residual stress characterization for laminated...
TRANSCRIPT
RESIDUAL STRESS CHARACTERIZATION FOR LAMINATED COMPOSITES
By
SHAO-CHUN LIU
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOLOF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1999
ii
ACKNOWLEDGMENTS
The author would like to thank Dr. Peter Ifju for his continuous support and
advice, and the other members of his graduate committee, Dr. C.-T. Sun, Dr. Cristescu,
Dr. Shankar, Dr. Vu-Quoc, Dr. Jenkins, Dr. Beatty and Dr. Brennan, for their guidance.
Thanks also go to Mr. Ron Brown for his helpful instruction in the machine shop, and his
colleagues Xiaokai Niu, Brian Kilady, Ali Abdel-Hadi, Leishan Chen, Jongyoon Ok,
Brian Wallace, Oung Park, and Scott Ettinger for their friendship and help.
The author would like to acknowledge the generous sponsorship of the National
Science Foundation (NSF) under the grant number CMS 9502678, and the National
Aeronautics and Space Administration (NASA) Space Grant Consortium of Florida.
Finally, the author would like to express the most sincere gratitude to his parents,
wife, and children for their patience and understanding in the occasions without the
presence of their son, husband and father.
iii
TABLE OF CONTENTSpage
ACKNOWLEDGMENTS................................................................................................... ii
LIST OF TABLES............................................................................................................... v
LIST OF FIGURES............................................................................................................ vi
ABSTRACT ........................................................................................................................ x
CHAPTERS
1 INTRODUCTION......................................................................................................... 1Background ................................................................................................................... 1Literature Review .......................................................................................................... 2
Experimental Techniques ........................................................................................ 3Destructive methods .......................................................................................... 3Nondestructive methods .................................................................................... 5Shrinkage measurement of polymers ................................................................ 7
Analytical and Computational Modeling ................................................................ 9Elastic modeling................................................................................................ 9Viscoelastic modeling ..................................................................................... 10
Research Objectives .................................................................................................... 12
2 HIGH TEMPERATURE MOIRÉ INTERFEROMETRY AND THE CUREREFERENCING METHOD .......................................................................................15High Sensitivity Moiré Interferometry ........................................................................15Non-Room Temperature Moiré Interferometry with the Optical Thermal
Chamber ................................................................................................................19Optical Thermal Chamber .....................................................................................20
General description..........................................................................................20Chamber calibration........................................................................................22
Cure Referencing Method ...........................................................................................27The Principle.........................................................................................................28Grating Replication ...............................................................................................28Composite Specimen Fabrication..........................................................................32Experimental Setup and Procedure.......................................................................33
Tuning of optical setup....................................................................................33Thermal loading for composite laminate specimens.......................................35
Data Analysis and Results.....................................................................................37Significance of the Tests.............................................................................................49
iv
3 POST-GEL CHEMICAL SHRINKAGE OF THERMOSET RESINS....................... 51Introduction ................................................................................................................. 51Experimental Methodology......................................................................................... 52Specimen Preparation and Experimental Procedure ................................................... 52
Master Grating Selection....................................................................................... 54Refining of Replication Procedure ........................................................................ 58
In-plane and Out-of-plane Deformation Measurement ............................................... 58Results and Discussion................................................................................................ 63Conclusion................................................................................................................... 71
4 VISCOELASTIC CONSTITUTIVE MODELING FOR POLYMER MATRIXLAMINATED COMPOSITES.................................................................................... 73Introduction ................................................................................................................. 73Standard Linear Solid Model and Correspondence Principle for Linear
Viscoelastic Materials ........................................................................................... 74Time-Temperature Superposition Principle ................................................................ 75Thermal Chamber Manufacture and Test Fixture Design........................................... 76Specimen Preparation and Testing Plan...................................................................... 78
Gage and Accessory Selection .............................................................................. 79Specimen Preparation............................................................................................ 80Experimental Procedure and Testing Plan ............................................................ 81
Results and Data Analysis ........................................................................................... 84General Behaviors of Materials............................................................................. 84
Loading phase.................................................................................................. 85Holding (creep) phase...................................................................................... 86Unloading and recovery phases....................................................................... 92
Curves Fitting for the Creep Tests ........................................................................ 92Model Formation....................................................................................................... 108
5 DISCUSSION AND CONCLUSION ....................................................................... 111Application of Linear Viscoelastic Model for Residual Stress Relaxation............... 111
Overview ............................................................................................................. 111Experiment Procedure and Results...................................................................... 112
Concluding Remarks ................................................................................................. 122
APPENDICES
A THE MATHCAD FILE FOR THE DATA ANALYSIS OF 90-DEGREECOMPOSITE TENSILE SPECIMENS WITH THREE-ELEMENTSTANDARD LINEAR SOLID MODEL.................................................................. 124
B THE MATHCAD FILE FOR THE DATA ANALYSIS OF 90-DEGREECOMPOSITE TENSILE SPECIMENS WITH FOUR-ELEMENTSTANDARD LINEAR SOLID MODEL.................................................................. 131
REFERENCES................................................................................................................ 152
BIOGRAPHICAL SKETCH........................................................................................... 158
v
LIST OF TABLES
Table page
1 Components of residual strain in different lay-ups of laminate specimens............... 502 Names and vendors of the resins used....................................................................... 533 Chemical curing shrinkage for five different resins. ................................................. 644 Time-Temperature Superposition shift factors (in log units), and reference
temperature (Tr) is 100 °C.......................................................................................1075 Comparison of apparent residual strains between newly made composite specimens
and two-year old specimens....................................................................................1166 Residual stresses calculated by linear viscoelastic model and laminate theory for the
same batch of composite specimens........................................................................121
vi
LIST OF FIGURES
Table page
1 NASA’s X-34 rocket plan...........................................................................................12 Schematic diagram of moiré interferometry..............................................................163 Photograph of a four-beam interferometer ................................................................184 Schematic diagram of the environmental chamber system setup..............................215 Photograph of the environmental chamber system setup..........................................226 Null field fringe patterns of distortion calibration. (a) U field without the chamber,
(b) V field without the chamber, (c) U field with the chamber, (d) V field with thechamber .....................................................................................................................24
7 Calibration tests in five different temperature settings. The curves are obtained fromthermal couple readings inside the remote chamber, and the markers are obtainedfrom analyzing moiré fringe patterns of aluminum control specimens which are themore precise representation of the specimen surface temperature............................26
8 The coefficient of thermal expansion for aluminum alloy 6061. The twodiscontinuities at -60 °C and 100 °C are due to different fitting functions fordifferent temperature ranges......................................................................................27
9 General grating replication procedure.......................................................................2910 Grating replication procedure for CRM ....................................................................3111 Vacuum bag lay-up for CRM ....................................................................................3312 AS4/3501-6 composite curing profile.......................................................................3413 Schematic diagram of a four-beam moiré interferometer .........................................3514 Collimation of incident beams. When the incident beams are perfectly collimated,
the frequency of the virtual grating will be space-wise constant (fA= fB = fC). Themoiré pattern due to the interference between virtual grating and the referencegrating will be insensitive to the position of the reference grating ...........................36
15 Fringe patterns for [0]16 unidirectional laminate at different temperatures...............3816 Fringe patterns for [02/902]2s symmetric cross-ply laminate at different temperatures
...................................................................................................................................4017 Fringe patterns for [03/90]2s symmetric unbalanced cross-ply laminate at different
temperatures..............................................................................................................4218 Fringe patterns for [0 2/452]2s symmetric angle-ply laminate at different temperatures
...................................................................................................................................4419 The strain-temperature curves of [1016] unidirectional laminated composite...........4720 The strain-temperature curves of [02/902]2s cross-ply laminated composite.............4821 The strain-temperature curves of [0/903]2s unbalanced cross-ply laminated composite
...................................................................................................................................4822 The strain-temperature curves of [02/452]2s angle-ply laminated composite.............49
vii
23 Apparatus to produce silicone rubber specimen molds: an aluminum pipe, analuminum rod, a piece of release-film-covered glass, and a specimen mold ........... 54
24 The original procedure for neat resin specimen replication ...................................... 5525 Distorted fringe patterns of a PC10-C specimen on the silicone rubber master grating
................................................................................................................................... 5626 Fringe patterns of a PC10-C specimen on the PC10-C master grating before
separation................................................................................................................... 5727 Illustrations for making (a) room temperature cure, (b) UV cure, and (c) high
temperature cure specimen gratings .......................................................................... 5928 Schematic diagram of a Fizeau interferometer.......................................................... 6129 Experimental setup for out-of-plane deformation measurement (Fizeau
interferometer). (a) Overview; (b) Close-up look of Fizeau interferometer and a neatresin specimen ........................................................................................................... 62
30 A high temperature specimen grating after being separated from its master grating ....................................................................................................................................... 63
31 Typical shrinkage vs. time curves for each kind of specimens ................................. 6632 In-plane chemical shrinkage and out-of-plane deformation fringe patterns for a
PC10-C epoxy specimen ........................................................................................... 6733 In-plane chemical shrinkage and out-of-plane deformation fringe patterns for a UV
cure resins (SR 448:SR 205=60:40).......................................................................... 6834 In-plane chemical shrinkage and out-of-plane deformation fringe patterns of 3501-6
epoxy at room temperature and 140 °C.....................................................................6935 Residual strain vs. temperature of AS4/3501-6 [0]16 laminate in transverse direction
...................................................................................................................................7236 Apparent strain of 3501-6 neat resin specimens vs. temperature..............................7237 Spring-dashpot arrangement for standard linear solid model....................................7438 Construction of the long-term “master curve” at reference temperature from short-
term testing data at different temperature..................................................................7639 Thermal chamber and test fixture for uniaxial tensile tests.......................................7740 Composite prepreg panel with Teflon release film template for applying strain gages.
The template for making 10-degree specimens is shown in this picture...................7841 Experiment setup from several viewpoints...............................................................8242 Testing profiles for (a) 0-degree specimens, (b) 10-degree and 90-degree specimens
...................................................................................................................................8443 Stress-strain curves for 0-degree specimens at different temperature settings during
loading phase. From Fig. 43 through 45, all strain values are the average of front andback strain gage readings...........................................................................................87
44 Stress-strain curves for 90-degree specimens at different temperature settings duringloading phase.............................................................................................................88
45 Stress-strain curves for 10-degree specimens at different temperature settings duringloading phase.............................................................................................................89
46 Strain-time curves for 0-degree specimens at different temperature settings duringholding phase.............................................................................................................90
47 Strain-time curves for 90-degree specimens at different temperature settings duringholding phase.............................................................................................................91
viii
48 Shear strain-time curves for 10-degree specimens at different temperature settingsduring holding phase ................................................................................................. 91
49 Estimating the values of initial guess for curve fitting.............................................. 9450 Curve fitting and experimental data for the longitudinal strain of 90-degree
specimens during the holding phase at different temperatures. (a) 22 to 80 °C; (b)100 to 177 °C; (c) 195 °C..........................................................................................94
51 Curve fitting and experimental data for the shear strain of 10-degree specimensduring the holding phase at different temperatures. (a) 22 to 100 °C; (b) 120 to 195°C...............................................................................................................................96
52 Estimating the values of initial guess for curve fitting with the fourth element .......9753 Curve fitting with four element model and experimental data for the longitudinal
strains of 90-degree specimens during the holding phase at different temperatures.(a) 22 to 80 °C; (b) 100 to 177 °C; (c) 195 °C..........................................................98
54 Curve fitting with four element model and experimental data for the shear strains of10-degree specimens during the holding phase at different temperatures. (a) 22 to100 °C; (b) 120 to 195 °C .......................................................................................100
55 log (S22(t)) versus log (t) for AS4/3501-6 unidirectional composite laminate atdifferent temperatures..............................................................................................102
56 log (S12(t)) versus log (t) for AS4/3501-6 unidirectional composite laminate atdifferent temperatures..............................................................................................102
57 log (S66(t)) versus log (t) for AS4/3501-6 unidirectional composite laminate atdifferent temperatures..............................................................................................103
58 The tensile creep compliance S22(t) master curve for 3501-6/AS4 unidirectionallaminate specimens in transverse fiber direction. Reference temperature is at 22 °C..................................................................................................................................104
59 The shear creep compliance S66(t) master curve for 3501-6/AS4 unidirectionallaminate specimens. Reference temperature is also at 22 °C. ................................105
60 The tensile creep compliance S22(t) master curve for 3501-6/AS4 unidirectionallaminate specimens in transverse fiber direction with consideration of physicalaging. Reference temperature is now at 100 °C. ....................................................106
61 The shear creep compliance S66(t) master curve for 3501-6/AS4 unidirectionallaminate specimens with consideration of physical aging. Reference temperature isalso at 100 °C. .........................................................................................................107
62 Comparison between directly calculated creep compliance S22(t) and curve-fittingS22(t). The curve represents long-term creep test is also shown to compare with themaster curve obtained from momentary creep tests (short-term creep). .................109
63 Fringe patterns for [016]T unidirectional composite laminate panel. Time lag frommanufacture date tlag= 702 days...............................................................................113
64 Fringe patterns for [02/902]2s balanced cross-ply composite laminate panel. tlag= 702days..........................................................................................................................114
65 Fringe patterns for [03/90]2s unbalanced cross-ply composite laminate panel. tlag=702 days...................................................................................................................114
66 Fringe patterns for [02/452]2s angle-ply composite laminate panel..........................115
ix
67 Continuous and piecewise S22 master curve in logarithm scale. In all the followingdiagrams, piecewise master curves are completely covered by continuous mastercurves and reference temperature is 22 °C..............................................................117
68 Continuous and piecewise S66 master curve in logarithm scale..............................11869 Continuous and piecewise S12 master curve in logarithm scale..............................11870 S22(t) for time from 0 to 107.78 seconds (702 days)..................................................11971 S66(t) for time from 0 to 107.78 seconds (702 days)..................................................12072 S12(t) for time from 0 to 107.78 seconds (702 days)..................................................120
x
Abstract of Dissertation Presented to the Graduate Schoolof the University of Florida in Partial Fulfillment of theRequirements for the Degree of Doctor of Philosophy
RESIDUAL STRESS CHARACTERIZATION FOR LAMINATED COMPOSITES
By
Shao-Chun Liu
December 1999
Chairman: Dr. Peter G. IfjuMajor Department: Aerospace Engineering, Mechanics and Engineering Science
With increasing applications of advanced laminated composites, process-induced
residual stress has drawn more and more attention in recent years. Efforts have been
devoted to understanding residual stress both quantitatively and qualitatively.
In the current study, a novel technique called the Cure Referencing Method was
developed which has the capability for measuring the residual stress on the symmetric
laminated composite plates. It can also differentiate residual stress into two components:
one is due to the mismatch of the coefficient of thermal expansion, the other is caused by
the matrix chemical curing shrinkage.
The chemical curing shrinkage of the polymer matrix was investigated in further
detail. A technique was developed to measure the post-gel chemical curing shrinkage
which is the portion of curing shrinkage that really induces the residual stress in the
polymer matrix composites.
xi
Time-dependent material property is another issue associated with polymer matrix
composite materials. The data of several short-term tensile creep tests run at different
temperature were used to construct a linear viscoelastic model for describing the behavior
of the composites over a long period of time. It was found that physical aging of the
polymer matrix needs to be taken into account in order to have a more accurate
representation of the long-term behavior. A fair agreement was obtained between the
result of the long-term creep test and the master curve constructed from several
momentary creep tests.
1
CHAPTER 1INTRODUCTION
Background
Significant engineering advances in structural design have historically depended
on materials. Today there is a need for high strength, lightweight materials for a wide
range of applications such as aerospace, civil infrastructure, automotive, sporting goods,
etc. As a result, fiber reinforced composites such as graphite/epoxy are currently
receiving a great deal of attention due to their favorable mechanical properties. For
example, the X-34 rocket plane (Fig. 1), a reuseable orbital plane for one of NASA’s
research projects, uses graphite/epoxy composites for both its primary and secondary
structures [1]. With the innovative use of these
materials, researchers will be able to reduce
the operating cost per unit payload by an order
of magnitude. In order to thoroughly utilize
these materials to their highest potential,
every aspect of their behavior must be
understood. One aspect that requires
characterization is the influence of the Fig. 1 NASA’s X-34 rocket plan
manufacturing process on the mechanical behavior of the material itself. Since many of
these materials require a high temperature cure, residual stresses often occur in the final
2
structure. These stresses are caused by thermal expansion mismatch of the constituents
and from chemical shrinkage of the polymers, and are inherently difficult to measure and
characterize. This work documents efforts to measure and characterize residual stress in
a common graphite/epoxy system (AS4 graphite fiber /3501-6 epoxy).
Literature Review
Residual stresses, sometimes called “process-induced stresses,” in composite
materials are always a serious issue. We can consider the residual stresses as a pre-load
to the manufactured structure. Depending on how we utilize the composite materials, this
pre-load can be useful. However, it usually significantly degrades the material strength
[2]. A great deal of work has been done to understand, measure, and try to reduce
residual stresses in composites. Among the different kinds of composites, such as
particulate composite, laminated composite, and woven composite, and among the
different kinds of material compositions, such as glass beads, carbon fibers, metal matrix,
and polymer matrix, it is the fiber reinforced polymer matrix laminated composite that we
are particularly interesting in. This kind of composite, sometimes called fiber reinforced
plastics (FRP), is the most used in advance applications such as in the aerospace industry.
For this type of composite, residual stresses exist on two different scales [3]: the
fiber scale and the ply scale. Several researchers have established analytical or
computational micromechanics models to obtain the residual stresses on the fiber scale.
However, so far no attempt has been made to experimentally measure the residual
stresses on the fiber scale for real composite structural elements. In this work, we focus
on the ply-scale residual stress measurement and modeling. Based on that, the available
3
literature will be divided into three different classifications: experimental approach,
analytical approach, and computational approach.
Experimental Techniques
For measuring residual stresses, there are two main categories of experimental
techniques, destructive methods and nondestructive methods.
Destructive methods
As it states in the name, the destructive methods render the composites unusable
after testing. After measurements are taken by using destructive methods, the specimens
are usually no longer suitable to be structural elements. The hole-drilling method,
cutting/sectioning method, and first ply failure method are several common methods in
this category. They all involve taking a portion out of the specimen to create a free
surface and release the stresses on the surface.
The hole-drilling method is a very popular method to measure residual stresses. It
was originally developed for isotropic materials [4,5] and later for orthogonal materials
[6]. Traditionally, a strain gage rosette is used with the hole-drilling method to record the
strain relief around the hole. However, in the case of composite laminates, we have
verified that the attenuation of the stress relief around the hole dissipates in a few ply
thicknesses. Hence, strain gages cannot be used on composite residual stress
measurement because of spatial resolution limitations. Even the whole-field strain
measurement method, such as high sensitivity moiré interferometry, may not have
adequate spatial resolution to measure the deformation due to the drilling. Furthermore,
Schajer and Yang [7] showed that the generalization of elastic relations from isotropic
materials to orthotropic materials, conventionally used for hole-drilling method, is not
4
valid. For different degree of orthotropy, it is necessary to make some correction in the
compliance matrix for gage calibration.
Similar to hole-drilling methods, cutting methods utilized the idea that producing
a free surface or separating plies releases residual stresses [8-10]. Lee et al. and
Gascoigne used moiré interferometry to record the change of the displacement field
before and after cutting. They first replicated a diffraction grating on the lateral surface of
the composite specimen, and then used a diamond saw to cut either parallel or
perpendicular to the ply interface. The surface residual stresses and the intralaminar
residual stresses can be obtained respectively. Lee and Czarnek [11] went one step
further: With the data obtained from moiré interferometry and a finite element method
(FEM) post-processor, they claimed that the distribution of the residual strain could be
accessed within a single ply. Sunderland et al. introduced another variant of cutting
methods called “successive grooving.” The authors applied strain gages on one side of a
thin composite laminate specimen. On the opposite side, they used a diamond saw to cut
a groove and let the saw incrementally advance. By monitoring the change of strain and
equilibrium conditions, using classical laminate theory, the internal stresses (residual
stresses) in each layer can be calculated. The cutting method, like the hole-drilling
method, also encounters the problem that the stress field around the cut diminishes
rapidly for the composite laminates. The strain reading may not be accurate enough to
represent the deformation due to the stress relief.
Ply sectioning is another destructive method reported in the literature [12-14].
Sectioning could be achieved by precise machining or embedding a separating film inside
the laminate (that Måson and Seferis called “process simulated laminate (PSL)”). After
5
the ply removal and stress relief, the unbalanced stresses create out-of-plane deformation
on the laminates. Residual stresses were then calculated from laminate theory. The
drawback of this method is the difficulty of ply-separation. Among these works, Lee et
al. [8], Gascoigne [9], and Joh et al. [14] chose moiré interferometry as the tool for strain
measurement and obtained the displacement contour map around the cuts or sections.
Those results become valuable information for verification and comparison to the results
of finite element method (FEM).
Hahn [15] and Kim and Hahn [16] introduced the first ply-failure method. They
claimed that the swelling due to moisture absorption could be used to compensate for the
effect of curing stress due to CTE mismatch between the matrix and the fibers. When the
specimens absorbed the moisture with different degrees of saturation, the tensile strength
of the specimen also varied. The relationship between the strength and the swelling could
be correlated to obtain the residual stresses of composites. However, waiting for the
specimens to become saturated takes a long period of time, and the authors’ goals were to
understand fracture initiation of composites rather than measuring residual stresses.
Nondestructive methods
Although the nondestructive methods are more favorable for mechanical testing,
the existing nondestructive methods for residual stress measurement are somewhat
limited as explained in the following section.
Measuring the warpage (i.e. the curvature) of an unsymmetrical composite
laminate is a widely used method [17-20]. There are two methods to produce warped
laminates. The most used method involves producing an unsymmetrical laminate. Upon
cooling the laminate warps as stresses increase. The second method is to produce a
6
symmetric laminate and machine off layers from one side to relieve the residual stresses
thus producing a warped laminate. Of course, this will be classified as a destructive
method. With classical laminate theory, the curvature can be related to the moment
resultant on the laminates; thus, from the stress-strain relationship, the residual stresses
can be calculated. The disadvantage of this method is that it is very unusual to make a
structural component out of an unsymmetrical laminated composite; therefore, the usage
of warpage measurement is greatly limited when dealing with practical situations.
Imbedding strain gages into laminates and allowing them to go through the
manufacture procedure while monitoring the strain change is another nondestructive
method. Two kinds of strain gages have been used for this application: traditional strain
gages [21] and optical fiber strain gages [22]. In order to maintain the reference for strain
measurement, the strain gages and strain indication instruments need to be connected to
each other throughout the whole composite curing procedure. Although the strain gages
are small (thin), their presence acts to redistribute the residual stress in the composite.
Even the optical fibers, which are even smaller in dimension than traditional strain gages,
have a diameter 10 to 100 times larger than the diameter of a carbon fiber, thus disturbing
the natural order of the composite.
X-ray diffraction was utilized to measure the residual stress of composites by
including metal filler particles into the matrix material [23-24]. Although the diameter of
the filler particles has been reported to be on the same order as the diameter of fibers, the
issue is that fillers are distributed throughout the composite and affect the mechanical
properties of the composite globally rather than locally as in imbedded strain sensor
methods. Additionally, the relationship between residual stress and the results from X-
7
ray diffraction need to be carefully correlated in order to obtain meaningful information.
Madhukar, Kosuri, and Bowles [25] measured the residual stress by monitoring
the tension development in a single graphite fiber imbedded in epoxy during the curing
process. In this paper, the authors clearly demonstrate the interaction between chemical
shrinkage and thermal expansion, and proposed the way to optimize the curing cycle to
reduce the residual stresses. Nevertheless, those measurements were not taken on a real
part or structural element; consequently, the usefulness of the technique is greatly reduced
but can still yield valuable information in an academic sense.
In order to overcome the drawbacks in the methods mentioned above, Ifju et al.
[26] proposed a novel technique to measure the residual stress for symmetric fiber-
reinforced composite laminates using high sensitivity moiré interferometry. A moiré
diffraction grating is attached to the composite panel during the manufacture process.
After cure and after separation from the tool the diffraction grating will deform with the
laminate and thus record the dimensional change. By comparing the specimen grating
with the reference grating on the tool, the strain information of the composite can be
retrieved. However, to obtain the residual stresses from residual strains, the authors used
linear elastic laminate theory, which may be an over-simplified theory for most of the
polymer matrix composites (PMCs). In their conclusion, they claimed possible errors due
to neglecting viscoelasticity in PMCs (the analysis will be discussed in more details in
Chapter 2).
Shrinkage measurement of polymers
Once a method has been established to measure the total residual stress, in order
to reduce or even eliminate them, it is necessary to identify and characterize the sources
8
of these residual stresses. Most of the literature assumes that the chemical shrinkage of
polymer matrix and the differences in the thermal expansion/contraction of each
composition (the coefficient of thermal expansion (CTE) mismatch) during the
manufacture procedure are the two primary mechanisms inducing residual stress.
However, there is limited literature that addresses these two issues individually.
There are two types of polymers: thermoplastic and thermoset. In this study, the
discussion will be limited to thermoset polymer resin, which is the matrix material of our
composite system. For thermoset resin, polymer chemical shrinkage occurs when the
polymer chains start to crosslink. As crosslinking of the polymer advances, the polymer
chains bond together with strong covalent bonds. The polymer becomes consolidated and
stiffer, and the specific volume of the polymer shrinks. Several papers have been
published on the measurement of polymer chemical shrinkage. Hodges et al. [27], and
Snow and Armistead [28] used glass-bulb dilatometers to measure the volumetric change
of thermoplastic and thermoset resins during the curing process. Essentially, they used a
glass bulb connected to a capillary tube. The apparatus was partially filled with the resin
to be tested and the confining fluid (usually mercury or silicone oil). Then the bulb was
put into a temperature-controlled bath. The level of the resin was monitored while it was
curing. The specific volume versus time profile was obtained.
Russell [29] illustrated a different dilatometer. Instead of a glass bulb, the author
used a piezometer cell connected to a metal bellow. When the resin in the piezometer
cell underwent a change in volume (the author also measured the volumetric change for
laminate prepregs), the deflection of the bellow was calibrated to record the difference in
specific volume. Both Hodges et al. and Russell utilized the volumetric measurement as
9
a tool to monitor the progress of the chemical reaction and compared the result with
different methods such as viscosity measurement, differential scanning calorimetry
(DSC), and dynamic mechanical analysis (DMA). Their purpose was different from
strain measurement and did not attempt to relate their results to residual stress
information.
In 1995, Wang et al. [30] presented a unique method to measure the thermal
induced strains during cure and cooling for epoxy resins. By using an epoxy-coated thin
aluminum film and a quartz displacement probe, the chemical shrinkage and thermally
expansion of the neat resin were derived from the deflection of the aluminum film and
Timoshenko beam theory, and monitored in-situ. The authors did not attempt to correlate
the results to the residual stresses in composite materials.
Analytical and Computational Modeling
In addition to the above-mentioned experimental efforts, there have been
numerous mathematical models developed to determine residual stresses in composite
materials.
Elastic modeling
Hahn and Pagano [31] proposed a stress-strain-temperature relationship to model
the process-induced residual stress based on classical laminated theory. This is the first
attempt in the literature to assess the residual stresses of composites with a mathematical
model. However, since this was the first effort on this topic, the authors made several
assumptions, which have proved to be invalid. They assumed the stress-free state was
located at the end of highest temperature (cure temperature) stage. Now we know the
stress-free state (the gel point) is usually at the beginning or prior to the final hold stage.
10
Their thermal-elastic model was correct for the cooling stage, but apparently could not
describe the viscoelastic behavior of the polymer at the cure temperature holding stage.
Also, the authors only considered the mechanical and thermal contributions and
completely neglected the irreversible chemical shrinkage of the polymer matrix materials.
Bogetti and Gillespie [32] used a one dimensional heat conduction equation and
finite difference analysis to stimulate the curing process of a thick thermoset composite
laminate. According to their parametric study, input values for volumetric shrinkage
significantly influenced the prediction of the residual stress distribution. This study gives
insight into how important the experimental shrinkage measurement can be in the residual
stresses. Additionally, their work did not take viscoelasticity into account.
Viscoelastic modeling
Strictly speaking, almost all commonly used PMCs more or less exhibit some
viscoelastic properties. When the service temperature of the composite is high, it is
usually necessary to cure the composite at an even higher temperature to achieve the
desired thermal stability at the service temperature. In this case, the viscoelastic
behaviors, such as stress relaxation and creep, will be highly accelerated due to the time-
temperature dependence of polymer matrix [33]. Since the mid-60s, the viscoelastic
properties of composites started to gain attention. Around the mid-60s, Schapery [34-37]
published several papers on the constitutive modeling of viscoelastic media under the
influence of temperature. In Schapery’s 1967 paper, by utilizing the corresponding
principle, the author clearly defined the general form of the stress-strain relationship for
anisotropic composite materials with a linear viscoelastic and thermorheologically simple
assumption. However, that work did not address the issue of residual stress in
11
composites, neither did it consider the effect on the composition of anisotropic composite
materials.
Weitsman [38] investigated the effect of the temperature profile on the curing
cycle both analytically and experimentally and attempt to obtain an optimal cooling path
to minimize the residual stresses. Harper and Weitsman [39,40] also employed the
corresponding principle by implementing the viscoelastic properties of the resin.
However, their work was only limited to the cooling stage of the curing cycle.
White and Hahn [41,42] made an in-depth discussion of the curing process model.
In the article, a model called LamCure was presented which combined the cure kinetics
with the residual stress model. Differential Scanning Calorimetry (DSC) was used to
obtain the parameters needed for the cure kinetics model, and several tensile and creep
tests were performed on partially cured and fully cured composite specimens to obtain
viscoelastic mechanical properties. Nevertheless, the authors neglected the contribution
of chemical shrinkage to the final residual stress. They claimed that at high temperature,
like the cure temperature of the composite, the viscoelastic properties of the material
dominates, and that stress relaxation was evident. The stress due to chemical shrinkage
would relax rapidly and cannot be developed. The present author believes that this is not
true. According to Liu et al. [43], the experiments show that, depending on the lay-up of
the composite laminate, there are from 2 to 22 percent of final residual stress induced by
the chemical shrinkage.
Wang et al. [44] used linear thermoviscoelastic laminate theory to calculate the
residual stresses and warpage of unsymmetrical woven-glass/epoxy laminates. The
formulation of stress-strain relationship was presented. Unlike White and Hahn’s
12
approach, the authors did not consider the matrix and fiber separately and did not
examine the cure kinetics of the system. Consequently, the model was not able to depict
the evolution of curing in terms of degree of cure. Instead, the time coordinate along a
certain temperature profile was used to describe the evolution of the chemical reaction.
Once the temperature profile was determined, the degree of cure should be a unique
function of time.
Kim and White [45] modeled relaxation modulus of 3501-6 epoxy as a
thermorheologically complex material. Shift functions were also obtained. The
experiments were done by using a dynamic mechanical analyzer (DMA). With the
information of the shift functions the creep master curves were obtained and the time-
dependent viscoelastic moduli could be formulated accordingly. At about the same time,
Adolf and Martin [46] presented a comprehensive constitutive model for a cross-linked
polymer to calculate the process-induced stresses. Experimental data for the viscoelastic
material properties and the cure kinetic model parameters were required for the
calculation. However, these papers used the data obtained from matrix resin specimens
to model the whole composite material system (fibers and matrix). It is doubtful that the
models could represent the composite systems well.
Research Objectives
In this work, there are two main objectives proposed.
First, to develop a series of methods that can measure the residual strains and
quantify the components of residual strains resulting from different mechanisms. The
mismatch of CTE between the fibers and matrix, and chemical curing shrinkage of
13
thermoset resins are two mechanisms we consider in this study. In order to achieve these
two purposes, this technique must allow us to perform the measurement in the elevated
temperature environments and be capable for “recording” the strain development during
the composites (and/or their matrix resins’) curing process. With the literature review
performed, modifying and improving the current methodology of moiré interferometry
are the most suitable for characterizing process induced residual stresses of polymer
matrix composites. There are many existing techniques that can be used to measure
residual strain of composites or the chemical curing shrinkage of polymers, but none of
them can accomplish both tasks at the same time. Moreover, not all of the chemical
shrinkage contributes to the residual stresses. The new technique is able to measure only
the portion of the shrinkage, which occurs after the gel point and is accountable for
residual stresses. In this work, the technique called “cure referencing method” (CRM),
based on high sensitivity moiré interferometry and an optical thermal chamber, is
implemented to achieve our first goal and is addressed in the following chapters.
Second, using a viscoelastic constitutive equation to describe the stress-strain
behavior of a polymer matrix composite laminate. Classical laminated theory and time-
temperature superposition (TTS) principle [47] were employed to construct the equation.
The long-term material properties, the elements in the creep compliance matrix, were
measured through a series of tensile tests on 0-degree longitudinal, 90-degree transverse,
and 10-degree off-axial unidirectional composite laminate specimens. To obtain the
long-term properties in a short-term test, another thermal chamber was used to elevate the
temperature of the testing environment to accelerate the time effect. With this
14
viscoelastic model, the relaxation of the residual stress can be predicted and compared
with the experimental measurement by the cure referencing method.
15
CHAPTER 2HIGH TEMPERATURE MOIRÉ INTERFEROMETRY AND THE CURE
REFERENCING METHOD
High Sensitivity Moiré Interferometry
Moiré interferometry is an optical method that utilizes moiré phenomenon of
optical interference to measure the deformation of objects. Geometric moiré, shadow
moiré, high sensitivity moiré interferometry, or even micro moiré interferometry are all
based on the same principle—the interference between two gratings of similar frequency.
Although the methodology of moiré interferometry is not new, the book published by
Post et al. in 1994 [48] opened a new era for the high sensitivity moiré interferometry.
Figure 2 illustrates the principle of high sensitivity moiré interferometry. The
angle of incidence α of the light beams into specimens can be determined by the
wavelength of coherent light source (λ), the frequency of virtual grating (f) and the
following equation
αλ
sin2 ⋅=f (2.1)
The frequency of specimen grating (fs) is required to be equal to f / 2 in order to
produce fringe patterns in the direction normal to the specimen surface. When the
specimen grating is undeformed, the reference virtual grating can be tuned by adjusting
the angle of the two incident light beams, so that it perfectly overlaps with the original
16
master gratings or undeformed specimen grating. In this case, if we view from the
camera, there will be no interference fringe. This is called null-field. However, when the
specimen is deformed in the x-y plane, the two first-order diffracted light beams carry the
distorted wave fronts and the interference between them will produce a fringe pattern.
α
1st order diffraction from incident beam 1 interference patterns
captured by camera
Incident beam 1
Incid
ent b
eam
2
0th order diffraction (reflection)from incident beam 2
0th order diffraction (reflection)from incident beam 1
1st order diffraction from incident beam 2
Specimen withdiffraction grating,grating frequency fs
Virtual reference grating,grating frequency f
Fig. 2 Schematic diagram of moiré interferometry
Once the fringe patterns are photographed, the in-plane displacement field can be
extracted according to the equations
),(1
),(
),(1
),(
yxNf
yxV
yxNf
yxU
y
x
⋅=
⋅=(2.2)
17
or with selected gage length, two in-plane normal strains and the in-plane shear strain can
be also calculated by the equations:
)(1
1
1
x
N
y
N
f
y
N
f
x
N
f
yxxy
yy
xx
∆∆
+∆
∆⋅=
∆∆
⋅=
∆∆⋅=
γ
ε
ε
(2.3)
where U (x, y) and V (x, y) are the displacements in x and y direction at the point of
interest (x, y) with respect to the chosen origin. Nx and Ny are the fringe numbers in x and
y direction from the origin to the point of interest. εx and εy are the normal strains in x
and y direction, and γxy is the shear strain. ∆x and ∆y are the selected gage length, ∆Nx and
∆Ny are the fringe order difference over the selected gage length respectively. The actual
sensitivity of the setup depends on the wavelength of our light source and the frequency
of reference grating. The optical setup in this study employs a 20 mWatt, class IIIb,
Helium-Neon laser (wavelength λ = 632.8 nm) from UNIPHASE, and the frequency of
the reference grating is 2,400 lines per mm (60,960 lines per inch). From equations (2.1)
and (2.2), we obtain the incident angle of 49.4 degrees, and the displacement sensitivity
of 0.417 µm per fringe order. Figure 3 is the photography of the moiré interferometer
setup used for residual stress characterization. A four beam system is employed to obtain
both U-field and V- field images. The laser beam is squeezed into a single mode optical
fiber using a laser coupler to produce a point light source at the fiber tip of the other end.
The divergenet light from the point light source is caught and collimated by a parabolic
mirror, and split into four beams by four 45-degree flat mirrors.
18
Fig. 3 Photograph of a four-beam interferometer
There are several advantages of high sensitivity moiré interferometry over
traditional geometric moiré or the other displacement/strain measuring methods. With the
much higher grating frequency than geometric moiré, it can achieve very high sensitivity.
Also the spatial resolution and signal-to-noise ratio are superior. From the pictures in the
next few chapters, even with very high density of fringes, the fringes are still well defined
and in excellent contrast. Whole field measurement is another advantage of moiré
interferometry. As in Fig. 2, the addition of two incident light beams in the y direction
will enable us to measure the horizontal displacement and shear strain components. This
capability makes moiré interferometry a very favorable experimental tool for finite
element analysis (FEA) validation. Material properties or boundary conditions required
for FEA input can be obtained or corrected. The displacement contours or strain contours
19
from moiré fringe patterns can also be used to compare with the computational results.
The most important advantage of moiré interferometry in this study is the long-
term capability of monitoring the strain field. Establishing a reference condition at some
point in time is the key for long-term strain measurements. The reference condition for
moiré interferometry can be obtained by tuning the master grating (parent grating) to a
specific specimen grating. The procedure of specimen grating replication will be
elaborated on a later section. The master grating is usually made on the substrate material
that can be treated as totally rigid within the time frame we are considering. Therefore, as
long as the master grating is not damaged or lost, the subsequent measurement can always
be performed. Unlike most of other methods we mentioned in the literature review, the
specimen will loose its reference once removed or disconnected from the experiment
apparatuses. One more benefit from its robust reference condition, moire interferometry is
also an outstanding method for investigating the hygrothermal effects on polymer matrix
composites. In the current study, thermal load was applied on composite laminate
specimens, but the moisture effect was not considered. The humidity effect is considered
negligible since we retain all specimens in a desiccant box.
Non-Room Temperature Moiré Interferometry with the Optical Thermal Chamber
Moiré interferometry can be much more versatile if the usable temperature range
can be extended well below and above room temperature. There are two issues
associated with non-room temperature moiré interferometry: First, one must create an
environment that can perform the moiré tests. For simplicity, a thermal chamber only for
retaining specimens (not entire interferometer optical setup) was designed and
20
manufactured. This issue is discussed in the next section. Second, the specimen gratings
must survive the specific environment without degradation or distortion. The second
issue can be resolved by carefully choosing the procedure and/or the materials used for
grating replication. This issue will be discussed later in this chapter.
Optical Thermal Chamber
In order to perform the moiré tests in a specific thermal environment, an optical
thermal chamber needed to be constructed to fulfill several requirements. First, after
mounted in the chamber, the specimen position must be adjustable. Second, distortion
caused by the window glass must be minimized in order to prevent errors. With these
conditions in mind, our first thermal chamber was constructed.
General description
To extend the thermal application of moiré interferometry, an environmental
chamber with a wide temperature range capability was necessary. To meet the
requirements of moiré testing, a remote thermal chamber was built and connected to a
thermal chamber (model number EC12 from Sun Electronic Systems, Inc.). The EC12
thermal chamber (will be referred as “the oven” hereafter) has a maximum allowable
temperature range of -173 °C to +315 °C when it is connected to a low pressure liquid
nitrogen tank, and serves as a heating or cooling source for the remote chamber. The
hot/cool air is drawn to the remote chamber by a 4-inch diameter centrifugal tangential
duct fan and flexible thermal tubing (McMaster-Carr Supply Company, Sure-Flow
silicone-coated fiberglass hoses, temperature range -60 °C to 260 °C) insulated with glass
fiber. The outer walls of the chamber were constructed with 1/4 inch thick aluminum
plates, and had dimensions of 165mm x 165mm x 165mm (6 ½” x 6 ½ ” x 6 ½ ”). For
21
chamber insulation, the interior of the remote chamber is lined with one-inch-thick high
temperature calcium silicate board (McMaster-Carr Supply Company, max. temp. 927
°C, thermal conductivity 0.8 Btu @ 427 °C), thus making the inner dimensions 102 mm x
102mm x 102mm (4” x 4” x 4”). The chamber window consisted of three pieces of 1/4”
thick glass with wedged aluminum spacers in between to reduce optical noise due to
multiple reflections. The inner most layer of window was made of low expansion
BorofloatTM glass to avoid cracking due to the high temperature gradient through the
thickness during heating or cooling. The system setup is shown in Fig. 4 schematically
and is photographed in Fig. 5.
Fig. 4 Schematic diagram of the environmental chamber system setup
22
Fig. 5 Photograph of the environmental chamber system setup
A thermal couple probe is extended from the oven through the air duct into the
remote chamber. By attaching the thermal couple onto the specimen surface, we are able
to set the temperature profile (thermal history) of specimens in the remote chamber.
Chamber calibration
There are two main concerns for the calibration of the environmental chamber.
First of all, since moiré interferometry is an optical method, any optical element we put
into the optical path will alter the results. The window glass of the chamber can be a
source of problems. To document the effect of the window glass, an aluminum coated
silicone rubber grating on 6061-T3 aluminum alloy substrate was made as a control
specimen for calibration. The aluminum control specimen was first tuned to null field
without the chamber (without glass), and a picture of the fringe pattern was taken. Then,
23
we replaced the regular specimen holder by the chamber. By only adjusting the position
of the specimen and without changing the interferometer, the new fringe patterns show
the distortion caused by the window glass. Figure 6 (a), (b), (c), and (d) show the pictures
of null field patterns of a control specimen without and with window glass. The size of
the square gauge mark at the center of the specimen is 9 mm by 9 mm, and there are
three-fringe difference in U field and one fringe difference in V field across the whole
area. According to this experiment, the distortion is small but not negligible. We can
minimize this problem by positioning the master grating inside the chamber when we
tune the moiré interferometer for the reference position. Also when replacing master
grating with the specimen, the tuning procedure needs to be done inside the chamber as
well.
The second concern of using the chamber is the temperature stability. To perform
the calibration, the aluminum control specimen was again put into the remote chamber.
Figure 7 shows the temperature change inside the remote chamber in the first 100 seconds
after we turned off the circulation fan to obtain a vibration-free condition, which is
essential for photographing moiré fringe patterns. From high temperature to low
temperature several settings were tested. The curves represent the thermal couple reading
from the oven with the probe almost touching the specimen, and the data points are the
results from analyzing moiré patterns. As we expected, the larger the temperature
difference (from room temperature), the faster the temperature drops (for high
temperature) or increases (for low temperature). Moreover, the actual temperature we
obtained from the fringe patterns was more stable than the temperature reading from the
24
(b)
(d)
(a)
(c)
Fig. 6 Null field fringe patterns of distortion calibration. (a) U field without the chamber, (b) V fieldwithout the chamber, (c) U field with the chamber, (d) V field with the chamber
25
thermal couple, since the air inside the chamber dissipates/absorbs heat much faster than
the specimen does. Also the thermal mass of the thermocouple was much smaller than
that of the specimen.
It is important that we make a correction for the coefficient of thermal expansion
(CTE) as a function of temperature for our aluminum alloy control specimen during
analysis. The CTE of the aluminum alloy is a function strongly depending on the
temperature [49] within our testing range (from -130 °C to +175 °C). From Fig. 8, we can
see the CTE at -130 °C is only about 62.5% of that at +175 °C. The actual temperature (
Ti ) on the control specimen is obtained from solving the following integration equation
dl l T dTl
l
T
Ti i
0 00∫ ∫= ⋅α ( ) (2.4)
here T0 is the room temperature, l0 is the original gauge length at room temperature, α(T)
is the function of the CTE in terms of temperature T, and
dll
li
0∫ = ∆l = fringe number N × 0.417 µm (2.5)
is obtained from fringe patterns. If we look at 175 °C data points in Fig. 7 where the
temperature drops fastest, the average temperature change in 10 seconds is 0.4 °C, which
corresponds to 1.24 µε on an AS graphite/epoxy laminate [50] in transverse fiber
direction, and 48 µε on the neat epoxy resin. These strain values are in about the same
order of magnitude as the sensitivity limit of our moiré interferometer, and the average
temperature change we used was obtained from our aluminum control specimen, which
has relatively high thermal conductivity. Therefore, the actual temperature change on
26
laminate specimens or neat resin specimens is less, and the strain error would be even
smaller.
Fig. 7 Calibration tests in five different temperature settings. The curves areobtained from thermal couple readings inside the remote chamber, and themarkers are obtained from analyzing moiré fringe patterns of aluminumcontrol specimens which are the more precise representation of thespecimen surface temperature.
27
Fig. 8 The coefficient of thermal expansion for aluminum alloy 6061. The twodiscontinuities at -60 °C and 100 °C are due to different fitting functions fordifferent temperature ranges.
Cure Referencing Method
With all the features stated before, moiré interferometry, combined with a novel
technique of grating replication during the composite specimen fabricating process,
constitutes a suitable methodology for the measurement of the process induced residual
strain. This methodology is called Cure Referencing Method (CRM). In order to explain
the concept of CRM better, first the principle of CRM will be given, and the following
sections will give a brief step-by-step description of how to perform residual stress
measurement by CRM.
28
The Principle
In moiré interferometry, displacement of the specimen is measured by the
deformation of the diffraction grating on its surface. For CRM, the grating is attached to
the composite specimen during the curing process, at the stress-free state. The stress-free
state exists before the matrix resin starts to gel. Because the ultra-low expansion substrate
material of the diffraction grating also serves as a tool for the composite curing
procedure, the diffraction grating remains undeformed until the curing procedure is
completed, and the laminate specimens are separated from the ultra-low expansion tool.
Before resin gelation, the epoxy resin is still free to flow around fibers and unable to carry
stress. At that point, the matrix is solidified and the grating is rigidly attached to the
laminate surface. Residual stresses arise from this point, but the in-plane dimensions of
specimen and the frequency of specimen grating remain the same due to the constraint
from the rigid tool.
Once the specimen and grating are separated from the master grating tool, the
accumulated residual stresses will deform the specimen as well as the diffraction grating
attached to it. If the interferometer is first tuned to the null field using the master grating
tool, and by replacing the master grating with specimen grating, the relative displacement
between the master grating and specimen grating can be measured accordingly.
Grating Replication
In order to perform a moiré test, a diffraction grating need to be replicated onto
the specimen surface. This is usually done by transferring the grating from a master
grating onto the specimen surface with an adhesive layer. Figure 9 shows the general
procedure for grating replication. Step 1: pour a pool of adhesive onto the master grating,
29
and slowly lower the specimen into the pool avoiding air bubbles. Step 2: apply uniform
pressure on the top of the specimen and clean up excessive adhesive. Step 3: separate the
specimen from the master grating after adhesive is cured.
VSHFLPHQ
XQFXUHGDGKHVLYH
PDVWHU�JUDWLQJ
PHWDOOLFILOP
(a) Step 1
ZHLJKWFRWWRQ�VZDE
XQFXUHGDGKHVLYH
(b) Step 2
FXUHGDGKHVLYH
PDVWHU�JUDWLQJ
PHWDOOLFILOP
VSHFLPHQDQG�JUDWLQJ
(c) Step 3
Fig. 9 General grating replication procedure
Like a regular moiré interferometry test, CRM starts from the replication of
diffraction grating. The difference is that the replication procedure takes place during the
30
composite manufacturing process. To survive the harsh environment of high temperature
and high pressure during the curing process, an alternate approach for grating replication
was developed and adopted [51]. In this procedure, all gratings were replicated on a kind
of ultra low expansion glass called Astrosital to maintain the dimensional stability of
gratings. First, we start with a silicone rubber (GE RTV 615) master grating. A 1,200
lines/mm, phase type, crossed line photo-resist master grating (A) was used for
replicating the silicone rubber master grating (B) in this study (refer to Fig. 10).
Second, an intermediate grating (C) made of Shell Epon 862 and curing agent W
was replicated from the silicone rubber master grating (B). For the best result the master
grating, substrate material, the epoxy compounds were heated to 130 °C, and the epoxy
was degassed in a vacuum oven. The intermediate grating was cured at 130 °C for ten
hours. After separation from the master grating, two layers of aluminum films with
diluted Kodak Photo-flo in between [48] were vacuum deposited on the intermediate
grating (C’).
This intermediate grating (C’) became the master grating of the grating tool for
autoclaving. The epoxy for replication of the grating tool is the same epoxy as the matrix
resin in our composite prepreg, 3501-6 epoxy from Hexcel. After it was cured at 130 °C
for ten hours and post-cured at 177 °C for two hours, the grating tool (D) was separated
from the intermediate grating, and two layers of aluminum were vacuum deposited on the
tool (D’) like the intermediate grating.
Finally, a thin layer of 3501-6 epoxy was cast on the top of the two layers of
aluminum on the tool grating with a special tool wrapped by Teflon film [51]. The tool
grating (D”) now is ready for the autoclave procedure.
31
Fig. 10 Grating replication procedure for CRM
32
Composite Specimen Fabrication
Four different prepreg lay-up sequences, [0]16 unidirectional, [02/902]2s cross-ply,
[0/903]2s unbalanced cross-ply, and [02/452]2s angle-ply were prepared for each autoclave
specimen fabrication. The original size of the panels were 6 inches (152.4 mm) square.
The prepreg panels were then put into a vacuum bag (Airtech International Inc.,
WN1500) together with release films (Airtech International Inc., Release Ease 234TFP-1
and 234TFNP), bleeder and breather fabrics (Airtech International Inc., Airweave Super
10), and the master grating tool. The vacuum bag lay-up for the laminate specimens is
illustrated in Fig. 11. The whole vacuum bag assembly was then put into an
electronically heated autoclave oven (Baron Blackesleee Inc., Model BAC-24, max.
pressure 110 psi, max. temperature 650 °F) for curing.
For AS4/3501-6 graphite/epoxy pregreg, the manufacturer's recommended curing
profile shown in Fig. 12 was used [52]. First, the temperature was raised at 5 °F per
minute to the first dwell stage of 225 °F for one hour accompanied with 15 psi pressure
outside the vacuum bag, and 30 inches Hg of vacuum inside the vacuum bag. The main
purpose of the first stage is to keep the matrix resin in a low-viscosity condition and allow
vacuum to eliminate voids inside the pregreg. After one hour in the first dwell stage, the
temperature is raised again at the same rate, 5 °F per minute, to the final curing
temperature of 350 °F. The outer pressure is then increased to 100 psi, and the vacuum
inside the vacuum bag is released to atmosphere when the outer pressure reaches 30 psi.
This final curing stage, with the high temperature and the high pressure conditions, allows
the crosslinking of matrix resin to develop and the consolidation of the whole composite
system to complete. After six hours in the final curing stage, we start
33
Fig. 11 Vacuum bag lay-up for CRM
decreasing the temperature also at 5 °F per minute until the autoclave temperature drops
to 100 °F. Pressure is released when the temperature reaches 175 °F. Finally, the
vacuum bag is removed from the autoclave, and the composite specimens are removed
from the vacuum bag and separated from the grating tool.
Experimental Setup and Procedure
Three batches and four different laminate lay-up specimens as mentioned before
for each batch were manufactured.
Tuning of optical setup
The total process induced residual strains were first measured at room
temperature. A four-beam moiré interferometer was used for this measurement.
A schematic diagram of the optical setup is shown in Fig. 13. The interferometer
is first tuned with respect to the intermediate grating, which has exactly the same pattern
as theundeformed specimen grating (unlike the grating tool, which is the mirror image of
the specimen grating due to the in-perfection of perpendicularity of the crossed-line
34
0
(°F)
50
100
150
200
250
300
350
400
1hr.
6 hr.s
10
20
30
40
50
60
7080
90
100
(psi)
Time
Temperature
Presure
vacuum (30 in. Hg)
vacuum released
Fig. 12 AS4/3501-6 composite curing profile
grating). The interferometer was tuned so that two incident beams were perfectly
collimated [51] (Fig. 14), the virtual grating was perfectly registered with the intermediate
grating, and a null field pattern appeared on the camera screen.
Once the null field image was obtained, the reference grating was removed, and
replaced by the specimen grating. The optical setting of the interferometer is required to
remain unchanged since the virtual grating is now our reference grating. The specimen
grating was tuned until the 0th order diffraction (reflection) light beams converged back
to the point light source on the tip of the optical fiber.
For the moiré tests at elevated temperature, the laminate specimens were cut down
to 3 inches (76.2 mm) square by a diamond impregnated saw to fit into the chamber. The
interferometer was first tuned to the null field with respect to the specimen grating at the
room temperature inside the thermal chamber. When the temperature inside the chamber
35
Fig. 13 Schematic diagram of a four-beam moiré interferometer
increased (closer to stress-free temperature), the fringe density also increased. By doing
so, if the specimen has 1% strain, the error can be easily calculated as 0.99%.
Thermal loading for composite laminate specimens
Starting from the room temperature, the specimen gratings were heated at a rate of
3 °C per minute to 15 °C increments. After the set temperature was reached, the
temperature was held for 15 minutes to allow the chamber and specimen grating to reach
thermal equilibrium. Then the oven and the circulation fan for the thermal chamber were
both turned off to obtain a vibration-free condition for photographing the moiré fringe
patterns. Both the U-field (horizontal) and V-field (vertical) images were taken in a
timely fashion to avoid a serious drop of temperature inside the chamber and specimen
36
perfect collimation
convergence
divergence
A B C
A B C
f f fA = = B C
f f fA > > B C
A B C
f f fA < < B C
Fig. 14 Collimation of incident beams. When the incident beams are perfectlycollimated, the frequency of the virtual grating will be space-wise constant(fA= fB = fC). The moiré pattern due to the interference between virtualgrating and the reference grating will be insensitive to the position of thereference grating.
37
grating. KODAK black and white 35mm TMAX 400 roll film and a single-lens reflex
(SLR) camera without a lens were used for photography. After photography, the
temperature was raised again to the next 15 °C increment until the curing temperature,
177 °C, was reached. The same procedure was repeated for cooling from 177 °C to room
temperature. Rotation carrier fringes were applied to several fringe patterns to make the
images more readable. Totally, twelve specimens, three autoclave runs, four different
lay-ups of laminates per run were tested. Figures 15 through 18 are some typical fringe
patterns taken with the thermal chamber during the tests at various temperature settings.
The circular marks shown on those pictures were drawn using a compass with 12.7mm
(1/2 inch) radius. The directions of the crossed marks were aligned with the fiber
direction and transverse fiber direction.
Data Analysis and Results
The process induced residual strain of [0]16 unidirectional specimen, {εuni}, was
directly obtained from moiré fringe images, and was treated as the result of free thermal
contraction and matrix resin chemical shrinkage, since all its fibers orient in the same
direction and there is no mutual constraint between plies. Equation (2.6) describes this
relationship.
shrinkage
yuni
xuni
thermal
yuni
xuni
yuni
xuni
xyuni
yuni
xuni
uni
+
=
=
=
000
}{ _
_
_
_
_
_
_
_
_
εε
εε
εε
γεε
ε (2.6)
The apparent strains {εlam}, which were obtained from the moiré fringe patterns of
the other three lay-ups of laminate specimens, were actually the result of free thermal
38
(a) -field, 23.2 °CU (b) -field, 23.2 °CV
(c) -field+ rotation, 23.2 °CU (d) -field+rotation, 23.2 °CV
(e) -field, 70 °CU (f) -field, 70 °CV
[0 ]16 T
0° fiber direction
Fig. 15 Fringe patterns for [0]16 unidirectional laminate at different temperatures.
39
[0 ]16 T
0° fiber direction
Fig. 15--continued
40
(a) -field, 23.2 °CU (b) -field, 23.2 °CV
(c) -field+rotation, 85 °CU (d) -field+rotation, 85 °CV
(e) -field, 177 °CU (f) -field, 177 °CV
0° fiber direction
[0 /90 ]2 2 2s
Fig. 16 Fringe patterns for [02/902]2s symmetric cross-ply laminate at differenttemperatures.
41
(g) -field+rotation, 100 °C(cooling)
U
(i) -field+rotation, 23.5 °C(cooling)
U
(h) -field+rotation, 100 °C(cooling)
V
(j) -field+rotation, 23.5 °C(cooling)
V
0° fiber direction
[0 /90 ]2 2 2s
Fig. 16--continued
42
(a) -field+rotation, 23.8 °CU
(c) -field+rotation, 100 °CU
(e) -field, 177 °CU
(b) -field+rotation, 23.8 °CV
(d) -field+rotation, 100 °CV
(f) -field, 177 °CV
0° fiber direction
[0 /90]3 2s
Fig. 17 Fringe patterns for [03/90]2s symmetric unbalanced cross-ply laminate atdifferent temperatures.
43
(g) -field, 115 °C(cooling)
U
(i) -field+rotation, 40 °C(cooling)
U
(h) -field, 115 °C(cooling)
V
(j) -field+rotation, 40 °C(cooling)
V
0° fiber direction
[0 /90]3 2s
Fig. 17--continued
44
(a) -field, 22.5 °CU
(c) -field, 100 °CU
(e) -field+rotation, 177 °CU
(b) -field, 22.5 °CV
(d) -field, 100 °CV
(f) -field+rotation, 177 °CV
[0 ]2 2s/452
0° fiber direction
Fig. 18 Fringe patterns for [02/452]2s symmetric angle-ply laminate at differenttemperatures.
45
(i) -field+rotation, 23.2 °C(cooling)
U (j) -field+rotation, 23.2 °C(cooling)
V
(g) -field+rotation, 115 °C(cooling)
U (h) -field+rotation, 115 °C(cooling)
V
[0 ]2 2s/452
0° fiber direction
Fig. 18—continued
46
contraction, matrix chemical shrinkage after transformation to the specific orientation
superimposed with the deformation due to residual stresses (constraint in between
adjacent plies). This relationship can be written as the following:
}{}{][}{ i.e.
}{}{][}{
1
1
lamunikkresidual
kresidualuniklam
T
T
εεε
εεε
−⋅=
−⋅=
−
−
(2.7)
[T]k-1 is the transformation matrix for the specific orientation of the kth ply.
Therefore, the residual stress in the kth ply {σresidual}k can be calculated from the residual
strain matrix {εresidual}k, the transformation matrix [T]k-1, and the laminate stiffness matrix
[Q] by the following expression:
}){}{]([][][][}{ 11lamunikkkkresidual TTQT εεσ −⋅⋅⋅⋅= −− (2.8)
More detailed explanation and analysis can be found in the previous work [53, 54].
To achieve another objective, separating the two components of residual strain
due to thermal contraction and matrix chemical shrinkage, the fringe patterns at every
temperature setting were analyzed. By gradually bringing the specimens back to the
stress-free temperature, the residual strain due to the thermal contraction was
compensated by thermal expansion. The amount of strain remaining on the specimens
would be solely due to the chemical shrinkage of matrix resin. Since the null fields were
tuned with respect to specimen gratings at room temperature, the strain values obtained
from the elevated temperature fringe patterns need to be subtracted from the total strains.
With the data points at different temperatures, and the assumption of constant coefficient
of thermal expansion (CTE), the CTE’s of composite laminate are the slope of strain
47
versus temperature curves. Figures 19 through 22 are the strain-temperature curves for all
four kinds of laminate specimens. The longitudinal fiber directions (0-degree direction)
are aligned with x- direction.
Fig. 19 The strain-temperature curves of [1016] unidirectional laminated composite
48
Fig. 20 The strain-temperature curves of [02/902]2s cross-ply laminated composite
Fig. 21 The strain-temperature curves of [0/903]2s unbalanced cross-ply laminatedcomposite
49
Fig. 22 The strain-temperature curves of [02/452]2s angle-ply laminated composite
Significance of the Tests
As seen in Fig. 19 through 22, at the cure temperature, those residual strain
components in the transverse direction (fiber dominated direction) did not come back to
zero. The difference is about 22.3% of the total residual strain for unidirectional
specimens, 3.2% for balanced cross ply specimens, 12.1% for the unbalanced cross ply
specimens, and 21.4% for the angle ply specimens. Table 1 summarizes the CTE’s and
the portion of residual strain due to matrix chemical shrinkage of the four different
composite specimens in both longitudinal and transverse fiber directions.
50
% of residual strain dueto chemical shrinkage
(%)
---
22.3
2.0
5.3
4.5
12.1
3.2
21.4
Residual strain duematrix chemicalshrinkage (µε)
---
1083.1
8.4
20.5
11.1
118.2
6.7
646.8
Residual strain due tothermal contraction (µε)
47.7
3768.4
406.6
365.0
232.5
857.8
203.3
2371.6
CTE(µε/°C)
0.31
24.47
2.64
2.37
1.51
5.57
1.32
15.40
Average totalresidual strain @
RT (µε)
36.1
4851.5
415.0
385.5
243.6
976.0
210.0
3018.4
Directions
longitudinal
transverse
longitudinal
transverse
longitudinal
transverse
longitudinal
transverse
Table 1 Components of residual strain in different lay-ups of laminate specimens
[0]16
[02/902]2s
[0/903]2s
[02/452]2s
51
CHAPTER 3POST-GEL CHEMICAL SHRINKAGE OF THERMOSET RESINS
Introduction
In Chapter Two, the results from the high temperature moiré tests indicate the
importance of characterizing the matrix chemical shrinkage of the polymer composite. It
is necessary to develop a more comprehensive and in-depth investigation of the curing
shrinkage. To measure the curing shrinkage of the matrix resin is the goal in this chapter.
For thermoset polymers, the crosslinking of the polymer chains result in chemical
shrinkage during the curing process. As the curing process progresses, the polymer
chains continue crosslinking until the asymptotic value of crosslinking density is reached
at the particular ambient temperature and pressure condition [55]. However, only the
shrinkage occurring after the solidification will contribute to the residual stresses in
composite materials. Although the post-gel shrinkage is only a small portion of total
curing shrinkage, we have proven the influence of post-gel shrinkage cannot be ignored.
Before gelation, the polymer resin is in its fluid state and is not capable of carrying
mechanical stress. After the gel point, the resin begins to solidify and stresses build up
between the fibers and matrix. Therefore, the shrinkage that occurs after the gel point is
more meaningful to the residual stress characterization. This chapter describes a new
technique, which focuses on measuring only the post-gel chemical shrinkage of thermoset
resins, and was developed particularly for the purpose of residual stress characterization.
52
Experimental Methodology
In these experiments, shrinkage measurement was taken using moiré
interferometry [48]. As mentioned in Chapter 2, one big advantage of moiré
interferometry is its long-term capability of strain measurement due to the ease of
preserving its reference condition. In this chapter, the advantage is exploited again to
record the deformation caused by resin chemical shrinkage during the curing process. An
investigation on the distortion of master grating was also documented to further support
the cure referencing method (CRM) developed previously.
The interferometer and optical setup used in the Chapter 2 was used again for
these shrinkage measurements, but this time, neat resin specimens were produced instead
of the composite laminate specimens. To demonstrate the usage of this newly developed
technique, five different resins (one room temperature cure epoxy, three ultra-violet light
cure arcrylates, and one high temperature cure epoxy) were used to produce neat resin
specimens. The names of the resins and their vendors are given in Table 2 below. The
high temperature cure epoxy was chosen to be the same epoxy of the composite system
used for the CRM.
Specimen Preparation and Experimental Procedure
The neat resin specimen gratings were directly replicated from master gratings
during the manufacture process of the neat resin specimens. In the process, the specimen
gratings solidify at the gel point of the resins. As the curing process continues, chemical
shrinkage and thermal effects cause stresses within the specimens. When the specimen is
53
Table 2 Names and vendors of the resins used
Roomtemperature cure UV cure1 High temperature
cure
Resins PC 10-CSR 3482, SR 205
2, Nadic methyl
anhydride3, camphorquinone4, and N,N-Dimethyl-p-toluidine4
3501-6
VendorMeasurementsGroup, Inc.
Sartormer Company, Inc. and Sigma-Aldrich Corporation
HexcelCorporation
1 Three different compositions were used for mixing the UV cure resins. The weight ratios are: (1)SR348:SR205 = 60:40, (2) SR348:SR205:Nadic methyl anhydride = 54:36:10, and (3)SR348:SR205:Nadic methyl anhydride = 48:32:20. Camphorquinone or N,N-Dimethyl-p-toluidine waslight sensitizing chemicals.2 SR348, ethoxylated bis-phenol A dimetharcrylate esters (Bis-MEPP), and SR205, triethylene glycoldimethacrylate from Sartormer Company, Inc.3 Nadic methyl anhydride from Sigma-Aldrich Corporation.4 Camphorquinone and N,N-Dimethyl-p-toluidine from Sigma-Aldrich Corporation.
separated from the master grating, the specimen undergoes deformation by the internal
stresses. From the specimen grating deformation we can measure the specimen
deformation from the gel point of the specimen to any time after grating separation
regardless the loading history of the specimens. This is the same principle as explored by
the CRM.
A section of an aluminum rod (60mm H x 25.4mm diameter) were put into a
section of an aluminum pipe (60mm H x 42.1mm OD, 37.5mm ID) and placed on a piece
of glass covered with release film to create a hollow cylindrical cavity for making neat
resin specimen molds. Two-part GE RTV627 silicone rubber compounds were mixed
and poured into the cavity to make hollow cylindrical specimen molds (Fig. 23).
The specimen grating replication takes place during the specimen curing process.
The specimen molds, with 25.4 mm inner diameter, and 6.4 mm (for room temperature
and high temperature epoxy) or 3.2 mm (for UV resins) height, were placed on top of the
54
Fig. 23 Apparatus to produce silicone rubber specimen molds: an aluminum pipe,an aluminum rod, a piece of release-film-covered glass, and a specimenmold
master gratings (as shown in Fig. 24). After degassing in a vacuum jar or a vacuum oven
(for high temperature cure epoxy), the resins were poured into the reservoir formed by the
master grating and silicone rubber molds (also preheated for high temperature epoxy).
The smooth and seamless contact between the silicone rubber molds and master gratings
prevented the resin from leaking. After curing, the neat resin specimens with diffraction
grating directly engraved on the surface can be separated easily from specimen molds
because of the weak adhesion of silicone rubber.
Master Grating Selection
For the first trial, GE RTV615 silicone rubber was used to make master gratings
because of its easy handling, transparency, and large working temperature range.
However, the experimental results showed that silicone rubber gratings are too compliant;
consequently, the master gratings were distorted by the shrinkage of the specimens during
the curing process. Figure 25 shows the distorted fringe patterns of a PC10-C specimen
55
Fig. 24 The original procedure for neat resin specimen replication
56
before it was separated from the master silicone rubber grating. The images were taken
from the back of the master grating. We can clearly see that the specimen area (inside the
circle) produced a lot of fringes due to distortion compared to the master grating area
(outside the circle) which is essentially a null field. Similar distortion occurred even
when the master gratings were made of a commonly used Envirotex Lite epoxy [48]
(from ETI Environmental Technology Inc.).
U field V field
Fig. 25 Distorted fringe patterns of a PC10-C specimen on the silicone rubbermaster grating
Much stiffer PC10-C epoxy was then chosen to make master gratings for room
temperature curing and UV curing resins, and 3501-6 epoxy from Hexcel Corporation
was chosen to make the high-temperature master gratings. Figure 26 shows the fringe
patterns before a PC10-C specimen was separated from the PC10-C master grating. The
pictures were also taken from the back of PC10-C master grating. The optical noise is
due to the multiple reflection through the glass substrate. In-plane rigid body rotation
was applied to the specimen in order to see the null field better. From the pictures with
rotation carrier fringes, the fringe density is very uniform both inside and outside the
57
circle. This indicates no distortion of the master gratings. The high-temperature master
gratings were specially made on ultra low expansion (ULE, Corning code 7971, tooling
grade) glass substrate to avoid error caused by thermal deformation of regular glass.
Corning ULE is more expensive than Astrosital used in the CRM, but has much better
optical properties allowing us to tune the interferometer through the substrate.
(a) U field (b) V field
(c) U field + rotation (d) V field + rotation
Fig. 26 Fringe patterns of a PC10-C specimen on the PC10-C master grating beforeseparation
58
Refining of Replication Procedure
Great caution was taken to avoid out-of-plane deformation of specimen gratings,
which can introduce considerable errors on the in-plane measurements. The keys to
reducing the out-of-plane deformation are having similar boundary conditions for both
the tops and bottoms of specimens and separating the specimens from the top and bottom
gratings/substrates at the same time.
After considerable trial and error, a standard procedure for replicating neat resin
specimen grating was developed to achieve the conditions stated above. The PC-10C
specimen gratings were made with two aluminized PC-10C master gratings on both the
tops and bottoms. For UV cure resins, the tops of the specimens were covered by a thin
piece of glass with a thin layer of GE RTV615 uncured silicone rubber compound in
between instead of an aluminized grating. This enables the UV light source to reach the
resin while allowing easy separation. For the high temperature cure specimen gratings
ULE glass substrates with two layers of aluminum were used on the top of specimens.
This also guaranteed the separation of the top covers from the epoxy specimens. Figure
27 illustrates the finalized procedure for specimen grating replication.
In-plane and Out-of-plane Deformation Measurement
Once the specimens were separated from their master gratings, they were
positioned on a grating alignment tool as described in the reference [48]. Two crossed-
line grating directions were determined and marked by razor blades. Then, the specimen
gratings were positioned in front of a moiré interferometer, which was tuned with the
master grating for the individual specimen. Both the U-field and V-field moiré fringe
59
patterns were captured using Polaroid 9cm × 12cm instant films Polapan 52 or 57. The
average of in-plane linear shrinkage can be obtained by analyzing the fringe patterns with
the equations (2.2) and (2.3).
(a) room temperature cure specimen
(b) UV cure specimen
(c) high temperature cure specimen
Fig. 27 Illustrations for making (a) room temperature cure, (b) UV cure, and (c)high temperature cure specimen gratings
60
For 3501-6 epoxy specimens, the strain values obtained at room temperature are in
fact the combination of their chemical shrinkage and thermal contraction. Therefore, in
order to extract the pure chemical shrinkage from the total strain values, the specimens
were placed into a pre-heated optical thermal chamber, which allowed moiré tests to be
performed at high/low temperature. The optical thermal chamber is the same chamber we
used for the CRM in Chapter 2. The temperature in the chamber was first set ten degrees
higher (T’) than the temperature that the fringe patterns were going to be taken (T) so that
after the specimens were put into the chamber, the designated temperature (T) can be
quickly reach. Fringe patterns were taken five minutes after the temperature reached the
setting (T) to avoid further corsslinking of the neat resin specimens. The specimens were
removed from the chamber as soon as the pictures were taken for the same reason. Two
elevated temperatures not near the glass transition temperature (Tg ≈ 195 °C in this case)
[56] of 3501-6 epoxy were chosen for photography. From these tests, the coefficient of
thermal expansion (CTE) of pure 3501-6 epoxy can be obtained. The curing shrinkage of
the epoxy was calculated by extrapolating the strain-temperature curves to the specimen
curing temperature.
For the verification of out-of-plane deformation, Fizeau interferometry was used.
Fizeau interferometry is a two-beam interferometry system. The biggest advantage of
Fizeau interferometer is the simplicity of the experimental setup and usage. An identical
He-Ne laser, optical fiber, laser coupler, and parabolic mirror as specified in Chapter 2
was used for the collimated light source. Figure 28 is the schematic diagram of a Fizeau
interferometer and Fig. 29 (a) and (b) are photographs of the experimental setup. An
optical-flat 10-degree wedge was used in the optical system to produce a reference
61
optically flat wave front. If the specimen has out-of-plane deformation, the wave front of
the reflected light from the specimen will become warped. When this warped wave front
interferes with the flat wave front reflected from one surface of the wedge, constructive
and destructive interference will occur. This forms a fringe pattern that represents the
topography of the specimen surface. The relative out-of-plane displacement can be
calculated by the equation (3.1) [48]
θλ
cos2
),(),(
⋅⋅= yxN
yxw (3.1)
where w(x,y) is the out-of-plane displacement, λ is the wavelength of the light source,
N(x,y) is the fringe order at a particular point (x,y), and θ is half of the angle between
incident light beam and reflected light beam.
collimated light source
2θ
specimen
optical-flatwedge
two waves interfere toproduce fringe patterns
reflection from wedge’sfront surface (will notgo into camera)
to lens andcamera
Fig. 28 Schematic diagram of a Fizeau interferometer
62
camera
laser
parabolic mirror
shutter controller
optical-flat wedge ( behind)specimen in
(a)
(b)
Fig. 29 Experimental setup for out-of-plane deformation measurement (Fizeauinterferometer). (a) Overview; (b) Close-up look of Fizeau interferometerand a neat resin specimen
63
Results and Discussion
At least four specimen gratings were produced for each kind of resin. Figure 30 is
a picture of a 3501-6 epoxy specimen grating and its ULE master grating after separation.
As seen in the picture, the vacuum-deposited aluminum films on the master grating and
the cover glass are now transferred to the specimen. This makes separation easier and
also improves the diffraction efficiency of the specimens. Among three different
thermoset resins, the high temperature cure, 3501-6 epoxy had the largest chemical curing
shrinkage upon separation (Table 3). One of the reasons is due to the high cross-link
density between polymer chains of high temperature cure epoxy.
Fig. 30 A high temperature specimen grating after being separated from its mastergrating
64
COV (%)
9.26
10.60
Average
-341
-644
Side B
-306
-645
Specimen #4
Side A
-295
-568
Side B
-339
-653
Specimen #3
Side A
-333
-680
Side B
-355
-750
Specimen #2
Side A
-381
-711
Side B
-366
-559
Specimen #1
Side A
-379
-593
PC10-C epoxy
upon separation
fully relaxed
COV (%)
4.27
12.87
5.70
7.03
0.98
3.95
Average
-1178
-2670
-1313
-3169
-1324
-3071
Specimen #4
-1244
-3031
-1222
-3057
-1319
-3200
Specimen #3
-1161
-2203
-1353
-3003
-1312
-2999
Specimen #2
-1124
-2730
-1391
-3121
-1324
-3144
Specimen #1
-1183
-2719
-1284
-3495
-1340
-2940
upon separation
fully relaxed
upon separation
fully relaxed
upon separation
fully relaxed
UV resins
48:32:20
54:36:10
60:40
COV (%)
11.06
13.21
Average
-1433
-1053
Specimen #4
-1239
-856
Specimen #3
-1536
-1114
Specimen #2
-1372
-1177
Specimen #1
-1586
-1065
3501-6 epoxy
upon separation
fully relaxed
Table 3 Chemical curing shrinkage for five different resins
(Unit is in µε)
65
When we separated the specimens from the grating molds, there was an
immediate contraction. As time past (2~3 months), the rate of contraction reduced to
near zero. Then we repeated the measurement again, but the results somehow surprised
us. Room temperature and UV cured specimens exhibited significant creeping as
expected. On the other hand, for 3501-6 epoxy specimens, the strain values remain about
the same or even decrease (specimens expanded) (also see Fig. 31). The reason why the
high temperature cure epoxy acts differently from the other kinds of resin is not fully
understood at this point. One possible reason is the hygro effect due to humidity, which
has the opposite effect of chemical shrinkage on specimens. Although we store all the
specimens inside a dry box, the relative humidity is at least 22% at all times. When we
open the box to access the specimens or when the specimens were outside the dry box for
measurement, the humidity is even higher. Room temperature cure and UV cure resins
should exhibit a similar effect, but were probably compensated by the large amount of
creep deformation. After the in-plane strains were measured, a Fizeau interferometer
was used to check the out-of-plane deformation. The fringe patterns showed that the out-
of-plane deformation was on the order of 20 wavelengths over the entire specimen
surface (25.4mm diameter), but only 1 to 3 wavelengths over the gage length (9.2mm
square). Therefore, the effect of the out-of-plane deformation on the in-plane curing
shrinkage is considered negligible. Figure 32 through 34 shows some typical fringe
patterns for each different kind of resin.
66
Fig. 31 Typical shrinkage vs. time curves for each kind of specimens
67
front, U field back, U field
front, V field back, V field
front, W field back, W field
Fig. 32 In-plane chemical shrinkage and out-of-plane deformation fringe patternsfor a PC10-C epoxy specimen.
68
U field V field
W field
Fig. 33 In-plane chemical shrinkage and out-of-plane deformation fringe patternsfor a UV cure resins (SR 448:SR 205=60:40).
69
U field V field
W field
(a) at room temperature
Fig. 34 In-plane chemical shrinkage and out-of-plane deformation fringe patterns of3501-6 epoxy at room temperature and 140 °C.
70
U field V field
(b) at 140 °C
Fig. 34-continued
In order to validate the methodology for the high temperature cure epoxy, the
following analysis was performed. From previous work [51], the strain-temperature
curves (in the transverse fiber direction) of 16-ply AS4/3501-6 unidirectional laminates
were measured as shown in Fig. 35. The strain-temperature curves of our current 3501-6
epoxy specimens are shown in Fig. 36. According to Agarwal and Broutman [3], using
the rule of mixture, the transverse CTE of a unidirectional laminate (αT), with fiber
volume fractions greater than 25%, can be represented by
mmmffT VV ⋅⋅++⋅= αναα )1( (3.2)
where αf and αm are the CTE of fiber and matrix material, Vf, and Vm are the fiber and
matrix volume fractions, and νm is the Poisson ratio of the matrix material. Using the
71
values αf ≈ 0, Vf = 0.65, and the νm = 0.36 from the manufacturer, and substituting average
αm = 52.4×10-6 m/m/°C from Fig. 36, we can obtain
6109.24 −×=Tα m/m/°C.
This value has excellent agreement with the average αT value, 25.0 × 10-6 m/m/°C, which
is calculated from Fig. 35.
Conclusion
A simple and effective technique was developed to measure the post-gel chemical
curing shrinkage of polymers, which plays an important role in characterizing the residual
stresses of polymer matrix composite materials. The curing shrinkage can be accurately
measured without knowing the exact gel point during the curing process. However, care
must be taken to avoid out-of-plane deformation of the specimen which will cause errors
in the in-plane shrinkage measurement. It has also been determined that depending on the
resin system, stress relaxation would dramatically influence the results of measurement
both at the room temperature and elevated temperature.
72
Fig. 35 Residual strain vs. temperature of AS4/3501-6 [0]16 laminate in transversedirection.
Fig. 36 Apparent strain of 3501-6 neat resin specimens vs. temperature
73
CHAPTER 4VISCOELASTIC CONSTITUTIVE MODELING FOR POLYMER MATRIX
LAMINATED COMPOSITES
Introduction
In the previous work [26, 51], we have been able to demonstrate the CRM to be
an effective tool for measuring residual strain caused by the manufacturing process of
polymer matrix composite laminates. In order to calculate the residual stresses from the
residual strains, a constitutive model of the laminate specimens is required. The authors
have used a simple model—linear classical lamination theory, to deduce the residual
stresses. For an orthotropic composite laminate plate, the stress-strain relation can be
represented and further simplified by Hooke’s law and classical lamination theory.
Equation 4.1or 4.2 describes this relationship for a special orthotropic and transverse
isotropic plate subjected to plane stress condition.
00
0
0
12
2
1
66
2212
1211
12
2
1
⋅
=
εεε
σσσ
C
CC
CC
(4.1)
⋅
=
12
2
1
66
2212
1211
12
2
1
00
0
0
σσσ
εεε
S
SS
SS
(4.2)
here Cij are the elements of stiffness matrix, Sij are the elements of compliance matrix.
74
Also from the study conducted in Chapter 3, it is desirable to employ a more
sophisticated model such as a linear viscoelastic constitutive model to describe the stress
relaxation behavior of polymer matrix composites (PMC) when time dependency is
involved. It is our goal to obtain the constitutive equation that can take into the account
of viscoelastic effects and replace the material constants in [C] matrix or [S] matrix of
equations (4.1) and (4.2) with time dependent functions.
Standard Linear Solid Model and Correspondence Principle for Linear ViscoelasticMaterials
To describe the behavior of a viscoelastic body, linear spring and linear dashpot
elements are usually used to construct the constitutive models [33, 57]. The Kelvin-Voigt
model (single spring-dashpot parallel arrangement) or Maxwell model (single spring-
dashpot series arrangement) alone is easy to formulate but is usually insufficient to
represent a material system. Standard linear solid model with the arrangement shown
below (Fig. 37) was the model we first chose for the AS4/3501-6 composite system.
k0
k1
Fig. 37 Spring-dashpot arrangement for standard linear solid model
The governing equations can be obtained from solving the following ordinary
differential equation:
01111
100 and , εεεεεµε −=⋅+⋅=⋅ kdt
dk (4.3)
75
where ε is the total strain of the assembly, k0, k1 are the spring constants of spring
elements, and µ1 is the viscosity of the dashpot element. Hence, when the whole
assembly is subjected to an instant constant load σ0 at isothermal condition, the solution
for the above equation becomes
)1()(1
0
0
0 τσσεt
ekk
t−−⋅+= (4.4)
here, τ = 1 / µ1 is the relaxation time. By finding the constants k0, k1 and µ1 in the
equation (4.4), the elements in the time-dependent creep compliance matrix can be
defined. In the current study, only orthotropic, transverse isotropic composite laminates
were considered. For this kind of material, four independent material constant are
required for constructing the creep compliance matrix, which are S11(t), S12(t), S22(t), and
S66(t). Uni-axial tensile tests were used for generating necessary data for analytical
modeling. Three different unidirectional tensile specimen configurations were chosen for
fabrication—[0]16, [90]16, and [10]16.
Time-Temperature Superposition Principle
When linear viscoelastic materials are subjected to a creep test or stress relaxation
test, a relationship between increasing ambient temperature and acceleration in time scale
exists. For thermorheologically simple viscoelastic materials, this relationship is even
more simplified. The long-term relaxation (or creep) behavior at a reference temperatures
can be constructed by shifting the short-term relaxation/creep curves at different
temperature parallel to the time axis (Fig. 38). This relationship to describe the
“speeding-up” behavior of the particular materials is called time-temperature
76
superposition (TTS) principle [50]. Equation (4.5a) or (4.5b) formulates this relationship
with a temperature-dependent shift factor aT.
)()( riTi TaT λλ ⋅= (4.5a)
)()( riTi TaT ρρ ⋅= (4.5b)
where λi = µi / ki is the relaxation time for ith Maxwell element, and ρi = µi / ki is the
retardation time for ith Kelvin-Voigt element in a multi-element model. In our standard
solid linear model, the subscript i vanishes, and only one λ and one ρ remain.
log(time) log(time)
log(
cree
p co
mpl
ianc
e)
log(
cree
p co
mpl
ianc
e)
T1
T2
T3
T4
aT2
aT3
aT4
Fig. 38 Construction of the long-term “master curve” at reference temperature fromshort-term testing data at different temperature.
Thermal Chamber Manufacture and Test Fixture Design
In order to utilize the time-temperature superposition principle, high temperature
tensile tests were designed to obtain the necessary data. The high temperature
environment was achieved by a newly designed and built thermal chamber. The regular
grips for tensile tests have large thermal mass; therefore, the thermal chamber was
designed to have just enough space to accommodate the active and dummy specimens,
77
necessary strain gage wiring, thermal couples, extension bars from grips, and two pairs of
small clamps. Two pinholes were drilled on each piece of the clamps. One is for the pin
connecting to the extension bar, and one is for the screw that goes through the pinholes
on the specimens and to fasten the clamps (Fig. 39).
Fig. 39 Thermal chamber and test fixture for uniaxial tensile tests
The chamber has folded aluminum sheet interior, 3/4” (19.1 mm) thick inner wall
made of high temperature calcium silicate insulation board (McMaster-Carr Supply
Company, max. temp. 927 °C, thermal conductivity 0.8 Btu @ 427 °C), and half inch
(12.7 mm) thick outer wall made of high strength calcium silicate board (McMaster-Carr
Supply Company, max. temp. 760 °C, thermal conductivity 1.16 Btu @ 427 °C). The
final dimensions of this chamber are 5.25” x 5.25” x 16” (133.4 mm x 133.4 mm x 406.4
78
mm) outside, and 2.5” x 2.5” x 13.5” (63.5 mm x 63.5 mm x 342.9 mm) inside. All the
pieces were bonded together with silicone sealant and connected to the EC12 oven used
in Chapter 2 by similar air ducts and the same circulation fan. This chamber was tested to
a have sustained working temperature of up to 210 °C.
Specimen Preparation and Testing Plan
Specimens were again prepared using the AS4/3501-6 unidirectional prepreg from
Hexcel Corporation. The aerial weight of the prepreg was 145, and the fiber volume
fraction was 65%. The prepreg was cut and stacked to sixteen plies. High temperature
strain gages (SK-06-250BA-500, Measurements Group, Inc.) were positioned at the
desirable location with Teflon sheet templates onto both sides of the prepreg panels (Fig.
40).
Fig. 40 Composite prepreg panel with Teflon release film template for applyingstrain gages. The template for making 10-degree specimens is shown in thispicture.
79
Gage and Accessory Selection
High temperature, high electrical resistance gages were chosen based on several
reasons. First, for the long-term experiments run at the elevated temperature, high
temperature resistance is essential. The SK series strain gages have a working
temperature range from 230 °C to -269 °C (450 °F to -452 °F). Matching solder (450-
20S-25) and wires (330-FTE) both from Measurement Group, Inc. were chosen to
accommodate the high temperature environment for our tests. Second, 500 ohm
resistance gages were selected to reduce the heat accumulation, since graphite/epoxy
composite does not have good thermal conductivity. Finally, for the high temperature
testing environment, high temperature adhesive was necessary to ensure proper load
transfer from the test coupons to the strain gages. High temperature adhesive usually
requires high temperature cure. However, in order to reduce the alternation of specimen
properties caused from the thermal history of strain gage adhesive curing, the best way to
attach the strain gages was to apply strain gages during the composite specimen
manufacturing process. Because the gages were attached to the composite laminates
during the curing process, there was no need to go through another thermal cycle.
For 0-degree and 90-degree specimens, a total of four gages were applied on the
surfaces (0° and 90° on both front and back surfaces) to obtain the normal strain and the
Poisson ratio. For 10-degree specimens, a total of six gages were applied on the surfaces
(0°, 90°, and 45° on both front and back surfaces) in order to obtain the shear strain and
shear modulus.
80
Specimen Preparation
The composite laminate lay-up was the same for the three different specimen
configurations (16 plies, unidirectional), but the specimens were cut along three different
angles to produce the three specimen types. After the prepreg lay-up and strain gage
application was completed, the same vacuum bag lay-up and the same autoclave curing
profile as stated in Chapter 2 for the CRM was used to cure the composite laminates.
After the autoclave curing process, the individual strain gages were inspected and tested
to ensure proper working condition. The composite panels were then cut using a low
speed diamond saw (Isomet, Bulter, Inc.) to prevent significant edge effect and
subsequent errors due to machining according to ASTM standard 3039/3039M [58].
Water was added as the coolant for cutting. Composite panels were mounted on a
platform which could slide down along a track parallel to the cutting blade. The
specimens were trimmed into the final dimensions of 1” x 10” (25.4 mm x 254 mm) for
tensile testing.
To prevent early failure of the unidirectional tensile specimens, end tabs were
bonded to the ends of the specimens. The end tabs were made of same composite
material system (AS4/3501-6) with a [02/902]2s cross-ply configuration with a dimension
of 1” x 1.5” (25.4 mm x 38.1 mm) without beveling. Shell Epon 828 resin and curing
agent 9552 were used as the bonding adhesive. In the trial tests the epoxy was able to
maintain adequate strength at the highest temperature setting, 195 °C, without end tab
failure.
After the epoxy for the end tabs was fully cured, holes for pins were drilled at the
center of end tabs through the specimens. A 3/8” (9.5 mm) outer diameter diamond core
81
drill bit (part no. 2868A25, McMaster-Carr Supply Company), and the matching adapter
were used to drill the pinholes. A slow drilling rate was used to prevent over heating of
specimens and tools. This also reduced the chance of delamination between end tabs and
specimens during drilling.
After drilling the holes, specimens were ready for strain gage wiring. Another
identical specimen was used as a dummy specimen in a half Wheatstone bridge
configuration since there is no self-temperature-compensation gage available for the
graphite/epoxy composite material system. Although it was more difficult than room
temperature application, high temperature solder and wire was required for our tests.
Strain gages were connected to a strain gage conditioner (2100 System, Measurements
Group, Inc.), where the Wheatstone bridge circuit was completed, excitation voltage was
supplied, and the output signals were amplified. The excitation voltage was 3.5 volts to
prevent local overheating. The gain setting varied depending on the gage orientation with
respect to the specimen fiber direction and the maximum strain values. After being
amplified by the strain gage conditioner, the output signals were input into the controller
of the testing machine for data acquisition.
Experimental Procedure and Testing Plan
The specimens and testing fixtures were assembled, and the thermal chamber and
EC12 oven were installed on a MTS servo hydraulic testing machine. The test machine
has both axial and torsional loading capability, and was controlled by a digital TestStar II
controller. Testing profiles and data acquisition were programmed, executed, and
recorded on a Compaq Deskpro 6000 computer. Figure 41(a) through (c) show the
experiment setup from several different viewpoints.
82
(a) MTS testing machine and EC12 oven
(b) TestStar II digital controller, strain gage conditioner, and the work station
Fig. 41 Experiment setup from several viewpoints
83
(c) thermal chamber, test fixture assembly, and strain gage conditioner
Fig. 41--continued
After several trial tests, load control creep tests were chosen to obtain time
dependent material properties. The thermal chamber, active and dummy specimens were
heated from room temperature to the desired temperature setting at the rate of 2.8 °C per
minute (5 °F /min.). The temperature inside the chamber was controlled by the controller
of EC12 oven with an extended thermal couple probe. Another thermal couple was
extended from the chamber to a thermocouple-to-analog converter (TAC80B-K, Omega
Engineering, Inc.), and connected to TestStar II testing machine controller as an input
signal for monitoring and recording the temperature during tests. These two
thermocouples (both CO1-K foil-terminal, Omega Engineering, Inc.) were stacked
together and directly attached to the surface of specimens with high temperature tape.
Test loading profiles are shown in Fig. 42. The loading rate was 2 lbf per second for all
three kinds of specimens. Once the load reached the final load level, the specimens were
held at that load value for 60 minutes, and then unloaded at the same loading rate. For 0-
84
degree specimens, final load was 700 lbf. For 90-degree specimens, 400 lbf. (about 80%
of the strength) was used for room temperature, 40 °C, 60 °C, and 80 °C settings, 250 lbf.
(about 50% of the strength) for 100 °C, 120 °C, 140 °C, 160 °C, and 177 °C settings, 50
lbf. (about 10% of the strength) for 195 °C setting in order to obtain enough strain
response. For 10-degree specimens, 150 lbf. final load was used. After unloading, the
specimens were held at 0 lbf. load for 30 minutes (0-degree) or 60 minutes (10- and 90-
degree) for recovery. During loading and unloading, the data sampling rate was 1 Hz,
and was 0.2 Hz for holding and recovery. At least four tests were continuously performed
in sequence for each temperature settings, and one specimen was repeatedly used for the
same temperature setting.
load
load
l oad
ing
l oad
ing
un loa di ng
u nlo ad in g
holding
holding
recovering recovering
time
(a) (b)time
60 min.
60 min. 60 min.30 min.
Fig. 42 Testing profiles for (a) 0-degree specimens, (b) 10-degree and 90-degreespecimens
Results and Data Analysis
General Behaviors of Materials
This section will describe the general behavior of our specimens during the creep
tests. For convenience, we will describe the testing procedure as loading, holding (creep),
unloading, and recovery phases.
85
Loading phase
For 0-degree specimens, the loading stress-strain curves look perfectly linear
(elastic), since this is the fiber-dominated direction. Also the curves did not show
noticeable temperature dependency. At the different temperatures, when the specimens
were subject to the same amount of load, they all had about the same value of strain (see
Fig. 43).
From Fig. 44 we can see that the stress-strain curves for the 90-degree specimens
deviated from the room temperature straight line as the temperature increased. There is
an obvious change in the elastic moduli in the transverse fiber direction (axial loading
direction) (from 1.54 Msi at room temperature to 0.93 Msi at 195 °C), since this is matrix
dominated direction.
For the 10-degree specimens, the axial and transverse strains were obtained
directly from readings of the strain gages. The shear strain values were calculated real-
time by the TestStar II controller during the tests with the following equation derived
from the strain gage rosette relationship [59].
°°° ⋅−⋅+⋅−= 9045012 598.0879.1282.1 εεεγ (4.6)
here γ12 is the shear strain in the 10-degree plane (fiber direction), ε0° is the read-out of
the strain gage along the loading direction, ε90°, ε45° are the read-outs of the strain gages in
the 90° and 45° loading direction respectively. Considering the 10-degree specimens
(Fig. 45), the specimens were still within the elastic range, but no matter axial or
transverse normal strains, or shear strain in 10-degree direction, they all exhibited
temperature dependence.
86
Holding (creep) phase
After the loading stages, the specimens were held at constant force for 60 minutes.
The longitudinal strain values for the 0-degree specimens remained essentially the same,
since these are fiber-dominated and we can assume that graphite fibers are elastic within
the testing temperature range (see Fig. 46).
For the 90-degree specimens (Fig. 47), even at room temperature, the longitudinal
strain had a 1% of increase (from 3370 µε to 3404 µε), and at 177 °C the increase was
24.5% (from 2823 µε to 3516 µε). At 100 °C, one specimen failed during the hold at the
final load of 400 lbf. Also another failed during loading at 195 °C when the load
exceeded 230 lbf.
Similar behaviors were observed for 10-degree specimens. There is a 2.6%
increase in shear strain value in 10-degree specimen at room temperature over the holding
period, and 7.0% at 177 °C (Fig. 48).
87
Fig. 43 Stress-strain curves for 0-degree specimens at different temperature settingsduring loading phase. From Fig. 43 through 45, all strain values are theaverage of front and back strain gage readings.
88
Fig. 44 Stress-strain curves for 90-degree specimens at different temperaturesettings during loading phase.
89
Fig. 45 Stress-strain curves for 10-degree specimens at different temperaturesettings during loading phase.
90
Fig. 45-continued
Fig. 46 Strain-time curves for 0-degree specimens at different temperature settingsduring holding phase.
91
Fig. 47 Strain-time curves for 90-degree specimens at different temperature settingsduring holding phase.
Fig. 48 Shear strain-time curves for 10-degree specimens at different temperaturesettings during holding phase.
92
Unloading and recovery phases
The unloading stress-strain curves remain linear and return to zero (the origin) for
the 0-degree specimens. For the 90-degree specimens, there was a considerable amount
of strain that remained when the stress returned to zero at every temperature setting.
Some of the strain gradually returned to zero during the one-hour recovery phase,
depending on the load level and temperature. The 10-degree specimens showed a similar
behavior. At the higher temperature, there was a large amount of shear strain remained
when the specimens were unloaded to zero. During the recovery phase, the shear strain
values increased (negative during the hold sequence) and even passed zero to positive.
Since shear strains were calculated from longitudinal and transverse strains, the drift in
the strain gage readings were enlarged due to the formula which we used to calculated
shear strains.
Curves Fitting for the Creep Tests
After tests, the cross-sectional area of each specimen was measured again to
ensure the change in cross-sectional area would not produce a noticeable increasing of
stress. The raw data for the holding phase of each test was extracted and imported into
MathSoft MATHCAD 8 program using its “READPRN” function, where all the curve
fitting and analysis were performed. It was necessary to make correction for the strain
values of 90-degree specimens due to the transverse sensitivity of strain gages and highly
biased strain field. The following equations were employed for making this correction
[60].
93
)K(K1
K1
)K(K1
K1
xxtyy2t
t0yy
yytxx2t
t0xx
εενε
εενε
′⋅−′⋅−
⋅−=
′⋅−′⋅−
⋅−=(4.7)
Here, ε’xx and ε’yy are the apparent strains from strain gage reading in the longitudinal and
transverse directions respectively, Kt is the transverse-sensitive factor of gages, ν0 =
0.285 is the Poisson’s ratio of the material used for gage calibration, and εxx and εyy are
the true strains in two directions after correction. It was found that the errors associated
with the transverse sensitivity of the strain gages could be as large as 35% of the final
values of creep compliance S12(t).
As stated in the previous section, a three-element standard solid linear model was
selected initially to describe the material behavior in the matrix dominated orientations.
In the MATHCAD program, the function “genfit” allowed us to fit an arbitrary function
to the data. By solving the governing equation of the standard solid linear model, the
function for fitting the data was
[ ])tvexp(1v)t( 010 ⋅−−⋅+= εε (4.8)
here ε0 = ε(0) is the strain value when the holding stage started, v0 and v1 are the
constants which curve fitting was used to solve for. With the non-linear fitting function,
v0 and v1 are very sensitive to the initial guess. The values of initial guess were obtained
from the experimental curves as illustrated in Fig. 49. After several iterations, when the
difference between the final values and the initial values was less than 1%, the iteration
process was terminated. The MATHCAD file for the curve fitting is attached in
94
Appendix A. The curve fitting results and experimental data are shown in Fig. 50 for 90-
degree specimens and in Fig. 51 for 10-degree specimens.
time
stra
in
v1
v0
1
Fig. 49 Estimating the values of initial guess for curve fitting.
(a)
Fig. 50 Curve fitting and experimental data for the longitudinal strain of 90-degreespecimens during the holding phase at different temperatures. (a) 22 to 80°C; (b) 100 to 177 °C; (c) 195 °C.
95
(b)
(c)Fig. 50--continued
96
(a)
(b)
Fig. 51 Curve fitting and experimental data for the shear strain of 10-degreespecimens during the holding phase at different temperatures. (a) 22 to 100°C; (b) 120 to 195 °C.
As seen in Fig. 50 and Fig. 51, the fitting curves do not match the experimental
data very well especially when the testing temperature was high. In order to improve the
97
fitting, a dashpot element was added to the model as the fourth element which is in series
with the other three elements. The MATHCAD file for curve fitting with the fourth
element added is attached in Appendix B. By solving the governing equation, the fitting
function used for MATHCAD “genfit” function was
[ ] tv)tvexp(1v)t( 2010 ⋅+⋅−−⋅+= εε (4.9)
where ε0, v0 and v1 are defined the same as in equation (4.8), and v2 is proportional to the
reciprocal of the viscosity, µ2, of Newtonian fluid in the additional dashpot. The initial
guess for v2 values is also obtained from the plots of experimental data by estimating the
slope of the final portion of the strain versus time curves during the hold sequence (see
Fig. 52). A similar iteration process was also applied to obtain the final values of v0, v1,
and v2.
time
stra
in
v1
∆ε∆t
v0
1
v =2 ∆t∆ε
k0
k1
Fig. 52 Estimating the values of initial guess for curve fitting with the fourthelement.
98
From Fig. 53 and Fig. 54, we can clearly see that the curve fit with the four-
element model fits the experimental data better compared to the fit with the three-element
model. Additionally, the fit is good, even when the test temperatures were high.
For the 0-degree specimens, the strain response in longitudinal direction (fiber
direction) was considered as a linear function (first order polynomial) of time. In
MATHCAD, functions “regress” and “interp” are used to find this polynomial.
(a)
Fig. 53 Curve fitting with four element model and experimental data for thelongitudinal strains of 90-degree specimens during the holding phase atdifferent temperatures. (a) 22 to 80 °C; (b) 100 to 177 °C; (c) 195 °C.
99
(b)
(c)Fig. 53--continued
100
(a)
(b)
Fig. 54 Curve fitting with four element model and experimental data for the shearstrains of 10-degree specimens during the holding phase at differenttemperatures. (a) 22 to 100 °C; (b) 120 to 195 °C.
101
A similar curve fitting procedure was applied for determining the Poisson’s ratio
versus time functions, ν21(t) (from 90-degree specimens) and ν12(t) (from 0-degree
specimens) at different temperatures. After all of the strain-versus-time functions (ε (t))
were obtained, the creep compliance matrix elements S22(t), S66(t), and S11(t) were
calculated by dividing ε (t) by the applied stress, and S12(t), and S21(t) were calculated
from the equations
)()()()()()( 211112222112 tStSttSttS =⋅−=⋅−= νν (4.9)
The statement, S12(t) = S21(t), comes from our orthotropic and transverse isotropic
material property assumptions. Figure 55 trough 57 show creep compliance S22(t), S12(t),
and S66(t) on a logarithmic scale. The calculation for the creep compliance matrix
element S22(t), and later the construction of the master curves for these elements are also
included in the files in Appendix B (90-degree specimens). For 10-degree specimens and
0-degree specimens, the data analysis were performed in the similar manner.
Using a thermorheologically simply assumption, we now shift the creep
compliance curves, S22(t) and S66(t), at the different temperatures along the horizontal
axis (time axis) until the initial values of the higher temperature curves meet the values of
the previous temperature setting curves. By connecting all curves at the different
temperatures into a single curve, we construct the master curve of the creep compliance
elements over a long period of time at a specific reference temperature (room temperature
(22 °C) in our experiments and calculation). Figures 58 and 59 show the master curves
for creep compliance matrix elements, S22(t) and S66(t). In both cases, noticeable kinks
occur when the curves at two different temperatures intersect.
102
log (psi-1) Creep Compliance S12(t) vs. Time (curve fit)
log(Time) (log(sec))
Fig. 55 log (S22(t)) versus log (t) for AS4/3501-6 unidirectional composite laminateat different temperatures.
log (psi-1) Creep Compliance S12(t) vs. Time (curve fit)
log(Time) (log(sec))
Fig. 56 log (S12(t)) versus log (t) for AS4/3501-6 unidirectional composite laminateat different temperatures.
103
log (psi-1) Creep Compliance S66(t) vs. Time (curve fit)
log(Time) (log(sec))
Fig. 57 log (S66(t)) versus log (t) for AS4/3501-6 unidirectional composite laminateat different temperatures.
This indicates an inadequate assumption of the thermorheologically simple material
properties. Additionally, when time value becomes very large, the v2 term in the fitting
function also becomes very large, which makes the fitting curves unreasonable.
To solve this problem, physical aging of 3501-6 epoxy is taken into account, since
most of our tests were done at the temperature much below the Tg of 3501-6 epoxy
(except at 195 °C). Physical aging is the process of reduction in free volume of
amorphous polymers after they are quenched to below their Tg and slowly return to the
thermal equilibrium state [50, 57]. Some work has been performed [61-65] to investigate
the influence of this phenomenon on the viscoelastic behavior of polymers. Since the
mobility of polymer chains decrease due to less and less free space available, the elastic
moduli of polymers tend to increase in general. As a result, the creep compliance versus
104
log (psi-1) Creep Compliance S22(t) Master Curve vs. Time
log(Time) (log(sec))
Fig. 58 The tensile creep compliance S22(t) master curve for 3501-6/AS4unidirectional laminate specimens in transverse fiber direction. Referencetemperature is at 22 °C.
105
log (psi-1) Creep Compliance S66(t) Master Curve vs. Time
log(Time) (log(sec))
Fig. 59 The shear creep compliance S66(t) master curve for 3501-6/AS4unidirectional laminate specimens. Reference temperature is also at 22 °C.
time curves in the matrix-dominated direction need to be shifted along the vertical axis.
By doing so, the new master curves are shown in Fig. 60 and Fig. 61. They look much
smoother and more reasonable with the combination of horizontal and vertical shift. The
shift amount for each temperature setting is listed in Table 4.
To compare with the curve fitting results, the creep compliance matrix elements
were also calculated point by point directly from the experimental data. From Fig. 62, the
curve fitting functions of the creep compliance matrix elements agreed with the values of
creep compliance directly calculated from experimental data fairly well.
Finally, to validate the creep compliance master curves which were constructed
from momentary creep tests at different temperatures, another creep test was performed
106
log (psi-1) Creep Compliance S22(t) Master Curve vs. Time
log(Time) (log(sec))
Fig. 60 The tensile creep compliance S22(t) master curve for 3501-6/AS4unidirectional laminate specimens in transverse fiber direction withconsideration of physical aging. Reference temperature is now at 100 °C.
107
log (psi-1) Creep Compliance S66(t) Master Curve vs. Time
log(Time) (log(sec))
Fig. 61 The shear creep compliance S66(t) master curve for 3501-6/AS4unidirectional laminate specimens with consideration of physical aging.Reference temperature is also at 100 °C.
Table 4 Time-Temperature Superposition shift factors (in log units), and referencetemperature (Tr) is 100 °C.
Creep compliance T-Tr (°C) Horizontal shift Vertical shift
S22 0 0 0
20 1.22 -0.0074
40 2.53 -0.0185
60 3.55 -0.0417
77 4.23 -0.0836
S66 0 0 0
20 0.86 0.0042
40 1.57 0.0079
60 2.25 0.0250
77 3.12 0.0623
108
specimen. The testing temperature was chosen to be a common application temperature,
100 °C, for this composite system. The load holding duration was increased from one
hour to seven days and was held at 250 lbf. In the first 20 minutes, data points were taken
every 5 seconds. From the 21st minute to the first five hours, sampling time was changed
to once every 60 seconds. Finally, from the sixth hour to the end of test, data points were
taken every 20 minutes.
As shown in Fig. 62, in the first three days of the long-term test, the result has
moderate agreement with the master curve constructed from the short-term creep tests. A
vertical shift in the log(creep compliance) axis was necessary to compensate the different
age of specimens, which were cut from the same piece of composite plate but tested at
different times. From around the fourth day of the test, the strain values separated.
Therefore, the data after 259,200 seconds (three days) were truncated.
Model Formation
Based on the creep compliance master curves constructed for unidirectional
laminated composites, and corresponding principle of linear viscoelastic materials, we
now will formulate the stress-strain relationship to describe the viscoelastic behaviors of
the composite laminate. First the equation (4.2) becomes
⋅
=
)t(
)t(
)t(
)t(00
0)t()t(
0)t()t(
)t(
)t(
)t(
12
2
1
66
2212
1211
12
2
1
σσσ
εεε
S
SS
SS
(4.10)
109
Fig. 62 Comparison between directly calculated creep compliance S22(t) and curve-fitting S22(t). The curve represents long-termcreep test is also shown to compare with the master curve obtained from momentary creep tests (short-term creep).
110
All terms in the equation are replaced by time-dependent functions except for S11.
From the experimental data, we can assume that S11(t) becomes a time-independent
constant. Second, these functions, in the material 1-2 coordinate system, can be
transformed into an arbitrary x-y coordinate system (usually the loading directions) by the
transformation matrix [T] according to the following equations
[ ] [ ]
⋅⋅
⋅=
−
)t(
)t(
)t(
)t(00
0)t()t(
0)t()t(
)t(
)t(
)t(
xy
y
x
66
2212
12111
xy
y
x
σσσ
εεε
T
S
SS
SS
T (4.11a)
or
⋅
=
)t(
)t(
)t(
)t()t()t(
)t()t()t(
)t()t()t(
)t(
)t(
)t(
xy
y
x
662616
262212
161211
xy
y
x
σσσ
εεε
SSS
SSS
SSS
(4.11b)
where
[ ]
−−−=
θθθθθθθθθθ
θθθθ
22
22
22
sincoscossincossin
cossin2cossin
cossin2sincos
T
By substituting the residual stress values we calculated from the Cure Referencing
Method, the strain functions of time can be calculated. In the next chapter, the relaxation
of residual stresses will be examined, and Cure Referencing Method was used again to
validate the time-dependent viscoelastic model for AS4/3501-6 composite laminates.
111
CHAPTER 5DISCUSSION AND CONCLUSION
Application of Linear Viscoelastic Model for Residual Stress Relaxation
Overview
In the last few chapters, research has been conducted to measure and characterize
the process-induced residual stresses and strains of the laminated polymer matrix
composite system—AS4/3501-6. A method called the “Cure Referencing Method”
(CRM) was developed, tested, and validated [66] to measure the residual strain for the
composites. During the analysis for calculating residual stresses from residual strains,
linear classical laminate theory was used.
Another method was developed to characterize the chemical curing shrinkage of
the polymer matrix. The residual strain due to the manufacture process can then be
separated into two parts: residual strain due to the CTE mismatch and residual strain due
to matrix chemical curing shrinkage.
A linear viscoelastic model was also constructed to describe the long-term
behaviors of the specific composite system based on the short-term (momentary) creep
tests at several different temperatures and time-temperature superposition principle.
Physical aging of polymer matrix was also taken into account to obtain smoother master
curves for the elements of the creep compliance matrix.
112
In this section, additional experiments were performed to verify our empirical
creep compliance master curves and the viscoelastic stress-strain relation. These tests
were done by using moiré interferometry. The composite specimens were from the same
batch of specimens which were been used in Chapter Two. After the manufacturing
process, these specimens were used for the CRM to measure the residual stresses. Since
moiré interferometry has the capability for long-term evaluation for in-plane deformation
of an object, we have stored the specimen gratings and their master gratings for about two
years (702 days = 107.78 seconds) in order to perform the necessary tests.
Specimens were kept in a desiccant box and room temperature ambient
environment. The fluctuation of room temperature is neglected in this study. The effect of
humidity is not believed to be negligible, but in the current study we have tried to
eliminate the influence as much as possible with laboratory humidity control and
desiccant substance. Further investigation can be found elsewhere [67]. The Astrocital
master gratings were kept in a cabinet to prevent damage from dust. The substrate of the
master gratings were considered dimensionally stable (no deformation) during this time
interval.
Experiment Procedure and Results
The optical arrangement was the same as describe in Chapter Two (please refer to
Fig. 13 in Chapter Two). At room temperature, the moiré interferometer was first tuned
with the intermediate epoxy master gratings for each individual specimen. After the null
fields were attained, master gratings were replaced by corresponding composite
specimens. After adjusting the position of the composite specimens, fringe patterns were
photographed using Polaroid instant sheet film. Figure 63 through 66 are the photographs
113
of fringe patterns for four composite specimens with different lay-up configurations. All
pictures were taken at room temperature.
Fig. 63 Fringe patterns for [016]T unidirectional composite laminate panel. Time lagfrom manufacture date tlag= 702 days.
114
0° fiber direction
(a) -field, 22 °CU (b) -field, 22 °CV
[0 /90 ]2 2 2s
Fig. 64 Fringe patterns for [02/902]2s balanced cross-ply composite laminate panel.tlag= 702 days.
(a) -field, 22 °CU (b) -field, 22 °CV
0° fiber direction
[0 /90]3 2s
Fig. 65 Fringe patterns for [03/90]2s unbalanced cross-ply composite laminate panel.tlag= 702 days.
115
Fig. 66 Fringe patterns for [02/452]2s angle-ply composite laminate panel
Since the diffraction gratings were attached to composite specimens via epoxy
grating films during manufacturing process, there was a concern about the integrity of the
gratings and specimens. From those photographs we can see that after almost two years,
the diffraction gratings on composite specimens are still very high quality. Most of the
116
grating area shows a very uniform fringe pattern without abnormal distortion. Grating
delaminated was observed in very small area only and primarily around the edge.
With s similar analysis was perfromed before, the residual strain information
could be extracted from the fringe patterns. Table 5 summarizes and compares the
apparent strains of four different composite specimens when they were just made (initial
time, ti = October 8, 1997) and about two years later (final time, tf = September 10, 1999).
Table 5 Comparison of apparent residual strains between newly made compositespecimens and two-year old specimens.
[016]T [02/902]2s [03/90]2s [02/452]2s
Date 0° 90° 0° 90° 0° 90° 0° 90° shear
10/08/97 -31.2 -4650.6 -428.1 -393.7 -241.1 -992.5 -200.1 -3034.8 2854.3
09/10/99 0 -4662.5 -369.4 -369.4 -192.1 -804.4 -147.8 -3037.2 1653.2
(unit in µε)
With the creep compliance master curves and strain values at the time tf, we can
calculate residual stresses of the laminated composite for a particular lay-up configuration
at the time tf by the following equation:
difff12
f2
f1
f1
residualf12
f2
f1
)t(
)t(
)t(
]T[)]t(Q[]T[
)t(
)t(
)t(
⋅⋅⋅=
−
γεε
τσσ
(5.1)
where
117
−⋅−⋅−
−⋅−
−⋅
=
)t(
100
0)t()t()t(
)t(
)t()t()t(
)t(
0)t()t()t(
)t(
)t()t()t(
)t(
)]t(Q[
f66
f2
12f22f11
f11
f2
12f22f11
f12
f2
12f22f11
f12
f2
12f22f11
f22
f
S
SSS
S
SSS
SSSS
S
SSS
S
(5.2)
and
⋅−
=
−
uni
f2
f11
lam
fxy
fy
fx
difff12
f2
f1
0
)t(
)t(
]T[
2
)t()t(
)t(
)t(
)t(
)t(
εε
γεε
γεε
(5.3)
From the piecewise master curves in Chapter Four, we can connect them into a
continuous master curve in logarithmic scale. Figure 67 through 69 show the continuous
master curves together with the piecewise master curves.
log (psi-1) Creep Compliance S22(t) Master Curve vs. Time
log(Time) (log(sec))
Fig. 67 Continuous and piecewise S22 master curve in logarithm scale. In all thefollowing diagrams, piecewise master curves are completely covered bycontinuous master curves and reference temperature is 22 °C.
118
log (psi-1) Creep Compliance S66(t) Master Curve vs. Time
log(Time) (log(sec))
Fig. 68 Continuous and piecewise S66 master curve in logarithm scale.
log (psi-1) Creep Compliance S12(t) Master Curve vs. Time
log(Time) (log(sec))
Fig. 69 Continuous and piecewise S12 master curve in logarithm scale.
119
With the continuous master curves on logarithmic scale, we can obtain a single
curve for each creep compliance matrix element as a function of time (S22(t), S66(t), and
S12(t)) over a long period of time (see Fig. 70, Fig. 71, and Fig. 72).
psi-1 Creep Compliance S22(t) Master Curve vs. Time
Time (sec)
Fig. 70 S22(t) for time from 0 to 107.78 seconds (702 days).
120
psi-1 Creep Compliance S66(t) Master Curve vs. Time
Time (sec)
Fig. 71 S66(t) for time from 0 to 107.78 seconds (702 days).
psi-1 Creep Compliance S66(t) Master Curve vs. Time
Time (sec)
Fig. 72 S12(t) for time from 0 to 107.78 seconds (702 days).
121
Now we can substitute the values of S22(t), S66(t), S12(t) and strain values from our
laminate specimens at t = 107.78 seconds into equations (5.1), (5.2), and (5.3) to obtain the
residual stresses at the final time. Again, using MATHCAD 8, the results are shown
below in Table 6.
Table 6 Residual stresses calculated by linear viscoelastic model and laminatetheory for the same batch of composite specimens. Unit is in psi.
Balanced Cross-ply, [02/902]2s
σx σy γxy
t = 0 t = 107.78 t = 0 t = 107.78 t = 0 t = 107.78
0° plies -6.582⋅103 -5.981⋅103 6.446⋅103 6.554⋅103 0 0
90° plies 6.485⋅103 6.554⋅103 -5.866⋅103 -5.981⋅103 0 0
0° plies -6.582⋅103 -5.981⋅103 6.446⋅103 6.554⋅103 0 0
90° plies 6.485⋅103 6.554⋅103 -5.866⋅103 -5.981⋅103 0 0
Unbalanced Cross-ply, [03/90]2s
σx σy γxy
t = 0 t = 107.78 t = 0 t = 107.78 t = 0 t = 107.78
0° plies -2.868⋅103 -2.403⋅103 5.63⋅103 5.95⋅103 0 0
90° plies 6.641⋅103 5.95⋅103 -1.851⋅104 -1.515⋅104 0 0
0° plies -2.868⋅103 -2.403⋅103 5.63⋅103 5.95⋅103 0 0
90° plies 6.641⋅103 5.95⋅103 -1.851⋅104 -1.515⋅104 0 0
Angle-ply, [02/452]2s
σx σy γxy
t = 0 t = 107.78 t = 0 t = 107.78 t = 0 t = 107.78
0° plies -2.886⋅103 -2.433⋅103 2.454⋅103 2.478⋅103 9.138⋅103 5.293⋅103
45° plies 8.956⋅103 3.188⋅103 -9.195⋅103 -1.531⋅104 -2.562⋅103 -9.237⋅103
0° plies -2.886⋅103 -2.433⋅103 2.454⋅103 2.478⋅103 9.138⋅103 5.293⋅103
45° plies 8.956⋅103 3.188⋅103 -9.195⋅103 -1.531⋅104 -2.562⋅103 -9.237⋅103
122
The residual stress calculation show considerable difference between initial time
and final time. In matrix dominated directions (90° plies and shear direction for angle-ply
specimens) the difference is even larger. This result is somehow surprising. Although
3501-6 epoxy is brittle and considered exhibiting more elastic behavior rather than
viscoelastic behavior, we can still clearly see that its viscoelastic properties have a large
influence on the process induced residual stresses.
Concluding Remarks
In conclusion, throughout this study, we have been trying to measure and
understand the process induced residual stresses for advanced laminated composites.
Numerous researchers have indicated that the residual stresses are innately difficult to
measure and characterize. With the newly developed methodology, Cure Referencing
Method (CRM), we were able to demonstrate the full advantages of high sensitive moiré
interferometry.
The most important feature of moiré interferometry for our application is its long-
term capability. In this study, we showed that the CRM can not only effectively measure
the residual stresses at the time when the composite laminates were manufactured, but
also well after.
Our study also proved that the chemical curing shrinkage of polymer matrix
contributes to a large portion of residual strains. With the aid of an optical thermal
chamber, the CRM enabled us to separate the residual stresses into two components:
stresses resulting from the CTE mismatch and stresses resulting from matrix chemical
shrinkage. Another method allowed us to accurately measure the post-cure
123
polymerization shrinkage of neat resins. The way by which the diffraction gratings were
transferred from master gratings to neat resin specimens eliminates the need of knowing
the gel point in polymerization process. Although the chemical shrinkage does not
completely relax during manufacturing procedure, the viscoelastic effects still cannot be
ignored, even for a brittle thermoset epoxy, such as Hexcel 3501-6.
The AS4/3501-6 composite system does not have a notable viscoelastic behavior
at room temperature in a short period of time. Therefore, the concern is how to accelerate
the time scale and still accurately represent the real material properties. With several
momentary tensile creep tests performed and considering physical aging of the polymer,
the time-temperature superposition principle allowed us to construct a function of time
for each material property.
In the future, this research can be further investigated. The moisture effect is
always a serious issue especially for polymer matrix composite systems. Our research
indicated that there is different level of influence of chemical shrinkage in different fiber
orientations or different ply sequence. Physical aging of polymer matrix alternates the
material properties considerably, and is a potential factor of our experimental error which
cannot be ignored.
124
APPENDIX ATHE MATHCAD FILE FOR THE DATA ANALYSIS OF 90-DEGREE COMPOSITE
TENSILE SPECIMENS WITH THREE-ELEMENT STANDARD LINEAR SOLIDMODEL
import data files:i 0 1, 680..cr90_22_1 READPRN "90_22c_2.txt"( ) cr90_40_1 READPRN "90_40c_1.txt"( )
cr90_22_2 READPRN "90_22c_3.txt"( ) cr90_40_2 READPRN "90_40c_4.txt"( )
cr90_22_3 READPRN "90_22c_4.txt"( ) cr90_40_3 READPRN "90_40c_6.txt"( )
cr90_60_1 READPRN "90_60c_2.txt"( ) cr90_80_1 READPRN "90_80c_2.txt"( )
cr90_60_2 READPRN "90_60c_3.txt"( ) cr90_80_2 READPRN "90_80c_4.txt"( )
cr90_60_3 READPRN "90_60c_4.txt"( ) cr90_80_3 READPRN "90_80c_5.txt"( )
cr90_100_1 READPRN "90_100c_5.txt"( ) cr90_120_1 READPRN "90_120c_2.txt"( )
cr90_100_2 READPRN "90_100c_6.txt"( ) cr90_120_2 READPRN "90_120c_3.txt"( )
cr90_100_3 READPRN "90_100c_7.txt"( ) cr90_120_3 READPRN "90_120c_4.txt"( )
cr90_140_1 READPRN "90_140c_2.txt"( ) cr90_160_1 READPRN "90_160c_2.txt"( )
cr90_140_2 READPRN "90_140c_3.txt"( ) cr90_160_2 READPRN "90_160c_3.txt"( )
cr90_140_3 READPRN "90_140c_4.txt"( ) cr90_160_3 READPRN "90_160c_4.txt"( )
cr90_177_1 READPRN "90_177c_2.txt"( ) cr90_195_1 READPRN "90_195c_2.txt"( )
cr90_177_2 READPRN "90_177c_3.txt"( ) cr90_195_2 READPRN "90_195c_3.txt"( )
cr90_177_3 READPRN "90_177c_4.txt"( ) cr90_195_3 READPRN "90_195c_4.txt"( )
average three test results:
cr220< > cr90_22_1
0< >cr90_22_2
0< >cr90_22_3
0< >
3
cr221< > cr90_22_1
1< >cr90_22_2
1< >cr90_22_3
1< >
3
cr400< > cr90_40_1
0< >cr90_40_2
0< >cr90_40_3
0< >
3
cr401< > cr90_40_1
1< >cr90_40_2
1< >cr90_40_3
1< >
3
cr600< > cr90_60_1
0< >cr90_60_2
0< >cr90_60_3
0< >
3
125
cr601< > cr90_60_1
1< >cr90_60_2
1< >cr90_60_3
1< >
3
cr801< > cr90_80_1
1< >cr90_80_2
1< >cr90_80_3
1< >
3
cr800< > cr90_80_1
0< >cr90_80_2
0< >cr90_80_3
0< >
3
cr1000< > cr90_100_1
0< >cr90_100_2
0< >cr90_100_3
0< >
3
cr1001< > cr90_100_1
1< >cr90_100_2
1< >cr90_100_3
1< >
3
cr1200< > cr90_120_1
0< >cr90_120_2
0< >cr90_120_3
0< >
3
cr1201< > cr90_120_1
1< >cr90_120_2
1< >cr90_120_3
1< >
3
cr1400< > cr90_140_1
0< >cr90_140_2
0< >cr90_140_3
0< >
3
cr1401< > cr90_140_1
1< >cr90_140_2
1< >cr90_140_3
1< >
3
cr1600< > cr90_160_1
0< >cr90_160_2
0< >cr90_160_3
0< >
3
cr1601< > cr90_160_1
1< >cr90_160_2
1< >cr90_160_3
1< >
3
cr1770< > cr90_177_1
0< >cr90_177_2
0< >cr90_177_3
0< >
3
cr1771< > cr90_177_1
1< >cr90_177_2
1< >cr90_177_3
1< >
3
cr1951< > cr90_195_1
1< >cr90_195_2
1< >cr90_195_3
1< >
3
cr1950< > cr90_195_1
0< >cr90_195_2
0< >cr90_195_3
0< >
3
curve fitting for each temperature:• 22 °C
A 0.0774 T0400
A
no_eta22 t v,( )
cr221< >
0 v1 1 ev0 t.
.
v1 t. ev0 t.
.
1 ev0 t.
vg_cr2.053 10
3.
3.307 105.
p22 genfit cr220< >
cr221< >, vg_cr, no_eta22,
126
k0_22T0
cr221< >
0
k1_22T0
p221
ρ221
p220
• 40 °C
A 0.0768 T0400
A
no_eta40 t v,( )
cr401< >
0 v1 1 ev0 t.
.
v1 t. ev0 t.
.
1 ev0 t.
vg_cr1.466 10
3.
4.331 105.
p40 genfit cr400< >
cr401< >, vg_cr, no_eta40,
k0_40T0
cr401< >
0
k1_40T0
p401
ρ401
p400
• 60 °C
A 0.0772 T0400
A
no_eta60 t v,( )
cr601< >
0 v1 1 ev0 t.
.
v1 t. ev0 t.
.
1 ev0 t.
vg_cr1.305 10
3.
4.932 105.
p60 genfit cr600< >
cr601< >, vg_cr, no_eta60,
k0_60T0
cr601< >
0
k1_60T0
p601
ρ601
p600
• 80 °C
A 0.0779 T0400
A
no_eta80 t v,( )
cr801< >
0 v1 1 ev0 t.
.
v1 t. ev0 t.
.
1 ev0 t.
vg_cr1.327 10
3.
8.344 105.
p80 genfit cr800< >
cr801< >, vg_cr, no_eta80,
k0_80T0
cr801< >
0
k1_80T0
p801
ρ801
p800
• 100 °C
A 0.0775 T0250
A
127
no_eta100 t v,( )
cr1001< >
0 v1 1 ev0 t.
.
v1 t. ev0 t.
.
1 ev0 t.
vg_cr1.656 10
3.
6.235 105.
p100 genfit cr1000< >
cr1001< >, vg_cr, no_eta100,
k0_100T0
cr1001< >
0
k1_100T0
p1001
ρ1001
p1000
• 120 °C
A 0.0774 T0250
A
no_eta120 t v,( )
cr1201< >
0 v1 1 ev0 t.
.
v1 t. ev0 t.
.
1 ev0 t.
vg_cr1.379 10
3.
1.003 104.
p120 genfit cr1200< >
cr1201< >, vg_cr, no_eta120,
k0_120T0
cr1201< >
0
k1_120T0
p1201
ρ1201
p1200
• 140 °C
A 0.0756 T0250
A
no_eta140 t v,( )
cr1401< >
0 v1 1 ev0 t.
.
v1 t. ev0 t.
.
1 ev0 t.
vg_cr1.335 10
3.
1.384 104.
p140 genfit cr1400< >
cr1401< >, vg_cr, no_eta140,
k0_140T0
cr1401< >
0
k1_140T0
p1401
ρ1401
p1400
• 160 °C
A 0.0771 T0250
A
no_eta160 t v,( )
cr1601< >
0 v1 1 ev0 t.
.
v1 t. ev0 t.
.
1 ev0 t.
vg_cr1.082 10
3.
2.78 104.
p160 genfit cr1600< >
cr1601< >, vg_cr, no_eta160,
128
k0_160T0
cr1601< >
0
k1_160T0
p1601
ρ1601
p1600
• 177 °C
A 0.0764 T0250
A
no_eta177 t v,( )
cr1771< >
0 v1 1 ev0 t.
.
v1 t. ev0 t.
.
1 ev0 t.
vg_cr1.283 10
3.
6.448 104.
p177 genfit cr1770< >
cr1771< >, vg_cr, no_eta177,
k0_177T0
cr1601< >
0
k1_177T0
p1771
ρ1771
p1770
• 195 °C
A 0.0777 T050
A
no_eta195 t v,( )
cr1951< >
0 v1 1 ev0 t.
.
v1 t. ev0 t.
.
1 ev0 t.
vg_cr1.542 10
3.
4.325 104.
p195 genfit cr1950< >
cr1951< >, vg_cr, no_eta195,
k0_195T0
cr1951< >
0
k1_195T0
p1951
ρ1951
p1950
generate curves:r 0 10, 3450..g22 r( ) no_eta22 r p22,( )0 g40 r( ) no_eta40 r p40,( )0
g60 r( ) no_eta60 r p60,( )0 g80 r( ) no_eta80 r p80,( )0
g100 r( ) no_eta100 r p100,( )0 g120 r( ) no_eta120 r p120,( )0
g140 r( ) no_eta140 r p140,( )0 g160 r( ) no_eta160 r p160,( )0
g177 r( ) no_eta177 r p177,( )0 g195 r( ) no_eta195 r p195,( )0
129
Experimental vs. Curve Fitting (Three Element Model)Strain (in/in)
Time (sec)
Experimental vs. Curve Fitting (Three Element Model)Strain (in/in)
Time (sec)
130
Experimental vs. Curve Fitting (Three Element Model)Strain (in/in)
Time (sec)
131
APPENDIX BTHE MATHCAD FILE FOR THE DATA ANALYSIS OF 90-DEGREE COMPOSITE
TENSILE SPECIMENS WITH FOUR-ELEMENT STANDARD LINEAR SOLIDMODEL
import data filesi 0 1, 680..cr90_22_1 READPRN "90_22c_2.txt"( ) cr90_40_1 READPRN "90_40c_1.txt"( )
cr90_22_2 READPRN "90_22c_3.txt"( ) cr90_40_2 READPRN "90_40c_4.txt"( )
cr90_22_3 READPRN "90_22c_4.txt"( ) cr90_40_3 READPRN "90_40c_6.txt"( )
cr90_60_1 READPRN "90_60c_2.txt"( ) cr90_80_1 READPRN "90_80c_2.txt"( )
cr90_60_2 READPRN "90_60c_3.txt"( ) cr90_80_2 READPRN "90_80c_4.txt"( )
cr90_60_3 READPRN "90_60c_4.txt"( ) cr90_80_3 READPRN "90_80c_5.txt"( )
cr90_100_1 READPRN "90_100c_5.txt"( ) cr90_120_1 READPRN "90_120c_2.txt"( )
cr90_100_2 READPRN "90_100c_6.txt"( ) cr90_120_2 READPRN "90_120c_3.txt"( )
cr90_100_3 READPRN "90_100c_7.txt"( ) cr90_120_3 READPRN "90_120c_4.txt"( )
cr90_140_1 READPRN "90_140c_2.txt"( ) cr90_160_1 READPRN "90_160c_2.txt"( )
cr90_140_2 READPRN "90_140c_3.txt"( ) cr90_160_2 READPRN "90_160c_3.txt"( )
cr90_140_3 READPRN "90_140c_4.txt"( ) cr90_160_3 READPRN "90_160c_4.txt"( )
cr90_177_1 READPRN "90_177c_2.txt"( ) cr90_195_1 READPRN "90_195c_2.txt"( )
cr90_177_2 READPRN "90_177c_3.txt"( ) cr90_195_2 READPRN "90_195c_3.txt"( )
cr90_177_3 READPRN "90_177c_4.txt"( ) cr90_195_3 READPRN "90_195c_4.txt"( )
week100 READPRN "90_100c_1w.txt"( )
average three test results:• time axis
cr220< > cr90_22_1
0< >cr90_22_2
0< >cr90_22_3
0< >
3
cr400< > cr90_40_1
0< >cr90_40_2
0< >cr90_40_3
0< >
3
cr600< > cr90_60_1
0< >cr90_60_2
0< >cr90_60_3
0< >
3
cr800< > cr90_80_1
0< >cr90_80_2
0< >cr90_80_3
0< >
3
132
cr1000< > cr90_100_1
0< >cr90_100_2
0< >cr90_100_3
0< >
3
cr1200< > cr90_120_1
0< >cr90_120_2
0< >cr90_120_3
0< >
3
cr1400< > cr90_140_1
0< >cr90_140_2
0< >cr90_140_3
0< >
3
cr1600< > cr90_160_1
0< >cr90_160_2
0< >cr90_160_3
0< >
3
cr1770< > cr90_177_1
0< >cr90_177_2
0< >cr90_177_3
0< >
3
cr1950< > cr90_195_1
0< >cr90_195_2
0< >cr90_195_3
0< >
3
• corrected strains
cr221< > cr90_22_1
1< >cr90_22_2
1< >cr90_22_3
1< >
3
cr401< > cr90_40_1
1< >cr90_40_2
1< >cr90_40_3
1< >
3
cr601< > cr90_60_1
1< >cr90_60_2
1< >cr90_60_3
1< >
3
cr801< > cr90_80_1
1< >cr90_80_2
1< >cr90_80_3
1< >
3
cr1001< > cr90_100_1
1< >cr90_100_2
1< >cr90_100_3
1< >
3
cr1201< > cr90_120_1
1< >cr90_120_2
1< >cr90_120_3
1< >
3
cr1401< > cr90_140_1
1< >cr90_140_2
1< >cr90_140_3
1< >
3
cr1601< > cr90_160_1
1< >cr90_160_2
1< >cr90_160_3
1< >
3
cr1771< > cr90_177_1
1< >cr90_177_2
1< >cr90_177_3
1< >
3
cr1951< > cr90_195_1
1< >cr90_195_2
1< >cr90_195_3
1< >
3
• Poisson ratio
cr222< > cr90_22_1
2< >cr90_22_2
2< >cr90_22_3
2< >
3
cr402< > cr90_40_1
2< >cr90_40_2
2< >cr90_40_3
2< >
3
cr602< > cr90_60_1
2< >cr90_60_2
2< >cr90_60_3
2< >
3
133
cr802< > cr90_80_1
2< >cr90_80_2
2< >cr90_80_3
2< >
3
cr1202< > cr90_120_1
2< >cr90_120_2
2< >cr90_120_3
2< >
3
cr1402< > cr90_140_1
2< >cr90_140_2
2< >cr90_140_3
2< >
3
cr1602< > cr90_160_1
2< >cr90_160_2
2< >cr90_160_3
2< >
3
cr1772< > cr90_177_1
2< >cr90_177_2
2< >cr90_177_3
2< >
3
cr1002< > cr90_100_1
2< >cr90_100_2
2< >cr90_100_3
2< >
3
cr1952< > cr90_195_1
2< >cr90_195_2
2< >cr90_195_3
2< >
3
• data smoothing for the Poisson ratioM22 medsmooth cr22
2< >11, M40 medsmooth cr40
2< >11,
M60 medsmooth cr602< >
11, M80 medsmooth cr802< >
11,
M100 medsmooth cr1002< >
11, M120 medsmooth cr1202< >
11,
M140 medsmooth cr1402< >
11, M160 medsmooth cr1602< >
11,
M177 medsmooth cr1772< >
11, M195 medsmooth cr1952< >
11,
r 0 10, 3450..
curve fitting for each temperature:• 22 °C
A 0.0774 T0400
A
w_eta22 t v,( )
cr221< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
vg_cr
2.421 103.
2.984 105.
1.334 109.
ν_22 t v,( )
cr222< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
ν_vg_cr
4.2 103.
1.3 103.
4.2 107.
p22 genfit cr220< >
cr221< >, vg_cr, w_eta22, ν_p22 genfit cr22
0< >M22, ν_vg_cr, ν_22,
134
k0_22T0
cr221< >
0
k1_22T0
p221
ρ221
p220
η0_22T0
p222
s22_22i
cr221< >
i
T0
• 40 °C
A 0.0768 T0400
A
w_eta40 t v,( )
cr401< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
vg_cr
2.568 103.
2.978 105.
4.68 109.
ν_40 t v,( )
cr402< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
ν_vg_cr
2.8 103.
1.9 103.
2.6 107.
p40 genfit cr400< >
cr401< >, vg_cr, w_eta40, ν_p40 genfit cr40
0< >M40, ν_vg_cr, ν_40,
k0_40T0
cr401< >
0
k1_40T0
p401
ρ401
p400
η0_40T0
p402
s22_40i
cr401< >
i
T0
• 60 °C
A 0.0772 T0400
A
w_eta60 t v,( )
cr601< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
vg_cr
3.021 103.
3.154 105.
6.588 109.
ν_60 t v,( )
cr602< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
ν_vg_cr
3.5 103.
1.6 103
1.3 107.
p60 genfit cr600< >
cr601< >, vg_cr, w_eta60, ν_p60 genfit cr60
0< >M60, ν_vg_cr, ν_60,
135
k0_60T0
cr601< >
0
k1_60T0
p601
ρ601
p600
η0_60T0
p602
s22_60i
cr601< >
i
T0
• 80 °C
A 0.0779 T0400
A
w_eta80 t v,( )
cr801< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
vg_cr
3.911 103.
5.136 105.
1.277 108.
ν_80 t v,( )
cr802< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
ν_vg_cr
1.7 103.
2.5 103.
1.7 107.
p80 genfit cr800< >
cr801< >, vg_cr, w_eta80, ν_p80 genfit cr80
0< >M80, ν_vg_cr, ν_80,
k0_80T0
cr801< >
0
k1_80T0
p801
ρ801
p800
η0_80T0
p802
s22_80i
cr801< >
i
T0
• 100 °C
A 0.0775 T0250
A
w_eta100 t v,( )
cr1001< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
vg_cr
4.289 103.
4.282 105.
8.123 109.
ν_100 t v,( )
cr1002< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
ν_vg_cr
5.8 103.
2.4 103.
3.2 107.
136
p100 genfit cr1000< >
cr1001< >, vg_cr, w_eta100,
ν_p100 genfit cr1000< >
M100, ν_vg_cr, ν_100,
k0_100T0
cr1001< >
0
k1_100T0
p1001
ρ1001
p1000
η0_100T0
p1002
s22_100i
cr1001< >
i
T0
• 120 °C
A 0.0774 T0250
A
w_eta120 t v,( )
cr1201< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
vg_cr
2.298 103.
7.342 105.
9.892 109.
ν_120 t v,( )
cr1202< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
ν_vg_cr
7.2 103.
1.6 103.
4.8 107.
p120 genfit cr1200< >
cr1201< >, vg_cr, w_eta120,
ν_p120 genfit cr1200< >
M120, ν_vg_cr, ν_120,
k0_120T0
cr1201< >
0
k1_120T0
p1201
ρ1201
p1200
η0_120T0
p1202
s22_120i
cr1201< >
i
T0
• 140 °C
A 0.0756 T0250
A
w_eta140 t v,( )
cr1401< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
vg_cr
3.168 103.
8.812 105.
1.898 108.
137
ν_140 t v,( )
cr1402< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
ν_vg_cr
4.1 103.
1.75 103.
6.2 107.
p140 genfit cr1400< >
cr1401< >, vg_cr, w_eta140,
ν_p140 genfit cr1400< >
M140, ν_vg_cr, ν_140,
k0_140T0
cr1401< >
0
k1_140T0
p1401
ρ1401
p1400
η0_140T0
p1402
s22_140i
cr1401< >
i
T0
• 160 °C
A 0.0771 T0250
A
w_eta160 t v,( )
cr1601< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
vg_cr
3.148 103.
1.522 104.
4.327 108.
ν_160 t v,( )
cr1602< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
ν_vg_cr
3.9 103.
3.4 103.
1.0 107.
p160 genfit cr1600< >
cr1601< >, vg_cr, w_eta160,
ν_p160 genfit cr1600< >
M160, ν_vg_cr, ν_160,
k0_160T0
cr1601< >
0
k1_160T0
p1601
ρ1601
p1600
η0_160T0
p1602
s22_160i
cr1601< >
i
T0
• 177 °C
A 0.0764 T0250
A
138
w_eta177 t v,( )
cr1771< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
vg_cr
2.813 103.
4.143 104.
8.712 108.
ν_177 t v,( )
cr1772< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
ν_vg_cr
3.1 103.
5.4 103.
1.4 107.
p177 genfit cr1770< >
cr1771< >, vg_cr, w_eta177,
ν_p177 genfit cr1770< >
M177, ν_vg_cr, ν_177,
k0_177T0
cr1771< >
0
k1_177T0
p1771
ρ1771
p1770
η0_177T0
p1772
s22_177i
cr1771< >
i
T0
• 195 °C
A 0.0777 T050
A
w_eta195 t v,( )
cr1951< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
vg_cr
2.922 103.
3.194 104.
4.569 108.
ν_195 t v,( )
cr1952< >
0 v1 1 ev0 t.
. v2 t.
v1 t. ev0 t.
.
1 ev0 t.
t
ν_vg_cr
1.1 102.
7.6 103.
1.2 107.
p195 genfit cr1950< >
cr1951< >, vg_cr, w_eta195,
ν_p195 genfit cr1950< >
M195, ν_vg_cr, ν_195,
k0_195T0
cr1951< >
0
k1_195T0
p1951
ρ1951
p1950
η0_195T0
p1952
139
s22_195i
cr1951< >
i
T0
generate curves:r 0 10, 3450.. j 0 1, 590..g22 r( ) w_eta22 r p22,( )0 g40 r( ) w_eta40 r p40,( )0
g60 r( ) w_eta60 r p60,( )0 g80 r( ) w_eta80 r p80,( )0
g100 r( ) w_eta100 r p100,( )0 g120 r( ) w_eta120 r p120,( )0
g140 r( ) w_eta140 r p140,( )0 g160 r( ) w_eta160 r p160,( )0
g177 r( ) w_eta177 r p177,( )0 g195 r( ) w_eta195 r p195,( )0
Experimental vs. Curve Fitting (Four Element Model)Strain (in/in)
Time (sec)
140
Experimental vs. Curve Fitting (Four Element Model)Strain (in/in)
Time (sec)
Experimental vs. Curve Fitting (Four Element Model)Strain (in/in)
Time (sec)
141
calculate creep compliance matrix elements:
S22_22 t( )1
k0_22
1
k1_221 e
t
ρ22. 1
η0_22t.
ν21_22 t( ) ν_22 t ν_p22,( )0 S12_22 t( ) ν21_22 t( ) S22_22 t( ).
S22_40 t( )1
k0_40
1
k1_401 e
t
ρ40. 1
η0_40t.
ν21_40 t( ) ν_40 t ν_p40,( )0 S12_40 t( ) ν21_40 t( ) S22_40 t( ).
S22_60 t( )1
k0_60
1
k1_601 e
t
ρ60. 1
η0_60t.
ν21_60 t( ) ν_60 t ν_p60,( )0 S12_60 t( ) ν21_60 t( ) S22_60 t( ).
S22_80 t( )1
k0_80
1
k1_801 e
t
ρ80. 1
η0_80t.
ν21_80 t( ) ν_80 t ν_p80,( )0 S12_80 t( ) ν21_80 t( ) S22_80 t( ).
S22_100 t( )1
k0_100
1
k1_1001 e
t
ρ100. 1
η0_100t.
ν21_100 t( ) ν_100 t ν_p100,( )0 S12_100 t( ) ν21_100 t( ) S22_100 t( ).
S22_120 t( )1
k0_120
1
k1_1201 e
t
ρ120. 1
η0_120t.
ν21_120 t( ) ν_120 t ν_p120,( )0 S12_120 t( ) ν21_120 t( ) S22_120 t( ).
S22_140 t( )1
k0_140
1
k1_1401 e
t
ρ140. 1
η0_120t.
ν21_140 t( ) ν_140 t ν_p140,( )0 S12_140 t( ) ν21_140 t( ) S22_140 t( ).
S22_160 t( )1
k0_160
1
k1_1601 e
t
ρ160. 1
η0_140t.
ν21_160 t( ) ν_160 t ν_p160,( )0 S12_160 t( ) ν21_160 t( ) S22_160 t( ).
S22_177 t( )1
k0_177
1
k1_1771 e
t
ρ177. 1
η0_177t.
ν21_177 t( ) ν_177 t ν_p177,( )0 S12_177 t( ) ν21_177 t( ) S22_177 t( ).
S22_195 t( )1
k0_195
1
k1_1951 e
t
ρ195. 1
η0_195t.
ν21_195 t( ) ν_195 t ν_p195,( )0 S12_195 t( ) ν21_195 t( ) S22_195 t( ).
142
the long-term test (one week):
A_1w 1.012 0.074. P_1w 250 T0_1wP_1w
A_1w
j 0 1, 1005..
S22_100wj
week1001< >
j
T0_1w
(psi-1) Creep Compliance S22(t) vs. Time (curve fit)
Time (sec)
143
log (psi-1) Log Creep Compliance S22(t) vs. Log Time (curve fit)
log(Time) (log(sec))
log (psi-1) Log Creep Compliance S22(t) vs. Log Time (experimental)
log(Time) (log(sec))
144
(psi-1) Creep Compliance S12(t) vs. Time (curve fit)
Time (sec)
log (psi-1) Log Creep Compliance S12(t) vs. Log Time (curve fit)
log(Time) (log(sec))
145
log (psi-1) Creep Compliance S22(t) Master Curve (no vertical shift)
log(Time) (log(sec))
log (psi-1) Creep Compliance S22(t) Master Curve (with vertical shift)
log(Time) (log(sec))
146
log (psi-1) Creep Compliance S12(t) Master Curve (with vertical shift)
log(Time) (log(sec))
connect piecewise master curves S22(t) into a single curvei 0 690..
t 1 1.1, 101.41.. t1 1 1.1, 10
1.17.. t2 1 1.1, 101.19.. t3 1 1.1, 10
1.12..
t4 1 1.1, 100.79.. t5 1 1.1, 10
0.91.. t6 1 1.1, 100.89.. t7 1 1.1, 10
3.35..m 0 1, 1150..
101.41 0( )
1
0.11 248.04= m1 0 1, 247.. 0 247 247=
logs22m1 log S22_22 m1 0.1. 1( )( )
logtm1 log m1 0.1. 1( )
102.58 1.41( )
1
0.11 138.911= m2 248 249, 248 137.. 248 137 385=
logs22m2 log S22_40 m2 248( ) 0.1. 1( )( ) 0.0054
logtm2 log m2 248( ) 0.1. 1( ) 1.41
103.77 2.58( )
1
0.11 145.882= m3 386 387, 386 144.. 386 144 530=
logs22m3 log S22_60 m3 386( ) 0.1. 1( )( ) 0.0101
logtm3 log m3 386( ) 0.1. 1( ) 2.58
104.89 3.77( )
1
0.11 122.826= m4 531 532, 531 121.. 531 121 652=
147
logs22m4 log S22_80 m4 531( ) 0.1. 1( )( ) 0.0205
logtm4 log m4 531( ) 0.1. 1( ) 3.77
105.68 4.89( )
1
0.11 52.66= m5 653 654, 653 51.. 653 51 704=
logs22m5 log S22_100 m5 653( ) 0.1. 1( )( ) 0.0397
logtm5 log m5 653( ) 0.1. 1( ) 4.89
106.59 5.68( )
1
0.11 72.283= m6 705 706, 705 71.. 705 71 776=
logs22m6 log S22_120 m6 705( ) 0.1. 1( )( ) 0.0475
logtm6 log m6 705( ) 0.1. 1( ) 5.68
107.48 6.59( )
1
0.11 68.625= m7 777 778, 777 69.. 777 69 846=
logs22m7 log S22_140 m7 777( ) 0.1. 1( )( ) 0.0589
logtm7 log m7 777( ) 0.1. 1( ) 6.59
101.5
1
0.11 307.228= m8 847 848, 847 306.. 847 306 1.153 10
3.=
logs22m8 log S22_160 m8 847( ) 0.1. 1( )( ) 0.0823
logtm8 log m8 847( ) 0.1. 1( ) 7.48
log (psi-1) Single Function for Creep Compliance log(S22(t))
log(Time) (log(sec))
148
(psi-1) Single Function for Creep Compliance S22(t)
Time(sec)
connect piecewise master curves S12(t) into a single curvei 0 690..
t 1 1.1, 101.05.. t1 1 1.1, 10
1.19.. t2 1 1.1, 100.98..
t3 1 1.01, 100.95.. t4 1 1.1, 10
0.72.. t5 1 1.1, 100.87..
t6 1 1.1, 101.22.. t7 1 1.1, 10
1.5..m 0 1, 984..
101.05 0( )
1
0.11 103.202= m1 0 1, 102.. 0 102 102=
logs12m1 log S12_22 m1 0.1. 1( )( )
logtm1 log m1 0.1. 1( )
101.19
1
0.11 145.882= m2 103 104, 103 144.. 103 144 247=
logs12m2 log S12_40 m2 103( ) 0.1. 1( )( ) 0.0397
logtm2 log m2 103( ) 0.1. 1( ) 1.05
100.98
1
0.11 86.499= m3 248 249, 248 85.. 248 85 333=
149
logs12m3 log S12_60 m3 248( ) 0.1. 1( )( ) 0.0459
logtm3 log m3 248( ) 0.1. 1( ) 2.24
100.95
1
0.11 80.125= m4 334 335, 334 79.. 334 79 413=
logs12m4 log S12_80 m4 334( ) 0.1. 1( )( ) 0.0774
logtm4 log m4 334( ) 0.1. 1( ) 3.22
100.72
1
0.11 43.481= m5 414 415, 414 42.. 414 42 456=
logs12m5 log S12_100 m5 414( ) 0.1. 1( )( ) 0.0196
logtm5 log m5 414( ) 0.1. 1( ) 4.17
100.87
1
0.11 65.131= m6 457 458, 457 64.. 457 64 521=
logs12m6 log S12_120 m6 457( ) 0.1. 1( )( ) 0.0002
logtm6 log m6 457( ) 0.1. 1( ) 4.89
101.22
1
0.11 156.959= m7 522 523, 522 155.. 522 155 677=
logs12m7 log S12_140 m7 522( ) 0.1. 1( )( ) 0.0149
logtm7 log m7 522( ) 0.1. 1( ) 5.76
101.5
1
0.11 307.228= m8 678 679, 678 306.. 678 306 984=
logs12m8 log S12_160 m8 678( ) 0.1. 1( )( ) 0.0401
logtm8 log m8 678( ) 0.1. 1( ) 6.98
150
log (psi-1) Single Function for Creep Compliance log(S12(t))
log(Time) (log(sec))
(psi-1) Single Function for Creep Compliance S12(t)
Time(sec)
compare with long term test (reference temperature at 100 °C):t 1 2, 10
1.22.. t1 1 10, 101.31.. t2 1 10, 10
1.02..
t3 1 2, 100.75.. t4 1 2, 10
2.5..
151
i 1 2, 10.. j 0 1, 590.. k 1 2, 579..
log (psi-1) Creep Compliance S22(t) Master Curve (curve fit, experimental, and long term)
log(Time) (log(sec))
152
REFERENCES
1. “X-34,” High-Performance Composites. May/June, p. 9 (1999).
2. Pagano, N. J. and Hahn, H. T., “Evaluation of Composite Curing Stresses,”Composite Materials: Testing and Design (Fourth Conference), ASTM STP 617, pp.317-329 (1977).
3. Agarwal, B. D. and Broutman, L. J., Analysis and Performance of Fiber Composites,Second Edition, John Wiley & Sons, Inc., New York (1990).
4. Mathar, J., “Determination of Initial Stress by Measuring the Deformations AroundDrilled Holes,” Transactions of ASME, Vol. 56, p.249, (1934).
5. Rendler, N. J. and Vigness, I., “Hole-drilling Strain-gage Method of MeasuringResidual Stresses,” Experimental Mechanics, Vol. 6, pp. 577-586, December (1966).
6. Bert, C. W. and Thompson, G. L., “A Method for Measuring Planar Residual Stressesin Rectangularly Orthotropic Materials,” Journal of Composite Materials, Vol. 2, No.2, April, pp. 244-253 (1968).
7. Schajer, G. S. and Yang, L., “Residual-stress Measurement in Orthotropic MaterialsUsing the Hole-drilling Method,” Experimental Mechanics, Vol. 34, No. 4, pp. 324-333 (1994).
8. Lee, J., Czarnek, R., and Guo, Y., “Interferometric Study of Residual Strains in ThickComposites,” Proceedings of the 1989 SEM Conference on Experimental Mechanics,Cambridge, MA, pp. 356-364 (1989).
9. Gascoigne, H. E., “Residual Surface Stresses in Laminated Cross-ply Fiber-epoxyComposite Materials,” Experimental Mechanics, Vol. 34, No. 1, pp. 27-36 (1994).
10. Sunderland, P., Yu, W. J., and Månson, J. E., "A Technique for the Measurement ofProcess-Induced Internal Stresses in Polymers and Polymer Composites,”Proceedings of ICCM-10, Vol. 3, pp. 125-132 (1995).
11. Lee, J. and Czarnek, R., “Measuring Residual Strains in Composites Laminates UsingMoiré Interferometry,” Proceedings of the 1991 SEM Conference on ExperimentalMechanics, Milwaukee, WI, pp. 405-415 (1991).
153
12. Chapman, T. J., Gillespie, J. W., and Pipes, R. B., "Prediction of Process-InducedResidual Stresses in Thermoplastic Composites,” Journal of Composite Materials,Vol. 24, pp. 616-643 (1990).
13. Månson, J. E. and Seferis, J. C., “Process Simulated Laminate (PSL)” A Methodologyto Internal Stress Characterization in Advanced Composite Materials,” Journal ofComposite Materials, Vol. 26, No. 3, pp. 405-431 (1992).
14. Joh, D., Byun, K. Y., and Ha, J., “Thermal Residual Stresses in Thick Graphite/EpoxyComposite Laminates-Uniaxial Approach,” Experimental Mechanics, Vol. 33, No. 1,pp. 70-76 (1993).
15. Hahn, H. T., “Residual Stresses in Polymer Matrix Composite Laminates,” CompositeMaterials, Vol. 10, pp. 265-277 (1976).
16. Kim, R. Y. and Hahn, H. T., “Effect of Curing Stresses on the First Ply-failure inComposite Laminates,” Journal of Composite Materials, Vol. 13, January, pp. 2-16(1979).
17. Jeronimidis, G. and Parkyn, A. T., "Residual Stresses in Carbon Fiber-ThermoplasticMatrix Laminates,” Journal of Composite Materials, Vol. 22, May, pp. 401-415(1988).
18. Hyer, M. W., “Calculation of Room-Temperature Shapes of UnsymmetricLaminates,” Journal of Composite Materials, Vol. 15, July, pp. 296-310 (1981).
19. Kim, K. S., Hahn, H. T., and Croman, R. B., "The Effect of Cooling Rate on ResidualStresses in a Thermoplastic Composite,” Journal of Composites Technology &Research, Vol. 11, No. 2, pp. 47-52 (1989).
20. Fakuda, H., Takahashi, K., and Toda, S., “Thermal Deformation of Anti-SymmetricLaminates at Cure,” Proceedings of ICCM-10, Vol. 3, pp. 141-148 (1995).
21. Unger, W. J. and Hansen, J. S., “The Effect of Thermal Processing on Residual StrainDevelopment in Unidirectional Graphite Fiber Reinforced PEEK,” Journal ofComposite Materials, Vol. 27, No. 1, pp. 59-81 (1993).
22. Carman, G. P. and Sendechyj, G. P., “Review of the Mechanics of Embedded OpticalSensors,” Journal of Composites Technology & Research, JCTRER, Vol. 17, No. 3,pp. 183-193 (1995).
23. Predecki, P. and Barrett, C. S., “Stress Measurement in Graphite/Epoxy Compositesby X-Ray Diffraction from Fillers,” Journal of Composite Materials, Vol. 13, pp. 61-71 (1979).
154
24. Fenn, R. H., Jones, A. M., and Wells, G. M., “X-Ray Diffraction Investigation ofTriaxial Residual Stresses in Composite Materials,” Journal of Composite Materials,Vol. 27, No. 14, pp. 1338-1351 (1993).
25. Madhukar, M. S., Kosuri, R. P., and Bowles, K. J, "Reduction of Curing InducedFiber Stresses by Cure Cycle Optimization in Polymer Matrix Composites,”Proceedings of ICCM-10, Vol. 3, pp. 157-164 (1995).
26. Ifju, G. P., Kilday, B. C., Liu, S.-C., and Niu, X., “Measurement of Residual Stress inLaminated Composites,” Proceedings of VIII International Congress on ExperimentalMechanics and Experimental/Numerical Mechanics in Electronic Packaging,Nashville, Tennessee, pp.198-199 (1996).
27. Hodges, J., Yates, B., Darby, M. I., Wostenholm, G. H., Clemmet, J. F., and Keates,T. F., “Residual Stresses and the Optimum Cure Cycle for an Epoxy Resin,” Journalof Materials Science, Vol. 24, pp. 1984-1990 (1989).
28. Snow, A. W. and Armistead, J. P., “A Simple Dilatometer for Thermoset CureShrinkage and Thermal Expansion Measurements,” Journal of Applied PolymerScience, Vol. 52, pp. 401-411 (1994).
29. Russell, J. D., “Cure Shrinkage of Thermoset Composites,” SAMPE Quarterly, Vol.24, No. 2, pp.28-33 (1993).
30. Wang, H. B., Yang, Y. G., Yu, H. H., and Sun, W. M., “Assessment of ResidualStresses During Cure and Cooling of Epoxy Resin,” Polymer Engineering andScience, Vol. 35, No. 23, pp. 1895-1898 (1995).
31. Hahn, H. T. and Pagano, N. J., “Curing Stresses in Composite Laminates,” Journal ofComposite Materials, Vol. 9, pp. 91-106 (1975).
32. Bogetti, T. A. and Gillespie, J. W. Jr., “Process-Induced Stress and Deformation inThick-Section Thermoset Composite Laminates,” Journal of Composite Materials,Vol. 26, No. 5, pp. 626-660 (1992).
33. Schapery, R. A., “Viscoelastic Behavior and Analysis of Composite Materials,” inMechanics of Composite Materials, G. P. Sendeckyi, ed., Academic Press, New York,pp. 85-167 (1974).
34. Schapery, R. A., “Application of Thermodynamics to Thermomechanical, Fracture,and Birefringent Phenomena in Viscoelastic Media,” Journal of Applied Physics, Vol.35, pp. 1451-1465 (1964).
35. Schapery, R. A., “A Method of Viscoelastic Stress Analysis Using Elastic Solutions,”Journal of the Franklin Institute, Vol. 279, pp.268-289 (1965).
155
36. Schapery, R. A., “Stress Analysis of viscoelastic Composite Materials,” Journal ofComposite Materials, Vol. 1, pp.228-267 (1967).
37. Schapery, R. A., “Thermal Expansion Coefficient of Composite Materials Based onEnergy Principles,” Journal of Composite Materials, Vol. 2, No. 3, July, pp. 380-404(1968).
38. Weitsman, Y., “Optimal Cool-Down in Linear Viscoelasticity,” Journal of AppliedMechanics, Vol. 47, March, pp. 35-39 (1980).
39. Harper, B. D. and Weitsman, Y., “Residual Stresses in an Unsymmetrical Cross-PlyGraphite/Epoxy Laminate,” AIAA Journal, Vol.19, pp. 325-332 (1981).
40. Harper, B. D. and Weitsman, Y., “On the Effects of Environmental Conditioning onResidual Stresses in Composite Laminates,” International Journal of Solids andStructures, Vol. 21, No. 8, pp. 907-926 (1985).
41. White, S. R. and Hahn, H. T., “Process Modeling of Composite Materials: ResidualStress Development during Cure. Part I. Model Formulation,” Journal of CompositeMaterials, Vol. 26, No. 16, pp. 2402-2422 (1992).
42. White, S. R. and Hahn, H. T., “Process Modeling of Composite Materials: ResidualStress Development during Cure. Part II. Experimental Validation,” Journal ofComposite Materials, Vol. 26, No. 16, pp. 2423-2453 (1992).
43. Liu, S.-C., Niu X., and Ifju, P., “Residual Stress Characterization by MoiréInterferometry.” Proceedings of the SEM Spring Conference of Experimental andApplied Mechanics and Experimental/Numerical Mechanics in Electronic PackagingIII, Houston, Texas, pp.176-178 (1998).
44. Wang, T.-M., Daniel, I. M., and Gotro, J. T., “Thermoviscoelastic Analysis ofResidual Stresses and Warpage in Composite Laminates,” Journal of CompositeMaterials, Vol. 26, No. 6, pp. 883-899 (1992).
45. Kim, Y. K. and White, S. R., “Process- Induced Stress Relaxation Analysis ofAS4/3501-6 Laminate,” Proceedings of the American Society for Composites, TenthInternational Conference on Composite Materials(ICCM10), Whistler, Canada, pp.281-292 (1995).
46. Adolf, D. and Martin, J. E., “Calculation of Stresses in Crosslinking Polymers,”Journal of Composite Materials, Vol. 30, No. 1, pp. 13-34 (1996).
47. Ferry, J. D., Viscoelasic Elastic Properties of Polymers, 2nd ed., John Wiley & Sons,New York (1970).
48. Post, D., Han, B., and Ifju, P. G., High Sensitivity Moiré: Experimental Analysis forMechanics and Materials, Springer Verlag, New York (1994).
156
49. ALCOA Aluminum Handbook, Aluminum Company of American, Pittsburgh (1967).
50. Gibson, R. F., Principles of Composite Material Mechanics, McGraw-Hill, Inc., NewYork (1994).
51. Ifju, P. G., Kilday, B. C., Niu, X., Liu, S.-C., and Ettinger, S., “Residual StrainMeasurement in Composites using the Curing Referencing Method,” to be publishedby Experimental Mechanics.
52. Boll, D. J., “Curing of 3501-6 Composite Systems,” Hercules Advanced Materials &Systems Company, Magna, February (1992).
53. Kilday, B. C., “Residual Stress Measurement of AS4/3501-6 Laminates Using MoiréInterferometry,” Master’s Thesis, University of Florida, Gainesville, May (1997).
54. Ifju, P., Kilday, B. C., Niu, X., and Liu, S.-C., “A Novel Method to Measure ResidualStresses in Laminated Composites,” Journal of Composites, Vol. 33, No. 16, pp.1511-1524 (1999).
55. Mallick, P. K., Fiber-Reinforced Composites—Materials, Manufacturing, andDesign, 2nd ed., Marcel Dekker, Inc., New York (1988).
56. Kim, Y. K., and White, S. R., “Stress Relaxation Behavior of 3501-6 Epoxy ResinDuring Cure,” Polymer Engineering and Science, Vol. 36, No. 23, pp. 2851-2862(1996).
57. Ferry, J. D., Viscoelastic Properties of Polymers, 3rd ed., John Wiley & Sons, NewYork (1980).
58. ASTM, 1997 Annual Book of ASTM Standards, Section 15, Vol. 15.03, pp. 98-108.
59. Chamis, C. C. and Sinclair, J. H., “Ten-deg Off-axis Test for Shear Properties in FiberComposites,” Experimental Mechanics, Vol. 17, No. 9, pp. 339-346 (1977).
60. Dally, J. W. and Riley, W. F., Experimental Stress Analysis, 3rd ed., McGraw-Hill,Inc., New York (1991).
61. Struik, L. C. E., "Physical Aging in Plastics and other Glassy Materials," PolymerEngineering and Science, Vol. 17, pp. 165-173 (1977).
62. Struik, L. C. E., Physical Aging in Amorphous Polymers and Other Materials,Elsevier, Amsterdam (1978).
63. Janas, V. F. and McCullough, R. L., "The Effect of Physical Aging on theViscoelastic Behavior of a Thermoset Polyester," Composites Science andTechnology, Vol. 30, pp. 99-118 (1987).
157
64. Ogale, A. A. and McCullough, R. L., "Physical Aging Characteristics of PolyesterEther Ketone," Composites Science and Technology, Vol. 30, pp. 137-148 (1987).
65. Sullivan, J. L., "Creep and Physical Aging of Composite," Composites Science andTechnology, Vol. 39, pp. 207-232 (1990).
66. Niu, X., Ifju, P. G., and Liu, S.-C., “Prediction of Process Induced Residual StressesUsing the Cure Referencing Method (CRM),” Proceedings of the SEM AnnualConference on Theoretical, Experimental and Computational Mechanics,Cincinnati, pp.739-742 (1999).
67. Niu, X. “Process Induced Residual Stresses and Dimensional Distortion in AdvancedLaminated Composites,” Ph. D. dissertation, University of Florida, Gainesville,August (1999).
158
BIOGRAPHICAL SKETCH
Shao-Chun Liu was born in 1967, Tainan, Taiwan, the Republic of China. He
obtained his Bachelor of Science degree from the Department of Physics at the National
Central University in 1989, where he had the most enjoyable memory of student life. He
served in Marine Corps of the R.O.C. for two years and came to U.S. in 1991.
He spent one year in the Intensive English Institute at the University of Illinois at
Urban-Champaign to improve his language, and arrived Gainesville, Florida in 1992 and
started his graduate study at the University of Florida.
Under the guidance of Dr. Chang-Tsan Sun, he finished the master’s program in
1994 and continued his doctoral study under the supervision of Dr. Peter Ifju. His field of
interest is residual stress of advanced laminated composites and experimental stress
analysis. In 1999, he received his Doctor of Philosophy degree.