- ’ “:0; v ,.‘
,. .

:3“; v‘

1‘"
-._. “1.1...

:2an; ‘ "
.| .
7;"; g”; ‘
'v- ‘va. rfl"
i

3‘31}
‘ "’§, (9.";

a???“ "
. ¢ .
I

3r.

7*?
I"! -

335’ yo» 7-
‘ 'V/“v ' .-

“4;, am I
an >1- T
_- Efyrrfigvfit
I. v .- ' “31-52: x; fir? "
‘f‘, “I ‘. . ‘ 'l ' ‘3’ £23.. I! ‘4‘ 1:}: r'J'Jn. "‘
I “Hawk"! : fiwrbr‘: J 'N ’7? " é: ' ‘ 4 ‘ ""
, - 1 , A ')
'AU'W' i. fit :1 . _ ‘ v
r“: g... .11.; = _ - . ,;
r, «’1'.

4..»

:‘ “ ér'i‘ L'L‘D‘I [m
r
u

dY-FE "' J!
-1 J” wig“

and;

.0.» $5,}
4: fifi‘rz&xy‘/Jréfr& q. A4 A35“
{:"JT'J. =3»

1 “f“ 35 Eg@

k.
42?.

:r .31.}
J.-
__ J. 1' i:

‘1.
.‘7'

”(5“;
”fit?!“ "3;“

A "’3 i AW; ‘ 55‘
J 413 if: n ng fi' e
' ,. a ‘
» - szgjgyexflg-g.
9"“ 'wu‘f - “@355 31,5 . :Y 4'

1;???

m“

.. an I

“.711.“
vua 23;"?

,.
I’thf‘”

“ flint-r
'~ ' 1‘7“!)
, .7“...1,).. fl
a gi’wg-Ag 1"; gag/"v
3'
a2, .1; (mull.

5):.“ ‘.... j. {.2.<:..1"‘".'-
‘4: ~"'.;,'-'§:3..g;..-,_. tr “1'
I”; WW“ 3“} malt ‘
‘ r . 2" gig, ’v j 4&5 . . “'7‘
mafia fr? z :2»; av?”

b-

fun“
“WM,
'1

”35’:
:07. 1‘3”?"

., ‘w‘

C
(:17 :"5’3 -': 13.13

.

‘I ‘ V
‘ fi I
r Auto.

'3.
ml

\
X

.Sé‘kjfla

 

 

 

 

 

 

 

 

 

VERSITY LIBRARIES

III III III

 

 

 

 

 

 

 

 

 

IIII II IIIIIIIIIIIIIIIIII III

3 1293

This is to certify that the

thesis entitled
Acoustic Emission, Internal Friction and Young's
Modulus in a Mechanically Fatigued GFRP Composite

and Macor, a Glass-Ceramic Composite

presented by
Karl Andre Tebeau

has been accepted towards fulfillment
of the requirements for

Master's Materials Science

degree in

 

 

flaggaCé/w

Major professor

Date January 30, 1991

 

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

 

LIBRARY

M'Chlgan State
University

*

 

- 0 ~l r." 5")! o o
PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before due due.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

QiI—JI
I I‘d

—_II I

MSU le An Affirmative ActiorVEquei Opportunity Institution
«ammo-m

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ACOUSTIC EMISSION, INTERNAL FRICTION AND YOUNG'S
MODULUS IN A MECHANICALLY FATIGUED GFRP COMPOSITE
AND MACOR, A GLASS-CERAMIC COMPOSITE
BY

KARL ANDRE TEBEAU

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE

Department of Metallurgy, Mechanics

and Materials Science

March 1991

Abstract

ACOUSTIC EMISSION, INTERNAL FRICTION AND YOUNG'S
MODULUS IN A MECHANICALLY FATIGUED GFRP COMPOSITE
AND MACOR, A GLASS-CERAMIC COMPOSITE

By
Karl A. Tebeau

The low cycle mechanical fatigue behavior of two types of modern
composite materials is explored. The damage induced changes in Young's
modulus and internal friction are compared to acoustic emission (AB) detected
from the specimen during four point bending. Changes in Young's modulus
and internal friction are measured by using the sensitive sonic resonance
method. Extensive qualification of the AE data was undertaken, to determine
the validity of the AE data taken during the fatigue testing. .

The specimens produced substantial AE throughout the flexural cycle.
Since it was shown that the AE detected at the points of load reversal was
contaminated by system noise, only the AB detected during the midstroke was
determined to be fit for analysis. The AE produced during midstroke was
noncontinuous, with some parts of the load range being more acoustically
active that others, indicating that damage production was also noncontinuous.

Several trends were observed in the data. The maximum changes in the
two monitored physical properties increased with increasing maximum fatigue
load. Changes of Young's modulus and internal friction for both materials
evolved in an unexpected manner, increasing and decreasing periodically
throughout fatigue life. It is believed that a fast acting, aggressive recovery
mechanism may be at work, periodically set back by continued fatigue flexure.

I am pleased to dedicate this thesis, the culmination of my efforts at M.S.U. to
my parents: for their support - both moral and financial, their understanding,
perseverance and for not selling off my stuff while I finished.

Acknowledgements

I would like to express my thanks and appreciation to Dr. Eldon Case for his
guidance, support, constructive criticism, intuitive pushes during analysis of
the data, thoughtful commentary, and especially for driving me to drive myself
harder than ever before.

Thanks to Dr. Fred Pink for his help in getting me started with, and the use of
the acoustic emission computer.

Thanks to Dr. John Martin for his advice concerning fatigue experimentation
and interpretations.

Thanks are also in order to my student colleagues of the 'Case Group',
Youngman Kim for my introduction to PlotIt ® and FrarneMaker ®, help with
the literature and many suggestions regarding the behavior of the materials,
and to Carol Gamlen, Chin Chen Chiu, and Won Jae Lee for their friendship,
advice and moral support.

Thanks also go to Subrato Dhar for his hospitality during my time to finish,
Chul Soo Kim, Rajendra Vaidya, Moti Tayal, Jae C. Lee, Mandar Hingwe,
Mondher Cherif, Satyanarayana 'Satya' Kudapa, Mahesh, Seungho Nam, and
Jim Nokes for their friendship and advice, and especially to Kristin
Zimmerman for her help in reading the final stages of the draft.

Table of Contents

List of tables.

List of figures.

1. Introduction.
2. Review of literature.

2.1.

2.2.

2.3.

2.4.

Inspection and testing techniques.

2.1.1. Nondestructive testing methods.
2.1.l.a. Internal friction.
2.1.1.b. Dynamic Young's modulus.

2.1.1 .c. Damage versus change in internal friction and
Young s modulus.

2.1.2. Destructive testing methods.

Acoustic emission.

2.2.1 . Physical background.

2.2.2. Acoustic emission parameters.

2.2.3. Crack growth and AE rate relations.

Fatigue in GFRP composites.

2.3.1 . Stresses in the four point bend test.

2.3.2. Fatigue damage modes.

2.3.3. Fatigue life relations.

2.3.4. Fatigue life modulus concepts.
2.3.4.a. Hwang and Han fatigue modulus.
2.3.4.b. Secant modulus concepts.

2.3.5. Acoustic emission in fatigue of GFRP composites.

2.3.6. Co arison of AEfrom three types of fiber
re creed composrtes.

Mechanical fatigue in glass-ceramic composites.
2.4.1 . Fatigue damage in polycrystalline ceramics.

V

OQNNNNHN§°

14
14
21
24
28

30
34
38
38
43
46
50

51
52

2.4.2. Ceramic fatigue theory.

2.4.3 . Fatigue models for ceramics and ceramic composites.

2.4.3.a. MicrOcracking model for polycrystalline
cerarmcs.

2.4.3.b. Fiber buckling model for ceramic composites.

3. Experimental procedure.

3.1.
3.2.
3.3.
3.4.

3.5.

3.6.
3.7.

Materials.

Specimen preparation.

Overview of experimental procedure.
Four point bend technique.

3.4.1 . Equipment and fixture.

3.4.2. Mechanical fatigue method.
Acoustic emission technique.

3.5.1 . Acoustic emission detection.
3.5.2. Machine noise quantification.
Sonic resonance technique.

Specimen preparation for S.E.M. fractography.

4. Results and Discussion.

4.1.

4.2.

4.3.

Thought ex riment model for AB during monotonic
mcreasmg l: exure m GFRP composrte.

 

Determination of uncertainty of intemal friction and
Young's modulus measurements made on the some
resonance system.

Machine noise qualifications.

4.3.1. 22.3 Newton maximum fatigue load test.
4.3.2. Stepped maximum load fatigue test.

4.3.3. Kaiser effect test.

4.3.4. Comparison of AE from load pin and specimen.

4.3.5. Results of machine noise tests.

’54

55
55

59
62
62
62
63
65
65
67
68
68
7O
72
73
76
76

77

81
81
89
91
92
93

4.4. Results of fractography.
4.4.1 . GFRP fractography.
4.4.2. Macor fractography.
4.5. Analysis of fatigue data.
4.5.1 . GFRP-AB fatigue analysis.
4.5.1.a. GFRP-AB fatigue analysis without load scaling.
4.5.l.b. GFRP-AB fatigue analysis with load scaling.
4.5.1.c. Common AB trends in GFRP composite.
4.5.1.c.1. Load reversal effect.
4.5.1.c.2. Active zone - dead zone phenomena.
4.5.2. Macor fatigue analysis.
4.5.2.1. Macor AB analysis with load scaling.
4.5.2.1.a. Specimen MA-12.
4.5.2.1.b. Specimen MA-l l.

4.6. Intemal friction and Young's modulus change over
fatigue hfe.

4.6.1 . Internal friction and Young's modulus versus
load cycles m GFRP specrmen GE-l .

4.6.2. Internal friction and Young's modulus versus
load cycles 1n GFRP specrmen GB-3.

4.6.3. Internal friction and Young's modulus versus
load cycles in GFRP specrmen GB-4.

4.7. Peak analysis on plots of internal friction versus number
of load cycles for GFRP spccrrnens.

4.8 Maximum changes in Young's modulus versus load
fractron of rupture strength.

4.9. Sensitivity ratio of changes in Young's modulus and
mtemal friction.

4.10. Possible mechanisms to account for variations in values
of mtemal friction and Young's modulus.

93
93
94
100
100

102
104
104
108
111
111
112
114
114

116

119

122

122

127

129

132

4.10.1. Damage recovery mechanism.
4.10.2. Damage progression mechanism.
4.10.3. Test for long term recovery

4.10. Internal friction and Young' s modulus versus
acoustic emission.

4.10.1. Possible trends between AB and internal friction.

5. Conclusions.
5.1. AB in load ranges.

5.2. Evolution of damage with changes in Young' s modulus
and internal friction.

5.3. Increasing variation of Young' s modulus with maximum
fatigue load.

5.4. Evolution of physical properties with number of
fatigue cycles.

5.5. Afterview of experiment.
5 .6. Recommendations for future work.
6. Appendices.

6.A. Mechanical perties and. com sition of Macor,
Corning cocieriium ber 9658 p0

6.B. Mechanical properties of 3-M type 1003 GFRP.

7. References.

132
134
134
136

136
153
153
154

154

155

155
156
157
157

158
159

List Of Tables

1&1:

3.a.

3.b.

3.c.

3.d.

Acoustic wave velocities for several ceramic materials

Uncertainty data for sonic. resonance measurements of Young's
modulus and mtemal fnctron.

Maximum chan es of Young's modulus and internal friction
for GFRP and acor fatigue specrmens, measured at 0.12
normalized distance from end of bar.

Maximum charlr‘ges of Young's modulus and internal friction
for GFRP and acor fatigue specimens, measured at 0.075
normalized distance from end of bar.

Maximum charlr‘ges of Young's modulus and internal friction
for GFRP and acor fatrgue specrmens, measured at 0.32
normalized distance from end of bar.

Maximum chan es of Young's modulus and internal friction
for GFRP and acor fatrgue specrmens, measured at 0.32
normalized distance from end of bar.

Sensitivity ratios for all fatigue specimens.

Long term value of Young's modulus.

18
79

84

85

86

87

130
135

MPP’N!‘

10.
ll.

12.

13.

14.

15.a.
15.b.

16.
17.

List Of Figures

'Free - free' vibrational modes of prismatic bars.

Sample ring-down output from specimen in sonic resonance.

Schematic and static force profiles for three point bend test.

Schematic and static force profiles for four point bend test.

Schematic of crack advancement sequence in a ceramic
material.

Definitions for acoustic emission elastic stress wave
parameters.

Acoustic frequency ranges for several mechanical sources.

Transition of failure modes as a frmction of interface .
condition for GFRP composites under four pornt bendrng.

Schematic of Hahn and Tsai secant modulus during nill
fatigue load cycle.

Hwang and Han fatigue secant modulus.

Schematic of fatigue in polycrystalline ceramic under
tension-tension loading.

Schematic of fatigue in polycrystalline ceramic under
compression loa

Schematic for fiber pullout fatigue mechanism in a fiber
reinforced ceramic composite.

Bag:

22

.23
31

43

45
56

58

61

Schematic of the four point bend fixture used for experiment. 66

Photograph of fatigue - AB experimental station.

Close up photograph of the four point bend fixture used
for the experiment.

Block diagram of sonic resonance system.

Acoustic emission counts per load range for GFRP 22. 3 N

machine noise test.

69
69

74
83

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

Scanning electron micrographs of the mid plane shear crack

 

(specimen GE-4) common to all GFRP fatigue
specrmens.

Scannin electron micrographs of lpartially opened midplacrlre
ge

crack 0 specimen GE-4, clearly s owing the fiber bri
zone.

Scanning electron micrographs of the fracture surface of
Macor specimen MA-2.

Scanning electron micrographs cf the fracture surface of
Macor specrmen MA-6.

Scanning electron rrricrograph of fracture surface of
Macor specrmen MA-9.

Histogram distribution of rise time divided b duration
per acoustic event, for Maror specrmen MA- 2, zero to
00 fatrgue cycles.

Acoustic emission counts sorted b load ran e for; GFRP
s cimen GB-l, fatigue loaded wi 111.3 maxrmum

ynamic load.

Acoustic emission counts sorted b load ran e for GFRP
cimen GE-3, fatigue loaded wi 155.8 maximum

ynanuc load.

Acoustic emission counts sorted b load ran e for GFRP
s cimen GE-4, fatigue loaded wi 111.3 maximum

ynamic load.
Acoustic emission counts sorted by load range for Macor

s cimen MA-12, fati e loaded with maxrmum dynanuc
lggd of 86.2 N, 0.75 rigrgdulus of rupture.

Internal friction versus fatigue load cles for GFRP
specrmen GE-l, maxrmum dynamic oad of 111.3 N.

Young's modulus versus fatcilgue load cycles for GFRP
specrmen GE-l, maxrmum ynamic load of 111.3 N.

Internal friction versus fatigue load c cles for GFRP
specrmen GB-3, maxrmurn dynanuc oad of 155.8 N.

Young's modulus versus fatcilgue load cycles for GFRP
specrmen GE-3, maxrrnum ynanuc load of 155.8 N.

Internal friction versus fatigue load c cles for GFRP
specrmen GB-4, maxrrnum dynarmc oad of 200 N.

Young's modulus versus fatigue load cycles for GFRP
specrmen GB-4, maxrrnum dynamic load of 200 N.

xi

95

97

98

99

103

105

106

107

113

116

117

120

121

123

124

34.

35.

36 a,b.
37 a,b.

37 c,d.

37 e,f.

37 g,h.

37 i,j.

38 a,b.

38 c,d.

38 e,f.

38 g,h.

38 i,j.

39 a,b.

39 c,d.

39 e,f.

39 g,h.

39 i,j.

Data from peak analysis of intemal friction verus load
cycles for glass epoxy specrrnens GB-l , GB-3 and GB-4.

Maximum chan es in Young's modulus versus fraction of
rupture strength or Macor specrmens, measured at a norm-
ahzed distance of 0.075 from the free ends end of the bar.

Sensitivity ratio versus maximum fatigue load for all fatigue
specrmens.

Internal friction versus acoustic emission counts
accumulated over 500 load cycles, GFRP specrmen GB-l .

Internal friction versus acoustic emission counts
accumulated over 500 load cycles, GFRP specrmen GE—l .

Internal friction versus acoustic emission counts
accumulated over 500 load cycles, GFRP specimen GB-l .

Internal friction versus acuostic emission counts
accumulated over 500 load cycles, GFRP specrmen GB-l .

Internal friction versus acoustic amission counts
accumulated over 500 load cycles, GFRP specimen GB-l .

Internal friction versus acoustic emission cormts.
accumulated over 1000 load cycles, GFRP specrmen GE-3.

Internal friction versus acoustic emission counts.
accumulated over 1000 load cycles, GFRP specrmen GE-3.

Internal friction versus acoustic emission counts.
accumulated over 1000 load cycles, GFRP specrmen GE-3.

Internal friction versus acoustic emission counts.
accumulated over 1000 load cycles, GFRP specrmen GE-3.

Internal friction versus acoustic emission counts
accumulated over 1000 load cycles, GFRP specimen GB-3.

Internal friction versus acoustic emission counts
accumulated over 1000 load cycles, GFRP specimen GB-4.

Internal friction versus acoustic emission counts
accumulated over 1000 load cycles, GFRP specimen GB-4.

Intemal friction versus acoustic emission counts
accumulated over 1000 load cycles, GFRP specimen GB-4.

Internal friction versus acoustic emission counts
accumulated over 1000 load cycles, GFRP specimen GB-4.

Intemal friction versus acoustic emission counts

accumulated over 1000 load cycles, GFRP specimen GE-3.

xii

125

128

131

A 137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

1. Introduction.

Mechanical fatigue is an important damage source in both polymeric and
ceramic composites. This experimentation explores the low cycle mechanical
fatigue behavior of two very different types of modem engineering composite
materials: 3 - M composition type 1003, a unidirectionally reinforced glass
fiber - epoxy composite and Macor, a glass - ceramic composite.

A comprehensive literature search and review covers the topics necessary
to provide a solid basis with which to evaluate and explain the behavior of the
materials under testing. A review of testing methods includes four point bend
stressing and sonic resonance determinations of Young's modulus and internal
friction. Acoustic emission is examined at length, including terminology,
damage emission and analysis methods. Fatigue processes in both materials
are extensively examined, to include relevant mathematical models and recent
qualitative mechanical models. _

Experimentally, a low cycle fatigue schedule was followed which
systematically increasd the maximum applied load per specimen until failure
regularly occured. Changes in Young's modulus and intemal friction were
monitored to characterize the evolution of damage as it accumulated under
continued fatigue loading. Acoustic emission was simultaneously recorded
and later compared with the changes in the physical properties.

Comparison to current work by colleagues and previous studies
involving thermal shock may enable us to draw an estimate of internal damage
and perhaps an equivalent thermal shock fatigue.

2. Review of literature.
2.1. Inspection and testing techniques.
2.1.1. Nondestructive test methods.

Nondestructive evaluation methods are available that can measure small
changes in physical properties. When used in conjunction with other tests,
the sonic resonance technique for determining the internal friction and dynamic
Young's modulus provides an estimate of a specimen's damage, crack growth
resistance, remaining strength and fracture energy [10-14].

2.1.l.a. Internal friction.

Any vibrating solid structure will consume its own vibrational kinetic
energy and eventually come to rest, even though it may be totally isolated from
its environment [14]. This energy loss effect is known as internal friction.
Internal friction characterizes an energy absorbing (or dissipating) mechanism
that provides displacement damping as an exponential 'ring-down' of the
specimen's vibration. Internal friction arises due to microcracks, grain
boundaries, second phases, voids, pores and other smaller contributions in the
host material [14].

Several methods, direct and indirect, exist for the measuring of intemal
friction [IO-21]. Specific loss or damping capacity is the most direct method,
described theoretically by Kolsky [18], as the ratio

AW/ W = Damping Capacity (1)

where

AW = energy loss in the specimen during one stress cycle
W = elastic strain energy stored in the specimen at
maximum strain.

Indirect methods for determining internal fiiction rely on the assumptions
that frictional or energy absorbing mechanisms are proportional to velocity.
This implies that the mechanical damping is of viscous nature, and that the
restoring forces are proportional to displacement from the neutral position,
further implying Hookean behavior [10-14].

A qualitative measure of internal friction is the relative spread or sharpness
of resonance [14]. A plot of the specimen's vibrational amplitude versus
driving frequency (at constant RMS amplitude), shows a local maximum at Fr,

the fundamental flexural resonant frequency [10, 11, 14, 17-20]. Figure 1
illustrates some vibrational modes of prismatic bars. Other resonant
frequencies also show local maxima, but the fundamental flexural mode of
vibration is the most easily detect'éd.

The sharpness of the peak is inversely proportional to the internal fiiction
[14]. A specimen with a sharper resonance than another can be said to have a
lower internal friction. Relative sharpness can be easily measured by
determining the spread of the frequency response at some chosen reference
level, usually the half power' point, i.e. 0.707, or 0.5 maximum amplitude.
This is known as the ‘full width at half maximum' method of peak analysis,
and is normalized to the resonant frequency [14]

Q '1 ~ AF/Fr (2)
where

AF = peak width in frequency, at chosen point of reference.

 

 

a) - %

 

 

 

 

13)
"‘\ I “ A
{\x/ "\‘\z”x"‘\ \/ I)
9 Dr Y xx x X1
( ’\ \—/I / \ \§./’ / 3

 

d

 

1
Lu]

Figure l. ‘Free-free’ vibrational modes of prismatic bars. a) quiescent,
b) flexural fundamental, c) first flcxural harmonic or overtone,
d) torsional fundamental.

Zener [14, 19] shows that this ratio is related to specific loss (Equation 1) as ”

AF / Fr = («13 / [2 n1) (AW/W). (3)
A typical definition of internal friction that is used today is the quantity

Q '1, derived by Kolsky [14, 18] from analogy with electrical theory, and is
defined as

l/Q = o-1 = AF/N/3Fr. (4)

For low internal friction (less than 102) measured at frequencies above
100 Hz [14], it is expedient to use Forster's free-decay method [10-16, 20].
Forster's relationship is written as

Q'1= 10(Ao/A¢)/1tft (5)

where
= intemal friction of specimen

driven vibrational amplitude of specimen
vibrational amplitude at threshold

driven frequency, usually fundamental flexural
time between driver cutoff and last output cycle to

cross threshold
number of times output crosses threshold.

Q-l
A0
A1
f

t

ft=N

Figure 2 illustrates a typical ring-down output from a driven specimen
after cutoff.

Dr iiiiiiiii

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2.1.1.b. Dynamic Young's modulus.

Dynamic Young's modulus can be calculated in several ways. One
method, used by some researchers [13, 14] is as follows

Ed = (9.467 x 10‘7)L3mf2 /d3b (6)

where, in cgs units

Ed = dynamic Young's modulus (dyne/cmz)
L = length of specimen (cm)
m = mass of specimen (g)

f = fundamental flexural frequency (Hz)

d = thickness of specimen (cm)

b = width of specimen (cm).

The constant premultiplier is a unitless shape factor.

A second and more widely used approach to dynamic Young's modulus
is derived from a relationship developed by Lord Rayliegh [21], and is of the
form [13, 14, 21]

Ed = (48 7:2 p L4 F,2)/(m4 d4) (7)
where, in cgs units

Ed = dynamic elastic modulus of the bar (dyne/cmz)

p = bulk density ofbar (g/cm3)

L = length of bar (cm)

Fr = fundamental flexural frequency of bar (Hz)
m = mass of bar (g)

d = thickness of bar (cm)

Note : 100 dyne/cm2 = 1 MPa.

In Equation 7, Lord Rayliegh made several simplifying approximations.
Most importantly, that the motion of the bar's elements perpendicular to its
length in the plane of flexure is assumed to be the only significant contribution
to the kinetic energy of the vibrating bar [14, 21].

Dynamic Young's modulus is typically about 0.1 to 5 percent greater than
the static Young's modulus. This phenomena is due to thermodynamic '
effects during measurement. In general, the difference amounts to adiabatic
(dynamic) versus isothermal (quasi-static) methods of measurement, i.e. sonic
resonance versus slow pull-test, respectively [14, 27]. In the dynamic test,
heat builds up during compression, especially near the free surface (where
maximum compression occurs). Compression at the fiee surface of the
specimen leads to heating, and the heat build up causes the lattice to expand
thermally. This effect negates, to a small degree, the effort of the driving
transducer to displace or vibrate the bar. The lattice expansion gradient results
in an apparent increase in stiffness, which is read by the pickup transducer as
an increase in the flexural fundamental frequency. The increased frequency is
interpreted through Equation 7 as an increase in Young's modulus [27 , 28].

2.1.1.c. Damage versus change in internal friction and
Young's modulus.

A change in internal friction or Young's modulus can be interpreted as a
measure of internal damage in a specimen [14]. Any change in internal
friction or Young's modulus during mechanical flexural fatigue is expected to
be due primarily to an increase in microcracking and void formation. This
damage will be developed and accumulated by cyclic mechanical loading in
four-point bend.

Internal friction is expected to increase as internal damage increases, and
thus track the evolution of damage, when plotted against the number of fatigue
cycles.

2.1.2 Destructive test methods.

Flexural strength or modulus of rupture (M.O.R.), is defined by the
American Society for the Testing of Materials as "the maximum stress in a
mode of flexure that a specimen develops at rupture; normally, the calculated

maximum longitudinal tensile stress at mid-point of the specimen test span
surface" [23-25].

Summary of methods:
Method 1: 3-point bend. The bar rests on two cylindrical support pins
and is loaded by means of a loading pin midway between the lower loading

pins. All. load pins should be identical and cylindrical to avoid specimen
indentation and to minimize stress concentration at the points of loading.

Method 11: 4-point bend. The bar rests on two cylindrical support pins

10

and is loaded at the two quarter points by means of two loading pins, each an
equal distance from the adjacent support point. The distance between the
loading pins is one half that of the support pin span.

For three- and four-point bend fixtures, the load bearing contact points
should all be identical and have cylindrical edges to minimize indentation of the
specimen, or failure due to stress concentration directly under the loading pins
may occur (Figures 3 and 4). Load bearing pill diameter should be a
minimum of 3.175 m (1/8 inch) and no greater than the specirnen's
thickness.

Recommended geometry specifications are [23-25]

L/d 16 Requires 20 2 L/d 2 2 i.e. support span should
be sixteen times the specimen thickness.

Ub = 4 Requires L/b 2 0.8 i.e. support span should be

four times the width.
b/d = 4 Requires b/d z 1 i.e. width should be four
times the thickness
where
L = support span, m or in
d = thickness of specimen, m or in
b = width of specimen, m or in.

Flexural strength in four point bend is calculated by the following equation
[23 -25]

o = 3l>1./4bd2 (8)

11

 

 

 

 

 

 

 

 

P/2 P/Z

 

 

 

 

 

 

I X3
C) ;
i all 613

Figure 3. Schematic and static force profiles for three-point bending test.
a) dimensional parameters, b) shear force (V) and bending moment (M)
distributions along X1 axis, c) flexural stress (0'11) and shear stress (0'13)
distributions in the thickness direction.

 

12

 

X3

1’)? P/2 X1

 

 

 

 

 

 

 

 

 

 

“I / +
Ixs _ Ixs

c
) ; —.O' 11 —' 0'13

Figure 4. Schematic and static force profiles for four-point bending test.
a) dimensional parameters, b) shear force (V) and bending moment (M)
distributions along X1 axis, c) flexural stress (on) and shear stress (013)
distributions in the thickness direction.

 

 

13

where

o = flexural strength, MPa or psi.
load at fiacture, N or lbf
L = distance between long span of supporting fixtures

(roller pins), in or in.

"0
II

b = width of specimen, m or in.
d = thickness of specimen, m or in.

Note: ifP is read in kgf, 6 will be calculated in kgf/rnmz. To obtain 0 in

MPa with P in kgf, multiply the right side of Equation 8 by 9.807. If 0 is
desired in ksi and P is recorded in lbf, divide the right side by 1000 [23-25].
The mean flexural strength value of the test lot should be calculated as

0: (ol+oz+...+on)/n (9)

where

Q
ll

mean value of flexural strengths for the test lot

01. . . o = individual specimen flexural strengths

n = number of specimens.
The sample standard deviation should be calculated as follows
5(n-1) = {[(ol - o)2 + (02 - o)2 + . . .+ (on - o)2] /(n - IN“2 (10)

where

S(n-1) = sample standard deviation, MPa or psi.

2.2. Acoustic emission background.
2.2.1. Physical background.

Acoustic emission (AB) analysis has developed over the past twenty years
as an important NDE tool. AB monitors and interprets the elastic stress waves
generated by the swift, local redistributions of stress associated with many
damage mechanisms.

AB methods may be preferable to the other NDT methods of transmission
X-radiography and ultrasonic C-scan, by virtue of real time analysis and the
ability to locate damage. Only after significant damage is detected
acoustically, in real time, may it be necessary to remove the specimen from the
test rig for a more detailed inspection. Multiple AB transducers can in some
cases, accurately determine the location of damage by time-of-flight
computations.

Acoustic signals are generated within a solid when a defect-containing
solid is mechanically stressed [5, 39-45], undergoes some phase change or
when a defect is created within an otherwise homogeneous solid [3 8-40] .

The spectral amplitude distribution of the acoustic signal depends on the
magnitude and character of the defect [2945]. Therefore, several intensity
thresholds or ranges may exist, depending on the defect structure and their
ability to generate AE within the solid [29-34, 36-40, 42, 44, 45].

Dislocation motion in metals exhibits the lowest theoretical level of
acoustic emission intensity, but with millions of dislocations sweeping
through a volume of material simultaneously, their total acoustic energy may
be detected as strain rate dependent [42]. In 1980, Dickinson showed that
what many people had previously interpreted as dislocation induced acoustic
emission, resulted from an oxide coating that was cracking and separating
[45]. Dickinson tested in tension two types of electropolished specimens

14

15

made of a 1350 aluminum alloy in a high vacuum (10'4 Pa). The specimens
were nominally identical, but some had the addition of a 30 m2, 300nm thick
patch of anodized oxide coating on the gauge section. Dickinson showed that
the AB from the patch area of the anodized specimen was over four orders of
magnitude greater than the AB from the clean specimen, which gave nearly
negligible AB above the background noise of the system [45] .

Other common examples of acoustic emission include the well known 'tin
cry', associated with Type I deformation twinning in tin single crystals under
tensile stress, which can sometimes be heard with the unaided car [42, 46].
Detectable levels of acoustic energy are also produced during the onset of
phase transformations such as retained austenite to martensite in HSLA steels
[40] and piezoelectric response and poling effects in PZT (lead zirconate
titanate) [38, 39]. A moving defect such as a crack opening or extending
[29-34, 36-45] may also generate acoustic emission. This is a manifestation
of the Kaiser effect [31, 46], which describes the release of acoustic emission
only when the threshold of the previous highest load is exceeded.

Under flexural loading, the surfaces of the crack grate against each other
(this is not crack growth, but mechanical contact of opposing crack faces).
Frictional asperity contact with grinding and pulverization of debris within the
crack may prevent closure [30, 34]. Crack rubbing may generate a level or
type of AB (different from that of cracking AB), which is continuously emitted
during flexure. Crack rubbing may allow the Kaiser effect to separate 'new
damage' AB from continuously-generated fiictional AB, if an unambiguous
acoustic signature for cracking can be determined.

When an internal rnicrocrack initiates or extends, part of the material is
fracturing and creating internal space. Crack extension may occur rapidly
with new intemal surface areas swept out by the advancing crack front
Alternatively, slow crack growth may fracture and separate a few grains at a
time. Both rapid cracking and slow crack growth can fiacture transgranularly

16

(along grain boundaries) or intragranularly (through the grain) [29-37].

During slow crack growth, intermittent bursts of AB energy are released
which may be detectable above the background noise of the detection system
[30-33]. A slow crack propagating through a matrix with dispersed second
phase particles may be modeled roughly by a dislocation in a metal with harder
second phases present. The advancing crack front is arrested by two nearby
second phase particles. Under continued loading, the crack tip stress
intensity rises and the crack front bows out between the two particles [31]
(Figure 5). Finally, the crack front either breaks around the particle, or
overcomes the strength of the binding particle which then fractures, and the
crack front advances to the next set of pinning points [31].

The minimum time required for a microcrack to extend only a few grains
is on the order of a few nanoseconds [31], but the acoustic signal is generated
for a much longer time due to vibration or ringing of the crack faces,
analogous to the ringing of a bell for a time after it is struck. The Rayliegh
velocity is the upper speed limit for crack propagation in a solid [31]. Table 1
lists wave velocities for several ceramic materials. Bmpirically, crack
velocities are often much slower [30, 34].

The frequency distribution for an initially sharp pulse broadens as a
function of propagation distance [35]. Also, higher frequencies attenuate
more with distance than lower frequencies [35].

Elastic stress waves are an inseparable mixture of several different wave
types. The wave types of relatively large amplitudes (highest energy) are the
longitudinal (compression) and transverse (shear) waves[41]. These two
wave types accompany each other but travel by different modes and at
different velocities [35, 38, 41]. Shear waves travel at about half the
compression wave velocity [38, 41] (Table 1). In practice, the amplitude of
the stress wave being sensed depends on where the AB transducer is located
and the distance it is from the source. Since AB is sensed on the suface of a

17

Direction of crack propagation

 

   

 

 

 

   

 

a)
b)
C)
® 0— Fractured
d) Particles

 

 

 

Figure 5. Schematic of crack advancement sequence in a ceramic composite
material. a) crack approaches hard second phase particles, b) crack is arrested
by particles, c) crack front bows out past particles, d) crack front breaks free
of pinning particles, advancing to next set of pinning particles.

18

Table 1. Acoustic wave velocities for several ceramic materials [38, 41, 44].

 

 

 

 

 

 

 

Material Wave type Wave velocity

Si3N4 compression 6 mrn/usec
shear 3 mm/usec
Rayliegh 4 qusec

A1203 * compression 11.0 mm/tlsec
shear 6.7 mm/tisec

Si02 ** compression 6 mm/usec
shear 3.8 mm/tlsec

PZT *** compression 3.9 to 5.0 mm/ttsec
shear 1.8 to 2.1 mm/ttsec

 

 

 

* Average of Voight and Reuss values.

** Values cited are for fused silica. Porosity will decrease wave
velocity, up to an order of magnitude for a highly porous specimen.

*** Lead Zirconate Titanate.

19

specimen, Rayliegh elastic surface waves, are important [27, 41, 44].

A timed piezoelectric transducer coupled to the surface may sense these elastic
stress waves as acoustic emission. The acoustic energy is transmitted by
atomic displacements.

Several other wave types are also present. Stonely waves travel along
interfaces such as grain boundaries and fiber/matrix boundaries [41], where
differences in elastic modulii exist across an interface. Lamb waves [41, 44] ,
plate and leaky plate waves [41] become important in specimens thin enough
to allow communication between opposite surfaces. Lamb and Stonely waves
typically have much lower amplitudes than longitudinal (compressive) or
transverse (shear) waves [41, 44].

The interaction of elastic stress waves is an extremely complicated
situation and is beyond the scope of this discussion. As a rninummn
comment, it is sufficient to say that elastic stress waves may interact with each
other and every inhomogeniety that they pass across, including, but not
limited to, specimen geometry, free surfaces, other cracks, grain boundaries,
other phases, voids and pores, inclusions, and smaller defects [29-41].
Elastic waves can be reflected, refracted, echoed, attenuated, diffracted,
transmitted and transformed into components of other wave types [41], always
seeking to minimize the specimen's potential energy.

Frequency attenuation and scattering has been shown to be sensitive to
the primary constituents of the microstructure, i.e. grain boundaries, second
phases, voids and pores [35]. Evans et al. showed that ultrasonic attenuation
in ceramics increases nearly linearly with frequency. Attenuations of 1 to 20
dB/cm are typical for MgO, RBSN, PZT and erS in the low MHz range [35].

For sintered and hot pressed SiC and hot pressed Si3N4 (a and B),

attenuations of 10 to 50 dB/crn were found for frequencies in the 100 to 400
MHz range [35]. An elastic stress wave is attenuated most when its

20

wavelength is on the order of the size of the scattering center or scattering
center separation distance [35].

As a given stress wave encounters a scattering center, it may emerge as a
wave or collection of wave components of totally diverse character [41]. This
effect means that the interpretation of AE signatures is very difficult. The AE
analysis software cannot recognize and differentiate the different waves that
stimulate the piezoelectric transducer. The system also cannot distinguish an
echo from the original pulse. Thus the 'envelope' that the software responds
to can be modified by echo pulses. Echo pulses are attenuated with respect to
the original pulse [31], but transmission distances vary with the location of the
site of the reflection. The presence of echoes may inflate the number of
counts received, potentially leading to an overestimation of the extent of
microcracking. Also, a finite dead-time for 'envelope reset' exists between
the times the system is able to receive a pulse. Dead time may reduce the
number of counts recieved, but this does not necessarily nullify the effects of
pulse reflection. .

Being aware of this potential misinformation is only the first step to
realizing an accurate interpretation of the crack/AB relation. Time of flight
data may be obtained from a transducer array to provide a distribution of
cracking activity across the specimen. Still, the echo problem is present and
must be treated with care. Echoes are attenuated to some degree as shown by
Evans [35], but the attenuation may not be enough to allow us to set an
amplitude threshold above the echo, and below the original A-E burst. The
difficulty comes when trying to distinguish between a nearby echo and a
distant pulse due to new damage. Experimental procedure will determine the
viability of the use of a threshold as a screen to separate new damage from
frictional emission. The business of interpretation of AB signals is certainly
challenging, but is still of worthwhile interest and pursuit.

2.2.2. Acoustic emission parameters.

There are several approaches to analyzing acoustic signals. In the
literature, the AE count rate is the most frequently used method. The AB
count rate is measured by the number of times the RMS voltage amplitude, or
intensity, crosses a voltage threshold. The threshold is set somewhat
subjectively or empirically by the researcher.

The voltage threshold is usually chosen as low as possible, to give the
highest degree of sensitivity without significantly compromising the
signal-to-noise ratio. A low amplitude threshold reads more of the lower
intensity AB events. If a certain level and higher AE activity is of interest, the
threshold setting may be adjusted to screen out the lower intensity events if
they detract from the desired data, as may be the case for frictional 'AE.

To avoid background noise, some threshold above that background noise
must be selected, but the threshold must remain low enough to not exclude the
expected AB signal [29-40]. Clearly, the choice of threshold level determines
the number of counts received during the AB burst envelope (Figure 6).
Background and machine noise must be known to enable one to know with
confidence, that the data being received is an actual specimen signal and not
extraneous acoustic activity. Researchers that report system gain have overall
system amplification between 80 and 130 dB [31, 33, 34, 45]. Often the
equipment used and control parameters and settings are not specified in detail
in the literature.

The exclusion of non-data is not as difficult as one may expect. Machine
noise is usually in the low kilohertz range [29]. Cracking in ceramics is
usually spread across the zero to 10 MHz range [32], with significant
components in the megahertz range (Figure 7).

The time resolution required for our AE equipment is on the order of 0.1
microsecond [42]. In 0.1 microsecond, only one event (though one is

21

22

 

Event Duration ___.,

Rise
Time

 

 

 

 

 

 

 

Peak
Voltage
Amplitude A
A I \ A Threshold
Reference V
Voltage

 

 

 

 

 

 

 

ichIIIIIIIIIIIII

Figure 6. Definitions for acoustic emission elastic stress wave parameters.

23

 

 

 

 

 

 

 

 

Mechanical failures
4 >
e. g. damaged bearings.
Structural dynamics leaks in pipelines,
f sersnnc disturbances
Ultrasonic NDT
of metals
<———>
Response of
.AE transducer
used for experiment
<—>
Rotor dynamics
< > o o 0
Acoustic emrssron from
growing cracks in
metals and ceramics
4 b
l l l I l l l
0.1 1.0 10 100 1K 10K 100x 1M 10M

Figure 7. Acoustic frequency ranges for several mechanical sources.

24

unlikely), or thousands of events may occur. Continuous emission can keep
an enve10pe open due to reflections and echoes.

Awerbuch et.al. presented a method of acoustic emission analysis
useful for correlating certain aspects of the AB with other measures of damage
[59-61]. The intensity of the AE signal, as measured by counts, rise time or
duration, may be screened by setting a lower threshold of counts, rise time or
duration. The AB events exceeding the threshold are then analyzed. By
knowing the number of fatigue cycles at which the AB was taken, we may plot
the AB parameter on the ordinate and the number of fatigue cycles on the
abcissa. The choice of an AB threshold depends on the damage that the
specimen has undergone.

2.2.3. Crack growth and acoustic emission rate relations.
Researchers report conflicting relations between emission rate and applied
stress intensity about the crack tip. The relation widely used as a starting
point for AB rate equations is [30, 31]
dN/dt ~ cKIn (11)

where

c = constant scaling factor
KI = stress intensity at the crack tip

11 = stress intensity exponent, varies with material.

The stress intensity exponent, n, can be as small as 2 or 5 for alumina,
and as large as 50 to 100 for the tougher ceramics such as SiC and Si3N4.

25

This range is attributable to a fimdarnental difference in the materials, and is
somewhat dependent on temperature and testing technique.
Crack growth rate may be approximated as

da/dt chi (12)

where

c constant scaling factor

KI = stress intensity for mode I Opening

11 stress intensity exponent.

The stress intensity exponent, n, has also been shown to vary with
temperature from 5 to 10 at room temperature, up to about 100 at

1300 °C for Lucalox [30]. To a first approximation, AB is related to crack
growth by combining Equations 11 and 12, which yields [30]
dN / da ~ q K12 (13)

where

dN / da = acoustic emission rate per unit crack area
KI = stress intensity for mode I opening

q = rnicrocrack density.

Evans and Linzer [31] develop an AB rate relation for slow crack growth

log (dN/dt) = (m - 1) log (1'a + (1 - m/r) log 0 + D' (14)

where

26

AB rate incounts perlmit time
particle strength distribution parameter
slow crack growth parameter

applied stress

stress rate
fracture site availability parameter.

The real time sequence of the acoustic emission begins when the stress
intensity exceeds a critical local threshold, advancing the crack. The elastic
strain energy stored in the crack tip zone then creates new surfaces, generating
elastic lattice vibrations as the crack front advances. The amount of strain

energy consumed by cracking has been modeled as [31]

U = [(no2a21) [E] + Uo (15)
where
Uo = strain energy of the uncracked body
0 = stress on body
a = crack surface created
1 = length of crack
E = Young's modulus of body.

The crack front advances under a far field tensile stress at an

exponentially decreasing rate [31 , 33-36] . Once sufficient strain energy has

27

been released, the crack tip process zone is able to elastically, and to a very
small degree plastically arrest the crack front. Acoustic emission is generated
during the cracking and for a time afterward as the crack faces vibrate. The
generation of AB by crack face vibration not only applies to slow crack
growth, but also for cyclic or rapidly applied loads. In most ceramics there is
very little or no microplastic deformation, therefore, cracking relieves much of
the stress. Toughening mechanisms, such as crack branching and crack tip
shielding, dissipate the strain energy with much less macrocracking [14, 31].
The acoustic emission associated with an event is proportional to the
strain energy released during that event [31]. Evans and Linzer developed a
relation based on strain energy in which AB count rate depends principally on
crack velocity, and is relatively insensitive to stress intensity or crack size [31]

dN/dt .. 3x105vlog(4x10‘5KI ‘Il/k) (16)
where
dN / dt = AB rate
v = crack front velocity
1 = length of crack front
k = crack front participation ratio, i.e. the number of active

cracks with respect to the total number of cracks.

These relations must be empirically calibrated. In practice, internal
rnicrocracks are not detectable during measurement, nor is their number
density or size known. We may be able to detect the acoustic emission and
determine the changes in internal friction and Young's modulus, with perhaps
a correlation between them without speaking directly to crack growth rate.
Crack growth rate (using a product of number and size) can be approximated
via a derived acoustic emission rate.

2.3. Fatigue in unidirectionally reinforced glass fiber -
epoxy matrix composites.

In-service flexural fatigue damage is common for engineered composite
structures. Flexural fatigue in glass fiber - epoxy matrix composites occurs
by the incremental accumulation of cracking damage under a random or
(usually) cyclic load profile. Fatigue damage results from three distinct
mechanisms which compete with one another for the demise of the specimen
[47]. The damage mechanisms are a flmction of the material properties (i.e.
stiffness, elastic modulus etc.), as well as fiber volume fraction, specimen
geometry, interfacial strength, test environment and pre-test treatments.

Mechanical damage mechanisms are (1) flexural tensile failure, which
causes transverse matrix cracking with fiber pullout and bridging, (2) flexural
compressive failure, which causes fiber buckling and (3) longitudinal matrix
splitting on the compressive side and interlaminar shear failure, in which the
fiber and matrix debond along the specimen's transverse-axial midplane.

F

2.3.1. Stresses in the four point bend test.

In four point bend loading, the specimen experiences a maximum flexural
tensile stress on the outer face of curvature and maximum flexural
compressive stress on the inner face of curvature, between the short span of
load - support pins. The maximum shear stress occurs along the mid-plane of
the specimen between the outer and inner load-support pins (Figure 4, page
11) [23-25, 47-49].

In the load span, the maximum flexural stress is given by [47-49]

0 = (3PS) / (4bd2) (17)

max

28

29

where, for Equations 17, 18 and 19

Omax = maximum flexural tensile or compressive stress (Pa)
Tm = maximum flexural shear stress (Pa)

P = total load applied (N)

S = span between load pins (m)

b = width of test specimen (m)

d = thickness (height) of test specimen (m).

Maximum shear stress is given by [47-49]

Tmax = (3P)/(4bd). , (18)

Equations 17 and 18 can be combined to yield [47-49]

Om = (48 Id) 13m. (19)

All three failure mechanisms operate on the test specimen simultaneously
(Equation 19). It is important to emphasize that different locations on the
specimen experience the three different maximum stresses. A test specimen

with a 4S / d value less than oi / ti favors shear failure and is known as the

'short - beam' test [47, 48].
The event of fracture may be conceptualized as failure of the weakest link
in a chain [48]. Therefore, fabrication parameters may be tailored to further

30

improve the performance of the composite. Also, the span to thickness ratio '
may be adjusted by varying the test fixture span to give either the flexural or
the short beam test.

2.3.2. Fatigue damage modes.

Shih and Ebert note that increasing the composite's fiber volume fraction
causes the failure mode to shift from flexural to shear. Both flexural and
shear strengths shift to a condition which encourages failure by shear mode
[47-49]. The shift of damage modes from flexural to shear occurs for three
reasons: (1) the availability of more fibers to bear flexural stress, reducing the
likelihood of flexural failure, (2) more fiber / matrix interfaces and (3) less
matrix material present.

Shih and Ebert tested the effect of interfacial strength ill several
combinations of fiber / matrix and coupling agents [47] . The fiber / matrix
interface was degraded by boiling the specimen in water for various time
durations, then breaking the specimen without prior fatiguing [47].
Modifying the interface 'quality', changed the failure mode from purely
flexural tensile failure in the undegraded condition, to flexural compressive
failure in the partially degraded condition. Shear failure occured for the most
degraded - longest boiling time exposure (Figure 8) [47].

In the flexural tensile failure mode, fiber ridging (not bridging of crack
wake) occured prior to matrix cracking. Fiber ridging occured on the tensile
face of flexure, as the matrix underwent Poisson contraction trying to pull in
between the fibers. Ridging gives the fibers the appearance of bulging out
from the surface of the specimen, under a thin skin of matrix material. Fiber
ridging is attributed to relaxation of residual stresses in the matrix material after
fiber / matrix interface debonding [47]. Fiber ridging was independent of the
type of coupling agent (fiber sizing) used [47], as long as failure occured in

31

 

 

 

 

 

 

 

 

 

 

 

 

0
g \
a *3
5 S
’5 0
3° :3 '2 3:; ‘25
g F a a
ca) ’5 o \ 0 E
'3 \ ct-t
K 9.. E “>9
o o .... :3
E .2 5 3 a
t: :2 :.-.1 2 m
a 2 s. a\
e is“ “:3 a
< 8 E
h 2
a a
c:
t2 8
Interfacial strength
Shear strength
Duration of boiling water immersion
(different scale for oomposrtes
wrth different couphng agents)
0—0
As molded (i.e. ) test ieces
with or with cougang aggnts

Figure 8. Transition of failure modes as a function of interface condition
for GFRP composites under four point bending [47].

32

the flexural tensile mode. Failure modes were observed in post test optical
and scanning electron micrographs.

In addition to fiber ridging, as the load is increased, transverse matrix
cracks formed between two or more fibers at a time. The transverse matrix
cracks linked up via short longitudinal cracks along the fiber / matrix interface,
as observed in scanning electron micrographs [47]. As the loading
proceeded, some fibers pulled out and were broken. The remaining nearby
unbroken fibers carried the load until they in turn were broken. Eventually,
the matrix and all the fibers were fractured. Under flexural tensile failure,
fiber / matrix bond strength is critical. Higher bond strength reduces fiber
ridging, but enables the matrix crack to penetrate the fiber ahead of the
advancing crack tip.

In the flexural compressive failure mode, similar fiber ridging occured
along with micro-buckling of the fibers. Shih and Ebert observed matrix
spalling over the ridged fibers, indicating fiber / matrix debonding, but
reported no matrix microcracking in the vicinity [47] . The failure mode was
also independent of the type of fiber sizing agent used. As seen in post test
S.E.M. micrographs, microbuckled fibers remained buckled after load
relaxation, indicating a permanent displacement between fiber plies [47].
Permanent buckling of a fiber implies that the fiber has been partially pulled
out, and resisted being forced back into its matrix hole upon stress relaxation.
Again, a stronger interfacial bond reduces micro buckling and fiber ridging,
but with the same tradeoff as for flexural tension.

In the shear failure mode, the shear crack propagates in a plane normal to
the applied force, splitting the composite through its mid - plane. A strong
interface increases shear strength [47] , which is characterized by a large
proportion of matrix material remaining bonded to the fibers [44]. These
markings are referred to as hackle' marks [47, 50, 51].

Shih and Ebert found one combination of fiber / matrix and coupling

33

agent that suffered very little interface degredation in boiling water. This
specimen was an Owens - Corning Fiberglas Corp. E - glass fiber and

Ciba - Geigy XU235 epoxy matrix with XU205 aromatic diamine hardener in
a weight ratio of 100:52, and coupling agent or fiber sizing AAPS (Dow -
Corning silane Z-6020) [47].

Fiber pullout and bridging increases interlaminar fracture toughness as
delamination grows. Sufficient fiber bridging can create a 'tied zone' [51],
which increases fracture toughness and critical load. The tied zone is created
when a shear or interlaminar crack propagates through the matrix and opposite
ends of a fiber (or group of fibers) remain embedded in opposing faces of the
crack.

Fiber nesting occurs when fibers are unevenly distributed throughout the
matrix, with clumps of fibers that have very little matrix material within the
clump. Fiber nesting occurs routinely in unidirectional layups in which fibers
migrate in the autoclave temperatme - pressure cycle [52]. Fiber nesting can
increase crack growth resistance [50]. Migration of the fibers during
autoclaving often twists the fiber bundle, such that the ends of some fibers end
up on opposite sides of the bundle.

Shear cracks do not propagate exactly down the center of the specimen.
Thus, the fracture surface is far from a perfect plane [47, 48, 54]. The instant
at which fiber bridging initiates can vary. Crack bridging by fibers may not
occur at all or bridging may initiate at the same instant the crack begins to
propagate [51]. Single fibers are usually broken, whereas bundles of fibers
tend to share and bear the stress and often peel out from the opposing crack
faces, thus creating the fiber bridged 'tied zone'.

Fractography indicates the complex nature of failure modes of
unidirectionally reinforced glass fiber/epoxy matrix composites under mode I
loading [47, 48, 51, 54]. Smooth and hackled grooves can show the
variation of interface bond strength and/or failure mode. The hackled grooves

34

suggest a strong interfacial bond and/or the stripping of the exposed fiber after
delamination. The smooth grooves suggest either low interfacial strength
and/or fiber pullout ahead of the crack front [47, 48, 51].

The 'tied zone' increases fracture energy. Under constant amplitude
cyclic displacement the crack growth rate decreases, due to a load reduction
with each increment in damage. However, the crack growth rate increases
under constant amplitude cyclic load [51], since the same load is applied with
each load cycle.

Hwang and Han [51] investigated fatigue threshold values of strain
energy release rates and the effect of overloads by load reduction method.
They reported that the delarnination growth rate returns to the pre-overload
values after the overload [51]. Mall et al. showed that the strain energy

release rate range (AGi) was the most important parameter for cyclic

debonding [53]. When no fiber bridging occured, the Paris' power law was
suggested to be a useful starting point in predicting the cyclic fatigue crack
propagation [51-53]. '

2.3.3. Fatigue life relations.

Hwang and Han [51] studied interlaminar fracture in a glass - epoxy
composite. A width tapered double cantilever beam (WTDCB) configuration

was used for two advantages over the plain DCB: (1) constant AGi, strain

energy release rate being easily controlled by load, and (2) crack growth rate,
da / dN remains constant within a determined constant strain energy release
rate range [51].

To predict fatigue crack growth accompanied by fiber bridging, Hwang
and Han [51] modified Paris' fatigue power law for isotropic materials such
that

35

da / dN = B A615 (20)
and A61 = GImax - GImin (21)
where
da / dN = crack growth rate
B, n = material constants
AGI = applied cyclic strain energy release rate range,
under mode I loading
N = number of loading cycles.

For the WTDCB specimen under constant amplitude cyclic load, the
cyclic strain energy release rate range is given by [51]

AGI = q (P2m " szin) (22)
where
q = 12 kz/be h3 (Beam Method) (23.a)
q = de/d(a2) (Compliance Method) (23.b)
where

k = taper of WTDCB (unitless)
B"x = effective bending modulus of the beam
along the x - axis (MPa)

36

5‘
I

- height (thickness) of specimen (m)
C = compliance of specimen
a = crack length (m).

Hwang and Han conclude that their experimental data agrees with the
normalization of Paris' power law [51], so that

da/dN

B (AGIIGRB(a))“ (24)

where

a crack length (m)

GRB = crack growth resistance due to fiber bridging.

For the WTDCB specimen under constant amplitude cyclic load range,
the crack growth resistance due to fiber bridging can be written in terms of
crack length, 'a' [51], such that

GRB = QIPCI+OL(AA)BJZ

= qua + 0({a2-a02} /21<)B]2 (25)

where
q = given in Equations 23 .a and b

PCI = initial critical load (not affected by fiber bridging)

AA = increase in fracture area, fiorn geometry of
WTDCB specimen

37

a, B = material constants, found to be 3270 and 0.442,
respectively, for 3-M type 1003 GFRP

AA = (1/211:)(a2 - 302)

(l / 2k)(Aa2 + ROM). (26)

Now, the increase in crack growth resistance may be written
AGRB = GRB - GIC (27)

where AGRB and GRB are defined as above, and

GIC = initial fracture matrix energy, not
affected by fiber bridging.

Integrating Equation 24 from initial crack length al to a later crack length
a2 gives [51]

a2

AN=N2-Nl= l/B I(GRB/AGRB)D d3. (28)

a1

Substituting Equations 22 and 25 into 28 gives the following fatigue life
expression for the WTDCB test specimen [51]

38

212

AN = N2 - N1 =1/BI{[1>CI + ct((a2 - a02)/2k)B]2/(szax - sz1,,)}9 da.
a1 (29)

Equation 29 can be evaluated numerically. Hwang and Han note that
Equation 29 gives the same result whether the beam or the compliance method
has been used to calculate the strain energy release rate [51].

A least squares fit of fatigue data for a glass fiber/epoxy composite
to Equation 24 yielded [51]

da / dN = 3.59 x lo-5 (AGI / GRB)13-5. (30)
2.3.4. Fatigue life modulus concepts.

Fatigue life prediction is further explored by Hwang and Han with the
introduction of a concept they call 'fatigue modulus' [51]. Experimental
fatigue data and fatigue life prediction are sometimes in good agreement for
models such as the S -N curve [55], the empirical Basquin's power law
relation [55], and the Coffin-Manson relation [55]. However, many material
characteristics such as crack length and number density, residual strength,
strain, elastic modulus, compliance and size of the fiber-bridged zone change
over the course of the fatigue life of the specimen. More research is
necessary to understand and better predict fatigue behavior of composites.

2.3.4.a. Hwang and Han fatigue modulus.

Hwang and Han define an overall fatigue modulus as the applied stress
divided by the overall strain at 'n' fatigue cycles [55]

39

F(n,r) = a, / 8(n) = 6,, (r/ 8(a)) (31)
where
F(n,r) = fatigue modulus at 11th loading cycle
8(n) = overall strain at 11th loading cycle
(ra = applied stress
r = ratio of applied stress (ca) to ultimate strength (on)

re. r = (5a / on.

Appropriate boundary conditions are

F(0,r) = Fo ~ BC (32.a)
where
B0 = Young's modulus of the undamaged specimen
and F(N,r) = Ff. (32.b)

According to the definition of the overall fatigue modulus, F(N, r) is
equal to the elastic modulus on the first cycle, and the fatigue modulus at
failure (Ff) is the value at N, the number of cycles to failure (Equation 32.b)

[55]. This approach requires that each period of the applied stress is
maintained at a constant maximum amplitude value, therefore, the fatigue

40

modulus is not a function of loading stress, but of loading cycle only.
The change in fatigue modulus as the specimen's life proceeds is given
by [55]

dF / dN - A c 11“”1 (33)

where

A, c experimentally determined material constants

A is on the order of Young's modulus / 10.33,
and c is about 0.15.

Integrating Equation 20 with respect to 11, number of fatigue cycles, fiorn

n1 to n2, gives

AF = F(nl) - F(n2) = - A (n°2 - n°1). (34)
Inserting n2 = n and n1 = 0 into Equation 34, we have [55]
F(n) - F(O) = - A n°. (35)
At failure, where n = N, we have [55]
Ff' F0 = - A NC. I “ I I f I (36)

Therefore, using the fatigue modulus concept, we can find N, the number
of cycles to failure, N, is [55]

41

N = [B(1-Ff/Fo)]1/° (37.a)
where

B = Fo/A. (37.b)

Using the linear stress / strain equation [55]

ca = Ff’a 8fa (38)
where
ca = applied stress
Ff’a = fatigue modulus at failure, under 0'a
8f = strain at failure, under (in
.a ,,_
it follows that
Ff,a / Fo = (5a / ou = r. . (39)
Inserting Equation 39 into Equation 37.a, we have [55]
N= [B(1-r)]1/°. (40)

Equation 40 now may be used to predict fatigue life as long as applied
stress levels, mode of stress and the material constants are known

42

(Equations 37 .b and 31 respectively).

Equation 40 is in much better agreement with Hwang and Han's
experimental data than either (a) the S - N curve or (b) Basquin's relation,
which are given by

(a) S - N curve [55]

r = klogN-I-d (41)
where

r = applied stress level

N = number of cycles to failure

k = slopeofS-Ncurveinr-logNspace

d = rinterceptofS -Ncurveinr-logN

space.

(b) Basquin's relation [55]

ca = oftzmg (42)

where

o = applied stress
2N = stress reversals to failure ( 1 cycle = 2 reversals)

fatigue strength coefficient

~9
II

fatigue strength exponent ( Basquin's exponent [55]).

00
II

2.3.4.b. Secant modulus concepts.

Hahn and Tsai introduced a secant (elastic) modulus [57], in which a line
is drawn from the zero stress - strain point to the maximum stress - strain point
(Figure 9), after 'n' load cycles. The primary loading, H(n), and secondary
unloading, H'(n), secant elastic moduli are drawn for the nth loading cycle.
Tangent elastic moduli are also illustrated.

Hwang and Han develop a geometric relation linking the four elastic
moduli and the secant modulus [55]. Considering Figure 9, we can state the
geometric relation

H(n) = E2(n) + t(n)[E1(n) - E2(n)] (43)

mm = E'2(n) - t'(n)[E'(n) - E'2(n)] (44)
where

t(n) = AB'IAC', 0<t(n)<1 (45)

t'(n) = D'E / CE, 0 < t'(n) < 1. (46)

Hwang and Han [55] describe a fatigue secant (elastic) modulus, in
which the secant modulus is degraded until fracture occurs (Figure 10).
Using Figure .10, the fatigue modulus may be written [55]

1/F(n)= Z [1/H(k)- l/H'(k)+ 1/H(n) (47)

43

Stress

 

 

 

 

 

 

 

 

 

 

 

C
_ E2 (n)
B
__ 5’1 (n)
E1(n)
D
/ H’(n)
A i D’ C’
Strain
Figure 9. Schematic of Hahn and Tsai secant modulus during the in fatigue

load cycle [57] .

45

 

 

 

 

stress

F(n)

E(n)

 

 

01234 strain nn+l

 

 

I‘ ‘ 8 (11)

Figure 10. Hwang and Han fatigue secant modulus [55].

 

46
1/F(n) - 1 /F(n - 1) = 1 /H(n) - 1 /H'(n - 1). "(43)

The above relations for fatigue life and fatigue modulus apply to the
unidirectional glass fiber reinforced epoxy matrix composite being used in this
thesis.

The material used for this thesis is the same as that used by Hwang and
Han [51], i.e. 3-M type 1003. See Appendix 6.A for 3-M composite type
1003 physical properties.

2.3.5. Acoustic emission in the fatigue of GFRP composites.

Fatigue induced acoustic emission in GFRP composites has been
investigated by several research groups [6, 7, 59-65].

To further understand composite damage modes, it is essential to
distinguish between the acoustic emissions of new damage (i.e. matrix
cracking and splitting and fiber pullout and breakage) and the emission due to
friction, rubbing and fretting between the newly created fracture
surfaces [6, 7]. Results from other researchers [59, 60] indicate that
substantial acoustic emission is generated by friction between fracture
surfaces. . . . .

The methods and data presentation formats of Awerbuch et a1. [59, 60]
appear to be most informative. There is a similarity of equipment between
Awerbuch's and MSU's MMM Department's Physical Acoustics Corporation
AE analysis computer and the 3-M type 1003 GFRP test specimens.
Awerbuch, with Ghaffari [59] and Eckles [60], discuss the acoustic emission
fron unidirectionally reinforced graphite fiber / epoxy composites stressed in
tension - tension fatigue loading, in which the ratio of minimum to maximum
tensile load is 0.1 (i.e. R = 0.1). A substantial amount of acoustic emission is
generated by friction between newly created internal crack surfaces, as made

47

evident by the continuous emission as the load is continuously increased and
decreased [59, 60].

Frictionally generated acoustic emission often exceeds and masks
emission due to new damage [59, 60]. Awerbuch et a1. developed a
methodology to discriminate friction generated emission from new damage
emission by correlating AB event intensity parameters [59, 60].

The acoustic emission apparatus used by Awerbuch et a1. [59, 60]
consisted of a Physical Acoustics Corporation (PAC) AB computer model
3000/3004. System operating parameters were listed as: 150 KHz resonance
transducers, model 1220A preamps with +40 dB fixed gain, system threshold
of 0.1 volt, postamplifier gain of +20 dB and dead time of 1 msec [59].

Two transducers were used for event location determination. Post test
spatial filtering analyzed only events within 20 to 80 percent of the gage
length. Damage progression was also monitored by a 150x closed circuit
television in real time. The specimen was X-rayed in transmission after each
test sequence to determine the extent of cracking and splitting damage [59].

Awerbuch et a1. [59-61] found that acoustic events that occur in the lower

load range (e.g. 10 to 95 percent of 0d) were primarily associated with

interfacial crack friction. Do not confuse this with the quantitative damage
measure of Q‘1 . Events generated in the upper load range (e. g. 95 to 100

percent of Gd) are primarily associated with new damage [59].

Although friction generated emission dominates the lower end of the load
range, some new damage is occuring, albeit generally masked by the
continuous frictional emission [59-61]. Conversely, at the high end of the
load range, where new damage is expected to dominate, a large amount of
frictional emission also occurs. The percentage of either kind of emission is
difficult to estimate, due to the nature of the specimens. Some AB events can
(and do) occur too rapidly for the PAC AB system to recognize and record.

48

However, the new damage can be indirectly observed by the increase in
friction generated emission when the AB counts data from the lower load
ranges is summed [59-60].

Awerbuch et a1. [59-60], investigated a single damage mode, i.e. matrix
splitting with very limited fiber breakage. Two distinct ranges of AE event
intensities were observed. Lower intensity AB events were recorded across
the entire load range. Only low intensity events were observed at the lower
load ranges. With increasing load, the number of events with higher count
rate and amplitude intensities increased [59—60].

Awerbuch etal. [59-60] introduced the method of plotting AB event
intensities as a function of load. Awerbuch et a1. [59, 61] plotted events
(ordinate) versus amplitude or duration (abscissa) as histograms, and
amplitude or duration (ordinate) versus load (abscissa) as point plots. A load
signal was fed to the PAC computer's parametric input as a DC voltage with a
full scale range zero to +10 VDC.

The number of AB events versus amplitude histogram appeared bi-modal,
implying that the lower amplitude group of events (centered at ~ 55 :I: 5 dB) is
primarily due to frictional emission. The higher amplitude group (centered at
~ 65 :l: 5 dB) was assumed to be primarily due to new damage [59, 60].

Specific graphical filters and parameters can be applied to recorded AB
data. Post test filtering can divide data on either side of a chosen threshold for
any parameter felt to be helpful in descriminating new damage emission from
frictional emission [59-60].

The accumulation of new damage has been shown to be a sporadic,
seemingly random process [59-65], analogous to natural processes such as the
initiation of corrosion sites in some metals [61]. The release of acoustic
energy by the specimen also varies fiom cycle to cycle. A given specimen
may be fatigued for several cycles without detectable AE, while large numbers
of AB events may be observed during subsequent load cycles [61]. The

49

apparently random nature of acoustic emission indicates that the progression of
damage is (macroscopically) a sporadic, random process.

As a specimen is fatigued, new damage accumulates rapidly at first. The
damage tapers down to a moderate increase until the damage reaches a
critical level At the critical damage level, AB increases as catastrophic
damage rapidly degrades the specimen [46, 59-61, 63]. The division of
acoustic emission into three ranges (initial rapid accumulation, lower steady
state accumulation and final rapid accumulation) suggests the operation of a
three regime damage curve analogous to the creep damage curve in metals and
some ceramics.

Damage saturation is found elsewhere in nature. Thermal shock damage
in alumina saturates with thermal fatigue, with the damage measured as
changes in elastic modulus and internal friction. When fatigued, certain metal
alloys show a saturation of persistent slip bands of dislocations which
accomodate plastic deformation induced by constant strain cycling.

Awerbuch et al. found that when sorted according to load range, AB data

in the range of 0.95 to 1.0 of cm tends to a steady state emission rate.

When the AE data, after sorting by load range, is screened off below a chosen
amplitude threshold, the rate of accumulation of AB events is even lower.
Awerbuch et. a1 terms that threshold as a 'friction emission threshold' or
'FRET' [5 9, 60].

In Awerbuch's experiment, the AB recorded in the lower range (below

0.95 dd), rises quickly at the start of the fatigue test, then lessens in rate of

increase only slightly as fatigue life progresses (as the number of loading
cycles increases), remaining the dominant source of AE [59-61, 63]. This
emission was largely attributed to frictional sources. Once frictional emission
began, it accumulated continuously and at an increasing rate. Frictional
emission (which encompasses crack surface rubbing, fretting, asperity contact

50

and debris pulverization) may be a continuous source, whereas new damage
occurs only once. Awerbuch et a1. [61] verified crack closure and
simultaneous frictional emission during load cycling by observation in situ at
150 X magnification by closed circuit television.

Awerbuch [61] explained that the Kaiser effect [5], present during fatigue
in GFRP composites, may provide insight regarding the initial rapid increase
of AB events from new damage at the beginning of a test. This effect is likely
due to a reduction in the rate of the creation of new damage [46].

The selection of the FRET level must consider AB event intensity
parameters. Awerbuch et al. found that the intensities of events in the lower
load range were generally characterized by less than 20 energy counts, less
than 40 counts per event and less than 250 microseconds in duration [59].
These parameters vary with material, maximum stress applied, cycling
frequency etc. and must be determined experimentally [59-61]. Elastic moduli
of the constituents, volume fractions, stress intenSity range, strain energy
release rates and rate of deflection all affect the emission rate dN/ da (AB
counts per increase in crack length or area) equations (Equations 20, 24 or 30).

The acoustic emission screening techniques [59-60] indicate a qualitative
conelation between acoustic emission and crack damage initiation and
progression. The link between AB and crack evolution allows the
construction of derived damage curves, which are related to fatigue life
[46—65] .

2.3.6. Comparison of AE from three types of fiber
reinforced epoxy composite materials.

Otsuka and Scarton [46] tested three fiber reinforced epoxy composites:
(1) unidirectionally reinforced glass fiber epoxy, (2) chopped glass fiber mat
epoxy and (3) unidirectional carbon fiber epoxy. A screw driven 'Shimazu'

51

general purpose testing machine was used, with the assumption that machine
noise was 'inherently small'. Specimens were tested in two methods: three
point bending and cantilever beam with a transverse saw cut at the depth of
one lamina. Otsuka and Scarton plot amplitude versus number of AB events
for all three specimens under each test combination. Although no clearly
defined ranges appear in the amplitude distribution, the authors assume matrix
crazing occurs at amplitudes up to about 33 dB, and that delamination and
fiber breakage occurs with amplitudes greater than 33 dB [46]. The
justification for this choice of threshold is the interpretation of a different test
on the unidirectionally reinforced glass fiber epoxy composite, in which a local
minimum of AB events appears at about 33 dB. Local peaks at about 24, 35
and 43 dB are assumed to result fiom the damage modes of matrix crazing,
interply delarnination and fiber breakage respectively.

A Kaiser effect experiment was performed by Otsuka and Scarton using
carbon fiber-epoxy under three point bending. The specimen was loaded and
relaxed, with loads of 245, 390, 490, 590 and 685 N. Virtually no AB was
observed during the load relaxation, but there was an exponential increase in
acoustic events during load-up, after the load exceeded the previous highest
load. The result of their Kaiser effect test was used to justify the 24 dB
peak's responsibility for matrix crazing, and not being a manifestation of
machine noise.

2.4. Mechanical fatigue in glass-ceramic composites.

Qualitative models exist for fatigue mechanisms for mechanical and
thermomechanical fatigue in both monolithic ceramics and ceramic matrix
composites [1, 6-8, 65 -69]. The failure of brittle ceramics under cyclically
varying compressive loads is of interest in a diverse number of engineering

52

applications [8, 65, 67, 70-85]. The driving force for the large-scale
engineering research effort in fatigue is the improvement in mechanical
properties and especially the reliability of ceramic structures. Reasonably
accurate qualitative predictions of lifetimes, strengths, wear resistance and
damage tolerance are said to be beyond the capabilities of current 'state of the
art‘ research [65, 67, 74]. Production quality and repeatability for ceramic
composite components must be significantly improved before accurate
predictions of mechanical properties can be made [65, 67, 74]. Nonetheless,
qualitative predictions are of great importance for making choices of materials
with superior physical properties.

Before about 1985, mechanical fatigue in ceramics had been detected and
acknowledged [9, 66] for only a few materials. Before 1985, mechanical
fatigue effects were labeled as fatigue-like, and were usually attributed to
environmentally assisted slow crack growth, also called stress corrosion
cracking [6-9, 66, 67, 69, 71, 73, 75, 80]. Molecular water (atmospheric
humidity) was named as the activating species for polycrystalline monolithic
silicon nitride, soda-linre glass and electrical porcelain containing ~20 micron
diameter dispersed quartz particles [7 , 8-10].

Ewart and Suresh, in 1986, reported mechanical fatigue fracture at room
temperature in chevron notched compression specimens of single phase
polycrystalline alumina subjected to fully compressive loads [8]. Mode I
fatigue crack growth occured without significant macroplasticity [8].

2.4.1. Mechanical fatigue damage in ceramics.

Several researchers [1, 2, 4-7] suggest that the materials which exhibit the
largest thermal-shock fatigue efl‘ect are those in which substantial preexistent
or stress-induced microcracking is found [1, 2, 6, 7]. Kim et a1. [3] and Case
[81] however found that thermal shock fatigue occurs in polycrystalline

53

alumina, Macor, and other ceramic oxides. Their specimens were prepared
and annealed to purposely minimize the number of preexistent rnicrocracks.

A correlation between thermal-shock fatigue and microcracking exists and
suggests the operation of microcrack based mechanisms. However, since
thermal stress is a body force, and mechanical stress is externally applied, it is
not clear how separate or separable are the roles of thermal [1, 2, 4, 6, 7, 11],
mechanical [1, 2, 4, 6, 7] and environmental [6 , 71, 78] stresses. The
separation of purely mechanical from thermomechanical fatigue effects is a
basic step toward defining any certain mechanism for thermal-shock or
mechanical fatigue [6, 7].

In the case of thermal shock fatigue, internal (body) stress is created
during the quenching and heating of the specimen due to macroscopic thermal
gradients imposed on the specimen. This situation arises due to thermal
strains which are set up by the thermal gradient.

The thermal expansion relation, ‘

AL = on L0 AT (49)

descnhes the dilatational behavior of a material under a thermal gradient. The
thermal expansion coefficient, of course, depends on direction within the
material. Equation 49 applies to macroscopic stresses as well as microscopic
stresses. Thus, thermomechanical fatigue may have additional mechanisms at
work, beyond the macroscopic stresses of purely mechanical fatigue created
by the thermal expansion gradient. However, thermal expansion mismatch
and thermal expansion anisotropy are not necessary requirements for thermal
shock fatigue [3, 81].

True fatigue crack propagation has been attributed to nricrocracking due to
residual stresses arising from several sources. These sources include grain
boundary misfit, crack wake asperity contact and other toughening

54

mechanisms such as crack tip interactions, crack bowing, phase
transformations and crack deflection [8, 65, 68, 80, 81, 85].

Crack front extension can be anested, restricted or diverted by various
toughening mechanisms. These mechanisms include crack tip shielding and
crack process zone rnicro-plasticity in some monolithic ceramics. For
composites add the toughening mechanisirns of matrix cracking, fiber
debonding and fiber pullout and crack diversion around hard second phases or
inclusions [6, 7, 74, 76, 78, 80, 81].

2.4.2. Ceramic fatigue theory.

The macroscopic fatigue effect results from micro-scale mechanisms
activated under cyclic loading [65 -79]. For ceramics, the ultimate
compressive fracture stress is often up to ten times that of the tensile ultimate
fracture stress [8, 65 , 70]. Ceramics still fail in time at loads substantially
lower than their rated ultimate load bearing capabilities [8].

Many ceramic fatigue theories or models utilize two important aspects:
(1) residual micro-stresses [6, 7, 68, 74, 76, 78] and (2) a mechanism to
enable damage and microstructural changes to occur during the unloading part
of the flexure cycle [6, 7, 9, 67, 80].

Significant residual stresses typically arise from grain orientation thermal
expansion anisotropy. For ceramic composites, interphase thermal expansion
mismatch is also present.

Lewis and Rice report that residual stresses may range as high as l to 2
GPa in alumina, and up to about 3 GPa in BN-containing particulate
composites [6, 7]. However, Lewis and Rice [6, 7] do not specify how they
measure the residual stresses. A propagating crack may interact with
grain-size scale residual (built-in) stresses [6, 7, 65]. Grain boundary
microcracking is frequently enabled by residual tensile stresses and low

55

fracture energy along grain boundaries [8, 65, 81].

Fatigue crack growth occurs during the load reversal transition from
tension to compression and especially from compression to tension. During
load reversal, the fracture mode changes from mode 11 or III during
compression to mode I in tension [6, 7, 9, 68, 69, 71, 74, 78].

Lewis and Rice proposed that progressive slow crack growth occurs on
each load cycle [6, 7]. Lewis and Rice examined Lucalox alumina, a
glass-ceramic Pyroceram code number 9606 (Coming Glass Works, Corning
NY), and alumina- and mullite- BN composites.

Lewis and Rice's model [6, 7] requires: (1) the pre-existence of residual
stresses and (2) a mechanism to allow damage to occur during both loading
and unloading. Far-field stresses take advantage of the residual stresses to
overcome the local fracture strength of grains or their boundaries. Incremental
fracture of single grain boundaries leads to grain relaxation and misfit between
adjacent and nearby grains during the flexural fatigue cycle. The
microcracking thus relaxes and redistributes the stresses on the nearby grains
and their boundaries. Lewis and_ Rice's [6, 7] self-feeding model describes a
plausible mechanism for true cyclic fatigue damage during successive load
cycles.

2.4.3. Fatigue models for ceramics and ceramic composites.
2.4.3.a. Microcracking model for polycrystalline ceramics.

A plausible tension-tension fatigue fracture process includes preexistent
residual stresses that enable an applied far-field stress to advance a crack
through a monolithic polycrystalline matrix (Figure 11) [6, 7]. Evans et al.
describe a similar scenario [9, 65 , 66], in which a polycrystalline ceramic
accumulates microcrack damage during cyclic loading.

 

 

 

 

Figure 11. Schematic of fatigue in polycrystalline ceramic under tension-
tension loading [6]. a) before loading, grain boundary one has residual
compressive and shear stresses, grain boundary two has small residual
tensile and large residual shear stress, b) on the first tensile load, grain
boundary one is fractured, c) unloading increases shear stress on grain
boundary two, fracturing it, (I) reloading in tension extends fracture
through grain boundary three by Mode 1 opening.

57

To model this cooperative grain boundary cracking process, a thought
experiment is developed. Grain boundary 1 is initially assumed to have a
residual compressive stress, and grain boundary 3 initially has small residual
tension and a large residual shear stress (Figure 11.a). This scenario may be
typical of a hypothetical polycrystalline ceramic material.

A far-field tensile stress (Figure 11.b) may produce a net tensile stress
which fractures grain boundary 1. Propagation of the crack along grain
boundary 2 may be suppressed due to the orientation of the boundary,
superimposing little additional tensile stress while reducing the shear stress.
Unloading then increases the shear stress at grain boundary 2 (Figure He).
The crack may then extend in Mode II opening (include Mode [[1 opening for
three dimensions). Reloading in tension subsequently extends the crack in
Mode I along the next favorably oriented grain boundary ( grain boundary 3 in
Figure ll.d).

Subcritical crack growth due to grain-misfit-induced tensile stress occurs
on each cycle until the crack grows to a critical length [6, 7]. When the crack
tip stress intensity exceeds the local critical value, the crack then extends
catastrophically on the next and last loading cycle [6-9, 65 , 66, 69, 71, 80].

During a load cycle of compression-tension or compression-
compression, a similar combination of residual and applied stresses can be
envisioned (Figure 12). Here, grain boundary one has a residual
compressive stress and grain boundaries two and three have small residual
tensile and shear stresses (Figure 12.a). On the first compression cycle
(Figure 12.b), the compressive stress on gram boundary one increases,
causing no harm to that grain boundary. Grain boundaries two and three p0p
Open by Mode 11 loading due to high shear stress, but do not propagate due to
high compressive stresses at grain boundaries one, four and five. Upon
unloading, the redistributed stress allows crack propagation along grain ,.
boundaries one and four or five by Mode I (tensile) loading (Figure 12.c).

 

 

 

 

Figure 12. Schematic of fatigue in polycrystalline ceramic under "
compression loading [7]. a) before loading, grain boundary one has a

residual compressive stress, and grain boundaries two and three both have
small residual tensile and shear stresses, b) on the first compressive load

cycle, grain boundaries two and three fracture in Mode 11 and are arrested
at grain boundaries one, four and five due to high compressive stresses on
those boundaries, c) on unloading, grain boundaries one, four and five pop
open under local tension.

59

The grains around the cracks now relax, assuming a position and shape that
minimizes their own strain energy. This new condition of stress relaxation
and shape change causes a local misfit of grains and induces additional
'residual' stresses upon nearby grain boundaries. With successive
compression- relaxation or compression-tension cycles, the fatigue cracking
proceeds until catastrophic failure occurs [6-9, 69, 71, 80].

2.4.3.b. Fiber buckling model for ceramic composites.

For whisker and fiber reinforced ceramic matrix composites, residual
stresses between the fibers and the matrix, while substantial, do not play a
significant role in the fatigue of these materials [6, 7]. Lewis and Rice [6, 7]
argue that these residual stresses are relatively minor when compared to other
processes which occur in the fiber composite that can lead to fatigue. For
example, a ten volume percent randomly oriented SiC fiber reinforced
mullite-BN composite does not have a longer fatigue life than an umeinforced
mullite-BN composite [7], although the SiC fiber reinforced composite is
tougher.

The primary toughening mechanisms in reinforced ceramic fiber
composites are matrix cracking, fiber debonding and fiber pullout [6, 7, 65,
80, 81]. The highest toughness possible requires fiber-matrix bond strength
to be optimized as neither too strong, nor too weak, and with minimal
compressive residual stresses on the fibers. Highest toughness may require
crack branching, fiber pullout and crack diversion around fibers. High
toughness implies the absorbtion of a large amount of energy during fracture
with less loss of material, thus leading to a higher margin of damage tolerance.
CMC's, in which the physical or chemical bond strength of the fiber to the
matrix is high, typically have touglmess values similar to that of the ceramic
matrix [65]. Lewis and Rice's fiber pullout fatigue model depends heavily on

60

the non-reversibility of the matrix-crackirrg and fiber pullout, rather than on
residual stress effects. For moderate stress levels, fiber damage during
pullout may be a fatigue mechanism in itself [6, 7]. Fiber abrasion during
pullout leads to weakening and eventual fiber fracture. As the fibers continue
to fracture into smaller lengths, the lengths of the fibers still embedded in the
matrix becomes less than the critical length for effective load transfer. The
load bearing ability of the composite decreases as the length of the fibers is
reduced to less than the critical length. The result is static fatigue, since the
remaining fibers are stressed to higher levels [6, 7, 65, 80, 81].

Extensive matrix cracking and fiber pullout may lead to fiber
buckling when the load is relaxed and the crack attempts to close (Figure 13)
[7, 48, 79]. As the matrix cracks close, the pulled-out fibers are forced back
into their respective holes in the matrix. Forcing the fiber back into its hole
requires stresses higher than for pullout, since the relaxation of the
fiber-matrix interfacial stresses create a substantial interference fit to overcome.
In addition to the interference fit, debris in the fiber's hole and asperities along
the fiber-matrix contact area substantially add to the force necessary to push
the fiber back into its hole [6, 7]. If the pulled out length of a given fiber
exceeds the critical length for buckling, given the particular stress geometries,
the fiber will buckle and fracture.

A computer simulation by Lewis and Rice indicates that a 90 um pullout

length for a 25 um diameter SiC fiber can be buckled by a crack closure force
of 500 MPa [7]. Fiber buckling likely produces more damage than
incremental grain boundary cracking with cyclic loading [7]. Rapid
degredation at high stress levels in tension-tension cycling of cerarnic-fiber/
ceramic-matrix composites, can occur. Once broken, the fiber can no longer
bear load [7, 48, 79].

The accumulated microcracking and fiber buckling fatigue models [6, 7,

61

 

 

 

 

 

 

 

 

 

 

3
H
3
"‘H
/

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6)

 

 

 

 

 

 

J K 1

Figure 13. Schematic for fiber pullout fatigue mechanism in a fiber
reinforced ceramic composite [7] . a) unbroken fiber in unloaded matrix,

b) matrix cracks under moderate tensile stress, c) at higher load, fiber fails
and debonds at weak point, d) at higher load, fiber is pulled out of matrix,
suffering some damage from abrasion, e) as the composite is unloaded, if
the fiber has been pulled out a sufficient distance, the fiber will buckle and
fracture, instead of being forced back into its matrix hole. Broken fiber
carmot bear any load, therefore the matrix crack advances, and the remaining
fibers must carry the load.

62

48, 79] qualitatively explain fatigue mechanisms, but quantitative equations for
fatigue life predictions are still developing.

The fiber buckling fatigue-model suggests that for reasonable reliability,
design stresses must be kept below the level at which fiber pullout occurs, or at
least below that in which fiber pullout exceeds a critical buckling length.

Lewis and Rice liken such a stress level to the true fatigue limit in ductile
metals [6, 7] .

3. Experimental procedure.
3.1. Materials.

Low cycle flexural fatigue loading was performed on two commercially
available composite materials. A unidirectionally reinforced glass fiber epoxy
composite, 3-M Company type 1003, was prepared and provided by Dr. D.
Liu of M.S.U. The second material, Macor (Corning code number 9658), a
machinable glass-ceramic composite, was purchased from Astro Met
Associates, Cincinnati Ohio, a licensed vendor of Conring Glass Works,
Corning NY. Macor is a mica-glass ceramic composite in which small (2 by
50 microns or less) platelet-shaped fluorophlogopite crystals are nucleated and
grown in a fluorine-containing parent glass [76, 77, 79]. See Appendix 6.A
for the material properties of Macor and Appendix 6.B for 3-M 1003 GFRP.

3.2. Specimen preparation.

Glass—epoxy test specimens were cut from a single parent plate, 7.5 cm
wide by 20 cm long by 0.2 cm thick, using a Beuhler low speed 150 diamond
grit saw, lubricated by Buehler cutting oil. The cutting direction was parallel
to the reinforcing fibers. After cutting, the width dimension was equalized

63

end to end by sanding the edges along their length by hand with 600 grit
wet-dry paper, laid on a flat surface. The comers of the specimens were kept
as square as possible. No further treatment was applied, leaving the surfaces
in the 'as received' condition.

The Macor specimens were cut from a single parent plate, 7.5 cm by
7.5 cm, using the same low speed 150 grit diamond saw. The widths were
equalized end to end as for the GFRP specimens. The cut edges were slightly
rounded (radius approximately 0.05 cm) with worn 600 grit SiC wet-dry paper
(water flushed). The final specimen dimensions were 7.5 cm long,
approximately 1.0 cm wide and 0.2 cm thick (as received). The Macor
specimen bars were then annealed in air, in an electric resistance 'Lindberg'
brand furnace. Specimens were placed on an alumina setter plate, with about
a 0.5 cm spacing between specimens. The setter plate was made of 'Coors'
brand AD96 alumina, 15 cm by 15 cm by 0.3 cm thick. The setter plate was
then set on several scrap pieces of AD96 on the furnace brick in the center of
the fumace. The anneal heat treatment began with a warm-up in the furnace to
200 0C for one hour, the temperature was then raised to 300 °C for one hour.
The temperature was then raised to 500 °C, and the specimens were left to
soak for fifteen hours. The furnace was then turned down and cooled to
300 °C over five hours, held for one hour, then cooled to 200 °C over three
hours and held at 200 °C for four hours. The furnace power was then turned '
off. The specimens were allowed to freely cool to room temperature without
opening the furnace for approximately twenty hours.

3.3. Overview of experimental procedure.
Prismatic rectangular specimen bars, approximately one centimeter i one

millimeter wide, seven centimeters long and two millimeters thick (as
recieved), were loaded in four point bending. Load cycling limits were

64

dependent on maximum and minimum load, rather than crosshead extension,
which is a necessary procedure for brittle materials. The maximum load was
set at a preselected percentage of the calculated rupture strength (Equation 17).
The absolute maximum load varied from specimen to specimen due to slight
variations in specimen width. The minimum load was set to zero.

Crosshead speed was determined to be one half inch per minute by manually
timing its displacement.

Load cycling was periodically airrterupted after a predetermined number of
cycles, in order to measure the specimen's internal friction and Young's
modulus. A logarithmic scale for measurement was chosen with frequent
interuptions early during the test, assuming that damage would occur rapidly at
first, then increase less rapidly as fatigue progressed.

As Macor specimens fiacturedtthe locations and conditions of fracture
were noted. The fracture surfaces were cut off for examination under a
scanning electron nricroscope. '

Acoustic emission was detected from the specimen with parametric
scaling by load for experiments using the glass-epoxy specimens. The
acoustic emission analysis computer's parametric input eventually
malfunctioned and temporarily did not respond to the load scaled parametric
input. Acoustic emission was however recorded for several Macor specimens
without load scaling. Subsequently, the load scaling capability was
recovered.

The accumulated damage was estimated using the number of counts of
acoustic emission given off during the loading process, as well as by changes
in internal friction and Young's modulus. The acoustic emission was sorted
into load ranges and summed up to the points at which load cycling was
intempted for sonic resonace measurements. Ultimately, the fatigue damage
was measured by the change in the measured physical properties and the
acoustic emission was related to those changes to determine the loads at which

65

the damage occured.
3.4. Four point bending technique.
3.4.1. Equipment and fixture.

The work was performed on an Instron Table Model low capacity screw
driven tension/compression testing machine, circa 1965.

The semiarticulating four point bending fixture was designed by the
author, based in part, on designs used by the Naval Research Laboratory [66],
and following the recommendations of ASTM standards [23, 24, 25]

(Figure 14). The fixture was built by machinists at the Physics Machine
Shop of M.S.U. The fixture consists of a pair of load platens, machined
from O-l type steel. The load pin support bars were bolted to the platens.
The upper load platen was bolted to the moving crosshead of the Instr-on.
The lower load platen was designed to rotate to accomodate misalignment
between the upper and lower load pins as they came in contact with the
specimen.

The lower half of the swivel bearing was also machined from type O-l
steel. The bearing seat in both halves was 45°, conically drilled and polished.
A one inch diameter, commercially available steel ball bearing enabled rotation
of the fixture. All O-l steel parts were heat treated after machining. After .
heat treating, the bearing seats were re-polished and mounting surfaces for the
load pin bars were flat-ground.

After receipt of the fixture, it was decided that further flexrhility in the
choice of location for the acoustic emission transducer was desired. The
author then redesigned the load pin support bars to allow adequate clearance to
mount the AE transducer above or below the specimen, and machined them
himself in the machine shop at Department of Engineering Research on the

66

 

Upper Load Platen

 

15cm

 

 

 

 

 

 

 

 

 

 

Lower Load Pin Support Bar —| I

’—

 

 

 

 

Lower Load Platen 4 ch
12 cm ./\-
//,z \ _I__
Bearing Ball 2.54 cm

 

 

 

Lower Bearing Base I

Figure 14. Schematic of four point bending fixture used for experiment.
Dimensions not to scale.

 

67

M.S.U. campus. The material used for the load pin support bars was 1018
carbon steel. The upper load pin support bar was machined with a pin span
of 20 mm. The lower load pin support bar had a pin span of 40 mm. Load
pin grooves (in the support bars) were machined with a 4.7 mm (3/16")
diameter and 1.5 m depth to allow for some rotation during loading. The
load pins were cut from a rod of commercially available 'die ejector pin' stock,
with length 3 cm and diameter 3.2 m (1/8"). Die ejector pin stock was
chosen for its high hardness, strength, highly polished finish and availability.

The load pins were secured to the support bars by rubber bands (zl inch
free diameter). The rubber bands looped around the ends of the pins and ran
under the support bar through a 3 mm deep relief groove cut into the load
platen.

3.4.2. Mechanical fatigue method.

Four point bending stress (R = -1, where R denotes the ratio of
maximum to minimum stress, with the negative sign indicating compression)
was applied to the specimen bar at room temperature, about 25 °C.
Laboratory humidity varied between 50 to 70 percent relative humidity during
the experimental procedure.

The glass-epoxy bars were fatigue cycled under four point loading at 25,
50 and 75 percent, respectively, of their static rupture strength of
approximately 36 MPa [51]. Cyclic frequency was approximately 0.67
cycles per second, based on a crosshead loading rate of one-half inch per
minute.

The Macor specimen bars were fatigue cycled in a similar manner, at
20, 30, 40, 50, 65 and 75 percent of its rupture strength of 100 MPa [69].
Maximum loads of 90, 85 and 80 percent of the theoretical rupture strength of
100 MPa were attempted, but every specimen fractured on the first or second

68

load cycle.

Once a particular fatigue testing run was started for the Macor and
glass-epoxy specimens, the maximum load and loading rate was held constant.
Effects of overloading, load changes or indentation crack growth are left for a

subsequent study.
3.5. Acoustic emission technique.
3.5.1. Acoustic emission detection.

Acoustic emission was detected and analyzed using equipment
manufactured by the Physical Acoustics Corporation, based in New Jersey.
The 'PAC' Model 3000/3004 AB computer/analyzer was used with PAC
1220A preamps and PAC 300 KHz resonance transducers (Figure 15.a).

Background noise was examined by initially setting the analog section of
the AB computer for absolute maximum sensitivity. Sensitivity was then
decreased by raising the voltage threshold by two detent steps and lowering
the overall amplification one detent step, so that random electrical noise no
longer triggered the AB detection system. The voltage threshold setting was
maintained at 0.3 x 10 volts. A sensitivity setting of 60 dB fixed gain on the
preamp and 18 + 20 dB gain on the 3004 analog unit gave an overall system
gain of 98 dB. The maximized sensitivity settings were maintained
throughout the experiment after it was determined that background noise could
confidently be avoided. Running the unit for up to thirty minutes in an idle
state gave no AE events.

Acoustic emission was detected via a transducer mounted directly on the
specimen's tensile side of flexure, in the center of the constant moment area.
The transducer, also a PAC product, measured 8 mm in diameter, 10 mm in
height and had an active 'Teflon pad' region of 6 mm diameter. The

69

 

Figure 15.b. Closeup photograph of the four point bending fixture used for
the experiment.

70

transducer was rated at a resonance frequency (highest response) of 300 KHz-
(Figure 7).

Consistent bonding of the transducer to the specimen was accomplished
by using a thin layer of Dow Corning brand 'High Vacuum' silicone grease
intended for glass to rubber vacuum fittings. The transducer was secured to
the specimen with a doubled rubber band, with the 'X' of the crossover point
across the back of the transducer. Common office type rubber bands were
used, with a free diameter of about 3 cm and a band width of about 0.5 cm.
The higher aspect ratio of width to thickness seats better on the back of the
transducer, exhibiting less slippage. Bonding and debonding the transducer
by a cement was considered impractical, as it would tend to degrade the
specimen and the transducer, and would interfere with the sonic resonance
measurements.

3.5.2. Machine noise quantification.

Machine noise also must be examined, quantified and determined to be
isolatable fiom the specimen's acoustic emission. Machine noise must be
known as accumme as possible, in order to qualify and give credibility to
other experimental results. Machine noise is considered part of the
background noise which is at least 50 dB below the threshold set for the
response of the AB system software. Typically, screw driven crosshead
machines are much quieter than the servo-hydraulic type [61, 64]. Machine
noise is also typically in the low kilohertz range, while ceramic cracking
acoustic emission is typically in the high kilohertz range, with components in
the megahertz range (Figure 7, page 23) [31]. The difference between
frequency ranges of acoustic signals allows for the electronic exclusion of the
machine noise by proper choice of transducer response ranges and signal
processor settings.

71

Three distinctly separate experiments were performed to show machine
noise under various operational conditions. For this series of
tests, two channels of AB aquisition were used, with identical control
settings, along with two nominally identical AB transducers. One transducer
was mounted to the specimen's tensile face and the other transducer was
mounted on the center of the front face of the upper load platen (the platen
mounted to the crosshead) (Figure 15.b).

To test for machine noise at the points of load reversal, a previously
undamaged glass-epoxy specimen (GE-5) was load cycled at a relatively low
load (22.27 N). 22.27 N (5 lb.) was approximately 1.7 percent of the rupture
modulus. Very little fatigue damage was expected to occur at that low load.
The test was interrupted every 100 cycles to check for changes in Young's
modulus or internal friction, by using the sonic resonance system. A
maximum of 500 load cycles were run. An initially undamaged specimen was
used.

Two tests for machine noise as a function of load were performed. In
one, the specimen was load cycled at a low load initially for 100 cycles, then
the load was increased for another 100 cycles. The fatigue load was stepped
up to the maximum of 222.5 N (50 lb.) used for the actual experiment. At
each stage of the teSt, the acoustic activity detected by both transducers was
compared to the AB observed in the prior stages. If machine noise was a
function of load, then the AB detected by the transducer located on the load
platen would increase proportionally with load, if not, only the AB detected by
the specimen mounted transducer should show an increase with load.

Isolation between the two transducers was also investigated. If the
specimen mounted transducer detected an AB event immediately after the load
platen mounted transducer did (or vice versa), then some degree of
communication between them would be indicated. The degree of
communication may be measured in terms of the relative differencs detected

72

between event amplitude or number of counts.

A third test combined elements of the first two tests. Load was applied
to the specimen, then released, in increasing load steps, with one application
of each given load. The procedure also explored the Kaiser effect of the
GFRP material, in particular, the number of acoustic events is expected to
increase only after the cunent load exceeded the highest previous load, and
there is no machine or system noise transferring into the specimen.

3.6. Sonic resonance technique.

The sonic resonance technique was used, to determine internal friction
(Equation 5), and Young's modulus (Equation 7). ASTM standard
procedures were followed [15-17].

The procedure used for the determination of internal friction consisted
of suspending the specimen by ordinary white cotton sewing thread from two
identical piezo-electric transducers (monaural phonogragh cartridges,
'ASTATTC' brand, part number 62-1, needle type N27-1). One transducer
was driven at the specimen's flexural fundamental resonant frequency
(ASTATIC cartridge maximum input = 3.0 VRMS)- The RMS voltage output
of the pickup transducer was then measured, after bandpass filtering (set at
one half and twice the resonant frequency) and amplification. The signal to
the driver was then cut off with a 'bounceless contact' type switch. The
number of times the voltage amplitude of the 'ring down' of the specimen

exceeds a preset reference level (V 2, Equation 5 ) was counted using a

universal digital pulse counter. The threshold or reference level (V 2) was set

by the experimenter to a level below the peak driven output voltage. This
level obtained a number of counts (N, Equation 5) in a range from 100 to 500.
The reference level was restricted to a minimum value at least 20 percent

73

above the peak noise level i.e. {1.20 (1.414 VRMS” (Figure 16).

Wachtrnan and Tefft's nonlinear least squares curve fitting program was
used to subtract the effect of suspension fixtures from the specimen's actual
internal friction [86]. For an acceptable correlation coefficient, at least five
data pairs (of ft = N and the point of suspension's normalized distance from
the fiee ends of the bar) must be read from the specimen. Data pairs were
read starting at about 2 mm hour the free ends of the specimen, and continued
by stepping inward two to five mm per step. At least two data pairs were read
from inside the fundamental flexural node. For each position at which a data
pair was read, a minimum of ten 'free decay' values were taken. At least ten
readings were necessary, due to scatter in the number of times the output
voltage crossed the reference.

3.7. Preparation for scanning electron microscope
fractography.

Macor specimens that fractured after 500 to 2500 fatigue cycles were
prepared for scanning electron microscope (S.E.M.) observation by cutting off
the fracture surfaces at about one centimeter from the fracture. The same low
speed saw used for initial specimen preparation was used for this operation.

The shear crack in the GFRP specimens was of great interest, due to its
suspected influence on Young's modulus and internal friction. The existence
of a fiber bridged zone between the crack faces was investigated by cutting
GFRP specimen GB-4 in half across the width. One of the half pieces was
cut in half again lengthwise to expose both midsections (transverse and
longitudinal) of the bar. One of the quarter pieces of GFRP specimen GE-4
was partialy split open from the short midsection, approximately five
millimeters, to provide a look into the fiber bridged zone.

74

 

 

 

Driver

Astatic
62-1

 

Pick-Up
Astatic

 

62-1

 

 

 

 

 

 

/ W]?

 

Prismatic Specimen

 

H/PB325A

 

Synthes’zer/Function _

 

 

 

 

 

Itham 4302 Dual 24 (TB/Octave Filter

 

 

I mghrass]

 

 

 

 

 

| Low-1

 

Input Output Input Output

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Generator @ L—IQ qi)
From Si From
Set V2 Genet-ati?” Amplifier Hitdclril
Oscilloscope
e "—9 " v-rroo A
100 MHz
Off IEree Decay Controller Unit I Decay HIP
5314 A
To Voltrneter To Exciter To Counter Universal
CD I—9 G_ 1 Counter
\J ’ Fluke 8050 A
Digital
Multimeter

Figure 16. Block diagram of sonic resonance system.

 

 

 

75

All specimens and mounting stubs were cleaned using methanol in an
ultrasound cleaner for 20 to 30 minutes. Both fracture surfaces (each side of
the fracture) from each specimen were mounted on the same stub, with
fracture surfaces oriented normal to the electron beam for good depth and field
of focus.

The mounting stubs used were prepared from 1018 carbon steel. Stub
dimensions were 1/2 inch in height and 5/8 inch diameter, to fit the receiver in
the stage of the S.E.M. Clean fractured specimen sections were glued to the
stub with 'DUCO' brand household cement' (Devon Company, Wood Dale,
IL 60191). The glued-up specimens were dried for at least 20 minutes under
two 150 watt light bulbs in close proximity.

The nonconductive specimens were sputter coated with gold to allow the
electron beam to be conducted away from the surface of the specimen without
spot charging on the surface. Sputter coating was accomplished at the Center
for Electron Optics, located in the Pesticide Research building, on the M.S.U.
campus. The sputter coating device used was an Bmscope Sputter Coater,
Model SC 500, manufacturedby Bmscope Company, Wotton Road, Ashford,
Kent, UK.

The sputtering procedure began with setting the prepared specimens
into the sputtering chamber. The chamber was evacuated to approximately
0.06 Torr. The chamber was then purged with argon gas and pressure was
maintained at approximately 0.1 Torr. The coating parameter of 20 ma
cunent gave a coating rate of about 7 nanometers per minute, per
’ recommendation in the Bmscope's operation manual. Exposure times of 30
seconds (Macor) and 90 seconds (GFRP) gave a coating thickness of 3 to 4
nm (Macor) and 10 to 12 nm (GFRP). The chamber was then flooded with
argon and pressure was equalized to atmospheric pressure. The cover of the
chamber was then raised, and the specimens were removed. After sputter
coating, the specimens were kept in a small, covered plastic box to prevent
contamination by handling and dust from the environment.

4. Results and discussion.

4.1. Thought experiment model for acoustic emission
during monotonic increasing flexure in a GFRP
composite.

Using a thought experiment, based on the literature [6, 7, 50-65], one
can envision the 'noise generating' mechanisms that are operative during the
loading of the unidirectionally reinforced GFRP composite in four point
bending. Upon initial loading, on the tensile face of curvature, the epoxy
matrix begins to crack transversely to the fibers. The crack spacing is
roughly proportional to both the maximum stress and the volume fraction of
fibers [52]. Matrix cracking likely produces acoustic emission events of
relatively high intensity, i.e. amplitudes of about 60 to 70 dB, high numbers of
counts per event (20 to 100), temporal durations of 200 to 300 microseconds,
and short to medium rise times of about 50 to 100 microseconds [59-63].

As the unidirectionally reinforced matrix fails, the reinforcing fibers
begin to break and pull out of their holes in the matrix, bridging the crack wake
[56]. As pullout and bridging occurs, an initial 'snap' may occur as the fiber
breaks and the interface breaks loose due to the high elastic modulus of the
fiber, followed by frictional AB as the fiber is dragged out of its hole [56, 59].
The level of frictional AB from pullout is likely quite low, however, in view of
the sheer number of fibers participating in pullout, we may detect this
emission as a general low level 'tail' following a high intensity burst as the
matrix cracks.

As an aside, for a randomly oriented fiber or whisker distribution, few
fibers will be oriented normal to the crack surfaces. Many fibers will break
immediately due to the bending moment imposed by geometry. The majority
of fiber breakage likely occurs simultaneously with the opening of the matrix

76

77

crack. Therefore, AB activity from fiber breakage and matrix cracking may be
recorded together, causing a larger number of counts per event to be detected
per AE burst. The combination of fiber breakage and matrix cracking is not
seen as a completely adverse situation because, although inseparable, both
sources contribute to 'new damage' emission.

With higher loading, the unidirectional GFRP composite's bridging
fibers begin to break and further matrix cracking, fiber pullout and shear
cracking or interply splitting (due to fiber-matrix debonding) may occur. At
this point, the dominant acoustic source is likely the shear crack, due to the
large surface area of its faces in frictional contact with each other [59, 60].
The magnitude of the fiictional AB generated by the shear crack may likely
overshadow or mask any other acoustic mechanism in operation.

Determining the active acoustic emission mohanisms is a matter of
some subjectivity and philosophy, based on one's understanding (or lack of
understanding) of the damage modes and their progression during fatigue
failure.

4.2. Determination of uncertainty of internal friction and
Young's modulus measurements made on the sonic
resonance system.

A series of measurements was performed on the sonic resonance-
system, to determine the uncertainty or error margin for readings made during
the course of the experiment. Measurements were made at two different
suspension points on the specimens, to check for any difference in data scatter
due to the relative positions on the bar. One set of readings was made at 0.12
normalized distance (n.d.) from the free end of the bar, and another set of
readings was made at 0.32 n.d. The fundamental flexural node was located at
approximately 0.223 n.d. from the fiee ends of the bar, thus the suspension

78

positions were equidistant on either side of the node.

The working standard specimens for each material were used for this
exercise. Both working standard specimens were cut from the same parent
plates for each material. The dimensions of the working standard specimens
were kept as close as possible to the average dimensions of the specimens
actually used for the experiment (width being the only dimension in control;
length and thickness remained as received).

To determine the error in Young's modulus, twenty individual
determinations of the fundamental flexural resonant frequency were made.
Since Young's modulus is related to the square of the resonant frequency

(Equation 5, Section 2.4.1), the values for F, were converted into values of

Youngs' modulus, then statistically analyzed for the mean and standard
deviation.

The uncertainty in internal friction was determined by making eight
groups of ten readings. Each group, consisting of ten individual readings of
V1, V2 and N (Equation 6, Section 2.4.2) were converted into a group mean,

and a standard deviation as the measure of scatter.

For all readings or groups of readings, the specimen was either
completely removed from the hanging threads and then rehung, or the threads
were loosened and moved about, simulating removal and then repositioned.

Table 2 lists uncertainty data for both Young's modulus and internal
friction, sorted by specimen material and suspension location. For the GFRP
material, uncertainty (one standard deviation) in Young's modulus at 0.12 n.d.
is about 0.027 percent, and at 0.32n.d. is about 0.048 percent. The error in
measurement of internal fiiction at 0.12 n.d. is about 2.3 percent and at 0.32
n.d. is about 3.4 percent, which indicates the readings made at 0.12 n.d., i.e.
outside the flexural node, may be more accurate by up to forty five percent for
Young's modulus and up to thirty percent more accurate for internal friction.

79

 

 

 

 

 

Table 2. Uncertainty data for sonic resonance measurements of Young's
modulus and internal friction.
Normalized Mean ** Std. Mean Std.
Specimen Hanging Young's Dev. Q'l Dev.
Position * Modulus ** *** ***
GE-2 0.12 43.694 0.0119 2.789 0.0645
(GFRP) (8 mm) (GPa) (GPa) x 10‘3 x 10'3
0.32 43.653 0.0209 2.762 0.0935
(24mm) x10‘3 x10‘3
MA-l6 0.12 62.148 0.0088 5.594 0.136
(Macor) (9 mm) x 10'4 x 10'4
0.32 62.084 0.0049 3.851 0.07 1
(24 mm) x 10'4 x 10‘4

 

 

 

 

 

 

 

* The nodal position for the fundamental flexural vibrational mode is
located at a normalized distance of 0.223 from the free ends of the bar. The
normalized distances listed here are also measured from the fiee ends of the

bar.

** The mean and standard deviation for the measurements of Young's

modulus were based on twenty measurements per suspension position per

specimen.

*** The mean and standard deviation for the measurements of internal friction

(Q'l) were based on eight groups of ten readings per suspension position per

specimen.

 

80

For Macor, the uncertainty in the measurements of Young's modulus
at 0.12 n.d.was about 0.014 percent, and at 0.32 n.d. about 0.008 percent.
The error in internal friction measured at 0.12 n.d. was about 2.43 percent and
at 0.32 n.d. was about 1.84 percent. This error indicates that the readings at
0.32 n.d. may be more accurate by about 43 percent for Young' modulus and
24 percent for internal friction.

No judgement or explanation can be made regarding the observation of
less data scatter at the measurement positions inside the flexural node for the
GFRP specimen and outside the flexural node for Macor. However, the
differences in absolute values of the two physical properties and the differing
types of reinforcement may have some influence.

Small errors are largely a function of the experience and ability of the
operator. In order to minimize error, great care must be taken to ensure the
accurate positioning of the specimen suspension threads. It is essential to
develop a consistent procedure for finding the reSonant frequency for Young's
modulus, and the setting of a threshold voltage for a similar number of 'N'
counts for internal friction. By showing experimental uncertainty to be
self-consistent and of a relatively small magnitude, the accuracy and reliability
of other measurements, as the detection of fine degrees of change, can be
' realistically interpreted.

To account for the possibility of transducer drift in the response of the
sonic resonance system, periodic measurements (made over five months),
using the GFRP working standard specimen, showed that the difference in
values for Youngs' modulus varied by approximately :1: 0.010 GPa (0.023
percent, approximately the same as the measurement scatter). The minor
variation indicated negligible change in the response of the transducers over
the time of the study. The average modulus of the GFRP control specimen
was 43.808 GPa.

Statistically, if a measured 'change' of Young's modulus or internal

81

fiiction falls within the standard deviation for measurement , it is less

likely that there will be a real difference in actual values. A measured
difference of two standard deviations implies a 68 percent probability that
actual change has occured. A difference of four or more standard deviations
implies that the measured change is of high probability to be real.

4.3. Machine noise qualifications.

Three separate tests were run to determine the effect of machine or
system noise on the acoustic emission detected from the specimen during
fatigue.

It is reasoned that, during mechanical fatigue, the quietest possible
specimen will be the most flee from extraneous noise. Machine functions,
such as gears and chains reversing direction and electrical relays clicking, are
potential sources of noise, as well as the action of physically fatiguing the
specimen, where the loading fixture contacts the specimen.

4.3.1. 22.3 Newton maximum load fatigue test.

A fatigue test of one thousand load cycles was run on GFRP specimen
GE-5, with a 22.27 N (five pounds) maximum load. This test provided two
important results for interpreting the data from the higher load fatigue tests of
the GFRP composite and the Macor composite.

Acoustic emission transducers were placed on the flour face of the upper
load platen and on the middle of the tensile face of the specimen. Two of the
four channels available with the PAC. acoustic emission computer were used
for signal processing. A load scaled voltage signal was connected to the
parametric input of the AB computer to allow sorting of AB by load. Both
channels would occasionally go for several load cycles without detecting any

82

acoustic events, indicating that machine noise, along with damage noise, is an
inconsistent source.

The AB observed during the 22.3 Newton maximum load test of GFRP
specimen GE-5 included substantial AB events at both ends of the load range
(Figure 17). Very little or no damage was expected to occur with the low
maximum load of 22.3 N. However, as indicated by changes in internal
friction and Young's modulus (Tables Ba and c), some damage probably
occured. A maximum change in Young's modulus of 0.07 GPa was
measured at 0.12 n.d., which is 0.17 percent of the mean value.
Experimental uncertainty (Table 2, page 80) of 0.012 GPa (smaller than the
measured change in Young's modulus by a factor of 5.8), measured at the
same suspension location, indicates that a real change has occured. This also
irnplyies that some amount of real damage has also occured. The implication
of damage is supported by a change of internal friction of 2.23 x 1033,
measured at 0.12 n.d. (Table 3.a). Experimental uncertainty in the
measurement of internal fiiction, determined to be approximately 0.065 x 10‘3
(Table 2), also measured at 0.12,n..d., indicated that the change was larger
than the uncertainty by a factor'of 34 times (Table 3.c).

If no damage had been indicated, then the accumulation of AE events at
the ends of the load range would suggest that machine noise is a function of
the number of loading cycles. Evidence of damage suggests that a percentage
of the AB events detected at the ends of the load range may be due to real
damage in the specimen.

A second result of the 22.3 N fatigue test indicates that machine noise
does not transfer to the specimen. Events of high relative amplitude and over
one hundred counts were detected by the load platen transducer, but in the next
few events recorded, no event was detected by the transducer located on the
specimen, or several seconds would elapse before the next event was detected
at either the specimen or load platen.

.533 32 can: no e352: coon—cash m< 3 dos—Moon» no
@8558 80:35: m< Am .23 came: 0:388 Z Qua ED c8 among 32 com 358 commence cum—38¢. .3 flamma—

 

 

 

 

 

g 50 . EHUQOZflz—ID‘IZGdN
ESE:— x 2”.me nae—mm Wmmmzé 33 3 «SEE: a 8 2.3 8:5» 9‘ .35.. 58 Q
388. 889.8823 2.8.8. 883-. 823.
1 IA ‘I7_I(Iii.m\..iul I1
9/ o . l
j cocoa fl
4 l. .2.
r 488 m , H
I 10.3 mm .. 0.8“ mm
. 4 \
5.88 n
B . / .. mm A mu
3 r 0.5 0..
3+8; r.
a L .4 mm . mm
.333 0 w.
mu 1 88 mm
a
a 12.9.. mm 4 mm
a
43.8.. m r 38
A a
e i... testing?
a 4. goo-8.5! . ..
888.83! 4 8+8.“ £883! .,
£23833 . figs-83! .
£8.85! of £8.83! a
F r insane: . sea manages: ..

 

 

84

Table 3 .a. Maximum changes of Young's modulus and internal friction for
GFRP fatigue specimens, measured at 0.12 normalized distance from end of
specimen bar. Data for working standard is provided for comparison of
measurement scatter to actual measured change in property.

 

 

 

 

 

 

 

 

 

 

 

Specimen Young's Modulus (GPa) Internal Friction

(max load) Max Min A Pet. Max Min A Pct.

[# cycles] [# cycles] [# cycles] change [# cycles] [# cycles] change

GE-S 41.240 41.170 0.07 0.170 4.95 2.72 2.23 58.15

(22.3 N) [0] [500] (no-3) (no-3)

[1000] [500] [150]

GE-l 41.643 41.486 0.157 0.378 4.98 2.35 2.36 64.40

k1 1 1.3 N) [20] [50] (x10-3) (no-3)

[10 K] [100] [250]

613-3 41.822 41.713 0.109 0.261 4.54 3.00 1.54 40.85

(155.8 N) [0] [250] (1:103) (1:103)

[10 K] [10] [500]

$6134 43.776 43.653 0.123 0.281 4.00 3.05 0.95 13.77

(200 N) [0] [150] (1:103) (1:103)

[10 K] [350] [3]

Error in Mean = 43.695 Std. C.O.V. Mean = 2.789 Std. C.O.V.
mt of (GPa) Dev. * (x10‘3) Dev. *

git-2 (Working Std.) .01191 .027% (Working Std.) .064 2.3%

 

 

 

 

 

 

 

* C.O.V. is the coefficient of variation, defined as the standard deviation

divided by the mean.

 

85

Table 3.b. Maximum changes of Young's modulus and internal friction for
Macor fatigue specimens, measured at 0.075 normalized distance from end of

specimen bar. Data for working standard is provided for comparison of

measurement scatter to actual measured change in property.

 

 

 

 

 

 

 

 

 

 

 

 

 

Specimen Young’s Modulus (GPa) Internal Friction

(max load) Max Min A Pct. Max Min A Pct.

[# cycles] [# cycles] [# cycles] change [# cycles] [# cycles] change
MA-12 62.67 62.51 0.16 0.256 15.80 11.90 3.90 28.16
(.75 MOR) [0] [250] (x104) (x104)

[5000] [500] [50]

MA—ll 64.09 63.98 0.1 1 0.172 15.30 12.40 2.90 20.94
(.65 MOR) [1000] [0] (x104) (x104)

[5000] [0] [1000]

MA-9 62.34 62.26 0.08 0.128 14.70 12.40 2.30 16.97
(.50 MOR) [0] [500] (x104) (x104)

[2500] [100] [500]

MA—7 62.31 62.28 0.03 0.048 11.65 10.55 1.10 9.91
(.40 MOR) [500] [1001 (x104) ' (x104)

[5000] [0] [1000]

MA—6 62.02 61.98 0.04 0.064 13.60 17.60 4.00 25.65
(.40 MOR) [100] [500] (x104) (x104)

[1000] [0] [500]

MA-4 62.13 62.04 0.09 0.145 10.725 8.454 2.271 23.68
(.30 MOR) [500] [0] (x104) (x104)

[5000] [5000] [0]

MA—2 63.58 63.52 0.06 0.094 10.70 5.60 5.10 62.58
(.20 MOR) [3] [250] (x104) (x104)

[500] [250] [0]

Error in Mean = 62.148 Std. C.O.V. Mean = 5.594 Std. C.O.V.
msmt of (GPa) Dev. (x104) Dev.

MA-16 (Working Std.) .0088 .015% (Working Std.) .136 2.43%

 

 

 

 

 

 

 

 

86

Table 3.c. Maximum changes of Young's modulus and internal friction for
GFRP fatigue specimens, measured at 0.32 normalized distance from end of
specimen bar. Data for working standard is provided for comparison of
measurement scatter to actual measured change in property.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Specimen Young's Modulus (GPa) Internal Friction

(max load) Max Min A Pct. Max Min A Pct.

[# cycles] [# cycles] [# cycles] change [# cycles] [# cycles] change
TEE-5 41.190 41.180 0.01 0.024 4.70 3.28 1.42 35.59
(22.3 N) [50] [0] (x10‘3) (x10'3)

[1000] [50] [150]

613-1 41.530 N/A N/A N/A N/A 2.02 N/A N/A
(1 11.3 N) [20] * (x10‘3)

[10K] [250]

"GE-3 41.720 41.618 0.102 0.245 4.00 2.43 1.57 48.83
(155.8 N) [0] [2000] (x103) (x10‘3)

[10K] [50] [2000]

613-4 43.651 43.577 0.074 0.170 6.05 2.45 3.60 84.7
(200N) [0] [150] (x10‘3) (x10'3)

[10K] [4001 [3]

Error in Mean = 43.653 Std. C.O.V. Mean = 2.762 Std. C.O.V.
msmt of (GPa) Dev. (x10'3) Dev.

GE-2 (Working Std.) .021 .048% (Working Std.) .094 3.38%

 

* N/A = not applicable, due to impossibility of taking a reading at that

position, at that time.

 

87

Table 3.d. Maximum changes of Young's modulus and internal friction for
Macor fatigue specimens, measured at 0.32 normalized distance from end of
specimen bar. Data for working standard is provided for comparison of
measurement scatter to actual measured change in property.

 

 

 

 

 

 

 

 

 

 

 

 

 

Specimen Young's Modulus (GPa) Internal Friction
(max load) Max Min A Pct. Max Min A Pct.
[# cycles] [# cycles] [# cycles] change [# cycles] [# cycles] change
MA-12 62.393 62.343 0.05 0.080 5.369 4.396 0.973 19.94
(.75 MOR) [0] [2500] (x104) (x104)
[5000] [500] [100]
MA-ll 63.826 63.794 0.032 0.050 4.655 3.892 0.763 17.85
(.65 MOR) [0] [2500] (x104) (x104)
[5000] [500] [0]
MA—9 62.123 62.088 0.035 0.056 5.434 4.708 0.726 14.32
(.50 MGR) [0] [1000] (x104) (x10'4)
[2500] [2500] [1000]
MA-7 62.152 62.127 0.035 0.056 5.311 4.309 1.002 20.83
(.40 MOR) [0] [2500] (x104) (x104)
[5000] [2500] [0]
MA-6 61.850 61.833 0.017 0.027 5.433 4.340 1.093 22.37
(.40 MOR) [0] [1000] (x104) (x104)
[1000] [500] [0]
MA-4 61.934 61.903 0.031 0.050 5.386 3.986 1.400 29.88
(.30 MOR) [500] [5000] (x104) (x104)
[5000] [100] [0]
MA-Z 63.413 63.380 0.033 0.052 4.523 3.059 1.464 38.62
(.20 MOR) [3] [500] (x104) (x104)
[500] ' [500] [0]
in Mean = 62.084 Std. C.O.V. Mean = 3.851 Std. C.O.V.
t of (GPa) Dev. (x104) Dev.
-16 (Working Std.) .0049 .008% (Working Std.) .136 2.43%

 

 

 

 

 

 

 

 

88

To determine the source of the high numbers of events at the ends of the
load range, two other runs of this experiment were performed, with slight
hardware differences. One thousand load cycles were performed in each run.

In one test, a 122 Newton weight was placed on the crosshead to
prevent rocking of the crosshead during load reversals. The accumulation of
AE events at the high end of the load range was substantially reduced to a level
even with the midstroke AE , while the number of AE events at the low end of
the load range was reduced by a factor of about three (from about 750 down to
about 260).

The addition of a weight to the crosshead implies that the majority of AB
events at the high end of the load range are likely to be due to the crosshead
rocking slightly during the load reversal from increasing to decreasing load.
At the low end of the load range, there was a significant addition of extraneous
AE by crosshead motion at load reversal. Therefore, the results of further
fatigue tests must be interpreted on that basis.

A second variation of the 22.3 N maximum load fatigue test involved
the same 122 N weight on the crosshead as before, but with the addition of
plastic sleeves to the load pins, and a thin coat of lubricant (the same silicone
vacuum grease used to couple the AB transducer to the specimen) between the
specimen and the sleeved load pins. Plastic sleeves two centimeters long were
made from the smooth insulation casing of a length of scrap electrical wire.
The casing had a slightly smaller inside diameter size than the load pin, and
was forced over and centered on the pin. Again, the specimen-mounted
transducer showed that the high end of the load range showed no more AE
events occuring than during the mid-range of loading (the same result as the
high end of the 'crosshead weight only' test). A substantial increase in the
AE events detected at the low end load reversal was noted, with a total
accumulation, during 1000 load cycles, of about 6200 counts. The addition
of lubricated plastic sleeves to the load pins evidently caused a detrimental

89

increase in the number of AB events at the low end load reversal. The
increase was possibly due to a 'squeaking' action of the specimen moving
against the sleeved load pins at that point.

The best of the test conditions indicated by the results of the three 22.3
Newton fatigue tests are: (1) add the weight to the crosshead, which prevents
its rocking, thereby quieting down the load reversals - especially the high end
load reversal, (2) do not add plastic sleeves to the load pins, which
significantly raised the number of AB events at the low end load reversal.

The interpretation of AB-fatigue data acquired with no weight on the
crosshead or sleeves on the load pins must acknowledge that the primary
source of AB at the points of load reversals is likely due to crosshead motion
relative to the specimen. AB detected during midstroke is likely due to
specimen damage.

There appears to be evidence supporting the possrbility flrat all AB
detected at the ends of the load range is not due to the fatiguing system. At
this point, it is uncertain what fraction of the AB is attributable to either the
specimen or to the system. It is not justifiable to subtract some fraction of AB
events or counts from those detected at the end ranges of load for other
specimens involving higher loads and more fatigue cycles. Therefore, we do
not place undue emphasis on the reliability of the AB detected at either the high
or low load ranges where load reversal occurs.

4.3.2. Stepped maximum load fatigue test.

An undamaged GFRP specimen was load cycled for one hundred cycles
at each of five load levels. Load levels started at 44.5 Newtons (ten pounds),
and increased by 44.5 N each change up, to a maximum of 222.5 N (fifty
pounds). Three experimental sets of conditions were tested. The first set of
test conditions used no weight on the crosshead and no plastic sleeves or

90

lubrication on the load pins. This setup is exactly the same as that used
during the actual fatigue runs of the GFRP and Macor specimens. The
second set of conditions was to add only the 122 N steel weight to the
crosshead to prevent its rocking. The third set of conditions used the
crosshead weight and added the same plastic sleeves to the load pins as
described in Section 4.3.1 . Acoustic emission transducers were placed on the
middle of the specimen's tensile face and on the front face of the upper load
platen.

Under the first set of conditions, acoustic emission detected on the
specimen at the low end of the load range increased with each increase in
applied load, to a total of about 165 counts. At the high end of the load range,
accumulation of AB occured slightly faster than in the middle of the load range,
but at a much lower rate than at the low end, totaling only about 50 counts.

AB detected by the transducer mounted on the load platen was
significantly higher across the entire load range. About 3300 counts were
observed at the low end of the load range, about equal in number from each
increasing load. About 1200 counts were observed at the high end of the load
range, decreasing in number of counts with each increasing step in load.

Under the second set of conditions, the transducer on the specimen
detected far fewer acoustic events at the high end of the load range, rising only
slightly from the level of AB detected throughout the middle of the load range.
At the low end of the load range, the specimen mounted transducer detected
about 50 counts during each step of load increase. In the middle of the load
range, the distribution of acoustic counts was quite smooth.

The AB transducer mounted on the load platen detected about 100 more
events at the high end than at the low end of the load range. The middle of the
load range exhibited several high intensity events in each step of increased
load, but at no consistent absolute load.

Under the third set of conditions, the AB detected by the transducer

91

mounted on the specimen was very similar throughout the load range from the
middle to the high end of the load range. AB detected at the low end of the
load range totaled 2715 counts with about 100 counts detected up through the
second load step, and about 1200 in the third step, 900 counts in the fourth
step and finally about 400 counts in the last (highest load) step.

AB detected at the upper load platen was higher at the high end of the
load range, totaling about 1100 counts, with about 80 counts observed in each
of the first four load steps, and about 800 counts detected during the highest
(222.5 N) load step.

The result of the series of stepped load fatigue tests is that: (1) acoustic
emission detected at the high end of the load range (under experimental
conditions of no crosshead weight or sleeves on the load pins) is largely due to
crosshead movement during load reversal (from increasing load to decreasing
load), and (2) there is little or no evidence of communication of acoustic
energy from the crosshead to the specimen.

4.3.3. Kaiser effect test.

An increasing series of loads was applied to GFRP specimen GE-7,
starting with 45.5 N, stepping up each time by 44.5 N, up to 222.5 N with
one application of each load level. Test conditions were the same as for the I
first set in section 4.3.2., i.e. no weight on the crosshead and no sleeves on
the load pins. AB was detected with one transducer, mounted on the
specimen. 9.

AB accumulated during the application of every load. Only during the
removal of the first (44.5 N) load was AB detected during load relaxation.
According to Awerbuch et al. [60], the lack of acoustic events (from the
specimen) at the low load condition implies that machine noise is undetectable
or not present.

92

The Kaiser effect test indicates that essentially no AB was detected
during load release, possibly implying low or no machine noise entering the

system through the specimen.
4.3.4. Comparison of AE from load pin and specimen.

As stated in Section 2.5, the use of the load pins of the four point bend
fixture as AB wave guides was investigated.

Adjustment of amplifier gain and threshold settings on the P.A.C.
amplifier, set to detect approximately the same number of AB events, indicated
that an amplitude difference of approximately -50 to -65 dB existed between
mounting the transducer on the lower span load pin and directly on the middle
of the tensile face of the specimen. The lower span load pin sits in the part of
the fixture that sits on the load cell, which in turn is bolted to the frame of the
Instron (Figure 14). The use of the load pin in the upper part of the fixture
was ruled out due to the results of tests involving mounting the transducer on
the upper load platen (see dicussion in Sections 4.3.1 and 4.3.2).

For the same number of events, the AB events/load distribution profile
taken from the load pin was drastically different from that taken directly from
the middle of the tensile face of the specimen. The distribution profile taken
from the loading pins showed a nearly flat profile with a few apparently
random spikes. The AB events/load distribution profile from the specimen-
mounted transducer showed apparently good response with lower sensitivity
settings, eliminating background noise, indicating a consistently poor acoustic
coupling between the load pin and the specimen. Thus, the load pin cannot be
used as an AB guide path since the pins attenuate AB signals by about 60 dB.

The poor acoustic coupling between the specimen and the load pins also
works to an advantage in reducing machine noise by 50 to 65 dB.

4.3.5. Results of machine noise tests.

The effect of mechanical fatigue loading on the AB detected from the
specimen is to inflate the number of events detected at the high end of the load
range, based on the reductions in AB seen when the crosshead is weighted.
The AB detected during the middle of the load stroke appears to be largely due
to the specimen. At the low end of the load range, where load reversal from
decreasing to increasing load occurs, no reduction in the number of AB events
detected could be made. Adding sleeves to the load pins significantly
increased the number of AB events detected, implying that the AE detected at
the low end of the load range may be actually generated by the specimen.

4.4. Results of fractography.
4.4.1. GFRP fractography.

A midplane shear crack (interply delamination) grew throughout all
fatigue loaded GFRP specimens, except specimens GB-S which experienced a
fatigue load of only 22.3 N, and specimens GB—6 and GB-7, which were used
for the stepped-load machine noise tests and saw only limited fatigue loading.

Examination of the thickness side of the fatigue loaded GFRP
specimens in a 10 to 250 X inverted stage optical metallographic microscope
did not reveal the shear crack length at any magnification. The specimen was
also viewed in a 10 to 150 X stereomicroscope, and again no crack was visible
under any lighting condition or magnification. It seemed implausible that the
shear crack did not intersect the side surface of the specimen, especially when
the unaided eye could obviously and clearly see the crack demarked by lighter
and darker halves of the specimen separated by the crack, when viewed near a
strong incandescent or fluorescent light. The only explanation for the

93

94

invisibility of the crack at the suface may be a tightly tied 'fiber bridged zone'
between crack faces, holding them together with the nearly perfect mating and
closure of the crack faces.

A scanning electron micrograph of the thickness side of the GFRP
specimen GB-4 (where the midplane crack was believed to intersect the
surface), confirms the intersection of the crack with the surface of the.
specimen (Frgure 18). The crack opening displacement is estimated at about
6.5 microns. The extensive tied zone is visible at a magnification of 250 X
(Figure 19). The ease at which the quarter piece of the GFRP specimen GB—4
was opened indicates that interply delamination had previously occured.

4.4.2. Macor fractography.

During fatigue cycling, Macor specimens typically fractured upon the
initial contact of a load pin on the first cycle in a new loading sequence.
Breakage always occured at a short span load pin, and initiated at the
edge on the tensile side of flexure, along a polishing or grinding mark in the
'as-recieved' surface. These fractures were all due to accidental overloads.

Representative scanning electron micrographs of the fracture surface at
the edge of the free surface and at about fifty microns from the tensile surface
of three Macor specimens were taken. Specimen MA-2 fractured after 500
load cycles at 0.20 M.O.R. (Figures 20.a and b). Specimen MA-6 fractured
after 1000 load cycles at 0.40 M.O.R. (Figures 21.a and b). The third
specimen, MA-9, fractured after 2500 load cycles at 0.50.M.O.R.

(Figures 22.a and b).

Edge-on micrographs at magnifications of 750 X show no recognizable
evidence of fatigue cracking (Figures 20.a, 21.a, and 22.a). There is also no
evidence of incremental fatigue cracking in the micrographs taken at about fifty
microns from the tensile surface at 750 X (Figures 20.b, 21.b and 22.b).

95

3,
_..
.TV

.2

.rfli n

...
. ...
. . . \

’ x-..

g .. . X
. illia .

. . .1. a .. . . nu
\ 7., \o n w
.. ... .V. . v .

«\‘L1 t N,

3.11:4;1utttét..: a.

1.. a
K.

. v; r.
. .4 in no .
.n‘« 9 ob.
.. r' I... .~
.2 .x
. we.
as. ..r,..\

.ue......§=..i....

a)

r \ ~ ~
\.l.v KO '9. 0".

.a‘dnfidgc.mmm:...1
- . .r . .. o

 

Figure 18. Scanning electron micrographs of the midplane shear crack
(specimen GB-4) common to all GFRP fatigue specimens. a) evidence of

multiple delarninations connected by transverse cracking, 185 X,

b) typical shear crack, 300x.

o

“”WH‘IW‘ U

.
r
.48,
. t.
>~>
'Ox
“fig 7
.u.
U
.2... .
4

uh

wen“

 

"‘ :vvhfimfii‘

*-
.p.
I

W‘ ’ 8385 25K ..
m

Figure 19. Scanning electron micrographs of the partially opened midplane
crack of specimen GE-4, clearly showing the fiber bridged zone.
Both micrographs taken as close to root of crack opening as possible.

a)

 

b)

Figure 20. Scanning electron micrographs of the fracture surface of Macor
specimen MA-2. a) view of fracture at edge of tensile face, 750 X,
b) view of fracture near tensile surface, 750 X.

98

 

Figure 21. Scanning electron micrographs of the fracture surface of Macor
specimen MA-6. a) view of fracture surface at edge of tensile face, 750 X,

b) view of fracture near tensile surface, 750 X.

a)

 

Figure 22. Scanning electron micrographs of the fracture surface of Macor
specimen MA-9. a) view at edge of fracture surface, 750 X,
b) view of fracture surface near tensile surface, 3000 X.

100

From the lowest to highest fatigue loads however, there is an increasing
angularity and sharpness of mica platelet definition in the fracture surface
morphology. The specimen with the lowest fatigue load appears to have the
smoothest surface, indicating that fracture occured through the matrix phase of
the microstructure, away from the mica platelets (Figure 20). The specimen
that experienced the highest fatigue load shows sharply defined mica platelets,
indicating that fracture occured along or closer to the mica platelet (Figure 22).
The specimen subjected to the median fatigue load, shows fracture surfaces
with an appearance less smooth as those of the specimen with the lower fatigue
load, but less shame defined as those of the specimen with the highest fatigue
load Figure 21 .

The trend in the appearance of the fracture surfaces (progressing from
smoother to sharper morphology with increasing fatigue loads) may, however,
be more a function of the maximum load applied to cause the fracture, possibly
increasing the rate of fracture.

4.5. Analysis of fatigue data.
4.5.1. GFRP - AE analysis with and without load scaling.
4.5.1.a. GFRP - AE analysis without load scaling.

When analyzing mechanical damage in terms of acoustic emission, one
must consider and exploit all means and types of data analysis available.
Without the ability to scale acoustic emission with its corresponding load, the
identification of some signature of the various modes of damage would be
highly valuable.

For the Macor specimen runs of MA-l through the first 2500 cycles of
MA-l l , the capability to scale the AB with load was temporarily lost.

101

However, the intensity of the AB events as measured by counts, energy,
amplitude, rise time or duration may be screened by setting a higher threshold
for those parameters and analyzing flre data above that threshold. Load scaling
of AB was used by Lewis and Rice who also described the frictional emission
level [5, 6].

By knowing the number of cycles at which the AB events occured,
Lewis and Rice suggest that plotting the AB parameter of interest on the
ordinate and the number of load cycles along the abscissa. This approach is
fatally flawed, however, since AE bursts are broadened in their frequency
distribution and attenuated with distance, as proven by Evans [35]. A distant
pulse originating from crack damage may appear (to the transducer) as a pulse
of lesser amplitude, but of longer duration and higher counts, etc. than an
event that occurs very close to the transducer which may be due to frictional
contact of existing crack surfaces. Thus, the use of a threshold is not a clear
means of identifying AB sources, but only gives attention to the higher
intensity events. A threshold to truncate or discriminate sorrre AB data may
perhaps be determined after understanding the damage that the specimen has
experienced. This is still a somewhat subjective or emperical approach, in
that the determination of a threshold is not a clearly indicated decision.
However, even when looking into a thick fog, some nearby details are still
visible, and to some degree represent the rest of the world shrouded in the
fog. Thus, the plotting of AB counts versus load cycle is still of interest, and
is described further in Section 4.5.2.1.

A reasonable description of a crack's acoustic signature must be
independent of its distance from the transducer. Therefore, the parameters of
energy, amplitude, counts and duration and rise time by themselves are not
indicative of the acoustic event's character. The ratio of rise time to duration
of an event may indicate the source's character. When a crack 'snaps' into
existence, it may have an acoustic energy burst with a relatively short rise time

102

with respect to the duration of the ringing of the crack faces. Further, each
type of damage may have a certain ratio of rise time to duration. This energy
envelope profile model may lead to an 'n'-modal distribution.

Data from the Macor specimen MA-12 was analyzed for this possibility
over the fatigue cycle range from zero to 100 load cycles, where most
microcracking damage is expected to occur. A histogram of rise time divided
by duration for each acoustic event (Figure 23) showed one possible division
in the distribution of events. The occurance of acoustic events with rise time
too short to measure may be indicative of cracking damage. The events with a
measurable rise time appear to be normally distributed about a mean of about
0.2, and may be due to friction. This distribution did not indicate any further
divisions. The characterization of an acoustic event as from cracking or
friction may be made using the separate tests of 'notched beam' and 'direct
pullout' respectively. The performance of these tests was beyond the scope
of this study, and the search for a 'damage mode' signature was summarily
abandoned.

A clear identification of an acoustic signature of specific damage modes
is not possible, again supporting the analogy of the ringing of a large bell.
You may hear the bell ring, but can not say with confidence what caused it to
ring, from the sharp 'ping' of a clapper strike to the dull 'thud' of a person's
fist.

4.5.l.b. GFRP - AE analysis with load scaling.

Awerbuch [59-61] presented the AB analysis method of separating the
acoustic emission information into narrower sections of the total load range.

During the cyclic loading of all three GFRP composite specimens,
acoustic emission was detected at different rates and in different amounts in
different regions of the load range. Segregation of AB data by load range

103

 

 

 

 

 

 

 

 

 

 

fl)
N=51 to lOOcyclesz214events
97events@0
5
8
§
0 !I!II!! _____II_I!_I!
.001 .01 0.1 1.0
N=0to 50 cycles : 133 events
62events@0
5
8
‘>’
o I II III I!I---IIII¢! I
1.0

.001

Figure 23. Histogram distribution of rise time divided by duration per
acoustic event, for Macor specimen MA-l2, zero to 100 fatigue cycles.

104

revealed several common features for all three load ranges tested. These
common features will be discussed first, followed by specific differences.
Acoustic emission counts from GFRP specimens GB-l load cycled at 111.3 N
(Figure 24), GB-2 load cycled at 155.8 N (Figure 25) and GE-4 load cycled at
200 N (Figure 26) are sorted by dividing the load range into ten equal parts.
No other discrimination such as increased threshold was used.

4.5.l.c. Common AE trends in GFRP composite.

4.5.l.c.l. Load reversal effect.

For all GFRP specimens, acoustic emission counts accumulated much
faster (with respect to fatigue load cycles) at the points of load reversal.
Considering the results of the machine noise tests of Sections 4.3.1 and 4.3.2,
the AB generated by the specimen near the point of load reversal during the
regular fatigue test, must be interpreted as a mix of both specimen noise and
system noise. The proportion of machine noise to specimen AB signal is not
estimatable at this point, without further testing and is thus deemed beyond the
scope of this study.

The load transition from increasing to decreasing was much more
acoustically active than the load reversal from decreasing to increasing. As
shown in Section 4.3, the proportion of machine noise and the specimen's AB
signal cannot be determined, therefore only a qualitative discussion can be
made. A significant amount of acoustic emission from the specimen is
expected during load reversal, suggesting that the occurance of damage,
whatever its mode, is more dependent on load reversal, rather than on the
absolute maximum dynamic load (up to the point of catastrophic failure of the
composite). The total number of counts at the load reversal points is
dramatically skewed, but with no apparent trend between specimens at the

105

.33 0883 8.888
2 m. ~ 2 55 882 0:35 . 7mm 808608 550 .6.“ omen.— 22 .3 v8.8 888 82880 038814. .vu 25E

“law zmfiuommm Enema mmjo
Bomb umber—.3“ .mo mug Q40..— Bzmommm 2m?

 

coo-.2
com-.2
fl ooulz
.. cod-.2
I oblz
oolz
omlz
. Onlz
emaokwnlym

 

O
3
Sth00 NOISSINEI cusnoov

OXD>O+I<M

 

 

106

.33 38.8%.. 858838
2 w. m2 5.? 333 2883. mama 58.8% EU .3 3:8 33 .3 388 58.8 8.880 3.283. .3 EEE

ml mo zmzmommm .Cnomm mmju
30>". HDUFSm ho HUS 204 Hzmommm 2H9

SI. pm on on 8.LI“.II on 2. on on . 20
.‘n (IIIIIIIIII .I -Inl',
\

 

I

III II

 

 

 

V

0

0

n

S

u

.63. 3

. m

flocculz a com S

Sorz n n .6.

3 83-2 6 . m
.1 ouulz + .600

. 8...... o . m

. 8.... o . n

.. 8:2 n .68. m

3 Owl-z X . S

. 0'2 0 .

 

 

p . . . . . r DON _.

107

.32 3885. 88888
2 8m .23 332 0:33 :15 586on E0 88 09.8 33 .3 388 $850 83880 05:23 .8 0.88m

filmo zmzuommm .63an mmju
wacky H9032...“ ho Hug n40..— .uzmummm 2mg.

 

   

 

o2 om on on 8 on 9 on on o. .
L I w - o
V
82 m
n
m
coon D
3
ocoalz < Iooon m
one-z . S
coo-z x j E
2312 g 189. m
Sana «j
Sun: + m
and l2 0 flooon n...
2:..2 p. m
8.2 a T88 S
8.869.; a
_ _ — P P — _ -

 

 

108

different maximum loads.

The variation in number of acoustic counts for nominally identical
specimens is on the order of the square root of the number of counts [31].
Typical error bars may be estimated: for example, at 400 counts the first
standard deviation is approximately 20 counts (5 percent variation), but for 5
counts, the error bar of one standard deviation must be 2.24 (45 percent
variation), which is comparatively large. Therefore, at low numbers of
counts or events, one cannot place a great deal of significance on any apparent
trend.

The GFRP specimen with the lowest load (GB-l, at 111.3 N) showed
more counts at the high end of the load range (Figure 24). At 500 load cycles
approximately 730 counts accumulated. For the same number of cycles, the
specimens that experienced maximum fatigue loads of 155.8 N (GE-3) and
200 N (GE-4), accumulated approximately 550 and 5500 counts respectively
(Figures 25 and 26). The gross difference for the highest load specimen may
be attributable to the exceeding of a strength threshold, where a larger amount
of damage has occured.

4.5.l.c.2. Active zone - dead zone phenomena.

At particular load levels during the four-point bending fatigue cycle, AE
events are generated with approximately the same characteristics, i.e.
amplitude, counts, duration and rise time. Excluding the end regions, there
are acoustically active zones where more AB is generated, and between them,
acoustically quiet or 'dead' zones with very little acoustic emission generated.
This behavior may imply the engagement of certain asperities in the mating
faces of existing internal cracks during the load cycle. Acoustic emission
during either loading or unloading may be a function of asperity geometry. A
saw - tooth configuration may 'catch' on itself while loading in one direction,

109

but not in the other. The crack opening displacement likely also contributes to
the depth of engagement, which would affect the amplitude, duration etc. of
the AE burst.

At certain numbers of fatigue cycles (also described in the next few
paragraphs), the dead and active zones are either more or less pronounced,
suggesting a nonlinear evolution of damage with fatigue life.

In the very first few load cycles, all three GFRP specimen's AB / load
range distributions appear quite similar. Three ranges of low acoustic activity
(dead zones) are observed in approximately the same load ranges of 10 to 20,
30 to 40 and 80 to 90 percent of maximum load. This may suggest that the
GFRP composite is similarly sensitive to the initial familiarization with load,
regardless of the loads absolute value, within the limit of the material's
breaking strength. At this low number of AE counts, the corresponding error
bar (as the square root of the number ofAE counts) is almost as large as the
data and therefore this trend may be lost as more specimens are tested.

As fatigue life progresses, a large amount of AE activity is seen in
varying load ranges (the active zones), as indicated by non-parallel lines on the
plots of Figures 24, 25 and 26. This suggests that some limited damage
occurs at various load levels and at various numbers of load cycles.

For the lowest load GFRP specimen (GE-l, Figure 24), four dead
zones appear with some degree of activity between them, at 10 to 20, 40 to 50,
60 to 70 and 80 to 90 percent of maximum load. In the active zones,the
evolution of AE with fatigue life is most pronounced, suggesting the occurance
of damage for that part of the load range. The dead zones at 10 to 20 and 80
to 90 percent of maximum load are most prominent in this specimen, possibly
lending support to the earlier evidence that machine noise is most sensitive to
load reversals. In the load reversal regions, the AE count distribution appears
skewed toward the high end of the load range by an end-range AE count of
about 300 counts more than at the low end of the load range.

110

GFRP specimen GE-3 (Figure 25), fatigued at 156 N (9.5 percent
M.O.R.) shows two or three dead zones that develop at different numbers of
fatigue cycles. One dead zone starts at 30 to 40 percent of maximum load then
shifts to the 40 to 50 percent load range at around 50 load cycles, remains there
until about 500 load cycles, then shifts to the 60 to 70 percent load range in the
last 500 cycles of fatigue life. A second dead zone of lesser prominence
appears
between60to70 and70to 80percentofmaximumloadthena1so shiftsto the
higher load range of 70 to 80 percent at around 50 load cycles, then to the
80-90 percent range as the number of load cycles approaches 1000. The least
prominent dead zone, at the 10 to 20 percent load range disappears after the
first 10 load cycles. The overall AE count distribution for specimen GE-3 is
skewed toward the low end of the load range by about 650 counts.

Specimen GE-4 (Figure 26), experiences significantly more damage
than specimens GE-l and GE-2, as evidenced by the high number of AE
counts observed across the load range. Load reversal AB is especially
skewed toward the high end by about 2700 counts, approximately double the
number of AE counts observed at the low end of the load range. Since
system noise detected at the specimen was shown to not increase with load
(Section 4.3.2), it can be inferred that the specimen produces substantial AE
(and damage) at the high end load reversal.

The absolute number of AE counts at the high end of the load range
through 500 load cycles for specimen GE-4 is larger than the number of
counts observed for the specimens of lower maximum loads. This difference
varies by a factor of about seven for specimen GE-l and about eleven for
GE-3. This does not appear to support a trend.

For specimen GE-4, one long dead zone exists through the 40 to 90
percent load ranges (Figure 26). The lower ranges from 10 to 30 percent of
maximum load increase between 150 to 350 load cycles, suggesting that

111

significant damage has occured in that load range during those load cycles.

Acoustically dead and active zones may result from the frictional contact
of asperities during flexure. During the experiment, it was frequently
observed that for several to 20 or 30 consecutive load cycles, acoustic events
were consistently observed either on load up or down, but rarely both. The
numerical data often showed that successive events occured at approximately
the same load (to within a few tenths of a volt on the main parametric scale,
which represented the load signal from the Instr-on).

A second mechanism for acoustically dead and active zones varying
with maximum load may be an arrest and progression sequence of damage that
occurs periodically during loading. At higher maximum loads, with the
specimen being taken directly to the higher maximum load, there is certainly
more bending deflection. The larger deflection likely causes a larger amount
of damage and possibly reduces any ability of the material to arrest and release
the crack front. Factors which may affect such a mechanism may include: (1)
sensitivity to strain energy release rates, (2) changes in stress intensity, (3)
initial loading history or (4) a critical load which, when exceeded, reduces the
damage tolerance of the material.

4.5.2. Macor fatigue analysis.
4.5.2.a. Macor AE analysis with load scaling.
Analysis parallel to that done with the GFRP specimens was performed

on the data obtained from tests with the Macor specimens. Similar trends
appeared in the Macor data, but with subtle differences.

4.5.2.1.a. Macor specimen MA-12.

Macor specimen MA—12 was fatigued to 5000 load cycles without
failure. AE data was sealed with load. Fatigue loading was run at the highest
percentage of breaking strength, at which the Macor would survive flexure.

Three quarters breaking strength (0.75 of) worked out to an absolute

maximum load of 86.2 N.

The AE counts accumulated over 5000 load cycles were divided into
ten equal load ranges (Figure 27). With respect to the evidence of system
noise, from Sections 4.3.1 and 4.3.2, there is a very large amount of AE at
the point of load reversal fiom decreasing to increasing load. The distribution
of AE counts is markedly skewed toward the low end of the load range.
About 7300 counts accumulated in the zero to 10 percent range after 5000 load
cycles, while about 2200 counts were observed in the highest load range of 90
to 100 percent maximum load. This is a difference of a factor of about 2.5.
The fact that the AE distribution is skewed toward one end of the load range
does not indicate the presence of if trend.

The most variation in AB counts between load ranges was observed
during the first few load cycles. From then onward, the end regions, where
load reversal occured, were the dominant AE production ranges. A relatively
major amount of damage occured in the range of 60 to 70 percent of maximum
load. The parallel lines about the small peak at the 60 to 70 percent range
indicate that the load range became less active after the initial fifty load cycles.
Where one load range produces more AE than another, the result is seen
graphically as a further displacement in the AE count direction.

The minor AE peak at the load range of60 to 70 percent, occurs at an
absolute load of about 55.8 N. The 60 to 70 percent load range in MA-ll
showed no hint of the same level of emission, nor did MA-l 1's level of

112

113

.8898 no 81608 who .2 New .8 e8— 8883 8.8888
55 332 0:38 .--<2 5.58% .582 .8“ 038 32 3 vote» 8850 8:280 cum—50¢. SN 0.8—ww—

.@QSmKdnuZNdmn;EQQEEEE§H
mal; zgommm moo;
macs graham ho mug 9404 Emommm 2H9

cop cm on on cm on 0*

P p L

 

 

 

 

WWII”: \".\ V
II +. O
‘. .fl\fl\ - n
- S
- l
I
. O
.. 3
m
. S
.. 83 m
. 83:: o N
182...: a coon m
.. comma + . fl
.. Sfirz o N

Srz o e

F p F — — F b —

 

 

coon

114

emission at maximum load appear at the same absolute point in MA-12.
Therefore, the exceeding of a strength threshold is not suspected. This
evidence supports the significance of the load reversal as the principle source
of damage generated acoustic emission.

4.5.2.1.b. Macor specimen MA-ll.

This specimen was cyclically loaded to 0.65 of (an absolute maximum

load of 88.6 N). Lacking the acoustic history prior to the 1000 load cycle
mark, it is still observed that the AE from 1000 to 5000 load cycles is as
markedly skewed to the low end of the load range as was specimen MA-12.
In the lowest load range, from 1000 to 5000 load cycles, specimen MA-12
produced about 2000 AE counts, while in the same range, MA-ll produced
nearly 7000 AE counts. This difference is in apparent contradiction to the
intuitive expectation of more counts for a higher maximum stress (as a higher
percentage of rupture modulus). At the high end of the load range, both
specimens produced approximately 2000 AE counts.

4.6. Internal friction and Young's modulus change
over fatigue life.

The standard procedure for the determination of internal friction
(Section 2.1.l.a.), had to be slightly modified in order to continue analysis of
the GFRP composite. Damage to the unidirectionally reinforced GFRP
composite became increasingly inhomogeneous as the fatigue experiment
progressed. The inhomogeneous damage made the use of Wachtmann and
Tefft's nonlinear least squares curve fitting routine impossible, with
correlation coefficients less than 0.95 (actually sometimes down to around

115

0.4). Therefore, internal friction readings were taken at only two locations, at
0.11 and at 0.33 normalized distance (n.d.) from the end of the bar. The
reading at 0.11 n.d. lies outside the flexural node and out of the regions of
flexural stress and shear stress imposed by the fatigue loading. Two locations
were chosen to monitor the relative changes in Young's modulus and internal
friction, and assumes that the effects of fixnlring remain constant.

An unexpected shear crack developed in the midplane of each GFRP
specimen, initiating in the short load span (constant moment) area, and
progressed throughout the entire specimen. The shear crack may account for
the large differences in internal friction, measured at the two locations.

The estimation of the extent of crack progression could not be more
accurately measured beyond using an unaided eye, and a metric caliper.
Transmission X-ray or ultrasonic methods such as C-scan would have been
much more accurate, and possibly could have provided a good conelation of
crack length or area with the changes in internal friction and Young's modulus.

4.6.1. Internal friction and Young's modulus versus load
cycles for GFRP specimen GE-l.

Internal friction versus fatigue load cycles for GFRP specimen GE-l
(Figure 28) shows a large peak at about 100 cycles, as measured at 0.12 n.d.
from end of bar. The corresponding data for 0.32 n.d. was estimated
between 75 to 150 load cycles, based on the similar plots for specimens GE-3
and GE-4.

Comparing Figures 28 and 29 (specimen GE-l's variation of internal
friction and Young's modulus respectively, with fatigue load cycling), there is
generally an inverse correspondence with changes of Young's modulus and
internal friction. Initially, from zero to 3 load cycles, there is an inverse
response between Young's modulus and internal friction, then as modulus

116

.85” 33838. 9588on 05 88898 83 8885 8828 05 no 88.5 .2 mg: no
93— 28858 88888 . m .m0 588on EC 88 86.8 32 2.qu 882, .8505 8885 .3 83E

Him—U zafiowmm Enema mmju
maowo 304 EDUHSE

¢O+m P 000—. GDP 0 _. .
-. b b L - N
.—

 

fi 1* T r
1' F)
(9-01 x ) NOIIOIHJ 'mraarm

T

 

l
T
If)

.92 «and B4 Gum—Dag a
.92 00...: ad magma o

 

 

 

117

.085 3388:: 9:280on 05 888%.. 88 8838 Mun—5.». 05 co 32g .2 n.2— .«e
32 08.8.8 88888 . fl -mo 58609. $50 88 8.28 32 oawufi 882, 88608 Mme—8% .3 8:8...—

le0 zmafiommm Enema mmjo
mauewo n40..— HDQBRE

 

40+”. coo. oo. o.
r t e. I I‘ 9.3.
,1 \v I . . ND. 7? m
m
1 18.; 9
. , . S
18.3 M
. a
n
i , 13.3 m
. .. S
.. . no.3. )
a 1 w
.. . . . tuni. m\
e 92 «on o .2 98848. a .
_ 3.2 as... .2 mama o

 

 

 

118

increases between 3 and 20 load cycles, internal friction goes through a local
maximum. A maxirna in Young's modulus (Figure 29) occurs at about 20
load cycles, accompanied by a significant local minimum on the internal
friction (Figure 29) at the same number of load cycles. As the major peak of
internal friction blooms and fades from the minimum at 20 cycles through its
maximum at 100 cycles then back down to its next local minimum at 250
cycles, the correspondence of Young's modulus goes through first a minimum
at 50 cycles then a maximum at 150 cycles then back to another minimum at
250 cycles. From 250 load cycles onward, there appears to be an
approximately direct correspondence between the two plots. However, the
changes in Young's modulus measured after 250 load cycles do not vary by
much more than the measurement uncertainty, thus precluding much strength
to this particular observation.

The relative correspondence between internal friction and Young's
modulus suggests a three regime damage activation process. The progression
of correspondence appears to be from somewhat proportional, through an
inversely proportional regime with some deviations, then back to
approximately proportional. This three regime behavior suggests a
correspondence with: (1) initial damage of matrix cracking transversely to the
fibers, (2) between 3 and 10 cycles, the shear crack initiates, and from 10
through 250 cycles the shear crack grows to the ends of the specimen and (3)
with further load cycling, continued matrix cracking and fiber pullout and
breakage may occur. Rough measurements indicate that the shear crack
passed the measurement suspension points between 50 to 200 cycles.

From the data and the specimen configuration, it is interesting to note
that in both Figures 28 and 29, there are five local minima and maxima, and
also five points of interest on the specimen. As the shear crack grows, it first
passes the short span load pins, then the suspension points at 0.32 n.d., the
long span load pins, then very closely comes the flmdamental flexural nodes,

119

the suspension points at 0.12 n.d, then finally the free ends of the specimen.
The close proximity of the fundamental flexural node and the long span load
pin was not planned, but may have had a strong influence on the sonic
resonance response. The passage of the shear crack across the long span load
pin and the flexural node probably occured during load cycles from 20 to 150.
This observation is based on rough measurements and the slope reversals of
the plots in that span of load cycles and suggests some influence of the shear
crack on the measured values of internal friction and Young's modulus.

4.6.2. Internal friction and Young's modulus versus
load cycles for specimen GE-3.

The most prominent feature in internal friction and Young's modulus
for GFRP specimen GE-3 (with maximum dynamic load of 15.88 kilograms)
is the large peak centered at about 260 load cycles (Figures 30 and 31
respectively). Sonic resonance readings were possible for both suspension
locations on this specimen, at the number of fatigue cycles in the range of the
peak, but were not possible early in the fatigue life of the specimen for reasons
unknown. The value read at the nearest usable position to 0.32 n.d. is plotted
for the sake of comparison in the low cycle range. Note that both internal
fiiction peaks are centered at about 260 cycles. Both curves tend to rise
monotonically after bottoming out after the large peak. This response may be
attributable to unifonnly increasing damage (no further major events) with
further load cycling after the shear crack has run out to the ends of the bar.

120

.38“ 3333: 95.88% 05 888m». .88 .835 3888 on. no 32.3 .2 wfifl we
33 388% 88838 .m-m0 5.83% 556 88 one—98 32 28:8 88? .8on 3885 .cm 85E

film—U zafiummm gonna mmjo
wacky Q40.” HDUERK

 

 

 

3+“: ooo. oo. o. .,
H! b b ? — N
3
M
n W
m
D
T u
0
I. N
r T )
e X
Tr
.. 1 o
u 8.2 mono .8 9838: o m hr
_ .62 mono 88838: o n (

 

 

 

121

.083 33838. 958283 2: 83298 23 22:38 Mun—5% 05 co oo2a> . Z wdfl .«e
32 288.3 88258 .m-m0 3.23% E0 .5..“ v.29? 32 o=m3u 889» 8838 99:5» Am 83E

mlmu zmaommm Enema r0930
macs BSA HD0532

 

 

 

 

3+“: 82 oo. o.
b L b P”. P‘
T
.A
O
n
N
.82. w
H
O
a
. m
1R: S
)
9
d
.. e
(
.3 wood .2 98:28 4
_ .32 83 .2 3:28. o

 

- mm. 3

4.6.3. Internal friction and Young's modulus versus
load cycles for specimen GE-4.

Internal friction versus the number of fatigue load cycles for GFRP
specimen GE-4 (maximum dynamic load of 200 N), shows a large peak,
centered at about 380 load cycles (Figure 32). In this specimen, the peak
measured at 0.32 n.d. rises high above that measured at 0.12 n.d. This may
indicate that the two peaks trade places based on maximum fatigue load.
Further experimentation with other values of maximum load may clarify the
possibility.

Young's modulus versus load cycles for GFRP specimen GE-4
(Figure 33), shows a similar variation as that of specimens GE-l and GE-3.
The curve for 0.012 n.d. deviates from a monotonic decrease more than the
curve for 0.32 n.d. measurement position, largely due to the smaller error bars
for the 0.012 curve.

4.7. Peak analysis on plots of internal friction versus
number of load cycles for GFRP specimens.

The location and full width at half maximum of peak (as number of
fatigue cycles) were determined for the most prominent peak from the plots of
internal friction versus number of fatigue cycles for the GFRP specimens used
in the experiment. ‘

Both peak measures show a linear or nearly linear relation
(Figure 34) with respect to the maximum stress used for fatiguing. This
behavior suggests that the two trends may be in the nature of the material.

Linear regression analysis showed the peak location to be the most
linear. The specimen that experienced the lowest load saw me peak occur
soonest in its fatigue life. The specimen that experienced the highest load saw

122

123

.88” comes—3:: Mao—58% 05 “cacao.— mna comet... 3:85 05 no 3:3, .2 8a «a
v8— umfiaiv 825388 .YmO 5.58% .550 you “£98 32 3E? 5:05 3:85 .Nm flamma—

filmo Zmzmommm ~68an mmjo
mmao>0 H9995" ho mmmgz

 

 

 

89 8. 9 .
— P .. - 0*.N
and mm
m
d
13.... m
0
. u
m
:8...
)
X
I
o.
8
(

 

.92 «and 94 g 0 109m
.92 Dead. 94 g o

 

 

 

124

.88..” 3383:: Managua» 05 “some? 83 3:62: page? 05 co 85:5 .2 8n .3
32 282% 82588 {m0 cofiooam EU c8 ”£26 32 3qu 38? 3:62: aha—g .3 25mm

«two zmufiommm 33mm mmju
mmqowo EDGE?“ ho mHmEDZ

 

 

 

 

 

 

coo, oo. o.
_ t 8.2. m
3
m
flow me D
N
m
/ S
N
, w m
- 4 row 2. fl
1
n
. S
6.2 «on... .2 magma a Mo;
.92 2:... .2 5:35 o d
r _ - 8.2. W

125

. .16 as 26 4.8 80.58%
.983 83m “8 ”£28 32 3e? .5on 1885 E 3qu 8?: mo $9258 Mann 59a Sun .3 9.5me

mzmwfiommm 9.553“ Enema mmju
wacky mbgafim ho mmmZSz

Dom 00¢ Don DON 00 P O
L a . L h b .

L L

 

1
1
ID

I

l
0
'-

 

c‘: J:
N .-
(ax) (IVO'I 3119mm omvuxa n

a a EBB .32 a
22.—.4on an o

 

 

b _ . p p —

 

IO
N

126

the peak occur the latest in its fatigue life. This trend at first seemed
counterintuitive, but it was compared to the effect of strain hardening in
metallic alloys. This analogy certainly implies that the GFRP composite gets
harder or stronger with fatigue, but the fact that the composite also appears to
periodically lose strength is apparently contrary to previous work showing that
GFRP composites may increase or decrease strength with strain, but not both
during the same test [SS-57]. The mathematical relations that describe the
effect in metals may be a useful model as a starting point in further analysis of
the effect in the GFRP composite. Further experimentation is necessary to
determine if the properties of Young's modulus and internal friction are
behaving in a clearcut trend.

The full width at half maximum also increased with maximum stress on
the specimen, with the curve approaching the fatigue cycle axis at lower
stresses. As the maximum dynamic load is reduced, the peak narrows, and at
higher maximum loads, the peak widens.

The indication of a maximum load axis intercept suggests that below
approximately 2 percent of rupture modulus (M.O.R.), there will not be a
prominent peak in internal friction, which is the case for GFRP specimens
fatigued at higher loads.

GFRP specimen GE—S, tested at a maximum load of 22.3 N (1.7
percent of M.O.R.), showed no evidence of a maxima of internal friction.
This supports the trend of a decreasing number of load cycles for both peak
location and peak width. If one makes the assumption that there is a hint of a
peak in the data measured at 0.11 n.d., and that it appears to occur at about 30
to 50 load cycles, it can be suggested that the peak location curve approaches
or intersects the fatigue cycle axis at lower maximum stresses.

It is reasonable to assume that for no load cycles, there will be no
change in physical properties. Since the data for the higher fatigue load
specimens falls into a nearly linear relation, and would indicate an intersection

127

with the ordinate, the existence of a 'damage nucleation' regime seems
possible. This regime is where the data point for specimen GE-5 falls, further
supporting the existence of the downward tail. It is further suggested that
there may be a fatigue limit, where no damage occurs below a certain stress (as
a fraction of M.O.R.), for any number of load cycles.

4.8. Maximum changes in Young's modulus versus load
fraction of rupture strength.

Young's modulus data from all specimen runs was analyzed for the
maximum overall change (Tables 3.a through (1, pages 85 through 88).

It was noted that for most data, grouped by material and by physical
property, a trend of increasing maximum change with increasing maximum
fatigue load can be seen. The trend was most clear for the Macor specimens.
Young's modulus, measured at 0.075 n.d., was plotted as maximum change
in Young's modulus versus maximum load as fraction of M.O.R. (Figure 35).

Linear regression analysis, for this data set, gave the relation

E (Y Ix) = -0.01023 + 0.002037(x) (50)

where

E (Y Ix) = expected variation in Young's modulus, given the percent
M.O.R. of fatigue load
Y = Young's modulus
x = percent M.O.R. fatigue load.

For both Young's modulus and internal friction, comparison of the

maximum changes (A and percent change columns of Tables 3) with possible

128

SB 05 00 3:0 020 05 880 2.0.0 .«o 853:. 3235.5: 3 3 08:32: .8220on 532
.80 Sweet» 23%: «o 53.5 32 2.qu «:29» 333:. Muss» 2 moms—«nu 82232 .3 Baum"—
.n.z 050.0 .5. 05593.22 .mZmaom—mm ~30;
90.5902 mmEmDm ho 203.053 931— @9295"
0... 0.0 0.0 5.0 0.0 m.0 +0 nd «.0 v.0 0.0
_ _ — P b F P p — 00.0

 

 

T w . T95

 

 

MDMDGOJ 902mg» Er ”02410 39.3 o

h P

 

cud

(was) sn'maon 3.9mm NI momma

129

error (as standard deviation or coefficient of variation respectively) for all data
groups indicates that changes in properties are in general, larger than possible
error by an order of magnitude or greater.

4.9 Sensitivity ratio of changes in Young's modulus and
internal friction.

The relative changes of Young's modulus and internal friction can be
compared using a sensitivity ratio, defined as

Sensitivity ratio = (AQ‘l [QC-1) / (AE / E0). (51)

Sensitivity ratio computations were based on the maximum overall
changes detected in Young's modulus and internal friction (Table 4). It is
indicated that the sensitivity ratio is highest for fatigue specimens which
experienced the lowest dynamic stresses. For the Macor, at low fatigue
stresses, the measurement of inté’rfial friction is up to 920 times more sensitve
than the measurement of Young's modulus. At the highest fatigue stress, the
sensitivity ratio fell to about 100 to 200 times greater normalized change of
internal friction than Young's modulus (Figure 36.a). For the GFRP
specimens, the sensitivity ratios measured at 0.104 n.d. decreased
monotonically (Figure 36.b) from about 330 to 60, but the measurements at
0.32 n.d. appeared erratic and were discounted as not informative.

At low stresses, Young's modulus' is affected much less than internal
friction. As maximum stresses are increased, the change in Young's modulus
grows much faster than the change of internal friction. The change in
sensitivity ratios may reflect a difference in the nature of the damage inflicted at
the different stress levels.

130

Table 4. Sensitivity ratios for all fatigue specimens.

 

 

 

 

 

Specimen Pct. M.O.R. # Cycles * Sensitivity Ratio **
at 0.075 n.d. at 0.32 n.d.
MA-2 20 500 918.7 919.5
MA4 30 5000 537.6 701.5
MA-6 40 1000 455.0 916.3
MA-7 40 5000 196.4 412.9
MA-9 50 2500 129.3 265.4
MA-ll 65 5000 111.7 391.0
MA-12 75 5000 112.7 263.6
at 0.104 n.d. at 0.32 n.d.
GE—S 1.7 1000 328.5 1571.8
GE-l 6.3 10 K 270.3 ***
GE-3 9.5 10 K 182.4 240.7
GE-4 14.2 10 K 59.1 793.1

 

 

 

 

 

 

* Number of cycles after which the maximum change had occured.

** Using maximum overall changes in Young's modulus and Q'l.

*** No data available for minimum Young's modulus or maximum Q'l.

 

131

 

 

 

 

 

 

 

 

mac ....-1.,.2fi,...,..3

3

:3

8

z 60- 1

rd

m .

:3

E 40.. -

:3

o:

rt.

0

E. 20- ~

[:1

0

n:

3'3 o 395°32‘39”“,2. . ,. .3 . . -
) o 200 - 400 600 800 1000
a

SENSITIVITY RATIO
MACOR SPECIMENS MEASURED AT 0.075 N.D.

m 15 1 r j 1 r l 1 1

3

:3

8 12- -+

2

a:

5 9‘ J

n.

:3

n: 5_ A

In

0

0

a:

‘53 c o cuss-non srzcnnms

r f Y T V 1 ‘ T j T r T '
o 50 100 150 200 250 300 350
SENSITIVITY RATIO

b) GLASS-EPOXY SPECIMENS MEASURED AT 0.104 NJ).

Figure 36. Sensitivity ratio versus maximum fatigue load for Macor and
GFRP fatigue specimens. a) Macor specimens, b) GFRP specimens.

4.10. Possible mechanisms to account for variations in
values of internal friction and Young's modulus.

4.10.1. Damage recovery mechanism.

The increases and decreases in the physical properties of both materials
may result from the closure and partial healing of microcracks. Such
recovery occurs in thermally shocked ceramics and mechanically loaded
glasses (both inorganic and organic),[87 - 90].

Kim et al. [87] studied the effect of one thermal shock in deionized
water at room temperature. Kim carefully monitored Young's modulus in
room temperature laboratory air as time progressed. A time constant for
recovery from thermal shock, analogous to that in capacitive and inductive
electrical circuits, was established for several ceramic materials. The time
constant is 30 to 100 minutes for Macor, depending on quench temperature
[87]. For polycrystalline alumina , the time constantfor recovery is about
twenty five minutes, and for NLS glass is about 4300 minutes [87].

Stravrinidis et a1. [88] showed significant compositional changes
beneath the fresh surface of an NLS glass within minutes after exposure to
moist laboratory air, using auger spectroscopy and progressive ion milling.

For the organic glasses, the strain energy release rate for repropagation
of a crack was independent of time, from 5 minutes to 4 months [88]. This
implies that all recovery occurs within the first five minutes after cracking.
The energy release rate for repropagation of a crack in Araldite CI‘ 200 epoxy,
with or without HT901 phthallic anhydride hardener (both Ciba - Geigy
products), was greater than for the propagation of the original crack [88].

The times involved in the fatigue tests of this study are about 1.5
seconds per load cycle, with the first major peak in the value of either physical
property occuring between 100 and 350 load cycles. The peak occurs over a

132

133

time span of about two to eight minutes. This time span is approximately
within the time determined by Stravrinidis et al. for full recovery in the Araldite
epoxy glass [88].

Although one fatigue cycle (as performed for this study) takes much less
time than the established time constants for Macor or the Araldite epoxy, there
is a significant difference here. The difference lies in the cyclic loading of the
specimen. Due to the fatigue loading, the repeated abrasion of internal
microcrack surfaces is likely to be occuring. Repeated abrasion is likely to
repeatedly expose a fresh, nascent surface ready to react with any atmospheric
species that can work its way to the crack. The transport mechanisms of
diffusion, convection and/or pressure differential may be involved in getting
the reactant species to the crack surfaces.

The increase inmodulus anddecrease inintemalfrictionmaybe
explained by recovery, but the'reversal of the recovery must also be
addressed. The material may only recover to a certain extent, and only so
many times throughout its fatigue life. For some number of load cycles, the
accumulation of a restoring 'chemical glue' may be overcome by repeated
damage (due to repeated load cycling) and the values of Young's modulus or
internal friction may be retumed to their damaged values. Further, it may be
possible that, as the material periodically recovers and damages, the internal
microcrack faces accumulate debris. The accumulation of debris may provide
two means by which recovery is reduced: (1) to restrict or prevent further
abrasion of the nascent surface and possible reaction with atmospheric species
and (2) to prevent the closure of the microcrack. The accumulation of debris
on the crack faces may explain the decreasing variations in Young's modulus
and internal friction after the initial large change.

The apparent periodicity of the change in properties for both materials
was plotted with very inconclusive results. This indicates that there is a need
for more data points. To have more data points, an in situ means of

134
measuring both Young's modulus and internal friction are required.
4.10.2. Damage progression mechanism.

The progression of damage in the GFRP specimen was inhomogeneous
in nature, as evidenced by the growth of a nridplane shear crack throughout the
specimen. Microcrack damage in Macor may progress similarly. The load
pin contacts on the specimen may play a role in the advancement of damage
across the specimen. If the damage was anested at the point where a load pin
contacts the specimen (where normal and shear stress states change
significantly), then recovery may be enhanced.

Since the variation of physical properties was similar between both
materials used for this study (with respect to the inhomogeneous progression
of the midplane crack in the GFRP composite), it can be argued that the
progression of microcracking in Macor followed a similar schedule.

If more reliable crack length data were available for the both types of
specimens, a correlation between damage and internal friction or Young's
modulus may be shown to exist. Reliable means of damage characterization
for the GFRP composite specimens, such as ultrasonic C-scan or transmission
X-ray, may provide an accurate monitor of damage development.
Microcracking in ceramics such as Macor is not yet quantifiable.

4.10.3. Test for long term recovery of Young's modulus.

After fatigue testing, all unbroken specimens were saved for future
measurements of Young's modulus and internal friction. The intitial values
were compared to the values after the final fatigue cycle and values determined
between 30 and 150 days later (Table 5). The latter values were determined ~

135

Table 5. Long term values of Young's modulus. *

 

 

 

 

 

 

Specimen # Cycles Young's Modulus (GPa)
** Initial 0 Final Fatigue 00 Long Term (Days)

GE-l 10 K 41.5394 41.6007 41.5925 (151)
GE-3 10 K 41.8252 41.7792 41.7419 (120)
GE—4 1000 43.7759 42.6243 43.6966 (112)
GE-S 1000 41.24 41.22 41.2828 (27)
GE-2 *** 43.820 N/A 43.797 (145)
MA-6 *** 62.1 10 N/A 62.1030 (64)
MA-3 1000 62.256 62.251 62.265 (94)
MA-4 5000 61.957 61.970 61.9971 (92)
MA-7 5000 62.200 62.189 62.1295 (88)
MA-ll 5000 63.913 63.878 63.8887 (72)
MA-12 5000 62.507 62.449 - 62.4925 (66)

 

 

 

 

 

* = GFRP specimens measured at 0.12 n.d. and Macor specimens
measured at 0.075 n.d.

Number of Cycles taken for min/max to occur.

0 = Prior to any fatigue damage.

00 = Immediately after the last fatigue load cycle.

*** = Working standards.

N/A Not Applicable.

#11:

 

136

for all specimens on the same day, accounting for the different times since the
final fatigue load. In the time scale of days, any real time constant for
recovery must be long exceeded. Considering full recovery to require five
time constants, the specimens were considered recovered to their full
capability.

All Macor specimens (except MA-7, fatigued at 40 percent M.O.R.)
exhibited recovery since their final fatigue load cycle. The specimens that
experienced stresses up to 20 percent of M.O.R. recovered to approximately
their initial values. Those specimens that saw the stresses of 30 percent of
M.O.R. and higher did not recover beyond their initial values.

4.11. Internal friction and Young's modulus versus
acoustic emission.

To the author's knowledge, no researcher has correlated the damage
measures of acoustic emission and changes in internal friction and Young's
modulus. This may be due to no one having looked into the matter, or
because no one has shown a strong relation and reported it in the open
literature.

4.11.1. Possible trends between AE and internal friction.

Acoustic errrission counts were sorted into ten equal sections of
maximum load, then plotted against internal friction (Figures 37, 38 and 39).
Each plot shows how AB counts accumulated during that range of load, with
respect to the change of internal friction. Again, the most prominent feature of
plotting AB with internal friction is the peak which developed early in the
fatigue life of all three GFRP specimens.

Short sections of some of the plots of internal friction versus acoustic

137

 

OI
J

r'r'TrIT—fifr

   
  
 

4‘
l

U
r

 

INTERNAL rmcrron (no-3)

 

 

M

b's'o froo' 150'2oor330’35o'330r4oo
ACOUSTIC EMISSION COUNTS

a) LOAD RANGE: 0 TO 10 PERCENT OF MAXIMUM
GLASS-EPOXY SPECIMEN GE-l

 

 

murmur. rmc'rron (no-3)

 

b ' 5101' 150'150'200T250
acousrrc amssron courrrs

LOAD RANGE: 10 TO 20 PERCENT OF MAXIMUM
GLASS—EPOXY SPECIMEN GE-1

 

N

b)

Figure 37 a,b. Internal friction versus acoustic emission counts accumulated over
500 fatigue cycles, GFRP specimen GE-l . a) load range from zero to 10
percent of maximum load, b) 10 to 20 percent.

 

 

 

 

 

6‘
'o
H
5
z
E
i
2 ' r 1 1 1 T r r 1 1
0 50 1 00 150 200 250

ACOUSTIC EMISSION COUNTS
LOAD RANGE: 20 TO 30 PERCENT OF MAXIMUM
C) GLASS—EPOXY SPECIMEN GE-l

 

 

INTERNAL rmcrrorr (no-3)

 

 

 

b's'o 'rt'ro'réo'zootzso
ACOUSTIC EMISSION COUNTS

LOAD RAéIGE: 30 TO 40 PERCENT OF MAXIMUM

d) USS—EPOXY SPECIMEN GE-l

Figure 37 c,d. lntemal friction versus acoustic emission counts accumulated
over 500 fatigue cycles, GFRP specimen GE-l . c) load range from 20 to 30
percent of maximum load, d) 30 to 40 percent.

139

 

 

marm. mrcrron (no-3)

 

 

2 l V I ' r T I V I r 4
0 50 100 150 200 250

ACOUSTIC EMISSION COUNTS
LOAD RANGE: 40 TO 50 PERCENT OF MAXIMUM
GLASS—EPOXY SPECIMEN GE-l

 

 

 

 

mama. rmcrroN (no-3)

 

 

 

O 50 100 150 200 2

ACOUSTIC EMISSION COUNTS
LOAD RANGE: 50 TO 80 PERCENT OF MAXIMUM
f) GLASS-EPOXY SPECIMEN GE-1

Figure 37 e,f. Internal friction versus acoustic emission counts accumulated
over 500 fatigue cycles, GFRP specimen GE-l . e) load range from 40 to 50
percent of maximum load. 1) 50 to 60 percent.

 

   
  
 
   

«L
l

U
r

 

INTERNAL FRICTION (no-3)

 

 

O ' 5'0 '130'130'230 ‘ 250
ACOUSTIC EMISSION COUNTS

LOAD RANGE: 60 TO 70 PERCENT OF MAXIMUM
GLASS-EPOXY SPECIMEN GE—l

8)

 

tn
_J

«fl
L

 

U

INTERNAL FRICTION (no-3)

 

 

N

6 11603603 360 ‘ 463T56F 560 V760
ACOUSTIC EmserN COUNTS

LOAD RANGE: 70 To so PERCENT or ruxnnnr
h) GLASS-EPOXY SPECIMEN OE-r

Figure 37 g,h. Internal friction versus acoustic emission counts accumulated
over 500 fatigue cycles, GFRP specimen GE-l . g) load range from 60 to 70
percentofmaximumload, h)70to80percent.

 

4.4

 

U

 

INTERNAL FRICTION (210-3)

 

 

N

6'130‘ zéo'afiooftéo'sioroéo’vbo
ACOUSTIC EMISSION COUNTS

LOAD RANGE: 80 TO 90 PERCENT OF MAXIMUM
GLASS-EPOXY SPECIMEN GE—1

 

  
  
 

.5
l

U
I

 

mTERNAL TRICTION (no-3)

 

 

N

T- 13? 2601550 -450 -553 WET-:30
COUSTIC EMISSION COUNTS

A .
-) LOAD RANGE: 90 TO 100 PERCENT OF MAXIMUM
J GLASS- EPOXY SPECIMEN GE- 1

Figure 37 i,j. lntemal friction versus acoustic emission counts accumulated
over 500 fatigue cycles, GFRP specimen GE—l . i) load range from 80 to 90
percent of maximum load, j) 90 to 100 percent.

 

5'0 ‘ T ‘ I 1 T w 1 1 1 1 I

 

 

INTERNAL FRICTION (nor-3)

 

 

 

 

i V V I

° I T I I I
O 1 00 200 300 400 500 600

ACOUSTIC EMISSION COUNTS
LOAD RANGE: 0 TO 10 PERCENT OF MAXIMUM
GLASS—EPOXY SPECIMEN GE-S

 

 

 

 

 

 

2!)
5.0 ......-rfi..,...,..-jfi,
6‘
| 1
3
"4.5- -
V
2 .
O
54.04 ..
E .
23.5 l
3.0 ,fijfij
a zoo 4oo soo 300 1000 1200

ACOUSTIC EMISSION COUNTS
LOAD RANGE: 10 To 20 PERCENT OF MAXIMUM
b) CLASS—EPOXY SPECIMEN CE-a

Figure 38 a,b. Internal friction versus acoustic emission counts accumulated over
1000 fatigue cycles, GFRP specimen GE-3. a) load range from zero to 10
percent ofmaximum load, b) 10 to 20 percent.

143

 

 

 

 

  

 

 

s-c wwifivpjn.+e.-.-..fifi
if? .
3
5 4.5? ‘I -
z A
0
E 4.0- .
g 3.5-
3.0 .-.j...,--.,.
6 100 zoo zoo 400 soo soo
ACOUSTIC EMISSION COUNTS
LOAD RANGE: 20 TO 30 PERCENT OF MAXIMUM
c) GLASS-EPOXY SPECIMEN GE-3
5.c~....fi--,...,...rfi,rnfir

 

INTERNAL FRICTION (210-3)

 

 

 

200

ACOUSTIC EMISSION COUNTS
LOAD RANGE: 30 TO 40 PERCENT OF MAXIMUM
d) GLASS-EPOXY SPECIMEN GE-S

Figure 38 c,d. Internal friction versus acoustic emission counts accumulated
over 1000 fatigue cycles, GFRP specimen GE-3. C) load range from 20 to 30
percent of maximum load, d) 30 to 40 percent.

I44

 

 

 

 

 

 

 

 

 

 

5.0...T—r.firr.fi.1..fir,...1...I
i‘
o
E 4.5-j -
z a
0
g 4.0. -
d 3.5~ ..
E 1
3.0 .-..,..rTj
O 100 200 300 400 500 600
ACOUSTIC EMISSION COUNTS
LOAD RANGE: 40 TO 50 PERCENT OF MAXIMUM
) GLASS-EPOXY SPECIMEN GE-3
e
5.0 .5.,---.-..1.r.j..jT.-.—T
6‘
I
3
5 4.5- -
z .
0
e .0- -
d 3.5- ..
E H .
3.0

 

 

 

 

6"'16o"'200 zoo"46€"56o"s60

ACOUSTIC EMISSION COUNTS
LOAD RANGE: 50 TO 60 PERCENT OF MAXIMUM
GLASS-EPOXY SPECIMEN GE-S

1)

Figure 38 e,f. Internal friction versus acoustic emission counts accumulated
over 1000 fatigue cycles, GFRP specimen GE—3. 0) load range from 40 to 50
percent ofmaximum load, 0 50 to 60 percent.

 

500 j j I 1 V U r ' T V 1 l V V "I I 'I r U 1 'fi T 1 V

4.5- A -

4.0-

 

3.5-

JOON j Tfi j i V V U V i

6' "160' "zoo zoo 460 560 660'
ACOUSTIC EMISSION COUNTS

LOAD RANGE: 60 TO 70 PERCENT OF MAXIMUM
GLASS-EPOXY SPECIMEN GE-3

 

INTERNAL FRICTION (no-3)

 

 

  

 

 

 

 

 

 

 

 

g)
sogz::.:..1:..1.::jjfi51r:...

'n‘ i

0

fl

:1

z

E

E

ACOUSTIC EMISSION COUNTS
LOAD RANGE: 70 TO 60 PERCENT OF MAXIMUM

h) GLASS-EPOXY SPECIMEN GE-S

Figure 38 g,h. Internal friction versus acoustic emission counts accumulated
over 1000 fatigue cycles, GFRP specimen GE-3. g) load range from 60 to 70
percent of maximum load, h) 70 to 80 percent.

 

 

 

 

 

 

 

 

 

 

           

 

5.0 H:,...,.:za,n:,n.,n:,
6‘
I
a
v4
‘5 u
z 4
E
222‘
3.0:::,fi:,.. :wa.‘
u too 200 300 400 500 600
ACOUSTIC EMISSION COUNTS
LOAD RANGE: 80 TO 90 PERCENT OF MAXIMUM
.) GLASS—EPOXY SPECIMEN GE-3
r
5.0 ::.j:.:,....n.1.~.—.,::—T.
6‘
I
o
E 4.5- -
z 4 J
o
E 4.0: ..
E t
3.0 : v . , .15 rfifi,frf,1 '
u too 200 300 400 500 600
ACOUSTIC EMISSION COUNTS
LOAD RANGE: 90 TO 100 PERCENT OF MAXIMUM
j) GLASS-EPOXY SPECIMEN GE—S

Figure 38 i,j. Internal friction versus acoustic emission counts accumulated
over 1000 fatigue cycles, GFRP specimen GE-3. i) load range from 80 to 90
percent of maximum load, j) 90 to 100 percent.

 

 

 

 

 

A

O)

I I

O

H

a -

z

E

g .

z.oo,f-f,-.. , -.-ffi. f

o 1000 2000 3000 4000

ACOUSTIC EMISSION COUNTS
LOAD RANGE: 0 TO 10 PERCENT OF MAXIMUM
GLASS-EPOXY SPECIMEN GE-4

 

 

 

INTERNAL FRICTION (no-3)

 

 

z.oc....rf.r..fir..f
o 1000 2000 3000 4000

ACOUSTIC EMISSION COUNTS
LOAD RANGE: 10 TO 20 PERCENT OF MAXIMUM
b) GLASS-EPOXY SPECIMEN GE-4

V

Figure 39 a,b. Internal friction versus acoustic emission counts accumulated over
1000 fatigue cycles, GFRP specimen GE-4. a) load range from zero to 10
percent of maximum load, b) 10 to 20 percent.

148

 

 

INTERNAL FRICTION (no-3)

 

 

3.00, - . f. r fit ++
O 400 800 1200 1600 2000

ACOUSTIC EMISSION COUNTS
LOAD RANGE: 20 TO 30 PERCENT OF MAXIMUM
GLASS-EPOXY SPECIMEN GE-4

 

 

INTERNAL FRICTION (no-3)

 

 

 

z.oo . f r f f r r f
6 zoo 400 560 coo 1000
ACOUSTIC EMISSION COUNTS

LOAD RANGE: 30 TO 40 PERCENT OF MAXIMUM

d) GLASS-EPOXY SPECIMEN GE-4

Figure 39 c,d. Internal friction versus acoustic emission counts accumulated
over 1000 fatigue cycles, GFRP specimen GE-4. c) load range from 20 to 30
percent of maximum load, d) 30 to 40 percent.

149

 

 

 

 

 

 

 

 

 

 

 

 

4.00 T W T V
6‘
l
3
5 3.75« -
z i
o
E 3504 -
g z.zsd -
z.oc,.r.,,,frm
o zoo 400 500 COO 1000
ACOUSTIC EMISSION COUNTS
LOAD RANGE: 40 TO 50 PERCENT OF MAXIMUM
) GLASS-EPOXY SPECIMEN GE-4
e
4.00 T . .
6‘
l
O
E z.75- -
z 4
o
E 3.50 ..
E 3.25 -
z.oc , . , . r . , .
6 zoo 400 can COO 1000
ACOUSTIC EMISSION COUNTS
LOAD RANGE: 50 TO 60 PERCENT OF MAXIMUM
f) GLASS-EPOXY SPECIMEN GE—4

Figure 39 e,f. Internal friction versus acoustic emission counts accumulated
over 1000 fatigue cycles, GFRP specimen GE-4. e) load range frcm 40 to 50
percent of maximum load, 0 50 to 60 percent.

150

 

 

 

 

 

 

 

 

 

 

 

 

 

4.00 . , - 1 ,
6‘
l
O
E z.754 -
z 3
o
E 3.50 -
z.oo . , . , . , . T f
6 zoo 4oo coo COO 1000
ACOUSTIC EMISSION COUNTS
LOAD RANGE: 60 TO 70 PERCENT OF MAXIMUM
g) GLASS-EPOXY SPECIMEN GE-4
450° 1
6‘
I d
2
5 3.754 ..
z 3
o
g 3.50 q
E z.zs -
z.oc . , - T f , r r T
6 zoo 400 zoo zoo Iooo
ACOUSTIC EMISSION COUNTS
LOAD RANGE: 70 TO 80 PERCENT OF MAXIMUM
h) GLASS-EPOXY SPECIMEN GE-4

Frgure 39 g,h. Internal friction versus acoustic emission counts accumulated
over 1000 fatigue cycles, GFRP specimen GE-4. g) load range from 60 to 70
percent of maximum load, h) 70 to 80 percent.

151

 

 

INTERNAL FRICTION (no-3)

 

 

 

z.oo f , f r .
u 500 1000 1500

ACOUSTIC EMISSION COUNTS
LOAD RANGE: 80 TO 90 PERCENT OF. MAXIMUM
GLASS-EPOXY SPECIMEN GE-4

 

 

 

 

 

 

4.oo-m,..-1.-.,.
6‘
I d
O
E 3.754
z 1
o
E 3.50--
E .
d 3.25% -
E J '
z.oc-...-......ji+fif...
II 1000 2000 zooo 4000 5000
ACOUSTIC EMISSION COUNTS
LOAD RANGE: 90 TO 100 PERCENT OF MAXIMUM
j) GLASS-EPOXY SPECDIEN GE—4

Figure 39 i,j. Internal friction versus acoustic emission counts accumulated
over 1000 fatigue cycles, GFRP specimen GE-4. i) load range from 80 to 90
percent of maximum load, j) 90 to 100 percent.

152

emission counts showed a piecewise linearity, but in general, appeared similar
to the plots of internal friction versus fatigue load cycles, with local maxima
and minima due to the variation of internal friction.

The significance of a linear increase of AB with internal friction is not
clear at this point, but clearly, in Figures 37, 38 and 39, the lower the relative
slope between examination points (since the internal friction coordinates were
fixed for the set of plots for each specimen), the more AB was produced, from
one load range to another. The observation that AB varied among load ranges
was examined in Section 4.5. This section attempts to tie AB counts into
Changes in internal friction; one of the initial goals of the study.

It is observed that in some load ranges (Figures 37 through 39), AB
counts accumulate at a rate which correlates linearly with the development of
internal friction. Only where there are at least three data points involved, can
this aspect of analysis be applied. The apparent linearity does not necessarily
imply that the load range in question has anything to do with crack growth.
On the contrary, those load ranges which have the most acoustic emission
during the larger Changes in internal friction or Young's modulus are likely
responsible for damage growth (Section 4.5).

Some load ranges produced more AB on the leading side of the major
peak, whereas other load ranges produced more AB on the trailing side of the
major peak. This Observation suggests that damage during recovery and
during degredation may occur in different ranges of the load cycle. This in
turn, suggests an evolution of the damage as recovery and degredation occur.

As example, Figures 37 .c, d and e, in load ranges 20 to 30, 30 to 40
and 40 to 50 percent of maximum respectively, show the lagest amormt of AB
on the leading side of the major peak. Figure 37.a, b and c, in load ranges O
to 10, 10 to 20 and 20 to 30 percent of maximum show the most AB on the
trailing side of the major peak. The remainder of Figures 38 and 39 are left to
the reader to examine and make similar comparisons.

153

There is some support for this observation. During the fatigue cycles
in which the largest Change of intemal friction was observed, the load range
responsible for the damage likely produces the most AB. This is with respect
to the confirmation that the AB detected at the ends of the load range was
contaminated by a substantial amount of machine or system noise
(Section 4.3).

For all GFRP specimens, the occurance of the most AB during the part
of the load range in which the physical property changes is likely in the load
range in which the most damage occurs. The occurance of AB in a part of
the load range in which the physical property does not Change may represent
the friction during which recovery occurs.

5. Conclusions.
5.1. AB in load ranges.

The end ranges of the load cycle produce most of the acoustic emission.
Machine noise tests revealed that substantial numme of AB events and counts
accumulated at the points of load reversal, precluding the reliability of AB data
at the ends of the load range.

Some intermediate load ranges produced more AB than their
neighboring ranges, due to frictional contact of asperities on the internal crack
faces, or crack propagation and arrest. However, the apparent upward drift
of the load ranges in which AB occured may attest to the evolution of that
particular AB site, as requiring a higher stress to force the contact of intemal
asperities, or the further growth of damage.

5.2. Evolution of damage as tracked by internal friction
and Young's modulus.

The increasing and decreasing behavior of internal friction and Young's
modulus possibly indicates the fast initiation, growth and recovery of a major
damage mode, probably the midplane shear crack which developed in the
GFRP composite specimens. The similar behavior of Young's modulus and
internal friction in the Macor specimens suggests that the progression of
microcracking damage and recovery may have followed a similar evolution.

The change of internal friction relative to the change of Young's modulus
(sensitivity ratio) was much greater at low loads than at the higher loads. The
damage that occured at the low loads had the greater effect on internal friction.
The higher loads had less effect on internal friction, and more of an effect on
Young's modulus.

5.3. Increasing variation of Young's modulus with
maximum fatigueload.

One set of specimen and measurement point conditions provided a fairly
Clear trend of increasing variation of Young's modulus with increasing fatigue
load. Seven Macor specimens, with Young's modulus measured at 0.012
n.d. contributed to the trend. Regression analysis gave the relation of
approximately 2.0 MPa increase of modulus variation per 10 percent increase
in fraction of M.O.R. (modulus of rupture), with a conditional standard
deviation of 0.031 MPa.

154

5.4. Evolution of Young's modulus and internal friction
with number of fatigue load cycles.

The Young's modulus and internal friction of both materials varied
unexpectedly. Both materials displayed similar behavior of both physical
properties versus loading cycles. Two possible damage mechanisms are:

( 1) the nonlinear progression of damage through the specimen and (2) an
aggressive, fast acting recovery process. Nonlinear damage progression is
not unlikely given the midplane shear crack that developed in the GFRP
specimens under four point bending. Both materials recover to some degree,
but mechanical flexural fatigue may periodically accelerate the damage.

Location of the source may be possible by using multiple transducers
placed on the specimen and the use of appropriate software in the analysis
computer. Source location may provide evidence of the nonlinear progression
of damage through specimens of both materials. '

The use of a 'guard transducer' is not necessary, due to the natural
attenuation of machine and system noise through the mechanical interfaces
between the specimen and the Instron.

5.5. Afterview of experiment.

The detection and analysis of acoustic emission is a delicate and
somewhat gray or subjective science. Much of the subjectivity is qualified by
knowledge of the simultaneous behavior of other parameters, such as cracking
or monitoring physical properties. Much care must be taken to assure one's
self of the best possible AB information.

The unexpected results of small positive and negative Changes in Young's
modulus and intemal friction were analyzed to pursue any trend or
explanation. A return to the literature supported the possibility that there may

155

156

have been some physical recovery of Young's modulus and internal friction, ‘
as a cyclically driven process.

5.6. Recommendations for future work.

As exploratory experimentation, this study points out several directions of
further work.

It is recommended to design an experiment to mechanically induce a
homogeneous damage in the specimen. Homogeneous damage may enable
the correlation of acoustic emission with the other prime measures of damage,
namely internal friction and Young's modulus.

It is also important to determine, in situ, the physical properties of
Young's modulus and internal friction. The measurement of Young's
modulus after every load cycle may allow the close comparison or correlation
with AB detected during damage, regardless of the type of apparatus forcing
the damage.

Appendices

6. Appendices.

A. Physical properties of Macor, Corning code 9658 [79].

Composition:

mum
Sio2
A1203
M80

K20

F

3203

Density:

Porosity:

Hardness:

Maximum use temperature:

Coefficient of thermal expansion:

Compressive strength:
Flexural strength:
Dielectric strength:
Volume resistivity:

Glass-ceramic, 55% mica crystal,
45% matrix glass

A roxirna wei t
46%

16%
17%
10%
4%
7%

2.52 g/cm3

0

250 Knoop

1000 °C, 1832 °F, no load
94 x 10-7 mm °C

52 x 10'7 in/in .F

50,000 psi, 333 MPa
15,000 psi, 100 MPa
1000 Volts-mil

>10l4 ohm-cm

157

6.B. Physical properties of 3-M GFRP composite

type 1003 [92].
Composition: B-glass fiber reinforced epoxy matrix,
55% fibers, 45% matrix epoxy
Maximum use temperature: 120 °C (250 0F)
For unidirectional layup, at 21 °C:
Four point bending strength: 1150 MPa
Flexural modulus: 38.6 GPa
Tensile strength: 965 MPa
Tensile modulus: 39.3 GPa
Compressive strength: 880 MPa

158

References

10.

ll.

12.

References

Hasselrnan, D.P.H. "Unified Theo of Thermal Shock Fracture
Initiation and Crack Pro a ation in rittle Ceramics."
J. Amer. Cer. Soc., 52 1 : 600 - 604, 1969.

Kingery W.D. "Factors Affectin Thermal Stress Resistance of Ceramic
Materials." J. Amer. Cer. Soc., 3 [1]: 3 - 15, 1955.

Kim, Y., Lee, W.J. and Case B.D. "Thermal Fatigue in SiC Fiber

Rernforced AlununosrlrcateOlass-Cerarmc Composrte." p . .479 - 486

In LIFE" flI'I figmrc ¥gfix gfimpgsrfirfié 2%,?ng my

Ha vrr. .y gat,.. e ,e and
ten ocrety, .

Rogers, W.P., Eme A.F., Bradt, R.C. and Koba ashi, A.S.,
"Statrstrcal Stud of ennal Fracture of Ce ramrc terIals In the Water
Quench Test." . Amer. Cer. Soc., 70 [6]: 406 - 412, 1987.

Evans, A.G. "Residual Stress Measurement Using Acoustic Emission."
J. Amer. Cer. Soc., 58 [ 5 - 6]: 239 - 243, 1975.

Lewis HI D. and Rice, R.W. "Com arison of Static, C clic and .
Thermal-Shock Fati e in Ceramic omposites." Cer. g. and Scr.
Proc., 3 [9 - 10]: 71 - 721, 1982.

Lewis III, D. "Cyclic Mechanical Fatigue in Ceramic - Ceramic
E‘I'4np8081'ei9-8?“ Update." Cer. Eng. and SCI. Proc., 4 [9 - 10]:

Ewart, L. and Suresh, S. "Crack Proggagation in Ceramics Under Cyclic
Loads." J. Mat. SCI., 22: 1173 - ll , 987.

Evans, A.G. "Fatigue in Ceramics." Int'l Journal Of Fracture, 16: 485 -
496, 1980.

Nishijirna, S., Matsushita K., Okada, T., Okamoto, T. and Hagihara, T.
" c Youn 's Modulus and Internal Frrctron In Composrte .

 

 

Materials." pp. 43 -151 in n
, Vol. 3. . y artwrg, . an vans,
enum ress, . . and London, 986.

Matsushita, K., Kuratani,S., Okamoto, T. and Shimada M., "Youn 's
Modulus and Internal Friction in Alumina Subjected to Thermal Shoo "
J. of Mat. Sci. Letters, 5: 345 - 348, 1984.

Suzuki, S. Takahashi M... Hickichi, Y. and Sakamura, H., "Internal
Fnctron of Network-F'orrmng Oxide Glasses in the Systems P205 -
(6e02, TeOz, Sb20

V 05).
. Amer. Cer. Soc.,aiO fit]: 254 - 256, 1987.

159 ‘

l3.

14.

15.

16.

17.

18.

19.

‘ 20.

21.

22.

23.

24.

160

Ono, T., Nagata, K., Hashiba, M. Miura, E., Nurishi, Y. and .
Shimada, T. "Internal Frrctron, CrackLength of Fracture Ongm and
Fracture Surface Ener 1n Alumma-erconra Composrtes."

J. of Mat. Scr., 24: 1 4 - 1978, 1984.

Marlowe M. "Elasticity and Kiloc cle Internal Friction of Y203."
Ill/gaggter's Thesis, Dept. of Ceranuc g. Iowa State Unrv., Ames, Iowa

ASTM C 623 - 71 ((1985). Standard Test Method for Young's Modulus,
Shear Modulus an Porsson's Ratro for Glassand Glass-Ceramrcs by
Resonance. Amer. Soc. for Test1n7g6of Matenals, Vol. 15.02, Glass:
Ceramic Whitewares, pp. 171 - 1 , 1988.

ASTM C 848 - 78 (1983). Standard Test Method for Young's Modulus,
Shear Modulus and Porsson's Rano for Cerarnrc Whrtewares b
Resonance. Amer. Soc. for Testmg of Matenals, Vol. 15.02, lass:
Cerarmc Whitewares, pp. 268 - 2 3, 1988.

ASTM C 747 - 74 (1979). Standard Test Method for Moduli. of Elasticity
and Fundamental Frequencres of Carbon and Graphite Matenals by Some
Resonance. Amer. Soc. for Testrng of Materials, Vol. 15.01
Refractorres: Carbon and Graphite and Products: Actrvated Carbon,
pp.199 - 206, 1988.

Kolsky, H. W. Oxford University Press, London
England, pp. - , .

Zener, C. . ° 2'1me University of Chicago
Press, Chicago, mors, pp. - , .

Forster, F. "Ein Neues Messverfahrer zur Bestimrnung des .
Elastzrtatmodulus und der Dampfung." ("A New Method for Determmg

 

 

 

the Elastic Modulus and Damping") Zertschrift fur Metalllcunde (Journal
of Metallurgy), 29, [4]: 109 - 1 , 1937.
Raylie , J .W.S. Second Ed. Vol. 1, (1877).

Dover blications, ew o , . .,pp.31 -37, 1945.

Pickett, G. "Efluations for Computing Elastic Constants from Flexural
and .Torsronal esonant F uencres of Vibration of Prisms and
4 hérilgrsggngggg Procee gs, Amer. Soc. for Testing of Materrals' ,

ASTM F - 417 - 78 (1984 Standard Test Method. for Flexural Strength
Modulus of Rupture) of lectronrc Grade Ceramrcs. Amer. Soc. for
estmg of Matenals, pp. 440 - 436, 1988.

ASTM C 203 - 85 Standard Test Method for Breaking Load and Flexural
Propertres of Block - T Thermal Insulation. .Amer. Soc. for Testing
of hgfiteréaélsiggé 4.06, rmal Insulatron: Envrronmental Acoustics,

PP- ' ’ -

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

161

ASTM C 158 - 84 Standard Test Method for Flexural Testing of Glass.
Amer. Soc. for Testin of Materials, Vol. 15.02, Glass: Ceramrc
Whitewares, pp. 19 - 4, 1988.

Mullin, J.V. and Knoell, A.C. "Basic Concepts in Composite Beam
Testing." Mat. Rsch. and Stds., 12: 16 - 33, 1970.

11351621,;ia, A.B. MW. Oxford Press, U.K., pp. 40 43,

Volterra, E. and Zachmano lou, E.C. gum.
C.E. Merrill Books, pp. 34 - 357, 19 .

Harnstead, M.A. "A Review: Acoustic Emission a Tool for Corn site
Materrals Studies." Experimental Mechanrcs, 24 [3]: 7 - 13, 19 .

Evans, A.G., Linzer, M. and Russel, L.R. "Acoustic Emission and
ggaickzlgrppf $2011 in Polycrystalline Alumina." Mat. SCI. and Eng, 15:

Evans, A.G. and Linzer, M. "Failure Prediction in Structural Ceramics
Using Acoustic Emission." J. Am. Cer. Soc., 56 [11]: 575 - 581, 1973.

Evans, A.G., Nadler, H. and Ono, K. "An Acoustic Emission Stud of
the Fracture of Zinc Selenide." Mat. Sci. and Eng., 22: 7 - 14, 19 6.

Safai, S., HennanJ-I. and Ono, .K. "Acoustic Emission Study of
32%6 Sprayed Oxrde Coatrngs." Am. Cer. Soc. Bulletm, 58 [3]:

Romrell, D.M. and Bunnell, L.R. "Monmitorin of Crack Growth in
Ceramic by Acoustic Emission." Mat. Eval., 1 , 267 - 271, 1970.

Evans, A.G., Tittman, B.R., Ahlberg, L. Khuri-Yalcub, B.T. and
Kmo G.S. 'Ultrasomc Attenuation 1n rarmcs." J. Appl. Phys., 49
[5]: 2669—2679,1978.

Evans A.G., Linzer, M., Johnson, H., Hasselmann D.P.H. and
Krpp, _ .E. [Thermal Fracture Studres 1n Ceramic S%stems using an
Acoustrc Emrssron Techmque." J. Mat. Sc1., 10: 1 8 - 1615, 975.

Kazuhiro E.,.Slrigekawa, H. and H odo, S. "Acoustic Location of
Fracture Ongrn 1n guard Glass F1 rs.‘ J. Am. Cer. Soc., 67 [6]:
C-104 - C-1 5, l9 ‘.

Kahn, M. ."Acoustic and Elastic Properties of PZT Ceramics with
Amsotroprc Pores." J. Am. Cer. Soc., 68 [11]: 623 - 628, 1985.

Jen, C-K., Chun , C-J., Shapiro, G., Monchalin, J-P. Langlois, P. and
Bussrere, J .F. " coustrc Characterization of Polinégffects in PZT
Ceramlcs." J. Am. Cer. Soc., 70 [10]: C-256 - C- 9, 1987.

41.

42.

43.

45.

46.

47.

48.

49.

50.

51.

52.

53.

162

Tseng, D., Long, Q.Y. and Tangri, K. "An Acoustic Emission Study of
nsitic Transformation of Retained Austemte 1n Intercntrcall
Annealed HSLA Steels." Acta. Met, 35 [7]: 1887 - 1894, 198 .

Kraut, E.A. "Review of Theories of Scattering. of Elastic Waves by
C6r221cks11.6"7 IE§%Tm. on Sonrcs and Ultrasomcs, SU - 23 [3]:

Dieter, G.E. WW, Third Ed. McGraw Hill Inc.,
N.Y., pp. 13 , .

tion Manual, PAC Acoustic Emission Com uter Model 3000/
3 . Physrcal Acoustics Corp. Sec. 5.5, p. 9 - 101, 1985.

Domany, E., Krumhansl, J .A. and Teitel, S. "Quasistatic

Awroxrmatron to the Scatterin of Elastic Waves by a Circular Crac "
J. pp. Phys., 49 [5]: 2599 - 604, 1978.

Dickinson, J.T. Snyder DB. and Donaldson, E.E. "Acoustic
Emission and Electron Emission Durin Deformation of Anodized
Alummum." J. Vac. Scr. and Tech., 1 [1]: 429 - 432, 1980.

Otsuka H. "Variations on Acoustic Emission Between Gra bite and
Glass Epoxy Composrtes." J. Comp. Mat., 5 [11]: 591 - 5 7, 1981.

Shih, G. C. and Ebert, L. J. "Flexural Failure Mechanisms and Global
Stress Plane for Unidirectional Com sites Sub ected to Four-Point
Bending Tests." Composites, 17 [4 : 309 - 3 , 1986.

Shih G. C. and Ebert, L. J. "The Effect of the Fiber/ Matrix Interface on
the Flexural Fati e Performance of Unidirectional Fibe lass
Composrtes." mposites Scr. and Tech., 28: 137 - 16 , 1987.

Rosensaft, M. and Marom, G. "Evaluation .of Bending Test Methods for
gtfiripolsite Ill/gatfgglf." Journal of Composrtes Technology and Research,

Davidovitz M. et a1. "Failure Modes and Fracture Mechanisms in
Flexure of Kevlar - Epoxy Composrtes." J. Mat. Sci., 19: 377, 1984.

Hwang, W. and Han, K. S. "Interlaminar Fracture Behavior and Fiber
Bnd g of a Glass - E xy Com srte Under Mode I Statrc and Cychc
Loa gs." J. Comp. at., 23 [4 : 396 - 430, 1989.

Johnson, W. S: and Mangal iri P. D. "Investigation of Fiber Bridging in
a Borilg§7Canulever Beam pecimen." J. Comp. Tech. and Rsch., 9: 10

Mall, 8., Ramamurthy, G. and Rezaizdeh, M. A. "Stress Ratio Effect on
Cychc Debondrng 1n Adhesrvel Bonded Composrte J omts."
Composrte Structures, 8: 31 - 5, 1987.

 

54.

55.

56.

57.

58.

59.

61.

62.

63.

65.

163

Wan , S. S. and Wang, H. T. "Interlarninar Crack Growth in Fiber
fielingcgce‘rileénfiosites During Fatrgue." J. Eng. Mat. and Tech., 101

Hwang, W. and Han, K. 8. "Fatigue of Composites - Fati e Modulus
1360510? g and Fatigue L1fe Predrctron." J. Comp. Mat., 20%]: 154 -

Hwan W. and Han, K. S. "Cumulative Damage Models and Multi -
Sgsegs atigue Life Prediction." J. Comp. Mat, 20 [3]: 125 - 153,

Han, H. T. and Tsai, S. W. "On the Behavior of Com site Laminates
After Initial Failure." J. Comp. Mat., 8: 285 - 305, 974.

Han, H. T. and Kim, R. Y. "Fati e Behavior of Composite Laminates."
J. Comp. Mat, 10: 156 - 180, 1 76.

Awerbuch, J. and Ghaffari, Sr "Monitoring Progressionof Matrix
Sphttrng Durin Fatigue Loadmg Through Acoustrc Emlssron 1n Notched

mdrrectronal raphrte / Epox Com osrte." Journal of Remforced
Plastrcs and Composrtes, : 2 5 - 2 , 1988.

Eckles, W. and Awerbuch, J. "Monitor-in Acoustic Emission in Cross -
Ply_ Graphite IEpoxy Larmnates Dunn7g an e Loadm ." Journal of
Remforced Plastrcs and Composrtes, : 26 - 283, 198 .

Awerbuch, J. Madhukar, M. and Gorman, .M. "Monitoring Damage
Accumulatron m Frlament - Wound Graphite lEpoxy Larmnate Coupons
Dunng Fanggre Loadm Through Acoustrc Ennssron." Journal of
Remforced lastrcs an Composrtes, 3: 2 - 39, 1984.

Favre, J. P. and Laizet, J. C. "Amplitude and Counts per Event Analysis
of the Acoustrc Emrssron Generated b the Transverse Crackmg of
Cross - Ply CFRP." Comp. Scr. and ech., 36: 27 - 43, 1989.

Guild, F. J ., Phillips, M. Grand Harris, B. "Amplitude Distribution
Analysrs of Acoustrc Emrssron from Corn sites: A New Method of Data
Presentation." J. Mat. Sci. Letters, 4: 13 5 - 1378, 1985.

Henrat, P. Vautrin, A. and Harris, B. "Mechanical and Acoustic
Emission Response of Unidirectional and Cross - Plied GRP
Composrtes." Comp. Scr. and Tech., 30: 263 - 277, 1987.

Nixon, J. A., Phillips, M. G. Moore, D. R. and Prediger R. S. "A
Study of the Development of Irn ct Damage in Cross - Ply Carbon
Fiber / PEEK Larmnates Usrng coustrc Emrssron."

Comp. Scr. and Tech., 31: 1 - 14, 1988.

67.

68.

69.

70.

71.

72.

73.

74.

75.

76.

77.

78.

79.

164

Lewis,D., Bulik, C. and Shadwell D. "Standardized Testingof .
Refracto Matron/Ceramic Frber Com osrtes." Ceramrc Engmeerrng and
Science goceedings, 6 [7 - 8]: 507 - 23, 1985.

Evans, A.G. "Perspective on the Devel ment of Iggy-Toughness
Ceramics." J. Am. Cer. Soc., 73 [2]: 87 - 206, 1 .

Evans, A.G. and Fuller, E.R. "Crack Propa ation in Ceramic Materials
Under Cyclic Loading Conditions." Met. rans., 5 [1]: 27 - 33, 1974.

Sm , K.K. and Ma ida, M.B. "Dynamic Fati e of a Machinable
Glays'ECeramic." J. . Cer. Soc., 66 [7]: 5 - 505, 1983.

Suresh, S. and Brockenbrough, J .R. "Theory and Experiments of .
Fracture in clic ComEressron; Single Phase Cerarrucs, Transformrn‘g
Cgegaérmcs an Ceramrc omposrtes. Acta. Met., 36 [6]: 1455 - 147 ,

Ewart, L. and Suresh, S. "Dagramic Faggue Crack Growth in .
Polycrystallme Alumrna un r Cychc ompressron." J. Mat. Scr.
Letters, 5: 774 - 778, 1986.

Rawlins, M.H., Nolan, T,A., Allard, L.F.. and Tennery, VJ. "Dynamic
and 8mm Fangue Behavror of Srntered Srlrcon Nrtndes: H,
Mrcrostructure and Failure Analysrs." J. Am. Cer. Soc., 72 [8]:

1338 - 1342, 1989.

Ewart, L. and Suresh, S. "Crack Pro agation in Ceramics under Cyclic
Loads." J. Mat. Sci., 22: 1173 - 11 2, 1987.

Han, L and Suresh, S. "Hi -Temperature Failure of an Alumina-Silicon
Carbide Composite under clis Loads: Mechanisms of Fati e
Crack-Tip Damage." J. Am. Cer. Soc., 72 [7]: 1233 - 123 , 1989.

Ritter, J .E. and Service, T.H. "Cyclic Fair'fue and Reliability of
lzifimgrglg-Bgrggd Abrasrve Materrals." at. Scr. and Eng., 82:

Brockenbrough J :R., and Suresh, S. "Constitutive Behavior of a
Mrerocrackmg Bnttle Solid in Cyclic Compression." J. Mech. Phys.
Solrds, 35 [6]: 721 - 742, 1987.

Kawakubo, T. and .Komeya K. "Static and Cyclic Fatigue Behavior of a
Srntered Srlrcon Nrtrrde at Room Temperature." J. Am. Cer. Soc.,
70 [6]: 400 - 405, 1987.

LewissD. III. "Observations on the Stren of a Commercial Glass-
Ceramrc." J. Am. Cer. Soc. Bulletin, 61 11]: 1208 - 1214, 1982.

Macor Product Technical Bulletin, N o. 1-R79. Coming Glass Works,
Commg N.Y., 14830, 1989 .

80.

81.

82.

83.

85.

86.

87.

88.

89.

91.

92.

165

Reece, M.J., Guiu, F. and Sammur, F.R. "Cyclic Fatigue Crack .
Propagation in Alumina under Dnect Tensron-Compressron Loading."
J. Am. Cer. Soc., 72 [2]: 348 - 352, 1989.

Jenkins, M.G., Kobayashi, A.S.,.White, K.W. and Bradt R.C. "Crack
Initiation and Arrest m a SrC Whisker / Alumrna Matrix Composrte.“
J. Am. Cer. Soc., 70 [6]: 393 - 395, 1987.

Lathabai, S., Nai, Y-W. and Lawn, B.R. "Cyclic Fangue Behavior of an
Alumina Ceramic with Crack-Resistance Characterrstrcs."
J. Am. Cer. Soc. Communications, 72 [9]: 1760 - 1763, 1989.

Ritchie, R.O. "Crack Ti Shieldin in Fatigue." %§§ha_niga% Behavigr of
W. Ed. b . an, M.G., (£3 .H. an eng :.. . .

rocee gs of the 1fth Intematronal nference, V012. Berjmg, Chrna,
June 3 - 6, pp. 1399 - 1417. Pergammon Press, 1987.

Green, DJ. "Micromechanisms of Brittle Failure in Ceramics."

hanr B vi r fM ri ls - .Ed. by Yan, M.G., Zhang, S.H.
an meng, .. . rocee s o e 1fth Intematronal Conference, Vol.
2. Berjmg, Chma, June 3 - , pp. 19 - 47. Pergammon Press 1987.

Case E.D. "The Effect of Mierocrackingzupon the Poisson's Ratio for
Bnttle Materials." J. Mat. Sc1., 19: 37 - 3712, 1984.

Wachtrnan J .B. and Tefft, W.E. "Effect of Sus nsion Position on
Apparent Values in Internal Friction Determin b Forster's Method."
Review of Scientific Instruments, 29: 517 - 520, 958.

Kim, Y. and Case E.D. "Time - (Legendent Elastic Modulus Recovery
Measurement on Themall Shock S 1C fiber/Alurmnosflrcate _
Composrtes, Machrnable lass Cerarmes and Pol’ybcrystalhne Alurmna."
Mrchrgan State Umversrty. To be pubhshed, 19 .

Stravrinidis, B. and Hollowa D.G. "Crack Healing in Glass." Physics
and Chemrstry of Glass, 24 [i]: 19 - 25, 1983.

Michalske TA. and Fuller, E.R. "Closure and Re ropa ation of Healed
Cracks in Silicate Glass", J. Amer. Cer. Soc., 68 [ 1]: 6 - 590, 1985.

Ina aki, M., Urashima, K., Tolyiomasu, S. Goto, Y. and Sakai, M.
" ork of Fracture and Crack ealing in Glass. ' J. Amer. Cer. Soc.,
68 [12]: 704 - 706, 1985.

Pulliam, G. "Preciiritaion in Microscopic Crac ". J. Amer. Cer. Soc.,
42 [10]: 477 -48 , 1959.

33gél‘echnical Bulletin, Material Type 1003. 3M Corp. Minnesota, MN