MULTI-SCALE MECHANISM BASED LIFE PREDICTION OF POLYMER MATRIX
COMPOSITES FOR HIGH TEMPERATURE AIRFRAME APPLICATIONS
by
PRIYANK UPADHYAYA
SAMIT ROY, COMMITTEE CHAIR ANWARUL HAQUE JAMES P. HUBNER MARK E. BARKEY NITIN CHOPRA
A DISSERTATION
Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Aerospace Engineering and Mechanics in the Graduate School of The University of Alabama
TUSCALOOSA, ALABAMA
2013 Copyright Priyank Upadhyaya 2013 ALL RIGHTS RESERVED ABSTRACT
A multi-scale mechanism-based life prediction model is developed for high-temperature polymer matrix composites (HTPMC) for high temperature airframe applications. In the
first part of this dissertation the effect of Cloisite 20A (C20A) nano-clay compounding on the thermo-oxidative weight loss and the residual stresses due to thermal oxidation for a thermoset polymer bismaleimide (BMI) are investigated. A three-dimensional (3-D) micro- mechanics based finite element analysis (FEA) was conducted to investigate the residual stresses due to thermal oxidation using an in-house FEA code (NOVA-3D).
In the second part of this dissertation, a novel numerical-experimental methodology is outlined to determine cohesive stress and damage evolution parameters for pristine as well as isothermally aged (in air) polymer matrix composites. A rate-dependent viscoelastic cohesive layer model was implemented in an in-house FEA code to simulate the delamina- tion initiation and propagation in unidirectional polymer composites before and after aging.
Double cantilever beam (DCB) experiments were conducted (at UT-Dallas) on both pristine and isothermally aged IM-7/BMI composite specimens to determine the model parameters.
The J-Integral based approach was adapted to extract cohesive stresses near the crack tip.
Once the damage parameters had been characterized, the test-bed FEA code employed a micromechanics based viscoelastic cohesive layer model to numerically simulate the DCB ex- periment. FEA simulation accurately captures the macro-scale behavior (load-displacement history) simultaneously with the micro-scale behavior (crack-growth history).
ii DEDICATION
This dissertation is dedicated to my sister Mrs. Ruchi Upadhyay.
iii LIST OF ABBREVIATIONS AND SYMBOLS
HTPMC High Temperature Polymer Matrix Composite
BMI Bismaleimide
C20A Cloisite 20A
DCB Double Cantilever Beam
FEA Finite Element Analysis
RVE Representative Volume Element
2-D Two Dimensional
3-D Three Dimensional
C Oxygen Concentration
Di j Orthotropic Diffusivity
R(C) Reaction Rate
W Weight of Specimen
DNC Diffusivity of C20A modified BMI
DBR Diffusivity of BMI
α Scalar Damage Parameter
δ Crack Opening Displacement
θ Rotation at the Load Pin
P Reaction Force at the Load Pin
Tg Glass Transition Temperature
iv Ti Traction
{ε} Strain Vector
{H} Hereditary Strain Vector
{σ} Stress Vector
[M(t)] Viscoelastic Stiffness Matrix
λcr Critical Principal Stretch
λ¯ Principal Stretch Measure
σr Radial Stress
σθ Hoop Stress
σcohesive Cohesive Stress
σ¯ ud Undamaged Stress
φ Oxidation State Variable
φox Threshold Value of Oxidation State Variable
ε Weight Fraction due to Oxidation in Air
γ Weight Fraction due to Oxidation in Inert Conditions
ζ Nano-clay Loading
ψ Aspect ratio of Nano-clay Platelets
v ACKNOWLEDGEMENTS
Foremost, my heartfelt gratitude goes to my advisor Dr. Samit Roy for his valuable guidance and continuous support at every step of this work, right from conception of the idea to its conclusion. I am grateful to Dr. Roy for providing me with this opportunity to study at The University of Alabama and for his constant support that was invaluable in the completion of this work. His sense of confidence and level of trust in me has led me to the successful completion of my dissertation work.
Besides my advisor, I would like to thank Dr. Mark E. Barkey, Dr. Anwarul Haque,
Dr. James P. Hubner, and Dr. Nitin Chopra, for their encouragement, guidance and patience while serving as members of my doctoral committee.
I greatly appreciate the contribution by Dr. Hongbing Lu and Dr. Mohammad H.
Haque at UT-Dallas towards this study. I am especially thankful to Dr. Mohammad H.
Haque for conducting experiments, needed to validate my dissertation work.
I am also grateful to all the faculty members of the Department of Aerospace Engi- neering and Mechanics at The University of Alabama, who helped me a lot during my course work and imbibed a learning spirit in me. The blessings of my parents and the love and affection of my sister (Ruchi) and brother (Vijay) has given me the necessary driving force to complete this work. In the end, a special thanks to all who could not be mentioned here, but have always had a positive influence in my life. The research project was funded by the
“Low Density Materials Program” of the AFOSR.
vi CONTENTS
ABSTRACT ...... ii
DEDICATION ...... iii
LIST OF ABBREVIATIONS AND SYMBOLS ...... iv
ACKNOWLEDGEMENTS ...... vi
LIST OF TABLES ...... x
LIST OF FIGURES ...... xi
1 INTRODUCTION ...... 1
1.1 Objective ...... 4
2 LITERATURE SURVEY ...... 6
2.1 Thermo-oxidative Behavior of Polymer Resin Systems ...... 6
2.2 Barrier Properties of Nanoparticles ...... 10
2.3 Damage Modeling for Composites ...... 15
2.4 Cohesive Layer Models ...... 22
3 MATHEMATICAL MODELING OF THERMAL OXIDATION ...... 25
3.1 Three Zone Oxidation Model ...... 26
3.2 Determination of Thermo-Oxidative Parameters ...... 27
3.2.1 Calculation of β ...... 28
3.2.2 Calculation of αR0 ...... 28
3.2.3 Calculation of φox ...... 29
vii 4 INFLUENCE OF NANO-CLAY ON THERMO-OXIDATIVE STABILITY AND MECHANICAL PROPERTIES OF BMI ...... 31
4.1 Materials and Manufacturing Process ...... 31
4.2 Thermo-Oxidative Aging Experiments ...... 32
4.2.1 Isothermal aging of BMI and C20A/BMI in air ...... 32
4.2.2 Isothermal aging of BMI and C20A/BMI in Argon ...... 35
4.2.3 Isothermal aging of BMI and C20A/BMI in 60% O2 ...... 36
4.3 Calculation of Thermo-oxidative Parameters from Experimental Data . . . . 41
4.4 Nano-indentation Test for Mechanical Properties ...... 42
4.5 Thermal Shrinkage Test ...... 44
5 FINITE ELEMENT SIMULATION OF RESIDUAL STRESSES DUE TO SHRINK- AGE IN IM-7/BMI COMPOSITE ...... 45
5.1 Finite Element Model ...... 45
5.2 Simulation Results ...... 49
6 MULTI-SCALE VISCOELASTIC COHESIVE LAYER MODEL FOR PREDICT- ING DELAMINATION IN HTPMC ...... 55
6.1 Viscoelastic Cohesive Layer Model ...... 55
6.2 Damage Evolution Law ...... 58
6.3 Model Calibration ...... 60
6.3.1 Flexure Experiment ...... 60
6.3.2 FEA simulation of flexure experiment ...... 60
6.4 Double Cantilever Beam Experiment ...... 67
6.4.1 Specimen preparation and DCB specimen geometry ...... 67
6.4.2 Experimental method ...... 68
viii 6.5 Determination of Cohesive Stress using J-integral ...... 69
6.6 Estimation of Damage Evolution Law ...... 74
6.6.1 Determination of critical principal stretch λcr ...... 75
6.6.2 Determination of damage parameters α0 and m ...... 77
6.7 Numerical Simulation of DCB Experiments ...... 81
6.8 Sensitivity study on model parameters ...... 87
7 CONCLUSIONS ...... 90
7.1 Conclusions ...... 90
7.2 Future Work ...... 92
REFERENCES ...... 95
ix LIST OF TABLES
4.1 Oxidation parameters for neat BMI and C20A modified BMI ...... 42
4.2 Comparison of mechanical properties for neat and C20A modified BMI(3 wt%) 43
5.1 Material properties used in FEA simulation ...... 49
6.1 Elastic properties used in FEA modeling for un-aged and aged (for 3000 hours at 250 °C) IM-7/BMI unidirectional laminate ...... 62
6.2 Creep Compliance of BMI resin (Reference Temperature = 204 °C) . . . . . 64
6.3 Micromechanical damage parameters used in FEA modeling for un-aged and aged (for 3000 hours at 250 °C) IM-7/BMI unidirectional laminate ...... 65
6.4 Damage evolution law parameters used in FEA model for IM-7/BMI . . . . 80
6.5 Elastic properties of transversely isotropic IM-7/BMI lamina[Andrews and Garnich, 2008]...... 81
x LIST OF FIGURES
2.1 (a) Cross section of a G30-500/PMR-15 unidirectional composite after 2092 hours of isothermal aging in air at 288 °C (b) Close-up view of oxidized com- posite showing fiber matrix debond [Ripberger, Tandon, and Schoeppner, 2005]9
3.1 Schematic diagram showing three zones during thermal oxidation ...... 25
3.2 Oxidized and un-oxidized regions depicted in section view ...... 29
4.1 Neat and C20A modified BMI specimens ...... 32
4.2 Specimens inside the oven during isothermal aging experiment ...... 33
4.3 Weight loss data for BMI and 3% C20A compounded BMI at 250 °C in air . 33
4.4 Experimental set up for weight loss experiment in Argon and 60% O2 envi- ronment ...... 36
4.5 Weight loss data for BMI and 3% C20A compounded BMI at 250 °C in Argon 37
4.6 Weight loss data for BMI and 3% C20A compounded BMI at 250 °C in 60% O2 37
4.7 Comparison of weight loss data for BMI and 3% C20A compounded BMI at 250 °C in (a) air (b) argon (c) 60% O2 environment ...... 40
4.8 Measured shrinkage strains for neat BMI and C20A modified BMI ...... 43
5.1 (a) Hexagonal arrangement of fibers and (b) In-plane displacement boundary conditions on the RVE with fiber-fiber interaction ...... 46
5.2 3-D finite element mesh of RVE showing boundary conditions ...... 47
5.3 3-D contour showing oxidation state variable φ (neat BMI case) ...... 50
5.4 3-D contour showing radial strain εr (neat BMI case) ...... 51
5.5 3-D contour showing radial stress σr (neat BMI case) ...... 51
5.6 3-D contour showing hoop stress σθ (neat BMI case) ...... 52
xi 5.7 FEA prediction of oxidation state variable φ at the interface along z direction along fiber-matrix interface at different aging times ...... 53
5.8 FEA prediction of σr at the interface along z direction along fiber-matrix interface at different aging times ...... 53
5.9 FEA prediction of σθ at the interface along z direction along fiber-matrix interface at different aging times ...... 54
6.1 Opening crack containing cohesive ligament (b) Reduction of RVE to cohesive zone by area averaging fibril tractions [Allen and Searcy, 2001]...... 56
6.2 Three-point bending fixture ...... 61
6.3 FEA mesh for flexure test specimen of IM-7/BMI unidirectional laminate . . 62
6.4 FEA simulation showing inter-laminar delamination for flexure test of un-aged IM-7/BMI unidirectional laminate ...... 65
6.5 Normalized load displacement curve for un-aged IM-7/BMI unidirectional laminate ...... 66
6.6 Normalized load displacement curve for aged IM-7/BMI unidirectional laminate 66
6.7 An image of the test configuration for delamination type DCB specimen hav- ing pre-crack under initial loading ...... 67
6.8 Schematic diagram of DCB specimen ...... 69
6.9 J-Integral versus COD (δ) for pristine IM-7/BMI ...... 72
6.10 Cohesive stress versus COD (δ) for pristine IM-7/BMI ...... 73
6.11 J-Integral versus COD (δ) for aged (1000 hours at 260 °C in air) IM-7/BMI 74
6.12 Cohesive stress versus COD (δ) for aged (1000 hours at 260 °C in air) IM-7/BMI 75
6.13 Scalar damage parameter for pristine IM-7/BMI ...... 79
6.14 logα˙ versus logλ¯ for pristine IM-7/BMI ...... 80
6.15 Scalar damage parameter for isothermally aged (1000 hours at 260 °C in air) IM-7/BMI ...... 81
xii 6.16 logα˙ versus logλ¯ for isothermally aged (1000 hours at 260 °C in air) IM-7/BMI 82
6.17 Meshed DCB specimen of IM-7/BMI with boundary conditions ...... 83
6.18 Load versus displacement for pristine IM-7/BMI DCB specimen ...... 84
6.19 Crack-length versus displacement for pristine IM-7/BMI DCB specimen . . 85
6.20 Load versus displacement for isothermally aged (1000 hours at 260 °C) IM- 7/BMI DCB specimen ...... 85
6.21 Deformed and un-deformed contour plots showing εy for isothermally aged specimen at 260 °C clearly showing crack propagation from 70 mm to 74 mm at 450 sec ...... 86
6.22 Crack-length versus displacement for isothermally aged (1000 hours at 260 °C) IM-7/BMI DCB specimen ...... 86
6.23 Load versus displacement for pristine IM-7/BMI DCB specimen for different λcr values ...... 88
6.24 Load versus displacement for pristine IM-7/BMI DCB specimen for different α0 values ...... 89
6.25 Load versus displacement for pristine IM-7/BMI DCB specimen for different m values ...... 89
xiii CHAPTER 1
INTRODUCTION
HTPMC used in high temperature applications such as supersonic inlet ducts, ad- vanced fan casings and engine exhaust washed panels are known to have a limited life due to high thermo-mechanical loads and environmental degradation. Although there has been a considerable amount of work investigating the response of metals under such loading condi- tions, such studies are still not mature for HTPMC. The kind of coupled physical, chemical and mechanical response these HTPMC exhibit under extreme hygro-thermal loads makes it a very challenging problem. Recently, considerable insight has been gained for polymers
(particularly amorphous polymers) undergoing physical aging, chemical aging and strain dependent aging [Regnier and Guibe, 1997], [Bowles, Jayne, and Leonhardt, 1994], [Colin,
Marais, and Verdu, 2002], [Colin and Verdu, 2005], [Pochiraju and Tandon, 2006], [Pochi- raju, Tandon, and Schoeppner, 2008]. Although there are models to predict the physical aging behavior of polymers, the problem of thermo-oxidative aging of HTPMC under cyclic mechanical/ hygro-thermal loading has not been addressed adequately.
Many new resin systems for HTPMC capable of sustaining temperature in the range
200°C to 350°C are providing new opportunities for using HTPMC in propulsion, airframe and space structure components. Some of these HTPMC resin systems are PMR-15, AFR
700B, DMBZ-15, Avimid N, bismalimide (BMI) and HTM 512. PMR-15 is one of the most widely used resins for bypass ducts, nozzle flaps, bushings, and bearings. However, PMR-15
1 is made from methylene dianiline (MDA), a known carcinogen and a liver toxin, and the
Occupational Safety and Health Administration (OSHA) imposes strict regulations on the handling of MDA during the fabrication of PMR-15 composites. Recent concerns about the safety of workers working to manufacture and repair PMR-15 components have led to the implementation of costly protective measures. These costs involved to avoid safety issues with PMR-15 have led to the development of new material systems. New resin systems are being developed to address both safety and other limitations in thermo-oxidative stability, hydrolytic stability and processing constraints of these resin systems. One of these new resins is BMI which is a polyimide with a 232°C maximum service temperature. BMIs have served on military aircraft for decades and into 5th generation fighter planes. BMI is also used for exhaust wash areas on Boeing aircraft.
The material qualification methodology currently followed by the aerospace industry is primarily empirical because design engineers are more comfortable with the enormous amount of experimental data available. For example, 16,000 coupon and element tests were required to specify the design allowables with the required fidelity and reliability for the
F-22 Raptor high temperature program. Moreover, a very small insignificant change in the material property, manufacturing process or service load requires additional costly and exhaustive set of experiments to ensure the safety. In the current methodology, there is no space for advanced models for material assessment before insertion. Structural analysis done after insertion of the component using FEA does not incorporate the underlying material system behavior. Consequently, additional costly experiments must be conducted to augment the structural analysis to accurately capture the response of environmental factors such as temperature and moisture content especially when coupled with high strain-rate effects.
2 Therefore, the primary hurdle to the introduction of a new improved HTPMC in de- sign allowable database is the overwhelming amount of experimental testing that is required.
The laboratory testing based design methodology is a direct consequence of the necessity to address the uncertainty involved in manufacturing and to accurately estimate the perfor- mance and failure of a component under service conditions. Because of limited large-scale tests in the emerging design environment of HTPMC, it is critical that a robust mechanism based multi-scale model evolves to capture the response and the failure mechanisms involved in HTPMC.
Therefore, the motivation for the current research is our limited ability to use ad- vanced composites in high temperature applications due to design requirements and knock- down factors. To fulfill these requirements we need to be able to predict end-of-life properties and durability of composites at elevated temperatures under mechanical and environmen- tal loading conditions. Durability and degradation mechanisms in composite materials are fundamentally influenced by the fiber, matrix, and interphase regions that constitute the composite domain. The failure modes in composite laminates mainly involve fiber-matrix debond, delamination, fiber failure and matrix cracking. Three of these failure mechanisms, namely fiber-matrix debond, matrix cracking and delamination are studied in the current research. It is envisaged that the high residual stresses due to thermo-oxidative shrinkage lead to micromechanical damage. From the collective review of previous research work it was found that high radial stresses at the fiber-matrix interface are potentially responsible for fiber-matrix debond. Similarly in the matrix region, high hoop stresses are present which can lead to matrix cracking. To investigate the possibility of micromechanical damage due to fiber-matrix debond and matrix cracking, a coupled 3-D diffusion-reaction and stress sim-
3 ulation is conducted for a repetitive unit cell. Material system that is being investigated during this research work constitutes of BMI polymer matrix and IM-7 fiber. These materi- als are currently gaining a wide usage in aircraft structures, especially in airframe and engine inlet casings [Luo, Lu, Roy, and Lu, 2012]. Neat BMI exhibits linear viscoelastic or time- dependent behavior and IM-7 fiber is transversely isotropic in nature. But, the response of a composite differs significantly from that of its constituents, especially after thermo-oxidative degradation. After thermal oxidation, composite laminate starts to exhibit anisotropy in mechanical and oxygen diffusion properties. Therefore, it is important to have a design based on micromechanical analysis rather than on experiments to enhance the affordability of insertion of new material systems. In this context, the current research is an attempt to present a micromechanics based model that is capable of predicting performance and failure of structural components under mechanical loads in extreme environmental conditions. The objective of this research is presented in next section.
1.1 Objective
The goal of this research is to present a multi-scale mechanism based model that is capable of handling the anisotropy, thermo-oxidative degradation and rate effects (due to viscoelasticity of the polymer matrix) to predict the response and degradation of HTPMC. It is envisioned that the proposed mechanism based multi-scale model will result in a HTPMC life prediction tool which will lead to the development of a reliable analysis-based design guideline for aerospace structures undergoing high strain rates and aggressive environmental loadings. Examples of the predictive capabilities of the model will be shown for both aged and un-aged test specimens made of IM-7/BMI. The model developed in this process could subsequently be incorporated in commercial FEA software, such as ABAQUS or ANSYS,
4 through user defined subroutines to enable user-friendly large-scale structural life-predictions by the Air Force Research Laboratory (AFRL) and the aerospace industry.
Although, the current research was conducted on IM-7/BMI, our methodology is quite general and expected to be valid for similar high temperature thermoset polymer matrix composites. However, it is feasible that the degradation mechanism-based models developed in this work might not apply to a novel high-temperature polymer system due to differences in polymer morphology (for example, cross-linked thermoset versus linear thermoplastic polymers). In that case, new mechanism-based modeling tool-set will need to be developed for that specific family of polymers. The next chapter will cover the recent advances in
HTPMC durability research.
5 CHAPTER 2
LITERATURE SURVEY
2.1 Thermo-oxidative Behavior of Polymer Resin Systems
For efficient introduction of HTPMC in the design allowable database, it is important to accurately predict the response of HTPMC under cyclic mechanical and environmental loading at elevated temperatures. The primary obstacle to the use of HTPMC is thermo- oxidative degradation and resulting damage evolution at elevated temperatures ranging be- tween 200 °C to 300 °C, as experienced, for example, by engine exhaust-washed structures.
Aromatic polyimides have proven especially useful for these high temperature applications because of their thermal and oxidative stability.
In view of this fact, Regnier and Guibe [Regnier and Guibe, 1997] performed dynamic thermogravimetric (TGA) experiments on K736 BMI resin at several heating rates in dif- ferent aging environments. These TGA experiments were performed in order to elucidate the thermal behavior and to obtain the data to study the degradation process. They con- cluded that the polymer degradation in air is the result of multiple mechanisms operating simultaneously.
In recent times, many researchers have tried to develop a model to capture the thermo- oxidative aging process. In most cases, it has been clearly established that matrix oxidation is the main aging process [Bowles and Nowak, 1988][Meador, Lowell, Cavano, and Herrera-
Fierro, 1996][Colin, Marais, and Favre, 1999][Colin et al., 2002]. However, there is not a
6 consensus on the possible contribution of fiber-matrix interface to the aging process. Usually, oxidation is diffusion limited process which leads to creation of an outer oxidized layer having different material properties than the inner pristine material. The oxidized region is typically characterized by increased density and embrittlement [Schoeppner, Tandon, and Pochiraju,
2008]. This can potentially lead to the development of cracks which provide additional pathways for oxygen diffusion and thereby assisting further oxidation into deeper regions
[Pochiraju and Tandon, 2006].
In general, the degradation of a polymer matrix composite depends on the rate of oxygen diffusion as well as the oxidation reaction rate itself. Bowles et al. [Bowles et al.,
1994] reported that weight loss rate for PMR-15/C6000 composites were less than that of both carbon fiber and the polymer individually. On the other hand, Alston [Alston, 1980] and
Wang et al. [Skontorp, Wong, and Wang, 1995] observed a higher weight loss for composite than for the neat polymer. Similar synergistic weight-loss data have been reported for
G30/PMR-15 composite by Tandon et al. [Pochiraju and Tandon, 2006],[Tandon, Pochiraju, and Schoeppner, 2006].
From a modeling perspective, Colin et al. [Colin and Verdu, 2005] proposed a kinetic model to predict the oxidized layer thickness for a BMI resin F655-2. Their strategy is based on a model which couples the O2 diffusion and reaction-consumption kinetics and provides access to the thickness distribution of all the chemical modifications involved in the aging process. They derived the mathematical representation of O2 reaction rate from the mechanistic scheme of branched radical chain reaction. Their model predictions of changes in weight density and oxidation zone thickness were in excellent agreement with the test results.
To extend their work for unidirectional composites they improved this model to predict
7 relative weight loss in T800H/BMI composites by assuming that the volatile formation results essentially from hydro-peroxide decomposition.
Pochiraju et al.[Pochiraju et al., 2008] studied the thermo-oxidative behavior of uni- directional of G30-500/PMR-15 composite. They found that the elastic modulus of PMR-15 is sensitive to temperature and oxidation state and it degrades by approximately 28% due to a temperature change from 25°C to 288°C. More recently [Tandon, Pochiraju, and Schoepp- ner, 2008], they concluded that the fiber axis is the preferred damage evolution direction, and the strong anisotropy in observed oxidation growth can be attributed to the fiber-matrix interface debond and/or matrix cracking.
Since thermoset polymers go through a high temperature curing cycle, cure shrinkage and mismatches in the coefficient of thermal expansion of the fibers and matrix during the composite cure process give rise to localized micromechanical residual stresses and damage.
Therefore, the highly stressed fiber-matrix interface regions and the interstitial (inter-fiber) matrix regions tend to oxidize and develop micro-mechanical damage at an accelerated rate.
Micro-mechanical damage can cause additional permeation paths for oxygen and moisture deep into the composite. Experiment conducted by Ripberger et al. confirmed the initiation
fiber matrix debond due to thermal oxidation in unidirectional G30-500/PMR-15 composite at 288 °C in air as illustrated in the micrograph presented in Fig. 2.1.
Recently, Andrews and Garnich [Andrews and Garnich, 2008] concluded from FEA models that the fiber ends at laminate free-edges experience high radial and shear residual stresses at the interface that could cause fiber-matrix debond or matrix crack initiation. Con- sequently, the highly stressed fiber-matrix interface regions and the interstitial (inter-fiber) matrix regions tend to oxidize and develop micro-mechanical damage at an accelerated rate.
8 (a) (b)
Figure 2.1: (a) Cross section of a G30-500/PMR-15 unidirectional composite after 2092 hours of isothermal aging in air at 288 °C (b) Close-up view of oxidized composite showing fiber matrix debond [Ripberger et al., 2005]
Micro-mechanical damage can cause additional permeation paths for oxygen and moisture deep into the composite, thereby providing a synergistic degradation mechanism. Roy et al.[Roy, Wang, Park, and Liechti, 2006],[Roy and Singh, 2010] experimentally observed and modeled the synergistic fiber-matrix debond due to thermal oxidation of unidirectional IM-
7/PETI-5 composite at 288 °C in air. Upadhyaya et al.[Upadhyaya, Singh, and Roy, 2011] presented a multi-scale mechanism based model to predict the thermo-oxidative degrada- tion in IM-7/PETI5. The multi-scale model incorporated micro-scale level damage such as inter-crosslink chain scission in a polymer due to isothermal aging and formulated an inter- nal state variable which is a function of remaining crosslink density of oxidized polymer in unidirectional composite laminate to predict the degradation in inter-laminar shear strength.
Therefore, to improve the performance of HTPMC, the barrier to permeation of gas should be improved. In this context, nano-clay reinforced polymers show considerable promise. As an added bonus, reinforced polymers also exhibit improved mechanical proper- ties. Next section presents a summary of related research done for enhance of barrier and mechanical properties of polymer matrix composites.
9 2.2 Barrier Properties of Nanoparticles
Inorganic particles, including carbon black, talc, and mica, have been compounded
with polymers as a reinforcing agent for many decades. Recently, dispersion of nano-clay
platelets has gained much interest for its ability to produce significant improvements in prop-
erties by inclusion of only a small weight percent of nano-clay platelets. A vast majority
of the work done on nano-clay reinforced composites is focused on thermoplastic polymer
systems. For thermoplastic polymers, significant increases in thermo-oxidative stability, spe-
cific strength, stiffness, and permeability, have been reported. For example, Hussain et al.
observed that a small addition ( 5 wt%) of dispersed nano-clay can result in an improve-
ment (by ∼100 %) in mechanical properties of thermoplastic [Hussain, Roy, Narasimhan,
Vengadassalam, and Lu, 2007]. Similarly, Yano et al. [Yano, Usuki, Okada, Kurauchi, and
Kamigaito, 1993], [Yano, Usuki, and Okada, 1997] studied thermoplastic polyimide systems
and found that well dispersed single silicate layers at 2-3 wt% in polyimide composites
improved stiffness (by ∼100%) and strength (by ∼50%). Similar improvements in tensile and barrier properties for polyimide/organoclay nanocomposites were found by Chang et al.
[Chang and An, 2002].
Tyan et al. [Tyan, Wei, and Hsieh, 2000] treated clay (0-7 wt%) with a reactive diamine, namely 4, 4’-oxydianilinediamine (ODA), which is used in synthesis of polyimide nanocomposites. Significant increases in modulus (210% for 7 wt% nano-clay loading) and maximum stress (52% for 7 wt% nano-clay loading) were observed in ODA treated nanocom- posites. Additionally, for nanocomposites, higher limits of elongation at break were observed as compared with that of the neat resin. Agag et al. [Agag, Koga, and Takeichi, 2001] con-
10 ducted a study focusing on both mechanical and thermal properties using similar materials.
They observed enhancement in tensile modulus by increasing the adhesion between polymer
matrix and nano-clay. An increment in Tg value was also reported for these polyimide films.
Wang et al. [Wang, Wang, Wu, Chen, and He, 2005] studied the effect of microstructure of nanocomposites on corresponding thermal and mechanical properties. The value of storage modulus (G’) was found to be 21% higher than that of neat polyimide in the glassy state.
In addition, an increase of 69% in the value of G’ was seen above Tg. However, no change in
Tg was reported.
In an effort to develop materials that can withstand high level of stresses coupled
with extreme hygrothermal conditions, researchers at NASA initiated work on nano-clay
reinforced thermoset polymers [Abdalla, Dean, and Campbell, 2002]. The desired material
characteristics for such structural components are high Tg value, better thermal stability at
high service temperature, and better mechanical properties over a broad range of temper-
ature. The following paragraphs presents a summary of recent works on development of
thermoset polymer based (PMR, epoxy) nanocomposites.
Abdalla et al. [Abdalla et al., 2002] investigated the effect of nano-clay platelets on
viscoelastic and mechanical properties of thermoset PMR-15. Dynamic mechanical anal-
ysis (DMA) testing showed that 2.5% (by weight) clay loading resulted in a significant
improvement in flexural modulus and strength with no reduction in the elongation at fail-
ure. However, doubling the clay loading (5%) resulted in degradation of flexural properties.
Higher Tg were measured for all the nanocomposites compared to the neat PMR-15, with
the highest values obtained for the 5% clay loaded samples. Thermo-mechanical measure-
ments suggested that coefficient of thermal expansion (CTE) improved for composites of
11 unmodified clay, but decreased for surfactant modified clays. Interestingly, no improvement in thermo-oxidative stability was observed for PMR-15, as measured by weight loss during isothermal aging at 288°C for 1000 hours.
Islam, et al. [Dean, Islam, Small, and Aldridge, Dean et al.] examined intercala- tion of both organically modified and unmodified montmorillonite clays with PMR-15 and observed an increase in Tg value of nanocomposites by approximately 28°C. Additionally, sig- nificant enhancement in thermo-oxidative stability substantiated by the reduction in weight loss (25%) recorded during isothermal aging experiments was reported in this research work.
Micrographs of these specimens suggested that an optimum clay loading exists for perfect exfoliation state in PMR-15. Nano-clay existed in a primarily exfoliated state for 1 wt% clay loading whereas 5 wt% clay loading showed nano-clay platelets in an intercalated state.
Similarly, Gintert et al. [Gintert, Jana, and Miller, 2007] have reported an optimum organic nano-clay exfoliation method to achieve high thermal stability, increased crosslink density, maximum clay exfoliation, and improved thermal properties in PMR-15 composites. Sim- ilar observations regarding exfoliation state in PMR-15 nanocomposites were reported by
Abdalla, et al [Abdalla et al., 2002].
Campbell and Scheiman [Campbell and Scheiman, 2002] investigated the thermo- oxidative stability of PMR-15 nanocomposites, specifically the effect of packing density and orientation of diamine on oligomer melt viscosity and oligomer crosslink enthalpy. Nanocom- posites of PMR-15 polyimide and a diamine modified silicate were prepared by adding the silicate to PMR- 15 resin. Their results showed that better dispersion of clay leads to en- hanced thermo-oxidative stability. As discussed in section 2, PMR-15 composites generally degrade through oxidation of the surface layer followed by micro-cracking in the polymer
12 matrix [Meador et al., 1996]. The micro-cracks in turn allow permeation of oxygen into the bulk of the sample, thereby promoting further oxidative degradation. But, in nano-clay reinforced composites with exfoliated nano-clay platelets, better thermo-oxidative stability is observed because nano-clay platelets act as barriers and reduces the permeability and thus decreases bulk thermal oxidation.
Campbell et al. [Campbell, Johnston, Inghram, McCorkle, and Silverman, 2003] investigated the effect of nano-clay on barrier properties of thermoplastic (BPADA-BAPP) and thermosetting (PMR-15) polyimide resin. Reductions in gas permeability and water absorption were observed in thermoplastic polyimide nanocomposites [Hussain et al., 2007].
The thermosetting polyimide showed a reduction in weight loss during isothermal aging in air at 288 °C. The addition of 5 wt% nano-clay to neat PMR-15 resin results in a 15% reduction in weight loss after aging for 1000 hours at 288 °C in air. While the weight loss of all samples aged in nitrogen was lower than those aged in air, addition of nano-clay reduced weight loss by 15% in each case. A greater decrease in weight loss was observed for carbon fabric reinforced composites. Exfoliation of only 1 wt% or 2 wt% nano-clay decreases the weight loss on oxidative aging by 25% and 20% for carbon-reinforced composites.
With a wide range of applications, including metal coatings, use in electronics/elec- trical components, high tension electrical insulators, fiber-reinforced plastic materials and structural adhesives epoxy is another conventional thermoset polymer. The wide applicabil- ity of epoxy has made it a very popular test platform for researchers conducting advanced studies such as nano-clay intercalation and exfoliation. In-situ polymerization of epoxy monomers, blending with monomer, and solution blending of monomer and clay using polar organic solvents has already been investigated. Brown et al. [Brown, Curliss, and Vaia,
13 2000] outlined the potential improvements in properties that can be obtained from epoxy resins with layered nano-clays. They concluded that layered nano-clays with proper organic surface modification compatible with the polymer matrix may result in enhanced ability to exfoliate clay layers and offer a significant increase in properties. Another study by Jiankun et al. [Jiankun, Yucai, Zongneng, and Xiao-Su, 2001] studied the intercalation and exfolia- tion behavior of organoclays in epoxy resin. Their work showed that the organically modified nano-clays can be easily intercalated by epoxy through a mild mixing at 70-80 °C to form a homogenous and stable organoclay/epoxy intercalated hybrid.
Messersmith and Giannelis [Messersmith and Giannelis, 1994] prepared an epoxy- silicate nanocomposite by dispersing an organically modified mica-type silicate (MTS) in an epoxy resin. The nanocomposite exhibited a broadened Tg at slightly higher tempera- ture compared to the neat epoxy resin. Furthermore, the dynamic storage modulus of the nanocomposite containing 4% (by volume) silicate was approximately 58% higher in the glassy region and 450% higher in the rubbery plateau region as compared with the neat epoxy resin.
Pinnavaia and Lan [Lan and Pinnavaia, 1994] studied epoxy/clay nanocomposites in a rubbery state. They developed a method of clay treatment which is now commonly used.
Tensile tests conducted in this work showed that inclusion of the intercalated clay increased tensile strength and modulus significantly over that of the pristine epoxy. In another study by Pinnavaia et al. [Lan, Kaviratna, and Pinnavaia, 1995] it was found that the extent of exfoliation of the clay in polymer depends on the accessibility of the epoxy and diamine to the clay galleries as well as the relative rates of intra and extra gallery network formation.
Preliminary mechanical measurements showed that exfoliated epoxy/clay nanocomposites
14 have higher moduli than intercalated clay composites. Similarly, there is evidence form research done by Choi et al. [Choi and Tamma, 2001] that exfoliated nano-clay platelets provide better thermo-oxidative stability by creating a more tortuous path for diffusion of oxygen into the polymer resin.
In summary, there is enough evidence supporting the idea of compounding nano- clay with thermoplastic as well as thermoset polymers to improve mechanical, thermal and, oxidative properties. In conclusion, it is now well established that properly exfoliated and dispersed nano-clay platelets can provide enhancement in both stiffness and strength of polymers with improved barrier properties. Recently, the costs involved to avoid safety issues with PMR-15 have led to the development of new material systems. One of these new resins is BMI resin which is a polyimide with a 232°C maximum service temperature. In the research presented in this dissertation, C20A nano-clay is compounded with BMI to study its effect on mitigating thermo-oxidative degradation and residual stresses due to thermal oxidation in a thermoset polymer.
2.3 Damage Modeling for Composites
Continuous fiber-reinforced composites possess high strength and stiffness in the fiber direction. The overall mechanical behavior depends on the constituent properties as well as the microstructure of the composite. The deformation and damage mechanisms of compos- ites are very different from the matrix material alone. Several possible failure mechanisms exist for fiber reinforced composites, such as interfacial fiber-matrix debonding, interlaminar delamination, matrix cracking, fiber breakage, fiber pull-out, and shear sliding of fibers.
Several techniques are available in the open literature to model the initiation and the propagation of damage in composite laminates. One way of modeling damage in com-
15 posites is based on continuum mechanics where reduced stiffness of damaged material is calculated with the help of effective damage parameters. In conventional damage models, reduction in load carrying capacity of the material is achieved through zeroing a few per- tinent stiffness terms or removing the failed elements from the FEM mesh. For example,
Chaboche [Chaboche, 1981] presented a stiffness reduction model in 1981. In this model,
3-D damage evolution was governed by an isotropic scalar damage parameter that repre- sented a macroscopic measure of actual crack density. Later, in order to incorporate the material characteristics of constituents and the composite parameters such as volume frac- tions, fiber shape and arrangements, Lemaitre and Chaboche [Lemaitre and Chaboche, 1994] proposed a combined approach which used micromechanics based analysis for the thermo- elastoviscoplasticity of the composite with damage evolution. Likewise, Tay et al. [Tay, Tan,
Tan, and Gosse, 2005] proposed a method where property reduction is effected by applying a set of external nodal forces such that the net internal nodal forces of elements adjacent to the damaged element are reduced or zeroed (for complete failure case).
Nemat-Nasser and Hori [Nemat-Nasser and Hori, 1999] presented a summary of var- ious analytical models to predict effective elastic properties of elastic solids with micro- cavities, micro-cracks and traction free surfaces. Similarly, Voyiadjis and Kattan [Voyiadjis and Kattan, 2006] applied continuum damage mechanics to composite materials within the framework of the theory of elasticity, and presented a directional data model of damage me- chanics for composite materials using fabric tensors. In this micromechanics based model, the behavior of composite materials is related to the damage effects through fabric tensors. To obtain the damaged properties of each constituent, damage mechanics was employed to each constituent separately. Then, proper homogenization approach was adopted to calculate the
16 damaged properties of composite. Also, a generalized thermodynamics based formulation of damage evolution was presented, which was used in solving different numerical problems successfully.
Damage and failure of composite materials is inherently a multi-scale phenomenon, coupling different scales of damage initiation and progression. The changes in the material structure, in general, are irreversible during the process of damage. Damage accumulation can take place under elastic deformation (high cycle fatigue), under elastic-plastic deforma- tion (ductile plastic damage and low cycle fatigue) or under creep conditions (creep damage)
[Kachanov, 1986]. The choice of damage parameter, as a law, is not simple. It can be done either by a physical microstructural study, or by a direct generalization of experimental data. From application point of view, it is very important that the evolution of damage parameter should be simple enough and must have an evident mechanical sense [Kachanov,
1986]. In this regard, two types of damage evolution models, namely phenomenological and micromechanical, have been proposed in the literature for modeling failure of composite materials. One of the phenomenological damage evolution model was given by Chaboche
[Chaboche, 1981] in 1981. In 1987, by examining experimental data, Simo and Ju [Simo and Ju, 1987] concluded that the amount of micro-cracking at a particular strain level is sensitive to the applied loading rate (high strain rate) in dynamic environments. There- fore, they proposed a phenomenological rate sensitive damage model which required only one additional parameter. Matzenmiller et al. [Matzenmiller, Lubliner, and Taylor, 1995] presented a strain controlled continuum model wherein damage variables are introduced for the phenomenological treatment of the state of defects, and its implications on the degrada- tion of the stiffness properties. To denote the loss area due to cracking within the material
17 ligament, they introduced two non-negative damage parameters quantifying the relative size of disk-like cracks. Similarly, Chan et al. [Chan, Cheng, Jie, and Chow, 2005] presented an anisotropic phenomenological damage model and damage criterion for localized necking.
To determine the damage parameter and hardening rule, experimentally measured material properties of damaged and pristine material were used in this model.
In all these models mentioned above, nature of phenomenological damage law is based on the macro level response of material and therefore, it did not explicitly account for the micro level damage. To study the effect of micro level damage, Lene and Leguillon [Lene and
Leguillon, 1982] investigated the effect of a tangential slip at the interface of the components
(fiber displacements in a matrix for instance). A linear law in terms of a scalar coefficient was used to describe this slip phenomenon. It was found that the equivalent medium is homogeneous, anisotropic material, and its elastic properties depend on the scalar parameter involved in the slip law. Similarly, in the work done by Wriggers et al. [Wriggers, Zavarise, and Zohdi, 1998], the damage variable is defined as the amount of open debonded interface area. It was concluded that the amount of debonded surface area should serve as a primary internal variable in a homogenized macroscopic constitutive model for damage in composites.
Fish et al. [Fish, Yu, and Shek, 1999] developed a non-local damage theory for brittle composite materials based on double-scale asymptotic expansion of damage. In this work, a closed-form expression relating local fields to the overall strains and damage has been derived. They introduced the concept of non-local phase fields (stress, strain, free-energy density, damage release rate, etc.) via weighting functions defined over the micro-phase. The capabilities of this model were confirmed by excellent numerical results.
Choi and Tamma [Choi and Tamma, 2001] conducted a finite element study and
18 modeling for the micromechanical predictions of the homogenized elastic properties of wo- ven composites. Then, they used a finite element model in a micromechanical damage analysis to predict the damage initiation and propagation in the unit cell, and the resulting stress-strain curves. The completely damaged properties were taken to be 1% of the original properties in each constituent of woven fabric. Ju et al. [Ju, Ko, and Ruan, 2006] developed a micromechanical elastoplastic-damage formulation to predict the overall elastoplastic be- havior and interfacial damage evolution in fiber-reinforced ductile matrix composites. In this work, concept of eigenstrains due to cylindrical inclusion in conjunction with ensemble area averaging was adopted to micromechanically estimate the overall damage accumulation.
Raghvan and Ghosh [Raghavan and Ghosh, 2005], have shown that the damage evo- lution in the microstructure of the composites significantly affects the material symmetry.
Also, different load paths would create a different damage evolution profile in the microstruc- ture which will result in a different change in material symmetry. The anisotropic stiffness tensor will couple normal and shear strain components in the elastic energy expression in the
fixed coordinate system. Coupling terms reduces to vanishingly small numbers if principal damage coordinate system (PDCS) is used to express the stiffness and the initial material symmetry is retained and the damage effect tensor has a diagonal representation. Determi- nation of continuously evolving PDCS requires determination of second order damage tensor and its Eigen vectors for each load step. Nonlinear least square minimization method was used to obtain the six independent components of symmetric damage tensor. Subsequently,
Eigen vectors are obtained to get the rotation matrix which transforms the global coordinate system (GCS) to the PDCS.
To examine the evolution of PDCS with different load histories, a micromechanical
19 analysis problem of a simple unit cell representative volume element (RVE) was used. For proportional loading (shear loading is proportional to the axial load and transverse axial load is zero) orientation of the PDCS with respect to the GCS is at some fixed angle and remains unchanged during loading and unloading. For non-proportional loading (axial and shear load is applied independently), in the first half of the loading when only axial load is applied keeping transverse load and shear to be zero, PDCS coincides with the GCS. For the second half of the loading (similar to the proportional loading) PDCS continuously rotates and gets a final orientation angle which is less than the angle obtained in the proportional loading case.
Most of the damage models discussed above do not account for the evolution of damage or the loading history. Only a few namely Voyiadjis and Kattan [Voyiadjis and
Kattan, 1992], Fish et al. [Fish et al., 1999], Kouznetsova et al. [Kouznetsova, Brekelmans, and Baaijens, 2001] and Raghvan and Ghosh [Raghavan and Ghosh, 2005] include damage evolution as well as loading history. Ignoring these factors can result in a significant error for the problems with independent (non-proportional) axial and shear loading. The solution for this shortcoming is possible through simultaneous RVE-based microscopic-macroscopic analysis in each load step [Fish et al., 1999],[Feyel and Chaboche, 2000],[Massart, Peerlings, and Geers, 2007]. However, this approach is computationally very expensive since detailed micromechanical analysis is needed for each load step at every gauss point in all the elements of macrostructure.
To overcome the shortcoming of the parallel micro-macro analysis Raghvan et al.
[Raghavan and Ghosh, 2005],[Ghosh, Bai, and Raghavan, 2007] developed a computationally efficient anisotropic homogenization based continuum damage mechanics (HCDM) model for
20 composites going through micro-structural damage. In this model fiber-matrix debonding is taken as the microstructural damage mechanism for a 2-D (two-dimensional) analysis using
Vornoi cell finite element model. In multi-scale modeling, the use of continuum damage mechanics in the non-critical regions can make the whole approach extremely efficient. Such models can avoid the extensive micromechanical analysis at each load step but the microme- chanical analysis is needed in the close vicinity of dominant crack or localized instability for better prediction of catastrophic failure. In an effort to incorporate loading effects, Ghosh et al. modified their 2-D model [Ghosh, Ling, Majumdar, and Kim, 2000] to a 3-D model
[Jain and Ghosh, 2008] by introducing a PDCS which evolves with loading path.
The general form of CDM models introduce fictitious stress acting on an effective resisting area. The reduction in the original resisting area is caused due to the presence of micro cracks in the degraded material. Work done by Simo and Ju, 1987 [Simo and Ju,
1987] relates the effective fictitious stress to the actual Cauchy stress through a fourth order damage effect tensor. The hypothesis of equivalent elastic energy was used to evaluate the damage tensor and establish a relation in damaged and undamaged stiffness. This hypothesis specifically assumes that the elastic complimentary energy in a damaged material with the actual stress is equal to that in a hypothetical undamaged material with a fictitious effective stress.
Murakami [Murakami, 1988] worked to find a relation between second order dam- age tensor and the damage effect tensor. Since any arbitrary damage tensor results in an asymmetric effective stress tensor, Voyiadjis and Kattan [Voyiadjis and Kattan, 1996] sug- gested a representation to maintain the symmetric nature of effective stress tensor. Using this representation, the damaged stiffness is updated in terms of undamaged stiffness.
21 The anisotropic CDM model proposed by Raghavan and Ghosh [Raghavan and Ghosh,
2005] involving fourth order damage tensor introduces a damage evolution surface to delin- eate the interface between damaged and undamaged domains in the strain space. The surface equation involves a fourth order symmetric negative definite tensor which corresponds to the direction of the rate of stiffness degradation tensor and a damage state variable.
2.4 Cohesive Layer Models
It is now well-established that in the presence of a large fracture process zone near the crack tip, the basic assumptions of linear elastic fracture mechanics (LEFM) are no longer valid [Kanninen and Popelar, 1985]. Specifically, in some polymers, the occurrence of void nucleation and growth ahead of the crack-tip results in a damage (process) zone that is not traction free. Further, for a crack in a polymer matrix composite, fiber-bridging may also be present within the damage zone. Therefore, in such cases, a cohesive layer modeling approach would be more accurate in accounting for the nonlinear processes that occur within the “damage zone”. Cohesive zone model was first introduced by Barenblatt [Barenblatt,
1962] and Dugdale [Dugdale, 1960] in the 1960s. In the 1980s, application of cohesive zone models to determine strength of composites and adhesive joints were introduced by Bäcklund
[Backlund, 1981] and Stigh [Stigh, 1987]. To model the Interfacial debonding in composite laminates, Swaminathan et al. [Swaminathan, Pagano, and Ghosh, 2006] developed a 3-D micromechanical model in which the interface behavior is described by 3-D cohesive zone with bilinear traction boundary conditions. In this model, the interface is represented by cohesive springs of infinitesimal length. Needleman [Needleman, 1987] and Stigh [Stigh,
1988] demonstrated how the cohesive zone model fits within the scope of conventional stress analysis using FEA. Cohesive zone models have seen an almost explosive increase in use
22 and applications during recent years. With cohesive modeling, no additional properties are necessary to simulate crack growth. Only the cohesive law is needed to analyze both initiation and growth of a crack. Typically, cohesive elements in FEA codes follow a pre- defined traction separation law that simulates the crack initiation and propagation. Initially the traction across the interface increases with the displacement then drops after reaching a maximum value and eventually vanishes indicating the failure of cohesive elements. Three different zones are identified to represent these tractions (normal and tangential) in terms of effective opening displacement. In hardening region traction increases with the opening displacement and decreases in the softening region and eventually drops down to zero when cohesive springs fail which is when complete debonding occurs. The unloading pattern is same as the loading path in the hardening region but it changes for the softening region demonstrating the irreversible nature of damage process. Another advantage of cohesive zone models is that these models can simulate different types of failure mechanisms, such as fiber matrix debond and interlaminar delamination. This is also a drawback in modeling flexibility; namely if the fracture toughness changes with crack growth, a conventional cohesive law cannot capture this phenomenon by itself. Another major issue involved in this kind of modeling work is the scale of crack opening length which needs to be chosen very carefully.
In order to develop a cohesive zone model that does not require a prescribed traction- separation law, Allen and Searcy [Allen and Searcy, 2001] proposed a viscoelastic cohesive zone model and demonstrated the use of this model by numerically solving example problems with different displacement boundary conditions and strain-rates. They also proposed a damage evolution law that was phenomenologically derived due to the absence of near-tip experimental data. In this context, the process zone ahead of the crack tip is usually very
23 small compared with the specimen size in most materials. Therefore, experimentally it is quite challenging to precisely determine the traction separation law in the cohesive zone. The work presented in this dissertation employs a modified version of the viscoelastic cohesive layer model prosed by Allen and Searcy [Allen and Searcy, 2001], but the damage evolution law is fully characterized based on actual data from experiments. Sorensen and Jacobsen
[Sorensen and Jacobsen, 2003] presented a review of existing experimental procedures to estimate the cohesive law and delineated two major approaches. The first approach is to use a direct tension test, with the assumption that a uniform damage state evolves across the ligament. In reality, it is very difficult to achieve uniform damage state in a ligament during direct tension test, and therefore this approach is impractical. The second approach is the J-integral approach where macro level J-integral data is used to extract the micro level constitutive (traction-separation) behavior in the process zone. Sorensen and Jacobsen
[Sorensen and Jacobsen, 2003] adapted the same approach to conduct their ongoing research.
Recently, Fuchs and Major [Fuchs and Major, 2011] used the J-integral approach to determine the cohesive zone models for glass-fiber reinforced composites and studied the effect of loading direction on the constitutive cohesive law. In the work presented in this dissertation, the
J-integral approach is employed to determine the cohesive law for delamination behavior of
IM-7/BMI unidirectional laminates, before and after isothermal aging at 260 °C for 1000 hours. For this purpose, double cantilever beam (DCB) experiments were conducted to acquire the macro level J-integral data, and the displacement and strain fields in the process zone were obtained using digital image correlation (DIC). From experimental data, cohesive law and damage evolution parameters were determined and used in the viscoelastic cohesive layer model to simulate delamination growth.
24 CHAPTER 3
MATHEMATICAL MODELING OF THERMAL OXIDATION
In this chapter, the use of a diffusion-reaction mechanism to represent the time depen- dent thermo-oxidation behavior of a polymer material is presented. Based on the work done by Tandon et al. [Pochiraju and Tandon, 2006], a three-zone model representing three dis- tinct oxidation zones was used. A schematic diagram showing three oxidation zones, namely, un-oxidized, active, and fully oxidized zone is presented in Fig. 3.1. A polymer state variable
φ is used to represent the oxidation state in a particular oxidation zone quantitatively, indi- cating the availability of active polymer site for oxidation reaction. Polymer state variable varies from its maximum value of 1 (in un-oxidized zone) to its minimum threshold value of
φ = φox (in the fully oxidized zone). A thin active oxidation layer is present in between the oxidized and un-oxidized zone where φox < φ < 1, as shown in Fig. 3.1
Oxidized Active Un-oxidized
O2
ϕ=ϕox 1>ϕ>ϕox ϕ=1 Three zone model with oxidation state parameter ϕ
Figure 3.1: Schematic diagram showing three zones during thermal oxidation
25 3.1 Three Zone Oxidation Model
The diffusion of oxygen in a polymer composite incorporating consumptive reaction
rate is governed by Fick’s law as
∂C ∂ 2C ∂ 2C ∂ 2C = D11 + D22 + D33 − R(C) (3.1) ∂t ∂x2 ∂y2 ∂y2
where, C(x,y,z,t) is the oxygen concentration, Di j are the orthotropic diffusivities, and R(C) is the polymer consumption rate. The rate of oxidation reaction is modeled with respect to the saturation reaction rate R0 when the reaction is not oxygen deprived. For BMI type polymer resins, the reaction rate reduces when the availability of oxygen reduces according to the equation,
R(C) = R0 f (c) (3.2)
here, R0 is the saturation reaction rate and f (C) is a function which simulates the situation when the amount of available oxygen is less than the saturation concentration, and is given by following expression [Pochiraju and Tandon, 2006]
2βC βC f (C) = 1 − (3.3) 1 + βC 2(1 + βC)
where βC is the normalized concentration and β is an oxidation reaction coefficient. If it is
assumed that the polymer weight loss is directly proportional to the rate of reaction R(C),
then, dW ∝ −R(C) (3.4) dt
26 The polymer state variable φ is defined as the ratio of the current weight of the polymer specimen to its original (un-oxidized) weight. Therefore time derivative of polymer state variable is proportional to the weight loss rate,
dφ dW ∝ (3.5) dt dt
Combining Eqns. 3.2, 3.4 and 3.5 gives,
dφ = −αR0 f (C) (3.6) dt where α is a proportionality constant. To model the thermo-oxidative behavior of a BMI type polymer, three oxidation reaction parameters (αR0, β and φox) need to be experimentally characterized. To determine these parameters, experimental data from isothermal aging of polymer specimen in conjunction with oxidation layer growth data is needed. The procedure followed to calculate each one of these parameters is outlined separately in the next section.
The aging parameters for nano-clay modified polymer can be determined in the same manner as described in next section.
3.2 Determination of Thermo-Oxidative Parameters
This section summarizes the methodology [Pochiraju and Tandon, 2006] adopted to determine the parameters needed to model thermo-oxidative behavior of thermoset polymer.
Definition of these parameters and the procedure adopted to extract each parameter from thermal aging experiments is outlined in following sections.
27 3.2.1 Calculation of β
To calculate β, weight loss data in two different aging environments is required.
Combining Eqn. 3.2 and 3.3 for two different oxygen concentrations C1 and C2, and equating
the reaction rates gives
2βC1 βC1 R(C1 2βC2 βC2 1 − = 1 − (3.7) 1 + βC1 2(1 + βC1) R(C2) 1 + βC2 2(1 + βC2)
Weight loss rate in a particular aging environments is proportional to the reaction
rate R(C). From the slopes of linear fits to the weight loss versus time data in two different
R(C1 aging environments, the ratio of reaction rates can be obtained. Here R(C1) and R(C2) R(C2) are reaction rates in different aging environment having oxygen concentrations C1 and C2
R(C1 respectively. Substituting the values of ratio , and C1, C2 in Eqn. 3.7, a cubic equation R(C2)
in β is obtained to be solved for only one physically permissible (positive) value of β.
3.2.2 Calculation of αR0
Once the value of β is determined, to calculate αR0 Eqn. 3.5 and Eqn. 3.6 can be
rewritten for a particular aging environment with constant oxygen concentration C as