 1-methly- 2-pyrrolidinone from the TGA vapor phase spectral library, showing a clear match. Figure 9-5b displays the FTIR spectrum of water, which can be identified at early time scales in the 3D plots in Figures 9-3b and 9-4b. An additional unidentified peak at approximately 1057 cm -1 was present in both adhesives, though it had evolved fully and was no longer present by 20-30 min into the cure cycle for both materials. 142 Figure 9-5. FTIR spectra of off-gassing components The results obtained from the TGA-FTIR analysis are consistent with the synthesis process for an imidizable condensation cure polyimide (see Figure 9-6) [101]. Commonly, the cure process for a polyimide occurs in two steps. During the first step, polycondensation of an aromatic dianhydride and an aromatic diamine occurs [101]. This process is performed by dissolving the diamine in an appropriate solvent, such as dimethylacetamide (DMAC) or NMP, followed by addition of the dianhydride [102]. The second step, usually at higher temperature (150-250°C), consists of the dehydration of the poly(amic acid) which results in the final polyimide structure. A 143 known drawback to such polyimide materials is formation of voids that can occur due to the release of moisture during the cure process [101]. Thus, for the film adhesives studied here, the liberation of moisture may occur during a high-temperature post-cure, resulting in defects. A schematic showing the synthesis and cure process of a typical polyimide material is presented in Figure 9-6. Figure 9-6. Example of the polymerization process for a polyimide [102] Knowledge of the synthesis process for the polyimide adhesives studied indicates that the primary off-gassing behavior during cure at 177° C is due to water and residual NMP solvent. In addition to identification of primary volatile components, the information obtained from TGA-FTIR experiments provides insight into the time and temperature at which gaseous species evolve. For both materials, using a ramp rate of 10° 144 C/min, approximately 60% of the total weight loss occurs during the temperature ramp, with the additional 40% taking place during the isothermal hold at 177° C. With a slower ramp rate of 2° C/min, however, 77% of sample weight loss is completed during the temperature ramp, indicating that a slower ramp rate can be used to more effectively remove volatile species. Calculations for total sample weight loss (in wt%) during each portion of the test (ramp and hold) are presented in Table 9-3. Table 9-3. Weight loss during temperature ramp and hold Adhesive A Adhesive B Sample # % Wt Loss Ramp % Wt Loss Hold Sample # % Wt Loss Ramp % Wt Loss Hold 1 60% 40% 1 60% 40% 2 60% 40% 2 62% 38% 3 77% 23% 3 77% 23% 9.3.3 Adhesive pre-treatment A goal of this work was to apply information obtained from TGA-FTIR testing to develop pre-treatments to reduce/eliminate adhesive foaming during the manufacture of sandwich structures. The purpose of a pre-treatment is to reduce out-gassing of volatile components prior to lay-up and cure. However, challenges arise when developing pretreatments for condensation curing polymers. For the film adhesives examined here, the release of volatiles can coincide with the polymerization process (cure reaction). Thus, removing volatile components, through high-temperature pre-treatment may reduce the processability of the material by increasing degree of cure and altering viscosity. As a starting point, a thermal treatment of 110°C for 30 min was implemented. The temperature was chosen to ensure removal of surface moisture. Based on the rate of 145 weight loss and dependence on ramp rate shown in Table 9-3, temperature was ramped to 100°C at a rate of 2°C/min. TGA plots for this pre-treatment are presented in Figure 9-7, along with total sample weight loss. Approximately 40% of the volatile components that evolve during cure at 177° C are removed by this thermal treatment. Figure 9-7. TGA weight loss curves during isothermal pre-treatment at 110° C for (a) adhesive A and (b) adhesive B To investigate the influence of thermal pre-treatments on processability, rheological testing was performed. Sample rheology was measured using a parallel plate rheometer and a temperature ramp from room temperature to 177° C at 2° C/min. Viscosity data is shown in Figure 9-8. Inset photographs show that the adhesive foaming that occurred during testing was more pronounced in adhesive A. 146 Figure 9-8. Viscosity during cure for (a) adhesive A and (b) adhesive B. Photographs show foaming At 110° C, the viscosity of both adhesives had already begun to increase (dashed lines in Figure 9-8). Thus, a lower temperature pre-treatment would be more appropriate from a standpoint of material processability. From the data obtained in this study, a pre- treatment involving a slower ramp (<1° C /m in) to 70 -75° C, followed by an isothermal hold would be appropriate. 9.4 Conclusions Coupled dynamic TGA-FTIR analysis was used to identify out-gassing components during the cure of two condensation polyimide film adhesives. The main 147 volatile component to evolve during a 177° C cure cycle was 1-methyl-2-pyrrolidinone, or NMP, a residual solvent from the synthesis process. An adhesive pre-treatment at 110° C was introduced to allow residual solvent to outgas gradually (without foaming). During this pre-treatment, 40% of volatiles were removed. However, results of rheological measurements indicate that a lower temperature pre-treatment is required to maintain low processing viscosity. The data obtained in this study also revealed greater volatile evolution and foaming behavior in adhesive A, indicating that adhesive B would be a better candidate for out-of-autoclave cure. However, adhesive B is much more expensive ($220 per square foot vs. $18 per square foot for adhesive A) [103]. 148 CHAPTER 10. Conclusions and future work 10.1 Conclusions The aerospace industry is moving toward the use of more composite parts, and the demand for air travel is increasing, making new composite manufacturing processes critical. Traditional autoclave cure methods are expensive and inefficient, and cannot meet projected production rates. For this reason, new out-of-autoclave processing methods have been introduced. One such method is vacuum bag only (VBO) processing of prepregs. With VBO processing, the maximum pressure differential during cure is 1 atm. The main concern with such low pressure processing methods is void formation. In this work, we have examined three leading sources of void formation in VBO processed parts; entrapped air, evolved gasses, and insufficient resin flow. The ability to produce VBO parts with autoclave equivalent mechanical properties was shown in Chapter 3. Further improvement of part quality with changes in material cure cycle was discussed in Chapter 4. Subsequent chapters addressed the influence of specific process parameters on defects in VBO cured composites. Impregnation, compaction, and air removal mechanisms were examined in Chapter 5 in unidirectional and woven prepreg. The mechanism of air removal was found to differ as a function of fiber architecture, with void removal occurring on a longer time scale for more complex woven fabric. This result indicates that proper time and temperature cure cycles are required to remove air and facilitate compaction, and that such cure cycles should be tailored to fiber geometry. Additionally, entrapped air was shown to lead to void 149 formation in the absence of adequate compaction pressure. Thus, full vacuum is required to remove trapped air and eliminate voids in VBO processed parts (Chapter 6). Another source of voids in VBO composites is the evolution of volatile components during cure. Because uncured epoxy resins absorb moisture from the air, the influence of resin moisture content on void formation was investigated (Chapter 6). Under autoclave pressures, moisture was held in solution, and no void formation was observed. With VBO processing, however, an exponential increase in void content with humidity exposure was noted. Relative humidity exposure of greater than 45% RH was shown to lead to voids in VBO processed parts, emphasizing the need to control the lay- up environment when low pressure processing methods are used. The evolution of volatiles was also examined for the cure of composite sandwich structures (Chapter 9). The out-gassing behavior of polyimide film adhesives used in the co-cure of sandwich panels was studied using a method of coupled TGA-FTIR analysis. Volatile species were identified, and the evolution of off-gassing components was tracked as a function of cure time and temperature. With such kinetic information, adhesive pre- treatments can be developed to reduce off-gassing, and diminish adhesive foaming and bond-line voids during sandwich structure co-cure. The final source of void formation examined was insufficient resin flow due to material aging. When exposed to ambient conditions, the degree of cure of epoxy resins will increase with time. This room temperature aging leads to increased resin viscosity, and reduced flow during cure, resulting in voids. The influence of aging on various material properties was presented in Chapter 7. Aging was shown to adversely influence 150 prepreg tack, material compaction, and mechanical properties. In Chapter 8, a method for monitoring prepreg chemical age was developed. This method involves monitoring changes in the B-stage glass transition temperature of the epoxy in prepreg, which increases linearly as a function of aging time. The age monitoring method developed is suitable for production floor testing, allowing for identification of material age and prediction of cured part quality. The work presented here provides an improved understanding of the mechanisms of void formation in low pressure VBO manufactured composite parts. By examining various process parameters that influence void content, optimized manufacturing protocols were developed. While VBO processing of void-free parts presents challenges, high-quality laminates can be produced when the proper lay-up and cure protocols are implemented. The experiments performed in this work provide valuable insight into the feasibility of VBO processing for large scale structural components. 10.2 Recommendations for future work The experimental work presented here was performed using small, flat panels. While flat panels are useful for examining fundamental issues of void formation, real parts contain complex geometries and additional features that pose manufacturing difficulties. Future efforts in the study of VBO processing of composites should include complex part geometries, and studies on the scale-up of VBO manufacturing to large part sizes. Also relevant to the scale-up of VBO manufacturing methods will be the development of robust process models. Experimental results and observations, such as 151 those presented here, provide critical insight into the mechanisms of void formation and growth. However, as VBO processing is applied to the manufacture of more complex parts, process models will be required. These models incorporate resin rheological behavior, material cure kinetics, thermal stability and reinforcement characteristics (fiber bed compaction, in-plane and through-thickness permeability, etc.), and can be used to develop appropriate temperature cure cycles for composite parts. 10.3 Broader implications The research presented here was focused on composite manufacturing for aerospace applications, as VBO manufacturing methods will enable increased use of composites in the aerospace industry, increased aircraft fuel efficiency, and reduced emissions. However, a greater understanding of the factors influencing void content and part quality in low pressure processed composites will have additional economic and technological impact. Eliminating the high capital investment and operating costs associated with autoclave manufacturing opens the field of composite processing to a larger number of manufacturers, increasing competition among part suppliers in the global market. Additionally, VBO methods eliminate the size constraints imposed by autoclave processing, allowing for the production of large composite components. Specifically, out-of-autoclave VBO processing could be used for the manufacture of wind turbine blades, contributing to green energy efforts. Additionally, lowering the cost of composite manufacturing will increase the use of composites outside of the aerospace industry, specifically in marine transportation, recreational equipment, wind energy applications, and civic infrastructure. 152 Bibliography 1. Tavares, S.S., V. Michaud, and J.A.E. Manson, Through thickness air permeability of prepregs during cure. Composites: Part A, 2009. 40: p. 1587- 1596. 2. Thomas, S. and S.R. Nutt, In situ estimation of through-thickness resin flow using ultrasound. Composites Science and Technology, 2008. 68: p. 3093-3098. 3. Tavares, S.S., V. Michaud, and J.A.E. Manson, Assessment of semi-impregnated fabrics in honeycomb sandwich structures. Composites: Part A, 2010. 41: p. 8-15. 4. Jackson, K. and M. Crabtree. Autoclave quality composites tooling for composite from vacuum bag only processing. in 47th International SAMPE Symposium. 2002. 5. Repecka, L. and J. Boyd, Vacuum-bag-only-curable prepregs that produce void free parts, in 47th International SAMPE Symposium. 2002. p. 1862-1875. 6. Liu, L., et al., Effects of cure cycles on void content and mechanical properties of composite laminates. Composite Structures, 2006. 73: p. 303-309. 7. Liu, L., et al., Effects of cure pressure induced voids on the mechanical strength of carbon/epoxy laminates. Journal of Materials Science and Technology, 2005. 21(1): p. 87-91. 8. Olivier, P., J.P. Cottu, and B. Ferret, Effects of cure cycle pressure and voids on some mechanical properties of carbon/epoxy laminates. Composites, 1995. 26(7): p. 509-515. 9. Tang, J., W.I. Lee, and G.S. Springer, Effects of cure pressure on resin flow, voids, and mechanical properties. Journal of Composite Materials, 1987. 21: p. 421-440. 10. Boey, F.Y.C. and S.W. Lye, Void reduction in autoclave processing of thermoset composites part 1: high pressure effects on void reduction. Composites, 1992. 23(4): p. 261-265. 11. Campbell, F.C. and A.R. Mallow, Porosity in carbon fiber composites, an overview of causes. Journal of Advanced Materials, 1994: p. 18-33. 12. Wood, J.R. and M.G. Bader, Void control for polymer-matrix composites (1): theoretical and experimental methods for determining the growth and collapse of gas bubbles. Composites Manufacturing, 1994. 5(3): p. 139-147. 153 13. Wood, J.R. and M.G. Bader, Void control for polymer-matrix composites (2): experimental evaluation of a diffusion model for the growth and collapse of gas bubbles. Composites Manufacturing, 1994. 5(3): p. 149-158. 14. Huang, H. and R. Talreja, Effects of void geometry on elastic properties of unidirectional fiber reinforced composites. Composites Science and Technology, 2005. 65: p. 1964-1981. 15. Costa, M.L., S.F.M. Almeida, and M.C. Rezende, The influence of porosity on the interlaminar shear strength of carbon/epoxy and carbon/bismaleimide fabric laminates. Composites Science and Technology, 2001. 61: p. 2101-2018. 16. Thorfinnson, B. and T.F. Biermann. Measurement and control prepreg impregnation for elimination of porosity in composite parts. in Fabricating Composites. 1988. Philadelphia, Pennsylvania. 17. Ridgard, C. Out of autoclave composite technology for aerospace, defense and space structures. in SAMPE. 2009. Baltimore, MD. 18. Molyneux, M., Prepreg, tape and fabric technology for advanced composites. Composites, 1983. 14(2): p. 87-91. 19. Thomas, S., Vacuum-bag-only processing of composites, in Materials Science. 2009, University of Southern California: Los Angeles. 20. Kardos, J.L., M.P. Dudukovic, and R. Dave, Void growth and resin transport during processing of thermosetting-matrix composites. Advances in Polymer Science, 1986. 80: p. 102-123. 21. Steele, M. and T. Corden, New prepregs for cost effective, out-of-autoclave tool and component manufacture. SAMPE Journal, 2004. 40(2): p. 30-34. 22. Bongiovanni, C., VBO Processing Fundamentals. 2010, Hitco Carbon Composites, Inc. 23. Cytec, CYCOM 5320 Information Sheet. 2009. 24. Grunenfelder, L.K. and S.R. Nutt, Void formation in composite prepregs - effect of dissolved moisture. Composites Science and Technology, 2010. 70: p. 2304- 2309. 25. Scola, D.A. and J. Vontell, Effects of ambient aging of 5245C/graphite prepreg on composition and mechanical properties of fabricated composites. Polymer Composites, 1987. 8(4): p. 244-252. 154 26. Jones, R.W., Y. Ng, and J.F. McClelland, Monitoring ambient-temperature aging of a carbon-fiber/epoxy composite prepreg with photoacoustic spectroscopy. Composites: Part A, 2008. 39: p. 965-971. 27. Crabtree, D.J. Quantitative analysis of resin matrix aging by gel permeation chromatography and differential scanning calorimetry. in ASC Symposium Series. 1980: American Chemical Society. 28. Miller, S.G., et al. Out-life characteristics of IM7/977-3 composites. in SAMPE tech. 2010. 29. Gu, Y., et al., Effects of resin storage aging on rheological property and consolidation of composite laminates. Polymer Composites, 2009: p. 1081-1090. 30. Frigione, M. and J.M. Kenny, Thermokinetic effects of the aging of epoxy matrix prepregs for high performance composites. Polymer Composites, 2002. 23(4): p. 530-537. 31. Akay, M., Effects of prepreg ageing and post-cure hygrothermal conditioning on the mechanical behaviour of carbon-fibre/epoxy laminates. Composites Science and Technology, 1990. 38: p. 359-370. 32. Yu, Y., H. Su, and W. Gan, Effects of storage aging on the properties of epoxy prepregs. Industrial Chemical Engineering Research, 2009. 48: p. 4340-4345. 33. Kratz, J. and P. Hubert, Processing out-of-autoclave honeycomb structures: Internal core pressure measurements. Composites: Part A, 2011. 42: p. 1060- 1065. 34. Kratz, J. and P. Hubert. Out-of-autoclave manufacturing of composite aircraft sandwich structures. in ICCM 17, 17th International Conference on Composite Materials. 2009. Edinburgh, UK. 35. Tavares, S.S., et al., Vacuum-bag processing of sandwich structures: Role of honeycomb pressure level on skin-core adhesion and skin quality. Composites Science and Technology, 2010. 70: p. 797-803. 36. Kamykowski, G. TA Training Seminars: Rheology. 2004. 37. Karl Fisher Titration: Good titration practice in Karl Fisher titration. Mettler- Toledo 2011. 38. Karl Fischer Titration: One step water determination. Simple and secure. Mettler-Toledo 2011. 155 39. Introduction to Fourier transform infrared spectrometry. Thermo Nicolet Corporation, 2001. 40. Breitigam, W.V. and R.S. Bauer, Novel processing and cure of epoxy resin systems. Polymer, 1992. 34(4): p. 767-771. 41. ASTM, D2344, in Standard test method for short-beam strength of polymer matrix composites and their laminates. 2000, ASTM International: United States. 42. Airbus, AITM 1-0002, in Fiber reinforced plastics determination of in-plane shear properties (+/- 45 tensile test). 1998, Airbus Industrie: France. 43. Airbus, AITM 1-0010, in Fiber reinforced plastics determination of compression strength after impact. 2005, Airbus S.A.S.: France. 44. Airbus, AITM 1-0008, in Fiber reinforced plastics determination of plain, open hole and filled hole compression strength. 2004, Airbus S.A.S.: France. 45. Airbus, AIMS05-01-013, in Carbon fiber reinforced epoxy prepreg twill 2 x 2 weave style/180C curing class epoxy out-of-autoclave cure/standard modulus fibre structural material. 2007, Airbus Industrie: France. 46. Hexcel, Hexply M21 datasheet, Publication FTA002d. 2008. 47. Cytec, Cytec 997-2 datasheet. 1995. 48. White, S.R. and H.T. Hahn, Cure cycle optimization for the reduction of processing-induced residual stresses in composite materials. Journal of Composite Materials, 1993. 27(14): p. 1352-1377. 49. Frank-Susich, D., D.H. Laananen, and D. Ruffner, Cure cycle simulation for thermoset composites. Composites Manufacturing, 1993. 4(3): p. 139-146. 50. ASTM, D2734, in Standard test methods for void content of reinforced plastics. 2009, ASTM International: United States. 51. Choi, W.J. Out of autoclave (OOA) process. in ASAP Technical Meeting. 2009. Toulouse, France. 52. Farhang, L. and G. Fernlund. Void evolution and gas transport during cure in out- of-autoclave prepreg laminates. in SAMPE Conference. 2011. Long Beach, CA. 53. Centea, T. and P. Hubert, Measuring the impregnation of an out-of-autoclave prepreg by micro-CT. Composites Science and Technology, 2011. 71: p. 593-599. 156 54. Centea, T. and P.Hubert, Modelling the effect of material properties and process parameters on tow impregnation in out-of-autoclave prepregs. Composites: Part A, 2012. 55. Loos, A.C. and G.S. Springer, Curing of epoxy matrix composites. Journal of Composite Materials, 1983. 17: p. 135-169. 56. Young, W.-B., Resin flow analysis in the consolidation of multi-directional laminated composites. Polymer Composites, 1995. 16(3): p. 250-257. 57. Young, W.-B., Consolidation and cure simulations for laminated composites. Polymer Composites, 1996. 17(1): p. 142-148. 58. Garschke, C., et al., Out-of-autoclave cure cycle study of a resin film infusion process using in situ process monitoring. Composites: Part A, 2012. 43: p. 935- 944. 59. Dave, R., J.L. Kardos, and M.P. Dudukovic, A model for resin flow during composite processing: Part 1 - General mathematical development. Polymer Composites, 1987. 8(1): p. 29-38. 60. Thomas, S. and S.R. Nutt, Temperature dependence of resin flow in a resin film infusion (RFI) process by ultrasound imaging. Applied Composite Materials, 2009. 16: p. 183-196. 61. Lee, W.I., A.C. Loos, and G.S. Springer, Heat of reaction, degree of cure, and viscosity of Hercules 3501-6 resin. Journal of Composite Materials, 1982. 16: p. 510-520. 62. Dave, R., J.L. Kardos, and M.P. Dudukovic, A model for resin flow during composite processing. Part 2: Numerical analysis for unidirectional graphite/epoxy laminates. Polymer Composites, 1987. 8(2): p. 123-132. 63. Ahn, K.J., J.C. Seferis, and J.C. Berg, Simultaneous measurements of permeability and capillary pressure of thermosetting matrices in woven fabric reinforcements. Polymer Composites, 1991. 12(3): p. 146-152. 64. Rudd, C.D., et al., Compaction and in plane permeability characteristics of quasi unidirectional and continuous random reinforcements. Materials Science and Technology, 1996. 12: p. 436-444. 65. Lekakou, C. and M.G. Bader, Mathematical modelling of macro- and micro- infiltration in resin transfer moulding (RTM). Composites: Part A, 1998. 29A: p. 29-37. 157 66. Griffin, P.R., et al., The effect of reinforcement architecture on the long-range flow in fibrous reinforcements. Composites Manufacturing, 1995. 6(3-4): p. 221- 235. 67. Ciriscioli, P.R., Q. Wang, and G.S. Springer, Autoclave curing - Comparisons of model and test results. Journal of Composite Materials, 1992. 26(1): p. 90-102. 68. Saunders, R.A., C. Lekakou, and M.G. Bader, Compression in the processing of polymer composites 1. A mechanical and microstructural study for different glass fabrics and resins. Composites Science and Technology, 1999. 59: p. 983-993. 69. Scriven, L.E., On the dynamics of phase growth. Chemical Engineering Science, 1959. 10(1/2): p. 1-13. 70. Subramanian, R.S. and M.C. Weinberg, The role of convective transport in the dissolution or growth of a gas bubble. Journal of Chemical Physics, 1980. 72(12): p. 6811-6813. 71. Delasi, R. and J.B. Whiteside, Effect of moisture on epoxy resins and composites, in Advanced Composite Materials - Environmental Effects. 1978, ASTM. p. 2-20. 72. Tung, C.Y.M. and P.J. Dynes, Relationship between viscoelastic properties and gelation in thermosetting systems. Journal of Applied Polymer Science, 1982. 27: p. 569-574. 73. Ahn, K.J., et al., Prepreg aging in relation to tack. Journal of Applied Polymer Science, 1992. 45: p. 399-406. 74. Sanjana, Z.N., W.H. Schaefer, and J.R. Ray, Effect of aging and moisture on the reactivity of a graphite epoxy prepreg. Polymer Engineering and Science, 1981. 21(8): p. 474-482. 75. Vuong, T. An out time study on PMR-15 polyimide resin. in 32nd International SAMPE Symposium. 1987. 76. Ji, K.J., et al., Evaluation of glass fibre/epoxy prepreg quality during storage. Polymers & Polymer Composites, 2002. 10(8): p. 599-606. 77. Kong, E.S., S.M. Lee, and H.G. Nelson, Physical aging in graphite/epoxy composites. Polymer Composites, 1982. 3(1): p. 29-33. 78. Cole, K.C., et al., Room temperature aging of Narmco 5208 carbon-epoxy prepreg. Part I: physicochemical characterization. Polymer Composites, 1989. 10(3): p. 150-161. 158 79. Cole, K.C., et al., Room temperature aging of Narmco 5208 carbon-epoxy prepreg. Part II: physical, mechanical, and nondestructive characterization. Polymer Composites, 1991. 12(3): p. 203-212. 80. Gill, P.S., S.R. Sauerbrunn, and M. Reading, Modulated differential scanning calorimetry. Journal of Thermal Analysis, 1993. 40: p. 931-939. 81. Ahn, K.J., et al., Analysis and characterization of prepreg tack. Polymer Composites, 1992. 13(3): p. 197-206. 82. Grunenfelder, L.K. and S.R. Nutt. Out time effects on VBO prepreg and laminate properties. in SAMPE. 2011. 83. Hagnauer, G.L., Quality assurance of epoxy resin prepregs using liquid chromatography. Polymer Composites, 1980. 1(2): p. 81-87. 84. Verdonck, E., K. Schaap, and L.C. Thomas, A discussion of the principles and applications of modulated temperature DSC (MTDSC). International Journal of Pharmaceutics, 1999. 192: p. 3-20. 85. Stutz, H., K.H. Illers, and J. Mertes, A generalized theory for the glass transition temperature of crosslinked and uncrosslinked polymers. Journal of Polymer Science: Part B: Polymer Physics, 1990. 28: p. 1483-1498. 86. Fox, T.G. and S. Loshaek, Influence of molecular weight and degree of crosslinking on the specific volume and glass temperature of polymers. Journal of Polymer Science, 1955. 15: p. 371-390. 87. Fraga, I., S. Montserrat, and J.M. Hutchinson, Vitrification during the isothermal cure of thermosets: comparison of theoretical simulations with temperature- modulated DSC and dielectric analysis. Macromolecular Chemistry and Physics, 2008. 209(19): p. 2003-2011. 88. Renn, D., et al., Composite honeycomb core crush in relation to internal pressure measurement. Journal of Advanced Materials, 1995: p. 31-40. 89. Rion, J., Y. Leterrier, and J.-A.E. Manson, Prediction of the adhesive fillet size for skin to honeycomb core bonding in ultra-light sandwich structures. Composites: Part A, 2008. 39: p. 1547-1555. 90. Harmon, B., J. Boyd, and B. Thai. Advanced products designed to simplify co- cure over honeycomb core. in SAMPE Conference. 2002. Long Beach, CA. 159 91. Butukuri, R.R., et al., Evaluation of skin-core adhesion bond of out-of-autoclave honeycomb sandwich structures. Journal of Reinforced Plastics and Composites, 2012. 31(5): p. 331-339. 92. Grove, S., E. Popham, and M. Miles, An investigation of the skin/core bond in honeycomb sandwich structures using statistical experimental techniques. Composites: Part A, 2006. 37: p. 804-812. 93. Yuan, C., et al., Experimental investigation on the co-cure processing of honeycomb structure with self-adhesive prepreg. Applied Composite Materials, 2008. 15: p. 47-59. 94. Hou, T., et al., Evaluation of sandwich structure bonding in out-of-autoclave (OOA) processing. SAMPE Journal, 2011. 47(1): p. 32-39. 95. Nagarajan, S., et al., Out-of-autoclave sandwich structure: Processing study. SAMPE Journal, 2012. 48(4): p. 24-31. 96. Pearce, P., et al., Cause and effect of void formation during vacuum bag curing of epoxy film adhesives. Journal of Adhesion Science and Technology, 1998. 12(6): p. 567-584. 97. Marini, A., et al., Factors affecting the spectral response in a TG/FT-IR experiment. Applied Spectroscopy, 1994. 48(12): p. 1468-1471. 98. Turk, M., et al., Evaluation of the thermal oxidative stability of polyimides via TGA techniques. Journal of Polymer Science: Part A: Polymer Chemistry, 1999. 37: p. 3943-3956. 99. Cervantes-Uc, J., et al., TGA/FTIR studies of segmented aliphatic polyurethanes and their nanocomposites prepared with commercial montmorillonites. Polymer Degradation and Stability, 2009. 94: p. 1666-1677. 100. Xie, W., W. Pam, and K. Chuang, Thermal characterization of PMR polyimides. Thermochimica Acta, 2001. 367-368: p. 143-153. 101. Fried, J.R., Polymer Science and Technology. 2nd ed. 2009, Upper Saddle River: Prentice-Hall, Inc. 102. Scola, D.A., Polyimide Resins. ASM Handbook Volume 21. 2001: ASM International. 103. Shiverdecker, K.M. Design of a field-level on-aircraft polyimide adhesively bonded repair process. in NASA/Ohio Space Grant Consortium Annual Student Research Symposium. 2003-2004. 160 Bibliography (Alphabetical) Ahn, K.J., et al., Analysis and characterization of prepreg tack. Polymer Composites, 1992. 13(3): p. 197-206. Ahn, K.J., et al., Prepreg aging in relation to tack. Journal of Applied Polymer Science, 1992. 45: p. 399-406. Ahn, K.J., J.C. Seferis, and J.C. Berg, Simultaneous measurements of permeability and capillary pressure of thermosetting matrices in woven fabric reinforcements. Polymer Composites, 1991. 12(3): p. 146-152. Airbus, AIMS05-01-013, in Carbon fiber reinforced epoxy prepreg twill 2 x 2 weave style/180C curing class epoxy out-of-autoclave cure/standard modulus fibre structural material. 2007, Airbus Industrie: France. Airbus, AITM 1-0002, in Fiber reinforced plastics determination of in-plane shear properties (+/- 45 tensile test). 1998, Airbus Industrie: France. Airbus, AITM 1-0008, in Fiber reinforced plastics determination of plain, open hole and filled hole compression strength. 2004, Airbus S.A.S.: France. Airbus, AITM 1-0010, in Fiber reinforced plastics determination of compression strength after impact. 2005, Airbus S.A.S.: France. Akay, M., Effects of prepreg ageing and post-cure hygrothermal conditioning on the mechanical behaviour of carbon-fibre/epoxy laminates. Composites Science and Technology, 1990. 38: p. 359-370. ASTM, D2344, in Standard test method for short-beam strength of polymer matrix composites and their laminates. 2000, ASTM International: United States. ASTM, D2734, in Standard test methods for void content of reinforced plastics. 2009, ASTM International: United States. Boey, F.Y.C. and S.W. Lye, Void reduction in autoclave processing of thermoset composites part 1: high pressure effects on void reduction. Composites, 1992. 23(4): p. 261-265. Bongiovanni, C., VBO Processing Fundamentals. 2010, Hitco Carbon Composites, Inc. Breitigam, W.V. and R.S. Bauer, Novel processing and cure of epoxy resin systems. Polymer, 1992. 34(4): p. 767-771. 161 Butukuri, R.R., et al., Evaluation of skin-core adhesion bond of out-of-autoclave honeycomb sandwich structures. Journal of Reinforced Plastics and Composites, 2012. 31(5): p. 331-339. Campbell, F.C. and A.R. Mallow, Porosity in carbon fiber composites, an overview of causes. Journal of Advanced Materials, 1994: p. 18-33. Centea, T. and P. Hubert, Measuring the impregnation of an out-of-autoclave prepreg by micro-CT. Composites Science and Technology, 2011. 71: p. 593-599. Centea, T. and P.Hubert, Modelling the effect of material properties and process parameters on tow impregnation in out-of-autoclave prepregs. Composites: Part A, 2012. Cervantes-Uc, J., et al., TGA/FTIR studies of segmented aliphatic polyurethanes and their nanocomposites prepared with commercial montmorillonites. Polymer Degradation and Stability, 2009. 94: p. 1666-1677. Choi, W.J. Out of autoclave (OOA) process. in ASAP Technical Meeting. 2009. Toulouse, France. Ciriscioli, P.R., Q. Wang, and G.S. Springer, Autoclave curing - Comparisons of model and test results. Journal of Composite Materials, 1992. 26(1): p. 90-102. Cole, K.C., et al., Room temperature aging of Narmco 5208 carbon-epoxy prepreg. Part I: physicochemical characterization. Polymer Composites, 1989. 10(3): p. 150-161. Cole, K.C., et al., Room temperature aging of Narmco 5208 carbon-epoxy prepreg. Part II: physical, mechanical, and nondestructive characterization. Polymer Composites, 1991. 12(3): p. 203-212. Costa, M.L., S.F.M. Almeida, and M.C. Rezende, The influence of porosity on the interlaminar shear strength of carbon/epoxy and carbon/bismaleimide fabric laminates. Composites Science and Technology, 2001. 61: p. 2101-2018. Crabtree, D.J. Quantitative analysis of resin matrix aging by gel permeation chromatography and differential scanning calorimetry. in ASC Symposium Series. 1980: American Chemical Society. Cytec, CYCOM 5320 Information Sheet. 2009. Cytec, Cytec 997-2 datasheet. 1995. Dave, R., J.L. Kardos, and M.P. Dudukovic, A model for resin flow during composite processing. Part 2: Numerical analysis for unidirectional graphite/epoxy laminates. Polymer Composites, 1987. 8(2): p. 123-132. 162 Dave, R., J.L. Kardos, and M.P. Dudukovic, A model for resin flow during composite processing: Part 1 - General mathematical development. Polymer Composites, 1987. 8(1): p. 29-38. Delasi, R. and J.B. Whiteside, Effect of moisture on epoxy resins and composites, in Advanced Composite Materials - Environmental Effects. 1978, ASTM. p. 2-20. Farhang, L. and G. Fernlund. Void evolution and gas transport during cure in out-of-autoclave prepreg laminates. in SAMPE Conference. 2011. Long Beach, CA. Fox, T.G. and S. Loshaek, Influence of molecular weight and degree of crosslinking on the specific volume and glass temperature of polymers. Journal of Polymer Science, 1955. 15: p. 371-390. Fraga, I., S. Montserrat, and J.M. Hutchinson, Vitrification during the isothermal cure of thermosets: comparison of theoretical simulations with temperature-modulated DSC and dielectric analysis. Macromolecular Chemistry and Physics, 2008. 209(19): p. 2003-2011. Frank-Susich, D., D.H. Laananen, and D. Ruffner, Cure cycle simulation for thermoset composites. Composites Manufacturing, 1993. 4(3): p. 139-146. Fried, J.R., Polymer Science and Technology. 2nd ed. 2009, Upper Saddle River: Prentice- Hall, Inc. Frigione, M. and J.M. Kenny, Thermokinetic effects of the aging of epoxy matrix prepregs for high performance composites. Polymer Composites, 2002. 23(4): p. 530-537. Garschke, C., et al., Out-of-autoclave cure cycle study of a resin film infusion process using in situ process monitoring. Composites: Part A, 2012. 43: p. 935-944. Gill, P.S., S.R. Sauerbrunn, and M. Reading, Modulated differential scanning calorimetry. Journal of Thermal Analysis, 1993. 40: p. 931-939. Griffin, P.R., et al., The effect of reinforcement architecture on the long-range flow in fibrous reinforcements. Composites Manufacturing, 1995. 6(3-4): p. 221-235. Grove, S., E. Popham, and M. Miles, An investigation of the skin/core bond in honeycomb sandwich structures using statistical experimental techniques. Composites: Part A, 2006. 37: p. 804-812. Grunenfelder, L.K. and S.R. Nutt, Void formation in composite prepregs - effect of dissolved moisture. Composites Science and Technology, 2010. 70: p. 2304-2309. 163 Grunenfelder, L.K. and S.R. Nutt. Out time effects on VBO prepreg and laminate properties. in SAMPE. 2011. Gu, Y., et al., Effects of resin storage aging on rheological property and consolidation of composite laminates. Polymer Composites, 2009: p. 1081-1090. Hagnauer, G.L., Quality assurance of epoxy resin prepregs using liquid chromatography. Polymer Composites, 1980. 1(2): p. 81-87. Harmon, B., J. Boyd, and B. Thai. Advanced products designed to simplify co-cure over honeycomb core. in SAMPE Conference. 2002. Long Beach, CA. Hexcel, Hexply M21 datasheet, Publication FTA002d. 2008. Hou, T., et al., Evaluation of sandwich structure bonding in out-of-autoclave (OOA) processing. SAMPE Journal, 2011. 47(1): p. 32-39. Huang, H. and R. Talreja, Effects of void geometry on elastic properties of unidirectional fiber reinforced composites. Composites Science and Technology, 2005. 65: p. 1964-1981. Introduction to Fourier transform infrared spectrometry. Thermo Nicolet Corporation, 2001. Jackson, K. and M. Crabtree. Autoclave quality composites tooling for composite from vacuum bag only processing. in 47th International SAMPE Symposium. 2002. Ji, K.J., et al., Evaluation of glass fibre/epoxy prepreg quality during storage. Polymers & Polymer Composites, 2002. 10(8): p. 599-606. Jones, R.W., Y. Ng, and J.F. McClelland, Monitoring ambient-temperature aging of a carbon- fiber/epoxy composite prepreg with photoacoustic spectroscopy. Composites: Part A, 2008. 39: p. 965-971. Kamykowski, G. TA Training Seminars: Rheology. 2004. Kardos, J.L., M.P. Dudukovic, and R. Dave, Void growth and resin transport during processing of thermosetting-matrix composites. Advances in Polymer Science, 1986. 80: p. 102-123. Karl Fischer Titration: One step water determination. Simple and secure. Mettler-Toledo 2011. Karl Fisher Titration: Good titration practice in Karl Fisher titration. Mettler-Toledo 2011. 164 Kong, E.S., S.M. Lee, and H.G. Nelson, Physical aging in graphite/epoxy composites. Polymer Composites, 1982. 3(1): p. 29-33. Kratz, J. and P. Hubert, Processing out-of-autoclave honeycomb structures: Internal core pressure measurements. Composites: Part A, 2011. 42: p. 1060-1065. Kratz, J. and P. Hubert. Out-of-autoclave manufacturing of composite aircraft sandwich structures. in ICCM 17, 17th International Conference on Composite Materials. 2009. Edinburgh, UK. Lee, W.I., A.C. Loos, and G.S. Springer, Heat of reaction, degree of cure, and viscosity of Hercules 3501-6 resin. Journal of Composite Materials, 1982. 16: p. 510-520. Lekakou, C. and M.G. Bader, Mathematical modelling of macro- and micro-infiltration in resin transfer moulding (RTM). Composites: Part A, 1998. 29A: p. 29-37. Liu, L., et al., Effects of cure cycles on void content and mechanical properties of composite laminates. Composite Structures, 2006. 73: p. 303-309. Liu, L., et al., Effects of cure pressure induced voids on the mechanical strength of carbon/epoxy laminates. Journal of Materials Science and Technology, 2005. 21(1): p. 87-91. Loos, A.C. and G.S. Springer, Curing of epoxy matrix composites. Journal of Composite Materials, 1983. 17: p. 135-169. Marini, A., et al., Factors affecting the spectral response in a TG/FT-IR experiment. Applied Spectroscopy, 1994. 48(12): p. 1468-1471. Miller, S.G., et al. Out-life characteristics of IM7/977-3 composites. in SAMPE tech. 2010. Molyneux, M., Prepreg, tape and fabric technology for advanced composites. Composites, 1983. 14(2): p. 87-91. Nagarajan, S., et al., Out-of-autoclave sandwich structure: Processing study. SAMPE Journal, 2012. 48(4): p. 24-31. Olivier, P., J.P. Cottu, and B. Ferret, Effects of cure cycle pressure and voids on some mechanical properties of carbon/epoxy laminates. Composites, 1995. 26(7): p. 509-515. Pearce, P., et al., Cause and effect of void formation during vacuum bag curing of epoxy film adhesives. Journal of Adhesion Science and Technology, 1998. 12(6): p. 567-584. 165 Renn, D., et al., Composite honeycomb core crush in relation to internal pressure measurement. Journal of Advanced Materials, 1995: p. 31-40. Repecka, L. and J. Boyd, Vacuum-bag-only-curable prepregs that produce void free parts, in 47th International SAMPE Symposium. 2002. p. 1862-1875. Ridgard, C. Out of autoclave composite technology for aerospace, defense and space structures. in SAMPE Conference. 2009. Baltimore, MD. Ridgard, C. Out of autoclave composite technology for aerospace, defense and space structures. in SAMPE. 2009. Baltimore, MD. Rion, J., Y. Leterrier, and J.-A.E. Manson, Prediction of the adhesive fillet size for skin to honeycomb core bonding in ultra-light sandwich structures. Composites: Part A, 2008. 39: p. 1547-1555. Rudd, C.D., et al., Compaction and in plane permeability characteristics of quasi unidirectional and continuous random reinforcements. Materials Science and Technology, 1996. 12: p. 436-444. Sanjana, Z.N., W.H. Schaefer, and J.R. Ray, Effect of aging and moisture on the reactivity of a graphite epoxy prepreg. Polymer Engineering and Science, 1981. 21(8): p. 474-482. Saunders, R.A., C. Lekakou, and M.G. Bader, Compression in the processing of polymer composites 1. A mechanical and microstructural study for different glass fabrics and resins. Composites Science and Technology, 1999. 59: p. 983-993. Scola, D.A. and J. Vontell, Effects of ambient aging of 5245C/graphite prepreg on composition and mechanical properties of fabricated composites. Polymer Composites, 1987. 8(4): p. 244- 252. Scola, D.A., Polyimide Resins. ASM Handbook Volume 21. 2001: ASM International. Scriven, L.E., On the dynamics of phase growth. Chemical Engineering Science, 1959. 10(1/2): p. 1-13. Shiverdecker, K.M. Design of a field-level on-aircraft polyimide adhesively bonded repair process. in NASA/Ohio Space Grant Consortium Annual Student Research Symposium. 2003- 2004. Steele, M. and T. Corden, New prepregs for cost effective, out-of-autoclave tool and component manufacture. SAMPE Journal, 2004. 40(2): p. 30-34. 166 Stutz, H., K.H. Illers, and J. Mertes, A generalized theory for the glass transition temperature of crosslinked and uncrosslinked polymers. Journal of Polymer Science: Part B: Polymer Physics, 1990. 28: p. 1483-1498. Subramanian, R.S. and M.C. Weinberg, The role of convective transport in the dissolution or growth of a gas bubble. Journal of Chemical Physics, 1980. 72(12): p. 6811-6813. Tang, J., W.I. Lee, and G.S. Springer, Effects of cure pressure on resin flow, voids, and mechanical properties. Journal of Composite Materials, 1987. 21: p. 421-440. Tavares, S.S., et al., Vacuum-bag processing of sandwich structures: Role of honeycomb pressure level on skin-core adhesion and skin quality. Composites Science and Technology, 2010. 70: p. 797-803. Tavares, S.S., V. Michaud, and J.A.E. Manson, Assessment of semi-impregnated fabrics in honeycomb sandwich structures. Composites: Part A, 2010. 41: p. 8-15. Tavares, S.S., V. Michaud, and J.A.E. Manson, Through thickness air permeability of prepregs during cure. Composites: Part A, 2009. 40: p. 1587-1596. Thomas, S. and S.R. Nutt, In situ estimation of through-thickness resin flow using ultrasound. Composites Science and Technology, 2008. 68: p. 3093-3098. Thomas, S. and S.R. Nutt, Temperature dependence of resin flow in a resin film infusion (RFI) process by ultrasound imaging. Applied Composite Materials, 2009. 16: p. 183-196. Thomas, S., Vacuum-bag-only processing of composites, in Materials Science. 2009, University of Southern California: Los Angeles. Thorfinnson, B. and T.F. Biermann. Measurement and control prepreg impregnation for elimination of porosity in composite parts. in Fabricating Composites. 1988. Philadelphia, Pennsylvania. Tung, C.Y.M. and P.J. Dynes, Relationship between viscoelastic properties and gelation in thermosetting systems. Journal of Applied Polymer Science, 1982. 27: p. 569-574. Turk, M., et al., Evaluation of the thermal oxidative stability of polyimides via TGA techniques. Journal of Polymer Science: Part A: Polymer Chemistry, 1999. 37: p. 3943-3956. Verdonck, E., K. Schaap, and L.C. Thomas, A discussion of the principles and applications of modulated temperature DSC (MTDSC). International Journal of Pharmaceutics, 1999. 192: p. 3-20. 167 White, S.R. and H.T. Hahn, Cure cycle optimization for the reduction of processing-induced residual stresses in composite materials. Journal of Composite Materials, 1993. 27(14): p. 1352-1377. Wood, J.R. and M.G. Bader, Void control for polymer-matrix composites (1): theoretical and experimental methods for determining the growth and collapse of gas bubbles. Composites Manufacturing, 1994. 5(3): p. 139-147. Wood, J.R. and M.G. Bader, Void control for polymer-matrix composites (2): experimental evaluation of a diffusion model for the growth and collapse of gas bubbles. Composites Manufacturing, 1994. 5(3): p. 149-158. Xie, W., W. Pam, and K. Chuang, Thermal characterization of PMR polyimides. Thermochimica Acta, 2001. 367-368: p. 143-153. Young, W.-B., Consolidation and cure simulations for laminated composites. Polymer Composites, 1996. 17(1): p. 142-148. Young, W.-B., Resin flow analysis in the consolidation of multi-directional laminated composites. Polymer Composites, 1995. 16(3): p. 250-257. Yu, Y., H. Su, and W. Gan, Effects of storage aging on the properties of epoxy prepregs. Industrial Chemical Engineering Research, 2009. 48: p. 4340-4345. Yuan, C., et al., Experimental investigation on the co-cure processing of honeycomb structure with self-adhesive prepreg. Applied Composite Materials, 2008. 15: p. 47-59.
Abstract (if available)
Abstract
Composite parts for commercial aircraft are traditionally manufactured using high-pressure autoclave processing of prepregs (carbon fiber pre-impregnated with epoxy resin). In recent decades, however, the use of composite parts for aircraft has increased, and aircraft markets have grown, creating pressure to increase production rates. To meet the growing demand for composite aircraft parts and to allow for the production of large composite components (i.e. wings and fuselage) alternative processing methods will be required. There are several drawbacks to autoclave processing, including a large capital investment, long cycle time, high cost of the nitrogen gas used to pressurize the vessel, size limitations, and poor energy efficiency. New out-of-autoclave processing methods have been developed to address these issues. One such method is vacuum-bag-only (VBO) processing of prepregs, a technique which uses atmospheric pressure alone to consolidate parts. ❧ VBO processing presents a potential solution for the manufacture of larger parts at faster rates using conventional layup and placement tools. However, before VBO methods can be used on primary structure, the quality of VBO processed parts must be shown to be equivalent to that of autoclave cured parts. The elimination of high external pressures during the cure cycle removes safeguards in the manufacturing process, resulting in the need for strict protocols in the layup and cure of VBO parts. To assess the feasibility of VBO processing for aerospace components a systematic study of the effect of process parameters on the quality of VBO parts is essential. Specifically, the mechanisms of void formation and growth in prepreg-processed carbon fiber composites are not well understood. The purpose of this work is to examine the potential causes of voids, to develop a complete understanding of the mechanisms of void formation. This knowledge will aid in the production of higher quality parts, and help to determine the feasibility of low-pressure VBO processing for large-scale structural components. ❧ As a starting point, carbon fiber/epoxy test laminates were manufactured using vacuum bag only methods as well as traditional autoclave cure cycles. Cured laminates were tested using aerospace qualification standards. Tests were performed on dry laminates as well as laminates that had been hot/wet conditioned. Mechanical properties were shown to be equivalent in vacuum bag only and autoclave processed laminates, and values for all test panels and test conditions exceeded the required level for structural aerospace applications. ❧ Cure cycle optimization was carried out to further improve the properties of out-of-autoclave processed parts. Variations in hold time and temperature were investigated for the first temperature dwell of the cure cycle. Several test panels were fabricated with a range of processing times and temperatures. Cured laminates were characterized for compaction (thickness), void content, and surface finish. Resin rheological properties were also examined. Based on experimental results, an optimized cure cycle was developed for the material system studied. ❧ The mechanisms of impregnation and compaction in vacuum-bag-only prepreg materials were investigated to determine the influence of fiber architecture on flow and void removal in out-of-autoclave processing. Material microstructure and laminate thickness were tracked as a function of cure time and temperature for a unidirectional material and a 5-harness satin woven prepreg featuring the same resin system. Isothermal resin flow was analyzed to determine the activation energy for resin impregnation in each prepreg, a quantitative measure of the influence of fabric type on flow behavior. Impregnation in the unidirectional material occurred early in the cure process, followed by additional ply compaction. In contrast, impregnation and compaction occurred on a longer time scale for the woven fabric. The impregnation, compaction, and air removal mechanisms observed for unidirectional and woven fabrics were different, indicating that cure cycles must be tailored to fiber architectures in prepregs, and possibly to part geometry. ❧ Void formation as a function of resin moisture content was studied to better understand and control process defects in composite parts made from prepreg. Uncured prepreg was conditioned at 70, 80 and 90% relative humidity and at 35°C. Conditioned prepreg was laid up into 16-ply laminates and cured using vacuum bag only processing, as well as partial vacuum and autoclave processing. Moisture uptake in the resin was measured using coulometric Fischer titration. Void content was measured by image analysis of polished sections of cured laminates. Void content increased substantially with increasing moisture content in vacuum bag only processed samples, and a strong pressure dependence was noted. Under autoclave cure conditions, void-free parts were produced even at high moisture levels. Experimental results were compared with trends predicted using a diffusion-based analytical model. ❧ Changes in vacuum bag only prepreg properties were tracked as a function of room temperature aging time (out-time). A modulated differential scanning calorimetry method was used to characterize the prepreg, and changes in prepreg tack levels were examined using an energy of separation technique. Laminates were cured from prepreg at various levels of aging using both traditional autoclave processing and low-pressure VBO techniques. Cured laminates were examined using ultrasound scanning, mechanical properties were tested, and laminates were sectioned for investigation of microstructure. Laminate quality as a function of out-time was examined in terms of prepreg properties and manufacturing technique. ❧ The results of the preceding study indicated that a need exists for an accurate and convenient method to monitor the extent of prepreg aging as a function of out-time. For this reason, a method to track prepreg age was developed, involving measurement of changes in glass transition temperature as a function of room-temperature aging time. Samples from three out-of-autoclave prepreg systems were aged in ambient conditions and tested periodically using modulated differential scanning calorimetry. A linear increase in glass transition temperature with prepreg age was noted. ❧ In addition to monolithic parts and flat laminates, prepregs are used for face sheets of sandwich structures, where they are bonded to low-density cores of honeycomb or foam. To reduce processing time and cost, co-cure of composite face sheets and sandwich structure adhesives is desirable. VBO processing is attractive for this application, as it allows for the use of lighter and less costly core materials, eliminating the risk of core crush that can result from high autoclave pressures. Under vacuum, however, sandwich structure adhesives often foam as gas species evolve from solution. The first step in eliminating adhesive foaming is to identify the volatiles evolving from adhesives during cure. An on-line coupled thermogravimetric analyzer-Fourier transform infrared spectrometer (TGA-FTIR) technique was employed to identify volatile components evolving during the cure of two polyimide film adhesives. ❧ Overall, the work presented here provides an analysis of the influence of various process parameters on voids, leading to an improved understanding of the mechanisms of void formation in VBO processed parts. The results of this investigation shed light on the cure protocols required for the production of high quality composite parts in the absence of autoclave pressures.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Void evolution in vacuum bag-only prepregs
PDF
Fabrication and analysis of prepregs with discontinuous resin patterning for robust out-of-autoclave manufacturing
PDF
In situ process monitoring for modeling and defect mitigation in composites processing
PDF
Efficient manufacturing and repair of out-of-autoclave prepreg composites
PDF
Sustainable manufacturing of out-of-autoclave (OoA) prepreg: processing challenges
PDF
Material and process development and optimization for efficient manufacturing of polymer composites
PDF
Vacuum-bag-only processing of composites
PDF
In situ process analysis for defect control during composites manufacturing
PDF
Processing and properties of phenylethynyl-terminated PMDA-type asymmetric polyimide and composites
PDF
Chemical recycling of amine/epoxy composites at atmospheric pressure
PDF
Sound transmission through acoustic metamaterials and prepreg processing science
PDF
Characterization, process analysis, and recycling of a benzoxazine-epoxy resin for structural composites
PDF
Developing efficient methods for the manufacture and analysis of composites structures
PDF
Deformation control for mask image projection based stereolithography process
PDF
Ageing and mechanical failure of fiber reinforced polymers
PDF
Development of composite oriented strand board and structures
PDF
Processing, thermal-oxidative stability and thermal cyclic fatigue of phenylethynyl-terminated polyimides
PDF
Metallic syntactic foams synthesis, characterization and mechnical properties
PDF
High temperature creep behaviors of additively manufactured IN625
PDF
Long term lunar radiation degradation of potential lunar habitat composite materials
Asset Metadata
Creator
Grunenfelder, Lessa Kay
(author)
Core Title
Defect control in vacuum bag only processing of composite prepregs
School
Viterbi School of Engineering
Degree
Doctor of Philosophy
Degree Program
Materials Science
Publication Date
10/31/2012
Defense Date
08/17/2012
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
composites,manufacturing,OAI-PMH Harvest,out-of-autoclave,porosity,voids
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Nutt, Steven R. (
committee chair
), Malmstadt, Noah (
committee member
), Povinelli, Michelle L. (
committee member
)
Creator Email
grunenfe@gmail.com,grunenfe@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-106508
Unique identifier
UC11288770
Identifier
usctheses-c3-106508 (legacy record id)
Legacy Identifier
etd-Grunenfeld-1261.pdf
Dmrecord
106508
Document Type
Dissertation
Rights
Grunenfelder, Lessa Kay
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
composites
out-of-autoclave
porosity
voids