University of Dayton
eCommons
Graduate Theses and Dissertations
Theses and Dissertations
2000
Development of nondestructive evaluation methods to
characterize anomalous microstructures in Ti-6Al-4V
Mark P. Blodgett
University of Dayton
Follow this and additional works at: https://ecommons.udayton.edu/graduate_theses
Recommended Citation
Blodgett, Mark P., "Development of nondestructive evaluation methods to characterize anomalous
microstructures in Ti-6Al-4V" (2000). Graduate Theses and Dissertations. 1604.
https://ecommons.udayton.edu/graduate_theses/1604
This Dissertation is brought to you for free and open access by the Theses and Dissertations at eCommons. It has
been accepted for inclusion in Graduate Theses and Dissertations by an authorized administrator of eCommons.
For more information, please contact mschlangen1@udayton.edu, ecommons@udayton.edu.
DEVELOPMENT OF NONDESTRUCTIVE EVALUATION METHODS
CHARACTERIZE ANOMALOUS
MICROSTRUCTURES
TO
IN Ti-6AL-4V
APPROVED BY:
Danigf
( ) James A. Snide, Ph. D.
Eylon, D. Sc.
isory Committee Chairman
Director and Professor,
Graduate Materials Engineering
Ofames 4. 8 Member
Professor Emeritus,
Graduate Materials Engineering
Sathish Shamachary, Ph. D.
Committee Member
Adjunct Professor,
Graduate Materials Engineering
Joseph P. Gallagher, Ph. D.
Committee Member
Professor,
Graduate Materials Engineering
Muhammad N. Islam, Ph. D.
Committee Member
Associate Professor,
Mathematics Department
Blake Cherrington, aN
Dean, School of WAN
Donald L. Moon, Ph. D.
Associate Dean,
Graduate Engineering Programs &
Research School of Engineering
il
ABSTRACT
DEVELOPMENT OF NONDESTRUCTIVE EVALUATION METHODS TO
CHARACTERIZE ANOMALOUS
MICROSTRUCTURES
IN Ti-6AL-4V
Author: Blodgett, Mark, Patrick
University of Dayton, 2000
Advisor: Dr. D. Eylon
main
The
of this
objective
dissertation
is to confirm
research
through
the
following hypothesis: “The use of nondestructive evaluation tools allows the detection
of different microstructure
allows the identification of microstructure
types and
anomalies (interior and surface) in metals and alloys.” The work was conducted on
Ti-6Al-4V forged bar stock, presenting a case study for a high performance structural
alloy.
Ti-6Al-4V
anomalies
is
model
a good
in demanding
material,
which
cannot
tolerate
microstructure
alloy is well established with extensive
applications. The
documentation on physical, chemical, and mechanical properties. It is also available in
many
different microstructures,
addresses
issues
concerning
readily
generated by heat treating.
characterization
microstructure
This
dissertation
and the identification of
microstructural anomalies. Specifically, this work includes i) background research on the
identification
of ultrasonic
and
electrical
characteristics
of five different
Ti-6Al-4V
microstructures; ii) an application of ultrasonic backscattering measurements to detect
ili
diffusion bonded Ti-6AI-4V microstructure changes, to simulate locally isolated remnant
cast structure for billet NDI; iii) original research on laser interferometric detection for
ultrasonic phase mapping
to characterize macroscopic
and iv)
texture in Ti-6Al-4V;
original research on eddy current electrical conductivity mapping in titanium alloys.
methods were developed to evaluate microstructure and
Three original NDE
microstructure
anomalies
in Ti-6Al-4V.
a forward
First,
scattering
measurement
technique was developed to spatially map the incoherent grain scattering in the forward
propagation
direction.
These
results showed,
for the first time, that mapping
of the
forward scatter provides a basis for characterization of texture in polycrystalline titanium
alloys. Second, a laser interferometric system was developed to map the signal amplitude
and
phase
of
the
transmitted
acoustic
crystallographic orientations. This new
field
in
development
samples
with
different
preferred
allows microscopic variations in
phase (and amplitude) to be experimentally measured, which is vital to understanding
statistical characteristics of NDE
data due to microstructure. The method is original as
this the first time the signal phase has been mapped with a sub-wavelength aperture for
materials characterization. The third method was discovered serendipitously, as a result of
applying an eddy current probe to the surface of different microstructures of titanium
alloy samples.
As a result, a new and original noncontacting
approach was developed.
The method
surface characterization
is based on the effects of grain anisotropy on
electrical conductivity in titanium alloys. This electrical property imaging method allows
for characterization of near-surface microstructure and processing anomalies.
iv
ACKNOWLEDGMENTS
Special appreciation is in order to my advisor, Dr. Daniel Eylon for his generous
support
and
direction of this dissertation topic,
and to the committee
members
for
following this work to conclusion. I would like to recognize the essential support of the
Materials
and
Manufacturing
Directorate
for
allowing
me
to
pursue
this
project,
especially to Dr. Walt Griffith, Mr. Tobey Cordell, and Dr. James Malas. A particular
debt of gratitude goes to Dr. Peter Nagy of the University of Cincinnati for his skillful
experimental support and also to Dr. Michael Gigliotti of the General Electric Company
for his helpful assistance in the generation of diffusion bonded composite microstructure
samples. I would also like to acknowledge the support given by Dr. Waled Hassan from
the United Technologies Company for his help with the analytical calculations and finite
element analysis of the eddy current data. Finally, my sincere thanks goes to my parents
and family for their patience and support throughout this research project.
TABLE OF CONTENTS
sO 0)
CO SO SOCOSOSOOOOOSS
i
EEE EERE EE EEE EEE ERE ETE
il
EEE
EE EEE
APPROVAL.
.o. cece cnc cece cece eee nee nee e nee een
ABSTRACT
oc ccccec ccc cc eee e eee ene nen een een EEE
ACKNOWLEDGMENTS.
EEE EEE EEE EEE EEE EEE EE EE EEE il
nena eats Vv
ce
eens e renee eens
e nent eee e cece
eee e eens cccc
........
LIST OF FIGURE CAPTIONS. ....... 0. ce cece cece een e nen e tence een n ene e eter e tenes Vili
LIST OF TABLES 0....... cc cece ccc e cece erence eee nee en nee nnn r enn e eee gees X1V
LIST OF SYMBOLS AND NOMENCLATURE
eee e eens XV
eee ee eeeee
..........:e:e
CHAPTER I. INTRODUCTION ........... cece ccc e cee ee eee e nen een entero nena ee en nase 1
1.1
1.2
1.3
1.4
1.5
1.6
OVErvicW oo. cece ccc ccc cece ee eee eee nner e een n nen nee eee EEE EOE reer e EERE EEE EE EE 1
The Problem ......... 0.0... ccc ce nsec eee eee renee eee teen Een ene EEE EEE EE EEE Ean 4
neces aeenaes 5
eceec
een ee een eeeennen
eee eee nee eee eeeccecc
eee ....c
The Hypothesis ..........
renee enn e aes 5
eee
renee
renee
... ccc cece cece cence eter eee
The Ti-6AI-4V Alloy...
e 7
anne eee nen eeaeaeags
een een eee eece
eee ee reece nentcccc
cee ee ....
Alloy Processing ........
9
eeee
eeneees
ee
neee
ne
nent
ne
.cecee
eee
sence
......
Ultrasonic Alloy Characterization ......
CHAPTER II. LITERATURE REVIEW
eeereas 14
ee ere ne nee eeeene
eeecececce
............:c
ee tenes 15
eee ee eee ce
...cce
eee etter
2.1 Fundamentals on Ultrasonic Scattering *...........
eens 19
seeee
nese
2.2 Wave Propagation in Inhomogeneous Materials ..............:sseee
2.3 Ultrasonic Backscattering for Structural Characterization .................:6665 20
2.4 Diffusion Bonding of Titanium Alloys
e 22
eee ee nent eee eeees
cece eee eeeeee
...........:
CHAPTER III. MICROSTRUCTURAL EFFECTS ON
ULTRASONIC PROPERTIES
...........c cece eee e cece cece eee ene en neeeeeseenaeeeeaenaes 27
Microstructures Evaluated
ccsees 27
ence essen ee eeneneeeeeee
cece cece eee eee e ...cc
..........
3.1
scenes 34
PART I: Conventional Ultrasonic Evaluation Methods ................:cceceee
3.2 Ultrasonic Velocity ........... ccc cc cece cece nee tenner n erent eet et teen eae 34
3.3
Ultrasonic Attenuation
eee eneeees 46
e eee eens een ee cce
ence teen ccecec
...........c
Vi
3.4 Ultrasonic Grain Scattering ............ ccc cece cece e eee eee e eee n eee ee ee tens 54
eet eeeees 62
eenseeeeeee
PART II: The New Forward Scattering Method ..............:cce
eee eee e ene nee 62
cece ence nee ee....
....
cece
3.5 Ultrasonic Forward Scattering ....
CHAPTER IV. SUBSURFACE MICROSTRUCTURE ANOMALY
DETECTION IN BULK TITANIUM ALLOY USING UT BACKSCATTERING 74
4.1
Introduction
..........ccceccc cece eect e eee eee nee e ene e eee ene
4.2 Sample Development
een E EEE EEE EEE Se nee es 74
ee es 76
anrenenetin
cceee
e eect eet eaeeenneeeenet
reece eee....c
..........
enna eeeees 80
.cc
cece eee eee
.....
4.3 Detection of Anomalous Microstructure .....cece
e nena eee e ene eee ee 85
ene ee eee eeecccc
...c
cece ence eee e near
4.4 Results and Discussion ........
4.4.1 Axially Bonded Samples
4.4.2 Radially Bonded Samples
eeeeneaes 86
eee nese ee eeeeee
eee eee eee..c:ec
............
eee ecees 92
senses en ee eec
:seceee
sense
.............:
AS SUMMALY .eecccceccsesecsccesseesecsceceeesecseceeceseeeeesscceseesesectecseeenesaeeass 101
CHAPTER V. THE INFLUENCE OF TEXTURE AND PHASE
DISTORTION ON ULTRASONIC ATTENUATION IN Ti-6AI-4V
5.1
5.2
5.3
5.4
5.5
Introduction
..........-604- 104
.......ccccccee cece cee e cnet cee ee ee en eee rene ee ee eee eee ene e ee en eee eee ene Ee ee es 104
teens 109
ence eee eeees
Ultrasonic Properties of Mill Annealed Ti-6AI-4V ..........eece
es 121
neeeeene
.:ceceeeeenee
.............
Laser Detection of Ultrasonic Phase Distortion
eees 125
ece
enereeeeereeen
cceee
ce eneeeneeeeeeen
.cccc
cence
Results and Discussion ..........
eats 134
Ener
e
een
TEE
n
eee
EEE
nee
n
eee
e
een
eee
e eee eee
SUMMAry oo... cece
CHAPTER VI. DEVELOPMENT
OF AN EDDY CURRENT MATERIALS
CHARACTERIZATION METHOD FOR TITANIUM ALLOYS..........-..-50005 138
6.1
6.2
6.3
6.4
6.5
Introduction 22.0.0... cece ccc cece eee e ence eee eee ee eee e eee e een ee eens nese renee enna ges 138
teense nes caies 146
cette eee e tennereeeeee
ees ee:::cee
Eddy Current Experiments ............
eeenenes 155
e eeeeene
nee
ee
:ceecee
renee
......:
Resolution of Eddy Current Imaging .......
e eee ee ene n enaetets 164
eee n een ee enecceee
erence ence..cce
Results and Discussion ..........
SUIMMALY 2... cece cece cece eee nee ee eee eee eee ee neste eee ee eee eeeeer eee eeceee eee ttty 175
CHAPTER VII. CONCLUDING REMARKS. ....... 00. ccece cee eee cnet een e eens
06 9
S100
7
7.2 CONCIUSIONS
179
SOSOOSOROOOOOOOCOOOOOOOCOOOOOOD 179
EEOC
1.0... cece cee cc cece eee e eee eee eee Eee Eee EEE ene DETTE EEE EEE EEE EERE 182
7.3 Comparison of Results for Microstructure Anomaly Characterization ......... 185
7.4 Future Nondestructive Materials Characterization Research .............-...5+5+ 189
7.4.1 Application and Development of Array Transducers ..............+++- 189
eee ee ee eeneees 190
esses cceeeee
7.4.2 Geometry Insensitive Techniques ............:s
7.4.3 Residual Stress Gradient Measurement ..............csceeeeeeeeeeeeeeees 191
7.4.4 Forward Scattering Using Dual-Transducer Articulation .............. 192
APPENDICES.
EE EEE 193
ener eee eset Eee
ne nee n enna e eensecen
ee ee ee ence cee ceec
eee ne eee ccce
......cc
A. Data Tables
.........c cece ccc cee ee eee cnet eee e een nee ne seas een eneeeeeeneens 193-203
Vii
LIST OF FIGURE CAPTIONS
3.1. The as-received condition (mill annealed) of Ti-6Al-4V
bar stock in the
transverse (a) and axial (b) orientation. Both microstructures shown at 200X.
3.2. Microstructures resulting from the duplex anneal (a) and the recrystallization
anneal (b). Both microstructures shown at 200X and both have nominally the
same primary o grain size of approximately 20 pm.
3.3. The beta annealed microstructures: (a) fine beta annealed structure shown at
200X, exhibiting approximately 300-500 ym prior beta grain size, and (b) coarse
beta annealed structure shown at 25X, with extremely large approximately 2-3
mm prior beta grain size.
3.4. The geometric configuration of the birefringence measurement with different
polarization and wave propagation directions.
3.5. Results showing the variation in average longitudinal wave velocity for the
various titanium alloy samples. The scatter on these measurements is generally
about + 0.5%.
3.6. Diagram showing the different Ti-6A1-6V sample sections cut from the 2.5"
diameter cylindrical bar (i.e., five samples having varying surface normals
relative to bar axis).
3.7. The hexagonal principal directions preferentially rotate to the radial direction
(x, y plane) in the as-received titanium bar stock, based on longitudinal velocity
and x-ray analysis.
3.8. Shear wave velocity data taken at different polarization angles consistent
with the slow and fast pure shear modes.
3.9. This diagram shows the fast and slow pure mode polarization direction in the
90° samples. Note the fast mode occurs when particle displacement is
perpendicular to the alignment bands in the axial direction of the bar.
3.10. Illustration of the ultrasonic wave propagating back and forth inside the
material and the associated waveform showing the multiple coherent echoes.
3.11. The
acoustic
field of a circular piston source
variations at discrete distances from the transducer.
Vili
showing
the pressure
3.12. Frequency dependent attenuation losses for the titanium alloy samples in
the (a) radial direction, and (b) axial direction.
3.13.
signal traces showing
Oscilloscope
the difference
in amplitude
of the
microstructure scatter from the mill annealed samples.
3.14. Schematic of the two different experimental set-ups used for microstructure
characterization using backscattering.
3.15. Spatial averaging of backscattered longitudinal waves in the time domain,
using broadband
excitation with a transducer having a center frequency of
approximately 8.5 MHz.
3.16. Schematic of the forward scattering experiment.
3.17. Backscattered signals from sample AR(90) showing the effect of increasing
the amplification by 40 dB to reveal the backscatter signals.
3.18.
Spatially averaged
backscattered
signals from the axial and transverse
orientations, using flat entry surfaces with beam focused on front surface.
3.19. Time-lapsed C-scan images mapping the amplitude and divergence
of
forward scattering in mill annealed Ti-6Al-4V, showing clearly distinguishable
differences between the axial versus radial directions.
3.20. Time-lapsed
C-scan
images
mapping
the amplitude
of
and divergence
forward scattering in coarse beta annealed Ti-6A1-4V, showing indistinguishable
differences between the axial versus radial directions.
3.21. Notional scattering cross-sections in the axial and radial directions for
anisotropic and random structures, which were generated based on the backward
and forward scattering results.
4.1. Schematic for the axial diffusion bonded Ti-6A1-4V samples.
4.2. The five axially diffusion bonded samples.
4.3. Schematic for the radial diffusion bonded Ti-6AI-4V samples.
4.4. The nine radially diffusion bonded samples.
4.5. Metallographic results comparing the original as-received microstructure
with that of the HIP'ed sample. Both images were taken at 200x from the same
orientation of the bar.
4.6.
Metallographic
AR/CBA)
comparing
results
from
an alpha
dissimilar
contaminated
microstructures
bond
samples
with a clean bond.
(i.
e.,
Both
images were taken at 100x from the same orientation of the bar.
4.7. A backscattered ultrasonic B-scan shown without the advent of color coding.
4.8. An illustration showing the relationship between the B-scan image, the
backscattered waveform, and the associated color scale.
4.9. An ultrasonic waveform which was collected through the curved surface on
the side of the radially bonded AR/CBA/0.9" sample.
84
bonded
85
4.11. An illustration showing how the transducer reference signal is removed
from the B-scan, leaving primarily only the material response.
4.12. Enhanced ultrasonic B-scans using equalized histograms from the axially
87
4.10.
The
B-scan
configurations
used
for the two
sets of diffusion
samples.
88
bonded sample series. AR/AR shows a clearly evident bondline signal. AR/BA
clearly demonstrates the typical change in the scattering characteristics between
two different microstructures.
4.13. Enhanced ultrasonic B-scans from the AR / X / AR sample series. Both the
FBA and CBA implant images show indistinguishable characteristics, with clear
90
indications of scattering changes between the joined microstructures.
4.14. B-scans in the axial direction through the curved surface on the side of the
93
AR/AR/X bonded cylinders focused at surface at 10 MHz.
4.15. Enhanced B-scans from the AR / AR / X / x sample
94
anomaly,
designates the diameter of the implanted
series, where X
and x indicates the scan
orientation (r - radial).
4.16. Enhanced ultrasonic B-scans from the AR / FBA / X / x sample series,
where X designates the diameter of the implanted anomaly, and x indicates the
96
scan orientation (a - axial).
4.17. Enhanced ultrasonic B-scans from the AR / FBA / X / x sample series,
where X designates the diameter of the implanted anomaly, and x indicates the
97
scan orientation (r - radial).
4.18. Enhanced ultrasonic B-scans from the AR / CBA / X / x sample series,
where X designates the diameter of the implanted anomaly, and x indicates the
99
scan orientation (a - axial).
4.19. Enhanced ultrasonic B-scans from the AR / CBA / X / x sample series,
100
where X designates the diameter of the implanted anomaly, and x indicates the
scan orientation (r - radial).
4.20. Correlation lengths of the radially bonded samples in the axial direction.
101
5.1. The as-received (mill annealed) condition of Ti-6Al-4V bar stock from
110
surfaces
of
the
(a)
axial
(0°)
and
(b)
transverse
(90°)
samples.
Both
microstructures are shown at 200x.
5.2. Longitudinal wave velocity as a function of orientation for 1.5"-thick mill
annealed Ti-6A1-4V samples.
111
5.3. Shear wave velocity as a function of orientation for 1.5"-thick mill annealed
112
Ti-6Al-4V.
5.4. Orientation dependence of the average attenuation loss versus frequency for
115
1.5"-thick mill annealed Ti-6Al-4V samples.
different
118
structure on
119
5.7. C-scans of the forward scatter in the axial (a) and radial (b) directions, taken
over a 1" x 1" area at 10 MHz with a 0.5"-diameter, 3"-focal length receiver,
120
5.5.
backscattering
Average
intensity
a
as
function
of
time
in
orientations of mill annealed Ti-6AI-4V.
5.6. Schematic
diagrams
showing the effect of the macroscopic
scattering for the 0 and 30 degree samples.
focused on the surface.
5.8. Schematic diagrams of the experimental set-ups used to map phase (a) and
122
magnitide (b).
5.9. Schematic diagrams showing (a) three regions on the specimen over which
the phase map is repeatedly folded back by 2n and (b) the variation in arrival
127
time at different positions across the wavefront.
5.10. Phase
through
(a) and corresponding
0.5"-thick
samples
magnitude
of the mill
(b) images taken at 9.7 MHz,
annealed
Ti-6Al-4V
from
128
the axial
direction, covering a 1" x 1" area. Same images are shown for the radial direction
(c), (d).
5.11. Phase maps taken through a 0.2"-thick Ti-6Al-4V sample which had been
heat treated to generate a coarse widmanstatten microstructure with very large
lamellar « + B colonies. Phase images taken at (a) 7.5 MHz and the same region
129
at (b) 9.7 MHz.
5.12. Phase (a) and corresponding magnitude (b) images over a 1" x 1" area
through a 1.5"-thick sample of mill annealed Ti-6A1-4V in the axial direction at
9.7 MHz. Same images are shown for the radial direction (c), (d).
5.13. Phase images taken through 1"-thick samples of the mill annealed Ti-6AI4V covering a 0.4" x 0.4" area using the 0.25"-diameter transmitter at 9.7 MHz.
130
132
Raw data from the axial direction (a) after removing the first 2n foldback, same
for the radial direction (b). High-pass filtered data from axial (c) and radial (d)
directions.
5.14. Comparison of attenuation measurements from a 0.5"-diameter unfocused
133
immersion transducer in pulse-echo, and from laser interferometric detection
using a spot size of approximately 50 ym in through-transmission.
6.1. Eddy current images of small fatigue cracks in 2024 aluminum and Ti-6Al4V specimens (0.5" x 0.5", 2 MHz, 0.060"-diameter coil).
Xl
144
6.2. Comparison of (a) optical, (b) eddy current, and (c) acoustic microscopic
images of a coarse-grained Ti-6Al-4V sample (1" x 1") from nearly the same
146
area on the sample.
6.3. Electrical resistivity probability distributions for three single crystal surface
151
orientations in a) aluminum, b) copper, and c) cadmium and d) on the surface of
polycrystalline Ti-6V-4V (solid lines are best fitting Gaussian distributions).
6.4. Scanned eddy current images of different Ti-6Al-4V microstructures;
a)
154
sample containing a severe microstructure anomaly (right side, middle); b) the
billet microstructure showing texture related features in the horizontal direction;
c)
a large
grained
sample;
and
d)
equiaxed
beta
annealed
microstructure
(dimension 1" x 1").
6.5. Magnetic field distribution produced by a small pancake coil in titanium at
158
four different frequencies.
6.6. Eddy current distribution produced by a small pancake coil in titanium at
159
four different frequencies.
6.7. Axial penetration depth versus frequency for a 1-mm-diameter pancake coil
in titanium. The symbols represent the numerical results calculated by finite
161
element (FE) simulation, the solid line represents the general trend of the FE
data, and the dashed line is the plane wave asymptote calculated from the
standard penetration depth according to Eq. (6.3).
6.8. Radial penetration p versus frequency for a 1-mm-diameter pancake coil in
162
titanium. The solid circles represent the numerical results calculated by finite
element (FE) simulation, the solid line illustrates the general trend of the FE data,
the empty circles represent the analytical results calculated by Dodd and Deed's
method, and the dashed line illustrates the general trend of the analytical data.
6.9. Corrected radial penetration
Pop,
= P — dep /2
(deg © 119 douter)
163
versus frequency for a 1-mm-diameter pancake coil in titanium. The solid circles
represent the numerical results calculated by FE simulation, the solid line
illustrates the general trend of the FE data, and the dashed line is the standard
penetration depth calculated from Eq. (6.3).
6.10. Experimental impedance diagram in titanium at 2 MHz. The rotation angle
167
was chosen so that the lift-off curve is horizontal.
6.11. Experimentally determined lateral resolution versus inspection frequency
by a commercial pencil-probe in titanium. The adjusted FE prediction for the
radial penetration is plotted to indicate the trend of the data.
Xii
168
6.12. Eddy current images (0.5" x 0.5") taken at three different frequencies to
demonstrate the effect on lateral resolution. These images were scanned from an
extremely large grained polycrystalline titanium alloy.
6.13. Impedance
diagrams
and resolution profiles at three different rotational
angles relative to the horizontal lift-off angle in titanium at 2 MHz.
6.14. Experimentally measured lateral resolution versus rotation angle curves for
three different frequencies in titanium.
6.15. Eddy current c-scan images of a coarse-grained Ti-6AI-4V specimen at 5
MHz and two different rotational angles (0.5" x 0.5").
CHAPTER I
INTRODUCTION
1.1
Overview
Many different nondestructive materials characterization techniques have been
developed and applied in recent decades to address grain scattering, texture, and the
detection of discrete manufacturing and service related flaws inside the bulk of different
forged and cast structural components [1, 2, 3, 4, 5, 6, 7, 8]. Likewise, a great deal of
work has been done to understand the physical and mechanical properties of titanium and
its many different alloys and microstructures for high performance applications [9, 10, 11,
12, 13, 23]. The primary objective of traditional NDE is the detection and quantification
of defects. However, the same nondestructive inspection approaches used to find flaws
can also provide
information
on microstructure
and microstructural
variations inside
metals, based on the characteristics of the NDE signals. Unfortunately, this NDE subject
is not well developed nor used.
The motivation
for this research is based on the need to develop new NDE
methods to address problems associated with the detection and quantification of interior
and surface-connected microstructural anomalies originating from alloy production and
processing. The main goal of this research is to develop new NDE
methods. Structure
related background noise is often a source of frustration to traditional nondestructive
inspection as these signals can mask over other genuine flaws, such as voids or porosity
[14, 15, 16]. This is especially true in cast and forged components due to the nature of the
solid in terms of local elastic property variations. What makes the evaluation of structural
damage such a difficult problem for cast and wrought components in general and titanium
alloys
in particular
stems
from
effects
the
the
material
has
on
the
NDE
signals.
Background signal noise is a problem that clearly hinders our ability to detect tiny, or
worse, microscopic damage due to incoherent grain scattering [17, 18, 19].
The innovative and unique aspect of this work is the development of new
NDE methods (not based on ultrasonic backscattering) to, rather than suppress the
background noise signals, magnify and evaluate them as a source of information to
characterize microstructure. NDE
signals require careful collection and interpretation
in order to relate to the microstructure
anomalies.
unique
Moreover,
probe,
used
and
identify the presence
these signals are the result of complex
to sense
elastic
or electrical property
of microstructure
interactions between a
variations,
and
a locally
changing grain structure.
The dissertation outline initially parallels the progress of sample development,
starting with the evaluation of simple individually categorized microstructures in Chapter
3 and leading to the study of more complex, joined microstructures in Chapter 4, where
standard ultrasonic backscattering was used as a basis for the detection and evaluation.
Since
from
the results
simulated anomaly
the
(implanted)
were
samples
somewhat
inconclusive (from the standpoint of actually discriminating the different anomaly types)
further original research was conducted to develop new NDE methods. In Chapter 5,a
new NDE method was developed to spatially map the signal phase of ultrasonic waves
propagating through different orientations of textured titanium bar stock. In Chapter 6, an
eddy current method was developed and applied for the characterization of grain structure
in titanium alloys.
The
of
development
nondestructive
inspection,
and
testing,
materials
characterization techniques is essential to evaluate the structural behavior, performance,
and life span of components, which are known to accumulate damage throughout the
[20]. This is true in titanium alloy components,
course of service
since they cannot
tolerate internal defects or microstructural anomalies due to their high level of loading
(21, 22] and the critical performance requirements in airframes and engines. Hence, these
structures
require
periodic
maintenance
and
nondestructive
careful
inspection
to
minimize the potential for defects, which could ultimately lead to catastrophic failures.
The detection and quantification of microstructure anomalies is a problem of major
importance
to
industry,
but
there
are
presently
no
definitively
effective
NDE
methods nor calibration standards available to do this.
New
phenomena.
NDE
methods
Experiments
have been developed based on two independent physical
were
designed
and conducted
to examine
both elastic and
electrical property variations in titanium alloy samples. Results have been collected from
conventional NDE
However,
techniques to benchmark the samples and from the new methods.
results from the different experimental
of experimental
the combination
categories into a single chapter did not make sense without a sound theoretical foundation
for doing so. Therefore, starting with Chapter 3, the organization of the dissertation is
its own
chapter has
such that each
independent
sections
covering
the introduction,
experiments, results, and discussion. This approach was taken in order to avoid confusion
regarding
what
versus
is new,
what
been
has
done
before
in each
of the
areas
investigated.
1.2 The Problem
Many
wrought alloys are prone to the development of interior microstructural
irregularities, remnant of the original ingot cast structure that did not get globularized
during the ingot breakdown
stage. These
may
ultimately
compromise
the material’s
structural integrity [23]. Such processing related problems could be easily overlooked due
to the difficulty, time, and cost of developing and applying nondestructive inspections
geared towards locating these, often subtle, microstructure signals. These anomalies could
lead to a number of different problems including: i) the introduction of processing defects
in subsequent operations due to stress induced porosity; ii) early introduction of damage
due to mechanical cycling; and iii) difficult challenges for flaw detection due to the
generation of background noise. The key reasons for the limitations of conventional NDE
approaches are : a) microstructure anomalies are generally weak scatterers and are easily
overlooked
[24,
signals
from
dependent
and often lack uniqueness
[16,
25];
b)
the
microstructure
anomalies
are
frequency
17, 18]; c) discriminating between benign
microstructure noise and genuine defects or damage is difficult due to lack of suitable
calibration samples; and d) most NDE methods are simply not capable of looking deep
inside materials for subtle microstructure anomaly indications.
1.3. The Hypothesis
The hypothesis for this dissertation is: “The use of nondestructive evaluation
tools
allows
identification
the
detection
of
of microstructure
different
microstructure
anomalies
(interior
alloys.” This work centers on the measurement
types
surface)
and
allows
the
in metals
and
and
of elastic and electrical properties in
samples processed to different microstructures, textures, and orientations. The work is
supported
by
experimental
of conventional
application
methods,
carried
out
on
numerous
the
and
techniques
samples
development
of new
of
uniform
comprised
microstructures and composite microstructures, generated via diffusion bonding.
1.4 The Ti-6AI-4V Alloy
Ti-6Al-4V, one of the first titanium alloy's to be developed, is an all-purpose
alpha + beta alloy commonly used in high performance structural applications such as
aircraft gas-turbine engine compressor blades and disks. The alloy also has an excellent
combination
of
strength
and
toughness,
excellent
corrosion
resistance,
and
biocompatibility for orthopedic applications [10]. The mechanical properties of titanium
alloys depend on the chemistry, microstructure, and proper design and selection. These
parameters influence the strength, toughness, environmental resistance, and the fatigue
crack initiation and propagation characteristics of the material.
Microstructure related noise is an important NDE consideration because it largely
determines
the
detectability
limits
of
conditions
anomalous
the
within
structure.
Ultrasonic grain noise stems from scattering in the polycrystalline structure as the wave
propagates through the material [26]. Scattering sites originate from the crystal-to-crystal
variations in density and elastic constants. For single phase materials, each crystal has the
same density and the same crystalline structure. Hence, scattering in these materials arises
mainly from the variation in velocity resulting from the anisotropy of the single crystal
elastic
constants
microstructural
and
from
the
random
noise
arises
due
to the
of
nature
constructive
grain
and
orientation.
destructive
Polyphase
interference
of
scattered signals from thousands of different combinations of individual grains and grain
colonies [27]. Grain scattering signals in such complex microstructures are generally used
only as an indication of structural differences, since the signals are not unique in terms of
relating quantitative features of the microstructure.
Ti-6Al-4V was developed at the U. S. Army's Watertown Arsenal Laboratories in
1954 [28] and is still the most commonly used titanium alloy all around the world. The
alloy is high strength, lightweight, corrosion resistant, relatively inexpensive, and can be
easily acquired from a number of vendors in various grades of purity. In stock materials,
the problem of microstructure anomalies is minimal because the sections are generally
small,
allowing
fabrication
processes
and
heat
treatments
to
refine
all remnant
microstructure anomalies. In titanium components requiring large cross-sections like in
gas-turbine engine fan disks or forged bulkheads, complete microstructural uniformity is
due to processing
difficult to achieve
complexities,
but advanced
techniques
geared
toward more complete homogenization and refinement are under consideration [29}.
1.5 Alloy Processing
All structural metals
begin by undergoing
generally
chemical treatments directed towards removing
composition
to
allow
best
the
combination
a series of thermal
and
impurities and adjusting the chemical
of performance,
reliability,
and
cost
effectiveness for a specified application. Cast metal ingots derived from the refining and
alloying processes ultimately yield stock materials (billet, bar, plate, sheet, etc.) via the
application of additional fabrication processes like rolling, press cogging, and extrusion
[9, 10]. The
ingot material is generally not used in its as-cast state due to obvious
performance limitations originating from the nonuniform, highly inhomogeneous nature
of the coarse cast structure both locally and throughout the ingot volume [13, 23].
The
presence
of
localized
microstructure
flaws
resulting
from
impurities,
inclusions, voids, or incomplete processing of the cast structure can compromise
the
ingots
are
strength
related
properties
of
stock
materials.
As
such,
titanium
thermomechanically transformed into billet, bar, and plate. Hot working is the only option
for the hexagonally symmetric Ti-6AI-4V alloy as lower temperature or room temperature
working
is difficult and leads to cracking. Hot working
is more
forgiving than cold
working in terms of processing related defect generation due to the ease of deformation,
a
but small inaccuracies in the processing time or temperature can result in structural
inadequacies
developed
subsequently
in
the
to
due
components,
presence
of
manufacturing defects.
For Ti-6AI-4V, the first breakdown of the large cast ingot structure is generally a
press cogging, performed above the beta transus (~1000 °C) which is geared towards
reducing the cross-section and grain refinement [9, 13, 23]. Critical aircraft components
such as aircraft gas-turbine engine fan and compressor disks demand specially developed
billet stock with higher
and uniformity
of the
to allow long-term reliable performance
of the
refinement
alloying purity and more
microstructure throughout its volume
component. Nondestructive evaluation of premium grade rotor stock material is crucial to
determine the suitability of material for further fabrication to avoid the generation of
manufacturing
defects.
This
incompatibilities
and
voids
is especially
can
form,
true
due
in
forging
operations
to microstructural
where
irregularities
strain
in high
deformation regions.
The processing industry is familiar with the pitfalls of titanium processing and the
current trend is to develop smaller [approximately 10" (250 mm) diameter] billet sections
in order to achieve structural uniformity throughout the volume [29, 30]. A component
may contain conglomerations
of grains with undesirable
extremes
in the mechanical
properties due to localized texture, which could lead to faster damage initiation in dwell
applications [31]. Similarly,
voids,
stress-induced
irregularities [30].
a component may contain manufacturing defects such as
porosity,
or
cracking
related
to
hard
alpha
compositional
1.6 Ultrasonic Alloy Characterization
A variety of microstructural features can account for scattering losses from both
ultrasonic and eddy current methods. Grain size, shape, and anisotropy for example are
key characteristics of the microstructure
and play a strong role in acoustic
[32] and
electrical grain scattering [33, 34]. Diffraction losses are due to interference effects from
edges, corners, and other features that generate wave fringes, phase shifts, or frequency
shifts. Beam spreading involves the widening distribution of energy in the propagating
wave and the transition from planar waves to spherical waves.
To further complicate the scattering problem, it is known know Ti-6AI-4V has a
relatively complex microstructure. A significant obstacle to detecting and characterizing
incipient
or low-level
damage
is grain noise,
which
evolves
due
to random
local
variations in the crystallography and density of the microstructure. In Ti-6Al-4V, the
microstructural complexity poses significant challenges for materials characterization,
stemming from the alloys two phase structure with different crystallographic symmetry,
density, and stiffness between the alpha and beta phases. Moreover, the range useful
titanium alloy microstructures is diverse and widely distributed in terms of the types of
features, which can generate scatter.
Grain scattering of ultrasonic waves occurs because most metallic materials are
simply not homogeneous. Virtually all structural metal components are polycrystalline,
being comprised of a large number of randomly oriented individual crystals. Each grain is
individually a homogeneous single crystal, which in general has anisotropic properties. In
some alloys, fabrication processes tend to impart long-ranging crystallographic texture.
Crystal discontinuities at grain and twin boundaries tend to deflect small amounts of
Undeformed
energy out of the main ultrasonic beam.
single crystals have a uniform
lattice structure that is generally characterized by three orthogonal axes, along which the
various properties, like elastic moduli, are unique. For the case of Ti-6A1-4V, each grain
of the primary constituent phase has hexagonal
symmetry
and is characterized by a
principal direction, i.e. normal to the basal plane (plane of isotropy). However for Ti-6Al4V, the five single crystal elastic constants (alpha) are not available in the open literature.
An interesting aspect of attenuation in polycrystalline materials has to do with the
influence of microstructure and texture on phase distortion of the propagating wave [35,
36, 37]. The phase distortion is related to the arrival time variations of rays on different
acoustic
paths,
through
different
grains
and
has
a major,
yet uncommon,
role
in
attenuation, especially in anisotropic materials. This attenuation factor is not very well
understood because the vast majority of experimental methods are based on amplitude
detection
and
the
detection
devices
generally
are
phase
sensitive
[38].
Usually,
attenuation is considered in the context of a loss of energy due to scattering, but losses
associated with phase distortion represents a separate attenuation mechanism, which is
addressed in detail by this work in Chapter 5.
10
Chapter 1, References:
1. E. P. Papadakis, “Ultrasonic Attenuation due to Grain Scattering in Polycrystalline
Metals,” Journal of the Acoustical Society of America, Vol. 37, p. 711 (1965).
2. E. P. Papadakis, “Ultrasonic Attenuation Caused by Scattering in Polycrystalline
Media,” Physical Acoustics Principles and Methods, Vol. IV, Part B, ed., Mason, W. P.,
(Academic Press, New York, 1968) p. 269.
3. J. Lewandowski, “Evaluation of the Texture of Polycrystalline Aggregate from
Ultrasonic Measurements,” Ultrasonics, March, p. 73 (1986).
4. C. B. Guo, P. Holler, and K. Goebbels, "Scattering of Ultrasonic Waves in Anisotropic
Polycrystalline Metals," Acoustica 59, pp. 112-120 (1985).
5. D. Beecham, “Ultrasonic Scatter in Metals, Its Properties and its Application to Grain
Size Determination,” Ultrasonics, April, p. 67 (1966).
6. J. Szilard, and G. Scruton, “Revealing the Grain Structure of Metals by Ultrasonics,”
Ultrasonics, May, p. 114 (1973).
7. J. H. Rose, “Ultrasonic Backscatter from Microstructure,” in Review of Progress in
ONDE,
Vol. 11, eds., D. O. Thompson, D. E. Chimenti (Plenum, New York 1992) p.
1667.
8. K. Goebbels, “Structure Analysis by Scattered Ultrasonic Radiation,” Research
Techniques in Nondestructive Testing, ed., by R. S. Sharpe, (Academic Press, New York,
1980), Vol. IV, pp. 87-157.
9. S. L. Semiatin et. al., “Hot working of titanium alloys — An overview,” Advances in
Titanium Alloy Processing, eds., 1. Weiss, et. al., Proceedings from the 125" TMS
Meeting held in Anaheim, California, 5-8 February, 1996, (TMS, 1997) pp. 4-26.
10. Materials Property Handbook on Titanium Alloys, eds., R. Boyer, G. Welsch, and E.
W. Collings, (ASM International, 1994) pp. 483-500.
11. W. F. Hosford, "The Mechanics of Crystals and Textured Polycrystals," (Oxford
University Press, New York, 1993) pp. 67-112.
12. J. F. Nye, "Physical Properties of Crystals - Their Representation by Tensors and
Matrices," (Clarendon Press, Oxford, 1985) pp. 12-17.
13. J. C. Williams and E. A. Starke, Jr., "The Role of Thermomechanical Processing in
Tailoring the Properties of Aluminum and Titanium Alloys," Deformation, Processing,
and Structure, G. Krauss, ed. (ASM International, 1984) pp. 279-300.
14. F. J. Margetan, R. B. Thompson, and I. Yalda-Mooshabad, "Backscattered
Microstructural Noise in Ultrasonic Toneburst Inspections," Journal of Nondestructive
Evaluation, Vol. 13, pp. 111-136 (1994).
15. J. H. Rose, “Theory of Ultrasonic Backscatter From Multiphase Polycrystalline
Solids,” in Review of Progress in ONDE, Vol. 12, eds., D. O. Thompson, D. E. Chimenti
(Plenum, New York 1993) p. 1719.
16. H. Willems and K. Goebbels, "Characterization of Microstructure by Backscattered
Ultrasonic Waves," Metal Science, Vol. 15, pp. 549-553 (1981).
17. C. B. Guo, P. H-ller, and K. Goebbels, "Scattering of Ultrasonic Waves in
Anisotropic Polycrystalline Metals," Acoustica, 59(2), pp. 112-120 (1985).
18. K. Goebbels, "Evaluation of the Structure of Steels by Ultrasonic Scattering,”
Materials Testing. 77(7), p. 231-233 (1975).
11
19. W. Hassan and P. B. Nagy, "Experimental Investigation of the Grain Noise in
Interferometric Detection of Ultrasonic Waves," Journal of NDE (submitted for
publication).
20. J. P. Gallagher, et. al., “USAF Damage tolerant design handbook: Guidelines for the
analysis and design of damage tolerant aircraft structures,” AFWAL-TR-82-3073, pp.
1.3.1-1.6.8, (1984).
21. D. Eylon and J. A. Hall, “Fatigue behavior of beta processed titanium alloy IMI 685,”
Met. Trans A, Vol. 8, pp. 981-990 (1977).
22. D. Eylon, “Fatigue crack initiation in hot isostatically pressed Ti-6Al-4V castings,” J.
Met. Sci., Vol. 14, pp. 1914-1922 (1979).
23. Titanium, A technical guide, ed., M: J. Donachie, Jr., (ASM International, 1988) p.
44.
24. F. J. Margetan and R. B. Thompson, "Microstructural Noise in Titanium Alloys and
Its Influence on the Detectability of Hard-Alpha Inclusions," in Review of Progress in
Quantitative Nondestructive Evaluation, Vol. 11, eds., D. O. Thompson, D. E. Chimenti
(Plenum Press, NY 1991) pp. 1717-1724.
25. P. J. Howard and R. S. Gilmore, "Ultrasonic C-Scan Imaging for Hard-Alpha Flaw
Detection and Characterization," in Review of Progress in Quantitative Nondestructive
Evaluation, Vol. 13, eds., D. O. Thompson, D. E. Chimenti (Plenum Press, New York,
New York 1994) pp. 763-770.
26. J. H. Rose, “Ultrasonic Backscatter from Microstructure,” in Review of Progress in
ONDE, Vol. 11, eds., D. O. Thompson, D. E. Chimenti (Plenum, New York 1992) p.
1677.
27. J. H. Rose, “Ultrasonic Backscattering from Polycrystalline Aggregates Using TimeDomain Linear Response Theory,” Review of Progress in QNDE, Vol. 11, eds. D. O.
Thompson, D. E. Chimenti (Plenum, New York 1992) p. 1677.
28. R. J. McClintick, G. W. Bauer, et al., “A New Titanium Alloy,” Materials and
Methods, Aug, pp. 90-92, (1955).
29. G. A. Salishchev, et. al., “Fine grained billet processing of titanium alloys,”
Proceedings, gt" World Conference on Titanium, 7-11 June 1999, St. Petersburg, Russia.
30. B. P. Bewlay and M. F. X. Gigliotti, “Hard alpha phase stability in nitrided Ti-64 and
Ti-17," Proceedings, 9 World Conference on Titanium, 7-11 June 1999, St. Petersburg,
Russia.
31. J. C. Chesnutt and N. E. Paton, “Hold Time Effects on Fatigue Crack Propagation in
Ti-6Al and Ti-6AI-4V,” in Titanium Science and Technology, ed. H. Kimura and O.
Izumi (The Metallurgical Society of AIME, Kyoto Japan, 1980), pp. 1855-1863.
32. F. E. Stanke and G. S. Kino, “A Unified Theory for Elastic Wave Propagation in
Polycrystalline Materials,” Journal of the Acoustical Society of America. 75(3) p. 665
(1984).
33. M. P. Blodgett, P. B. Nagy, “Anisotropic Grain Noise in Eddy Current Inspection of
Noncubic Polycrystalline Metals,” Appl. Phys. Let., 72(9), pp. 1045-1047 (1998).
34. M. P. Blodgett, W. Hassan, P. B. Nagy, “Theoretical and Experimental Investigations
of Lateral Resolution in Eddy Current Imaging,” Materials Evaluation, 58(5), 2000.
35. M. P. Blodgett and D. Eylon, “The influence of texture and phase distortion on
ultrasonic attenuation in Ti-6AI-4V,” J. of NDE, (approved for publication).
12
36. P. D. Panetta, F. J. Margetan, I. Yalda, and R. B. Thompson, "Observation and
Interpretation of Microstructurally Induced Fluctuations of Back-surface Signals and
Ultrasonic Attenuation in Titanium Alloys," Review of Progress in Quantitative
Nondestructive Evaluation, Vol. 16, Eds., D. O. Thompson and D. E. Chimenti, (Plenum
Press, NY, 1997) pp. 1547-1554.
37. T. Seldis and C. Pecorari, "Scattering-induced Attenuation of an Ultrasonic Beam in
Austenitic Steel," Journal of the Acoustical Society of America (submitted for
publication).
38. M. R. Hollard and J. G. Miller, "Phase-insensitive and Phase-sensitive Quantitative
Imaging of Scattered Ultrasound Using a Two-dimensional Pseudo-array," Ultrasonics
Symp. Proc. IEEE Cat. No. 88CH2578-3, pp. 815-819 (1988).
13
CHAPTER If
LITERATURE REVIEW
This research draws heavily upon the works of previous researchers. In fact, there
exists
an
enormous
wealth
of
information
in
archival
journals
and
conference
proceedings dealing with subject matter similar to this research. Not surprisingly, the
search for "new ground" in the subject area has required a substantial investment of time
and
effort in reviewing
the literature.
Much
of the documented
literature
shows
similarities to activities undertaken during this project. This is especially true in Chapter
3, which reviews the conventional ultrasonic measurements, including longitudinal and
shear wave velocities, attenuation, and backscattering. While similarity to previous works
clearly exists in Chapter 3 and 4, Chapters 5 and 6 are completely original and based on
refereed publications, resulting from this dissertation. The results from Chapter 4 are
based
on
the
well
known
ultrasonic
backscattering
approach.
All
of the
samples
developed are original and unique to this dissertation. To demonstrate originality, this
research has yielded the publication of three refereed journal papers, with at least one
more to follow. To the best of my knowledge, this work is not duplicated nor examined
as such in any of the open literature.
14
2.1
Fundamentals on Ultrasonic Scattering
The fundamental scattering problems have been around since the days of Lord
Rayleigh [1] in the study of sound wave interaction with discrete inhomogeneities, which
laid the groundwork
for developing the single-source scattering theory. This is an
important starting point and it is well understood that when an oscillating wave traveling
in a homogeneous medium like pure water impinges on an inhomogeniety, the response
of the wave to the obstruction will depend on certain physical and mechanical aspects of
the scatterer and propagating medium. The disturbance presented by the inhomogeniety
will
result
in scatter that will
inside
characteristics
propagation
depend
on the
the
medium
direction
and
the
of the
incident
wave,
the
on
the
and
inhomogeniety,
boundary conditions across the interface of the scatterer, which must satisfy continuity of
stress and displacement across the boundary.
While
scattering
from
a discrete,
single
source
of a certain
characteristic
geometric shape is an important fundamental acoustical problem, the main issues for the
current research stem from the consideration of a material whose
scatterers.
Here,
incoherent
scattering
arises
from
the
elastic
volume
anisotropy
is full of
and
phase
heterogeneity of the generally microscopic single crystals which comprise the material.
An abundance of analytical work has been conducted in the recent past to characterize the
grain-to-grain variations of elastic stiffness, relative to the average material stiffness, for
untextured single phase materials with cubic symmetry [2, 3, 4]. Past research has also
delivered expressions for the effective wave speed and attenuation in what are now called
the Rayleigh and Stochastic scattering regimes, where the wavelength is much larger and
15
comparable
size of the
to the nominal
scatterer,
respectively.
work
Subsequent
by
Huntington [5] in the Diffusive (or geometric) scattering regime, where the wavelength is
much smaller than the scatterer, completed the fundamental theoretical development of
the problem.
Microstructural disturbances generate ultrasonic scattering and mode conversion
at discontinuities associated with changes in phase, density, or stiffness because of slight
differences
sound
in acoustic
travels
in
a
impedance
velocity and acoustic
given
direction
across the boundary.
When
there
is a
a polycrystalline
through
material,
discontinuity in the wave speed at each grain boundary, due in the simplest case, to local
elastic
property
new
each
With
differences.
unique , entity
crystallographically
encountered by the propagating wave, scattering takes place at the boundaries when the
mean
linear
wavelength.
dimension,
D,
of the
grains
is comparable
with,
or smaller
than
the
The scattering represents a loss of energy to the propagating wavefront,
generally providing the main source of attenuation in metals. In the case of grain sizes of
1/1000th to 1/100th of the wavelength,
scatter is considered negligible [6]. However,
when the dimensions of the scatterer are 1/10th to the full value of the wavelength the
scattering increases very rapidly, approximately as the third power of the grain size.
Three
different
regimes
for scattering
of ultrasonic
waves
can
depending on the ratio of the mean grain size to the wavelength (i.e. D /):
be
specified
@) the low
frequency (Rayleigh) region with scattering induced attenuation proportional to the fourth
power of the frequency and to the cube of the mean grain diameter, (ii) the medium
16
frequency (stochastic) region with attenuation proportional to the square of frequency and
to the
mean
grain
diameter,
and
(iii) the
high
frequency
region
(geometric)
with
attenuation independent of frequency. For the geometric scatter case, conditions suitable
for testing may no longer exist, due to the enormous degree of scatter, especially if the
test material is anisotropic.
Other theoretical studies have been conducted to model more complicated two
phase material systems [7, 8]. A simplified approach was used to model the material as if
it were composed of spherical scatterers in a matrix with different density and stiffness
between the two phases. This simple model cannot however, predict the attenuation and
velocity in complicated structural alloys of titanium. Titanium microstructures, owing to
their elastic anisotropy, the nature of the slip, and the effects of deformation, are complex
and wave propagation and scattering cannot be described in simple mathematical terms.
This complication is normal
processing,
for materials subjected to thermomechanical
which often result in the alignment of the second phase particles and grain boundaries. In
addition,
preferred
crystallographic
orientation
adds
another
of
level
analytical
complexity due to the directionally dependent macroscopic elastic properties.
Multiple scattering is an important subject that has been investigated in the recent
past [9, 10]. Single scattering theories assume there is no interaction between scattered
signals.
scattering
However,
is of major
titanium
alloys,
types
of materials
including
importance
as the neglect
of this phenomenon
in many
multiple
can result in
significantly misleading interpretations of the materials ultrasonic characteristics. The
17
microstructure dependent ultrasonic attenuation of a material can be generally resolved
by isolating the energy carried off by scattering, and correlating this fractional quantity of
the incident power to grain structure. This correlation works nicely when the assumption
of single scattering holds (as for an equiaxed media), but tends to breakdown
more
becomes
microstructure
scatter becomes
Multiple
complex.
a more
as the
significant
problem in highly scattering materials. In multiple scattering materials, the influence of
on ultrasonic
microstructure
energy
attenuation
is generally
overestimated
scattered again and again. Moreover,
is subsequently
as the
scattered
scattering can
multiple
interfere with determining the origination of the scatter source inside the material, which
can lead to anomalous defect indications where in fact there is no defect of significance.
The grain scattering problem occurs with many different types of structural alloys,
due to the natural solidification process and other changes that take place in the solid as a
result of fabrication and heat treatment processes. Papadakis [11, 12, 13] has conducted
research in developing
extensive
applied
to
different
types
of ultrasonic
the understanding
media.
of inhomogeneous
His
work
grain scattering as
concentrated
on
understanding the frequency dependence of attenuation, the influence of microstructure
on velocity
and attenuation,
and on developing a
correlation between
the ultrasonic
response and grain size in cubic, monophase, polycrystalline materials having equiaxed
grains
with
a
narrow
distribution
of
grain
sizes
and
shapes.
This
work
also
comprehensively reviews related research going back to the late 1950’s. Equiaxed single
phase aluminum and steel alloys have been the subject of a great deal of research in the
ultrasonic grain scattering arena. These materials are nicely suited for studies on the
18
effect of grain size on ultrasonic properties, due to the ease of generating samples with
uniform, narrowly distributed grain sizes. Many
made
researchers [14, 15, 16, 17, 18] have
long lasting contributions to this field, having concentrated on modeling the
interaction of the ultrasonic waves with the solid. Others [19, 20, 21] have taken the more
practical
approach
by
empirically
correlating
ultrasonic
measurements
with
microstructural features like grain size.
2.2 Wave Propagation in Inhomogeneous Materials
The literature on wave propagation in inhomogeneous or heterogeneous media is
extensive. Composite materials are considered heterogeneous and wave propagation is
dramatically influenced by the structure of the material. Many general studies concerning
ultrasonic propagation in heterogeneous material have been made in the recent past (22,
23]. Wave propagation and scattering theory in weakly anisotropic, equiaxed polycrystals
in the absence of preferred orientations has been extensively studied [24, 25] for both
single and multiphase materials.
A unified approach to the solution to the wave equation, based on ultrasonically
detected variations
of elastic properties, has been
developed
[26]. The
influence
of
texture and elastic anisotropy on wave propagation in polycrystalline metals has been
also the subject of extensive research [27, 28, 29, 30]. Here, the theoretical work has
concentrated on allowing the prediction of scattering coefficients, longitudinal and shear
wave
velocities, and attenuation as a function of frequency
in textured polycrystals.
However, these models only apply to cubic metals. Much of the work on theoretical wave
19
propagation is mathematically intensive and goes well beyond the scope of this research.
The wave propagation in polycrystalline metals involves solving the wave equation,
taking
into
account
for
calculations
scattering,
directional
dependencies,
frequency
variations, elastic properties, different types of waves, mode conversion and so on.
It is important to realize that much work has been conducted in this research
subject; however, most of these studies are based on assumptions and criteria that are
simply not supported for titanium alloys. For example, the fact that the single crystal
elastic constants for Ti-6AI-4V do not exist in a rigorously determined and reproducible
manner makes it difficult to apply any of these extensively developed theoretical models
to the
material.
Moreover,
wave
propagation
in Ti-6AI-4V
is complicated
its
by
hexagonal symmetry and differences in density, elastic constants, and lattice structure of
its two phase microstructure. Likewise, developing an understanding of the influence of
microstructure on wave propagation in the absence of a preferred orientation is very
difficult, due to the relative ease in imparting texture and the difficulty in getting rid of it
without otherwise affecting the microstructure.
2.3 Ultrasonic Backscattering for Structural Characterization
The occurrence of grain structure in metal alloys gives rise to randomly occurring
ultrasonic
signals, which can be in some
Microscopically
homogeneous
but
cases easily mistaken
randomly
oriented
individual
for genuine
grains
make
defects.
up
a
macroscopically isotropic but inhomogeneous medium which produces incoherent wave
scattering commonly
referred to as "grain noise." This ultrasonic scatter generated in
structural alloys is basically the result of local variations of either stiffness or density in
the material. Features which promote scatter are both microscopic, due to grain structure,
second phase particles, precipitates, crystal defects, microporosity, and microcracks, and
macroscopic, due to the alignment of second phase particles or grain boundaries, the
formation of large colonies of similarly oriented grains, and deformation flow lines. In
Ti-6Al-4V, the microstructural complexity is such that the ultrasonic scatter originates
from grain-to-grain differences in crystallographic orientation, density, elastic properties,
the formation of macroscopic structure; and the formation of features originating from a
prior phase state (i.e. prior beta grain boundaries). In light of the multitude of scattering
sources and microstructural variations, titanium alloys represent a truly challenging class
of materials for ultrasonic characterization.
Acoustic
backscattering.
ultrasonic
flaw
observed
in many
materials
using
ultrasonic
grain noise has an obvious
adverse,
often prohibitive,
effect on
and
it can
be
grain noise
The
detection
is readily
[31,
32,
33,
34]
exploited
for ultrasonic
characterization of the grain structure (35, 36]. A major source for the generation of
ultrasonic scatter is grain boundaries. The factors primarily responsible for the ultrasonic
scatter in most titanium alloys are the alignment of grains and grain boundaries, second
phase particles, and grain colonies. Changing the ultrasonic inspection parameters, such
as reducing the frequency or increasing the transducer aperture, causes a reduction in the
generation of grain noise, but these tradeoffs often reduce the detectability threshold of
the inspection. Likewise, a variety of different averaging techniques can be employed to
reduce grain noise and improve the microstructure discrimination capability, but these
techniques are not favorable for detection of localized defects [37, 38].
2.4 Diffusion Bonding of Titanium Alloys
While
the
Ti-6Al-4V
alloy
not
is
ideal
for
analytical
microstructure
characterization from the standpoint of its many microstructural features responsible for
scattering and its diverse array of useful microstructure types, the alloy does have some
uniquely attractive properties. One of the most useful aspects of this alloy, aside from its
extensive processability, is its joining capability given the right combination of pressure,
time, and temperature [39, 40]. A beneficial feature of Ti-6AI-4V is that it is capable of
being welded and diffusion bonded. In diffusion bonding, the mating parts are heated to
about
one
half of the melting
point,
then
pressurized
to a stress
level
below
the
macroscopic yield point, and held at temperature and pressure for a given time period.
The procedure commonly practiced, and used for development of bonded samples for this
research, was conducted in a HIP (Hot Isostatic Pressure) chamber with temperature,
pressure, and dwell time conditions of 950 °C, 30 ksi, for4 hours, respectively.
The diffusion bonding process involves three basic steps: (i) local yielding of
initially contacting surface asperities, (ii) creep deformation of the bond plane yielding
discontinuous voids, and (iii) closure of the voids by vacancy diffusion [41]. In practical
situations, oxide free surfaces are rare. Oxidized surfaces generally interfere with the
bonding process due to impedance of the rate and extent of diffusion across the bond
plane. This is generally not the case for titanium alloys, which at temperatures above 850
°C readily dissolve minor amounts of absorbed gases and thin surface oxides, allowing
22
diffusion to carry them
away
from
the bond
interface.
Another
aspect of diffusion
bonding unique to this research is the joining of dissimilar microstructures of the same
chemical
composition.
While
many
researchers
have
investigated
the
bonding
and
characterization of dissimilar metal compositions, it is evident from the limited number
of publications found that bonding of dissimilar microstructures has been given very little
attention in the literature [42]. A great deal of research has been conducted to better
understand how to nondestructively assess diffusion bond characteristics and to address
the determination of processing parameters that provide optimal bond strength (43, 44,
45, 46]. Here, considerable effort has gone into the development of theories to predict the
relationship between ultrasonic scattering from diffusion bonds and the bond conditions,
based on experimental NDE measurements.
23
Chapter 2, References:
1. L. Rayleigh, “The Theory of Sound”, (Macmillan, 1945) pp. 147-152.
2. I. M. Lifshits and G. D. Parkhomoskii, “On the Theory of the Propagation of
Supersonic Waves in Polycrystals,” Zhur-Experim. I. Theoret. Fiz., Vol. 20 (1950).
3. A. B. Bhatia, “Scattering of High Frequency Sound Waves in Polycrystalline
Materials,” Journal of the Acoustical Society of America, Vol. 31, pp. 16-23 (1959).
4. A. B. Bhatia and R. A. Moore, “Scattering of High Frequency Sound Waves in
Polycrystalline Materials II,” Journal of the Acoustical Society of America, Vol. 31, pp.
1140-1144 (1959).
5. H. B. Huntington, “On Ultrasonic Scattering by Polycrystals,” Journal of the
Acoustical Society of America, Vol. 22, p. 362 (1950).
6. Ultrasonic Testing of Materials, eds., J. Krautkramer and H. Krautkramer, (SpringerVerlag, NY, 1990), pp. 425-430.
7. C. F. Ying and R. Truell, “Scattering of a Plane Longitudinal Wave by a Spherical
Obstacle in an Isotropically Elastic Solid,” Journal of Applied Physics, Vol. 27, pp.10861097 (1956).
8. P. C. Waterman and R. Trvell, “Multiple Scattering of Waves,” Journal of
Mathematical Physics, Vol. 2, pp. 512-537 (1961).
9. Ultrasonic Methods in Solid State Physics, eds., R. Truell, C. Elbaum, and B. B. Chick,
(Academic Press, New York, 1969).
10. R. H. Latiff, and N. F. Fiore, “Ultrasonic Attenuation and Velocity in Two-Phase
Microstructures,” Journal of the Acoustical Society of America, Vol. 57, pp.1441-1447
(1975).
11. E. P. Papadakis, “Ultrasonic Attenuation due to Grain Scattering in Polycrystalline
Metals,” Journal of the Acoustical Society of America, Vol. 37, p. 711 (1965).
12. E. P. Papadakis, “Ultrasonic Attenuation Caused by Scattering in Polycrystalline
Media,” Physical Acoustics Principles and Methods, Vol. IV, Part B, Ed. Mason, W. P.,
(Academic Press, New York, 1968) p. 269.
13. E. P. Papadakis, “Influence of Preferred Orientation on Ultrasonic Grain Scattering,”
Journal of Applied Physics, 36(5), May, p. 1738 (1965).
14. F. J. Margetan, R. B. Thompson, and I. Yalda-Mooshabad “Modeling Ultrasonic
Microstructural Noise in Titanium Alloys,” Review of Progress in Quantitative
Nondestructive Evaluation, Vol. 12, eds., D. O. Thompson and D. E. Chimenti, (Plenum,
New York, 1993), p. 1753.
15. B. Fay, “Theoretical Considerations of Ultrasonic Backscatter,” Acoustica 28, p. 354
(1973).
16. J. Lewandowski, “Evaluation of the Texture of Polycrystalline Aggregate from
Ultrasonic Measurements,” Ultrasonics, March, p. 73, (1986).
17. C. M. Sayers, “Ultrasonic Velocities in Anisotropic Polycrystalline Aggregates,”
Journal of Applied Physics D, 15, p. 2157, (1982).
18. C. B. Guo, P. Holler, and K. Goebbels, "Scattering of Ultrasonic Waves in
Anisotropic Polycrystalline Metals," Acoustica 59, pp. 112-120 (1985).
19. D. Beecham, “Ultrasonic Scatter in Metals, Its Properties and its Application to Grain
Size Determination,” Ultrasonics, April, p. 67 (1966).
20. J. Szilard, and G. Scruton, “Revealing the Grain Structure of Metals by Ultrasonics,”
Ultrasonics, May, p. 114 (1973).
24
21. R. L. Smith, “The Effect of Grain Size Distribution on the Frequency Dependence of
the Ultrasonic Attenuation in Polycrystalline Materials,” Ultrasonics, September, p. 211
(1982).
22. J. M. Perdigao, A. Ferreira, and C. Bruneel, “Ultrasonic Anomalous Behavior in
Composite Samples,” Acoustica, Vol. 63, pp. 106-110 (1987).
23. D. Sornette, “Acoustic Waves in Random Media, Experimental Situations,”
Acoustica, Vol. 68, pp. 15-25 (1989).
24. §. Hirsekorn, “The Scattering of Waves by Polycrystals,” Journal of the Acoustical
Society of America, 72(3) (1982).
25. S. Hirsekorn, “The Scattering of Ultrasonic Waves by Multiphase
Journal of the Acoustical Society of America, 83(4) (1988).
Polycrystals,”
26. F. E. Stanke and G. S. Kino, “A Unified Theory for Elastic Wave Propagation in
Polycrystalline Materials,” Journal of the Acoustical Society of America. 75(3) p. 665
(1984).
27. S. Hirsekorn, “The Scattering of Ultrasonic Waves in Polycrystalline Materials with
Texture,” Journal of the Acoustical Society of America, 77(3), (1985).
28. S. Hirsekorn, “The Dependence of Ultrasonic Propagation in Textured Polycrystals,”
Journal of the Acoustical Society of America, 79(5), (1986).
29. C. M. Sayers, “Angular Dependent Ultrasonic Wave Velocities in Aggregates of
Hexagonal Crystals,” Ultrasonics, Vol. 24, pp. 289-291 (1986).
30. S. Ahmed and R. B. Thompson, “Propagation of Elastic Waves in Equiaxed
Stainless-Steel Polycrystals with Aligned [001] Axes,” Journal of the Acoustical Society
of America, 99(4), (1996).
31. J. H. Rose, “Ultrasonic Backscattering from Polycrystalline Aggregates Using TimeDomain Linear Response Theory,” Review of Progress in QNDE, Vol. 11, eds D. O.
Thompson, D. E. Chimenti (Plenum, New York 1992) p. 1677.
32. J. H. Rose, “Ultrasonic Backscatter from Microstructure,” in Review of Progress in
ONDE, Vol. 11, eds D. O. Thompson, D. E. Chimenti (Plenum, New York 1992) p. 1677.
33. J. H. Rose, “Theory of Ultrasonic Backscatter From Multiphase Polycrystalline
Solids,” in Review of Progress in QNDE, Vol. 12, eds D. O. Thompson, D. E. Chimenti
(Plenum, New York 1993) p. 1719.
34.
F.
J.
Margetan,
R.
B.
Thompson,
and
I.
Yalda-Mooshabad,
“Backscattered
Microstructural Noise in Ultrasonic Toneburst Inspections," Journal of Nondestructive
Evaluation, Vol. 13, 111-136 (1994).
35. K. Gobbles, “Structure Analysis by Scattered Ultrasonic Radiation,” Research
Techniques in Nondestructive Testing, Vol. IV, edited by R. S. Sharpe, (Academic Press,
New York, 1980), pp. 87-157.
36. A. Hecht, R. Thiel, E. Neumann, and E. Mundry, "Nondestructive Determination of
Grain Size in Austenitic Sheet by Ultrasonic Backscattering," Materials Evaluation, Vol.
39, pp. 934-938 (1981).
37. H. Willems and K. Goebbels, "Characterization of Microstructure by Backscattered
Ultrasonic Waves," Metal Science, Vol. 15, pp. 549-553 (1981).
38. Y. K. Han and R. B. Thompson, “Ultrasonic Backscattering in Duplex
Microstructures:
Theory
and
Application
to Titanium
Materials Transactions A, Vol. 28A pp.91-103 (1997).
25
Alloys,”
Metallurgical
and
39. H. G. Kellerer, and L. H. Milacek, “Determination of Optimal Diffusion Welding
Temperatures for Ti-6Al-4V,” Welding Research Supplement, 49(5), pp. 219s-224s,
(1970).
:
40. R. J. Rehder, and D. T. Lovell, “Process Development for Diffusion Welding Ti-6Al4V Alloy,” Welding Research Supplement, 49(5), pp. 213s-218s, (1970).
41. C. H. Hamilton, “Pressure Requirements for Diffusion Bonding of Titanium,” in
Titanium
Science
and
Technology,
eds., R. I. Jaffee
and H. M.
Burte
(Plenum
Press,
1972), pp. 625-647.
42. A. A. Gelman, A. A. Kotelnikov, and N. I. Kolodkin, “The Effect of Diffusion
Bonding Variables on Structure and Properties of the Bonded Alpha + Beta Titanium
Alloy,” in Titanium and Titanium Alloys, Eds. J. C. Williams and A. F. Belov, (Plenum
Press, 1976), 1209-1220.
43. D. D. Palmer, C. D. Roberts, et al., “Strength and Ultrasonic Characterization of
Metallic Interfaces,” in Review of Progress in Quantitative Nondestructive Evaluation,
Vol 7, Eds. D. O. Thompson and D. E. Chimenti, (Plenum Press, 1988), pp. 1335-1342.
44. J. H. Rose, “Ultrasonic Characterization of Solid-Solid Bonds from Microstructural
Changes,” in Review of Progress in Quantitative Nondestructive Evaluation, Vol. 7, Eds.
D. O. Thompson and D. E. Chimenti, (Plenum Press, 1988) pp. 1311-1318.
45. G. T. Thomas and J. R. Spingarn, “Ultrasonic Determination of Diffusion Bond
Strength,” in Review of Progress in Quantitative Nondestructive Evaluation, Vol. 3, Eds.
D. O. Thompson and D. E. Chimenti, (Plenum Press, 1984) pp. 1243-1250.
46. B. H. Hosten, L. A. Ahlberg, B. R. Tittman, and J. R. Spingarn, Ultrasonic
Characterization of Diffusion Bonds,” in Review of Progress in Quantitative
Nondestructive Evaluation Vol. 6, Eds. D. O. Thompson and D. E. Chimenti, (Plenum
Press, 1986) pp. 1701-1706.
26
CHAPTER III
MICROSTRUCTURAL EFFECTS ON ULTRASONIC PROPERTIES
3.1 Microstructures Evaluated
One
of this research
was
influence
the
to determine
of
of the
main
goals
on
the
ultrasonic
response
in
investigated,
including:
(i) the initial as-received mill annealed
microstructure
microstructures were
material, (AR); (ii) coarse beta annealed, (CBA);
and
(RA);
(v)
duplex
Hence,
Ti-6Al-4V.
annealed,
(DA)
annealed,
Microstructures
Evaluated).
microstructures
of interest since the bar is highly anisotropic.
as-received
different
(iv)
(ili) fine beta annealed, (FBA);
recrystallization
The
five
material
really
As
(see
Table
has
two
3.1
—
different
such, each of the
different microstructures were examined in both the axial and transverse orientations.
However, only the mill annealed Ti-6A1-4V clearly showed a distinction between the two
orientations
studied,
as
shown
in Figure
3.l.a
and
3.1.b.
The
commonly produced in the mill annealed condition, where it has
strength,
toughness,
ductility, and fatigue properties,
Ti-6Al-4V
alloy
is
a good combination of
but the properties
are strongly
orientation dependent [1, 2]. Approximately 2 meters of 2.5” (63.6 mm) diameter forged
bar stock was acquired as a basis for the different microstructure samples, which were
sectioned, heat treated, and polished for flat and parallel surfaces.
27
87
‘SOINJOMASOIOIU UOTIPPe OJ pue ‘UOTIPUOD poataool-se oY} JO SUOTIEIUSLIO a[dryNUI Jo pajstsuod YOrYM porprys
OE-ST
Or-0€
06 ‘09 ‘Sb “LE “OE “0
a
Te e198L
UleIS [BUIWION
(ur) 9218
pure poyersue8 osom sopdueg ‘sydesSoso1w Surpnyjout ‘Apmis WoNeyossip Sty} SULINP pojenyead SoMjon.ysOIONW sy YL,
00S-00€
06 0
bes a
000€-0007
,
SUOTJCJUSTIO
_, | (seeudap) POTPTS
.
06 0
soiniee
fe «
06 0
saluojoo
,
,
Jo suorSau Je]/auey
urexd ‘sures Jo
—ures3 pu ‘saepunoq
,
06 0
q+
Jo Sunsisuoo ‘ejaq]}
3
pouojsuey paynqlsysip
Joiid yo aouapiaa | Jord Jo souaplAa
yuOTUeO yeOHT
SulYysInsurjsiq
nyeoj
sured eydje pexembs|
Sulels 8}0q
Bjoq IBjNUEIsIO}U!
surels B}9q
eydye poxeinbs surly
OV ‘OD 0€2 “Ub
dINJON.YSOISTYA
SITYepOUSUION,
dv
(PSAT999I-SB)
poereouue [TIA
yuowustye jerxe Suoys
DAROMO¥sSeq
Jeg] pue satucjoo | = — SWOs ‘anjon43s
uieId JEppaweE] 93Ie]
Smid ‘OW
ayIsusTeU SAOge JOY
‘OD geo “AY | “dutay els
OV ‘D POL “84
DUISUSLEU MOfeq JeaY
‘9.006 TY p ‘dus eis |
ov ‘ovo. | ov ‘DO rol UN zsnid SOV]
sy zsnd oa}
OV ‘DO H0L
‘> 000154 S‘0
vd
peyTeouue xojdnq
poyenjeay SompoNQSOIJAl
UOZT[VISAINIY
Va
poyeouue
‘ay Z snd ‘oq
Vdd
peTeouue
ejoq Our]
“> LEOLIY 9
Vado
peyeouue
BIJ9q BSIBO7)
Figure 3.1
The as-received condition (mill annealed) of Ti-6Al-4V bar stock in the
transverse (a) and axial (b) orientation. Both microstructures shown at
200X.
Thermomechanical breakdown of the original ingot is generally done via cogging,
upsetting, or (for smaller sections) extrusion [3]. In cogging, the initially round ingot is
pressed on its sides between flat dies and incrementally rotated and indexed forward,
resulting in a long, relatively thin (billet) product. Upsetting
cogging
to introduce
additional
strain energy
to promote
is often a precursor to
recrystallization upon
hot
working in the a +B phase field where the self diffusion coefficient is quite low (e.g. at
500 °C, Dar © 107!9m? /s). As a result, the grain structure remains fine upon cooling,
with
nominally
15-20
pm
primary
a
grains
[4]
and
approximately
5-7%
Vr
of
intergranular B in either orientation. The mill annealed microstructure in the transverse
orientation takes on a distinctive "wavy" lamellar appearance sometimes called a partially
transformed widmanstiatten structure. In the axial orientation, the primary alpha grains of
the mill annealed structure have nominally a 5:1 aspect ratio and large colonies of
similarly oriented alpha grains tend to agglomerate
into distinctive bands, which are
thought to behave essentially like large single crystals. The mill anneal conditions the
29
alloy into a relatively soft (Rc ~ 34), easily machinable
state. Typical mill annealing
involves heating the alloy to approximately 730 °C (in the a + B phase field) where it is
held for approximately 4 hours, followed by air cooling to room temperature. This heat
treatment is generally given to all titanium mill products and often the annealing is
intentionally left incomplete, leaving telltale traces of the heavily worked structure as
seen in Figure 3.1b. This conditioning leaves the billet in a highly anisotropic state.
The duplex and recrystallization annealed microstructures were also investigated.
It is important to point out that neither of these samples demonstrated any significant
recognizable
difference
in microstructural
features
between
the axial and transverse
orientations, except at low resolution (~25X) optical microscopy images, which reveals a
banded structure similar to that pbserved
in the mill annealed condition. The duplex
microstructure results from heat treating the mill annealed alloy above the martensite start
temperature. Different variants of the duplex anneal process are available to vary the
relative amounts of « versus B, but the process chosen for this work involved heating the
alloy to 938 °C for 1 hour, followed by air cooling, then reheating to 704 °C for 2 hours
and air cooling to room temperature. The duplex microstructure dramatically transforms
the mill anneal to appear as is shown in Figure 3.2.a. This microstructure is comprised of
equiaxed a grains of nominally 30-40 pm with distributed transformed 8 appearing as
fine lamellar regions throughout the structure. The annealing temperature determines the
amount of a and transformed B according to the phase diagram lever rule. During
annealing, the primary o phase retains the morphology of the originating structure from
thermomechanical processing. On the other hand, the transformation of the beta phase
30
depends on the cooling rate, whereas slower cooling results in the commonly observed
lamellar packets of secondary alpha, shown as the dark regions.
The recrystallization annealed (RA) microstructure is also result of heating the
mill annealed material below the martensite start temperature, leading to the evolution of
fine equiaxed a and B grains as shown in Figure 3.2.b. This treatment involves heating
the alloy to 900 °C for 4 hours followed by air cooling, then reheating to 704 °C for 2
hours and air cooling to room
temperature.
The
microstructure
consists of equiaxed
primary o grains of approximately 25-30 jm and intergranular 8. The recrystallization
annealed condition results in high fracture toughness, good hot formability, and good
fatigue initiation resistance.
Beta annealing relies on slow cooling from above the beta transus where the alloy
is transformed to a cubic lattice structure. As the hot worked alloy is cooled from the B
phase region, « crystallizes in the form of plates with a known crystallographic regularity
‘in which the o plates form with their basal planes parallel to the {110} planes in the B
phase.
Similarly, the body diagonal in the B phase forms parallel to the basal plane
diagonal
in the
a
phase.
These
orientation
rules
are
relationship [5, 6], expressed in terms of Miller indices as:
{110}, //(0001),
and
(111), / (1120)
consistent
with
the
Burgers
Eso yc + eb ete!¢
Bee aura
Figure 3.2
Above
Gora
a Se
ub Bhd
t.
(a) and_ the
anneal
the duplex
resulting from
Microstructures
recrystallization anneal (b). Both microstructures shown at 200X and both
have nominally the same primary « grain size of approximately 30 jm.
the beta transus
(~1000
°C)
each B grain has six possible
combinations
of
independent {110} planes on which a plates can form, resulting in the now commonly
known Widmanstatten microstructure. This microstructure consists of a plates separated
by thin films of the B phase (approximately 10-15 ym thick), which form with a known
crystallographic relationship to the prior B phase from which they originated.
The fine beta annealed microstructure examined in this work is the result from
heating the alloy to 1000 °C for 0.5 hours followed by furnace cooling, then reheating to
704 °C for 2 hours and air cooling to room temperature, resulting the microstructure as
shown in Figure 3.3.a. This heat treatment drastically changes the appearance of the
microstructure from the highly anisotropic mill annealed condition. The banded structure
evident at low magnifications in the AR, DA, and RA
samples is almost completely
removed, leaving a structure with a relatively small prior beta grain size of 300 to 500
um,
but broad lamellae because
of the slow furnace cooling rate (~100 °C/h)
32
from
approximately
1000
°C.
The
relatively
slow
(air) cooling
rate results
in diffusion
controlled partitioning between the « and B phases as the temperature falls below the B
transus. The microstructure is lamellar and the size of the microcolonies depends on the
prior beta grain size and on the cooling rate. To reiterate, for this FBA microstructure the
transverse orientation is virtually indistinguishable from the axial orientation.
The coarse beta annealed microstructure is heated to 1037 °C for 6 hours followed
by furnace cooling, then reheated to 704 °C for 2 hours and air cooled, as shown in Figure
3.3.b. The prior beta grain boundaries are clearly evident in the coarse beta annealed
microstructure. The prior beta grain size is determined by the dwell time above the beta
transus and the morphology of the « /B combination is strongly dependent on the cooling
rate. Quenching from above the beta transus results in a martensitic transformation where
the beta phase becomes needlelike. Slower cooling allows diffusion between the phases,
resulting in a lamellar structure with broad o plates alternating with fine B phase. This
lamellar structure partitions into colonies of similar orientation. The size of the colonies
depends on the prior beta grain size and the cooling rate. Another interesting feature,
often occurring
in alloys that have been cooled from
boundary « (GBa), which is clearly evident in the CBA
above the B transus is grainmicrographs of Figure 3.3.b.
GBa nucleates at prior B grain boundaries and its thickness and continuity depends on the
cooling rate and on alloy composition.
Again,
no difference was
observed
in the
appearance between the two CBA orientations studied. However, a huge difference is
seen between the fine versus coarse beta annealed samples in terms of both the prior beta
grain size and microcolony size.
(b)
Figure 3.3
The beta annealed microstructures: (a) fine beta annealed structure shown
at 200X, exhibiting approximately 300-500 1m prior beta grain size, and
(b) coarse beta annealed structure shown at 25X, with extremely large
approximately 2-3 mm prior beta grain size.
PART I: Conventional Ultrasonic Evaluation Methods
3.2 Ultrasonic Velocity
This section centers on an assessment of the influence of texture, orientation, and
wave
microstructure
on
ultrasonic
microstructures.
For
fundamental
in
propagation
understanding
of the
several
common
interaction
of
titanium
propagating
ultrasonic waves within a titanium solid, many reliable techniques have been developed
in the
past
for measurement
of ultrasonic
velocity,
attenuation,
and
backscattering
characteristics. While these fundamental techniques are not new, they provide a baseline
capability to evaluate materials and provide useful information on the elastic properties
and microstructure characteristics. This section relies on NDE experimentation to study
different structures, especially since theoretical models have not been developed which
are capable of predicting the complex interactions that take place as an ultrasonic wave
passes through a complex material. Moreover, the theoretical understanding of elastic
34
wave propagation in an inhomogeneous anisotropic solid is mathematically very intensive
this research.
and goes beyond
an
objective of this section is to develop
The main
understanding of the basic ultrasonic response characteristics of different microstructures
of the titanium alloy and to determine what parameters primarily influence the behavior
waves.
ultrasonic
of propagating
into three basic
is broken
section
The
categories:
velocity, attenuation, and backscattering.
The longitudinal velocity of sound, including ultrasound, depends on the elasticity
and
density
propagation
of the
The
medium.
where
compressional wave of a small amplitude is c= JE/p,
and
p_
is the density. For fluids,
E
of a
for velocity
expression
general
E is the elastic modulus
is the bulk modulus of elasticity, but is equal to
Young's
modulus
for the case of solids. Crystallographic texture and anisotropy
integrally
linked
to the
elastic response
ultrasonic
velocity
due
to directional
are
of the material
and
significantly
influence
[7,
8, 9].
The
elastic
property
dependencies
variations in polycrystalline solids are evident from the ultrasonic velocity and depend
fundamentally on a number of different factors including: (i) the single crystal elastic
properties of the metal, (ii) the crystallographic structure, and (iii) the effects (if any) of
deformation and processing, which may bring about residual stresses or crystallographic
texture. Acquiring single crystal elastic properties of metals has been a subject of interest
to
many
researchers
[10,
frequencies to the sample
algorithms
to
deduce
the
11].
Resonance
techniques
introduce
a broad
range
of
and record the response which feeds different computer
independent
elastic
moduli
of anisotropic
single
crystal
materials [12, 13, 14]. However, fundamental technical challenges still inhibit our ability
to experimentally determine the independent elastic constants of Ti-6Al-4V.
Accurate velocity measurements require sophisticated hardware and the use of a
as pulse-overlap
excited via tone-burst,
which
results in an
reliable technique
such
accuracy of +0.5%.
Also, samples are required to be flat and parallel with a sample
diameter to transducer aperture ratio of approximately 3:1, to minimize errors introduced
via sidewall
interference.
An
extensive
effort has been taken to satisfy the sample
requirements, consistent with guidelines specified in the literature for the practice of
measuring ultrasonic velocity via the pulse overlap technique [15, 16]. All samples were
measured in both the axial and transverse directions to assess the effects of texture,
orientation, and microstructure on the velocity. The nomenclature used to describe the
samples is shown in Table 3.2 -Sample Nomenclature, Appendix A.
Anisotropy can be observed through a variety of phenomena such as orientation
dependent
acoustic
velocity
changes
and
birefringence.
Experimentation
involving
orientation dependence of longitudinal and shear wave propagation provides a method to
assess texture and microstructural influence in bulk materials. Birefringence means the
refraction of the ultrasonic shear wave into two waves of slightly different velocity.
Figure 3.4 shows the geometrical configuration for the birefringence measurement. At the
pure mode polarization directions both fast and slow modes can simultaneously travel at
two separate, slightly different times. However, the phase interference resulting from the
difference in arrival times often causes significant amplitude reductions in the received
radial
:
‘=>
“~
AR(90)
Figure 3.4
axial polarization
The geometric configuration of the birefringence measurement
different polarization and wave propagation directions.
with
signal in-between the pure mode directions.
Velocity measurements have revealed only minor variations due to microstructure,
except in the coarse beta annealed sample. However, some dramatic velocity variations
results were observed due to the presence of a preferred crystallographic orientation,
shows
the
average
stock.
Figure
3.5
velocities (see Table 3.3 — Longitudinal
Wave
Velocity, Appendix
especially
in the
as-received
bar
longitudinal
A). To minimize
experimental error, the thickness and time delays between signals on each sample was
measured
several times and averaged. Thickness calibration blocks were also used to
make certain the delay times were measured accurately.
The data clearly shows the average longitudinal wave velocity is highest in the
transverse direction of the mill annealed sample, AR(90). A difference of about 2% is
observed in the wave velocity between the axial versus the transverse direction of the as-
received bar stock. The longitudinal data also shows a linear decrease in velocity as a
37
6300 7
6050 +t
| 90 degree
L] 0 degree
6200 --
_
—
[|
6150 ++
6100 4
6050
|
+
cog JM yt tit AR(30)ttt
AR(0) DA(0) DA(90) RA(O) RA(90) FBA(0)FBA(90)CBA(0) CBA(90)
AR(90) AR(60) AR(45) AR(37)
sample
Figure 3.5
Results showing the variation in average longitudinal wave velocity for the
various titanium alloy samples. The scatter on these measurements is
generally about + 0.5%.
function of the angle between the surface normal and the axial direction in the six asreceived
samples.
Special
samples
were
developed
to
examine
the
influence
of
macroscopic texture on the ultrasonic response. These as-received samples were cut such
that the surface normals were at known angles (0, 30, 37, 45, 60, and 90 degrees) relative
to the axis of the mill annealed bar stock, shown in Figure 3.6. A steady reduction in
velocity of about 22 m/s is observed as the surface normal changes from transverse (90°)
to parallel (0°), relative to the bar axis. Beyond this uniform reduction in velocity as the
incidence angle changes
in the as-received samples, we see measurable variations in
velocity between the other the axial and transverse directions for samples AR, DA, and
RA, all showing approximately a 2% difference. Samples FBA and CBA also show this
38
-
ay
Lo)
ON
rom)
o
‘©
Cnt:
ee
it
~ ~,
Figure 3.6
Diagram showing the different Ti-6Al-6V sample sections cut from the
2.5" diameter cylindrical bar (i.e., five samples having varying surface
normals relative to bar axis).
directional dependence effect, but to a much lesser degree. The reason for this behavior is
because of recrystallization, which removes the texture, based on the dwell time above
the beta transus. Sample DA(0) has the lowest velocity, closely followed by samples
RA(0) and AR(0) with FBA(0) and CBA(0) having slightly higher velocity then the rest
in the axial group. The data shows the influence of microstructure has a measurable, but
modest, effect on the longitudinal acoustic wave velocity. What is evidently much more
influential on the velocity is the effect of texture. The
appears
between
(lowest velocity,
sample AR(90)
~6,110
m/s),
(highest velocity,
a difference
largest difference in velocity
~6,250
of approximately
m/s)
and
2.25%.
sample
We
DA(0)
see lesser
velocity variations between orientations in the beta annealed samples, due to a reduction
in the originally present texture. The directional velocity dependence in samples AR, DA,
and RA is due to the presence of macroscopic texture. Even though samples DA and RA
were heat treated to transform the microstructures, residual macroscopic texture is clearly
still present in the structures. This is also evident from the metallographically prepared
DA and RA samples which reveal a heavily banded post heat treated structure, similar to
39
that observed in the AR sample. Beta annealed samples FBA
have
trace
amounts
of texture
still
present
even
though
and CBA
the
evidently also
are
microstructures
dramatically altered from the as-received material condition.
If we were to assume the material isotropic, then the measurements on the as-
received sample demonstrate that the combination of Lame's constant are higher in the
transverse direction. Hence, Young's Modulus is clearly higher in the transverse direction.
This result comes from the relationship between acoustic velocity and stiffness, which
can be expressed as:
G.1)
¢) = Ae 2e
p
for an isotropic solid. This velocity dependence in the anisotropic as-received material is
not too surprising since it has been known for many years that titanium often develops a
preferred orientation due to deformation. The type of texture observed in the as-received
bar stock is consistent with that of a wire texture with a uniform reduction in the crosssection of the bar.
The
Ti-6Al-4V
bar
material
studied,
has
sustained
the
largest
strain
and
deformation in the axial direction, therefore, we see the formation of a heavily banded
macroscopic structure along the axis of the bar. What is not so obvious, is the orientation
of the basal plane normals.
The
material
is clearly highly textured
40
in terms
of the
alignment of features in the structure and the velocity data suggests the presence of a
preferred crystallographic
orientation in which the principal
axis of the hexagonally
symmetric alpha grains tend to preferentially lie in the transverse direction of the bar.
This fact was corroborated by x-ray analysis, using Shultz back-reflection.
The development of texture in the Ti-6Al-4V bar is due to a combination of
intrinsic crystallographic and structural anisotropy [17]. The longitudinal velocity data
indicates the basal normal directions are preferentially oriented to the transverse direction
of the bar as depicted in Figure 3.7. This figure shows the principal directions tend to
rotate to the x-y plane, but are not aligned uniformly in the radial direction of the bar.
Similarly, in a rolled plate the principal crystallographic directions tend to align laterally
to the high deformation directions used in the processing [1, 2, 18, 19]. We can see from
the velocity data that the heat treatments have developed microstructures that are much
alike in terms of the overall elastic properties, except in the beta annealed samples, which
effect the longitudinal velocity significantly more
normals
tend
to a preferred
processing, but what is more
orientation
in the
due to recrystallization.
transverse
direction
as
The basal
a result
influential to the velocity of propagating waves
of
is the
presence of large colonies of similarly oriented crystals in the axial direction. These large
oriented colonies tend to essentially behave as if they were single crystals that have very
high
longitudinal
aspect
ratio.
Consequently,
as the
sound
propagates
in the
axial
direction of the textured material, the ultrasonic wavefront becomes distorted because of
local velocity extremes. This phase distortion causes the coherent backwall echoes to vary
in amplitude due to interference and can result in misleading measurements of the
4]
Figure 3.7
The hexagonal principal directions preferentially rotate to the radial
direction (x, y plane) in the as-received titanium bar stock, based on
longitudinal velocity and x-ray analysis.
ultrasonic properties if one is not careful to do significant averaging.
Shear
wave
velocity
and
birefringence
[20,
21]
measurements
allow
more
extensive understanding of the effects of texture and microstructure as these waves are
generally polarized,
Shear
waves
allowing greater flexibility for materials
incident
normal
to the
face
of a sample
will,
characterization studies.
thanks
to birefringence,
generally be resolved into two transverse waves vibrating at right angles to each other and
propagating at different velocities. This effect is more dramatic in single crystals and
anisotropic polycrystalline metals due to the directional dependence
of velocity. The
shear wave velocity data of Figure 3.8 shows significant velocity variations between the
various samples, especially evident in the as-received samples (see Table 3.4 — Shear
Wave Velocity, Appendix A). The most dramatic evidence of this influence is illustrated
42
sample, with the difference between the slow and fast pure mode
in the AR(90)
of
approximately 4.8%. Similar to the longitudinal velocity data, we see in the as-received
samples a steady decrease in the birefringence delta as the incidence angle goes from
transverse to axial, with an axial birefringence delta of less than 1%. When evaluated in
the axial orientation, all of the samples exhibit a modest (Ag
< 1%) difference in
velocity between the slow and fast pure shear modes. These low values for birefringence
deltas
are
indication
an
that
the
macroscopic
texture
influencing
the
shear
wave
propagation in the axial direction is essentially axisymmetric. That is, the shear velocity is
basically independent of the polarization direction for propagation in the axial direction.
On the other hand, for wave propagation in the transverse direction, the shear polarization
direction has a major influence on the wave velocity. It was interesting to note that for
each sample evaluated in the transverse direction, with the exception of sample CBA(90),
3350 T
;0)wWww
§
ww wo 0° gg nq
31501NJ
NAAN
iS
S
slow mode
A
72 47 3
ne
in
310 +NZ
\Z “Z «7GF
A427
YZ
sooo
N40.
WCQ
shear velocity (m/s)
3300 +
7G
62,7,
tast mode
i
.
y
y
j
jy
G
4,4 y
,
wnrsgnunvyv
4NA NZ V9
NZ ZN
GW
[GAM
YGNYGYVENMYVYWV
TNGNWGYGNWVYNMYGV
,NA NA NZ NZ, NANA
AR(90) AR(60) AR(45) AR(37) AR(30) AR(0)
SANZ
DA(0) DA(90) RA(0) RA(90) FBA(0) FBA(90)CBA(0)
sample
Figure 3.8
Shear wave velocity data taken at different polarization angles consistent
with the slow and fast pure shear modes.
43
the direction
consistently
supporting
was
the fast mode
polarized
transversely
with
respect to the macroscopic inhomogeneity, as shown in Figure 3.9. Unfortunately, sample
CBA(90) was too attenuative to take accurate velocity measurements.
Wave
propagation in the transverse direction gives rise to the fast mode
only
when the particle displacements are perpendicular to the macroscopic inhomogeneities.
This finding makes sense in light of the longitudinal velocity data, which also exhibited
faster velocities for waves traveling perpendicular to the aligned features, which are in the
axial direction. Conversely, for transversely directed wave propagation, the slow mode
occurs when the particle displacements are parallel to the inhomogeneity, the less stiff
direction. The shear data is another indication that samples AR, DA, and RA demonstrate
S
more stiffness in the transverse direction.
+ -———-)
“os
'
me
~
ra
Ft
Pra
————
wv” fast mode
of
SK
Figure 3.9
This diagram shows the fast and slow pure mode polarization direction in
the 90° samples. Note the fast mode occurs when particle displacement is
perpendicular to the alignment bands in the axial direction of the bar.
44
In terms of elastic properties, if we assume the polycrystalline titanium alloy is
isotropic,
then the
shear modulus
is clearly higher
when
the polarization
is in the
transverse direction as compared to when the polarization is in the axial direction. This
result comes directly from Lame's elasticity relationship expressed as:
c=
(3.2)
|=
Pp
for shear waves, where 11, is the same as the shear modulus, G.
The RA and DA heat treatments appear to have only a minor influence on shear
velocity.
The
two
beta
the
same
velocity
sample
has
the
and
annealed
samples
have
essentially
The
coarse
beta
annealed
.5%.
This
is consistent with the idea that the
birefringence
characteristics.
birefringence
delta of approximately
smallest
relatively long time (6 hours) above the beta transus has effectively wiped out most of the
initial texture, leaving behind a coarse grained, but randomly oriented microstructure. The
other axially oriented measurements
in AR(0), DA(0), and RA(O) all show nearly the
same birefringence values, with minor variations in shear velocity between samples. The
scatter in the shear velocity is slightly higher than in the longitudinal velocity data. This is
primarily due to the fact that proper coupling is more difficult to achieve in shear
measurements, requiring a viscous coupling agent, compared to water coupling for the
longitudinal measurements.
45
3.3 Ultrasonic Attenuation
Attenuation
the loss of a signal's amplitude
generally means
with increasing
propagation distance. The loss is defined as the ratio of two amplitudes and expressed in
logarithmic units, Neper or decibel (dB) [22]:
(3.3)
L(Neper) = inh or L(dB) = 20 log AL,
Ay
Ay
where A, and 4, denote the amplitude with and without attenuation, respectively. In
some
cases
loss
the
JZ
occurs
locally
as a result of interaction
with
a material
discontinuity. These losses could be due, for example, to reflection and transmission at an
interface or scattering losses at a rough surface. There are other losses which occur over a
given distance as the wave propagates in the medium, but not necessarily proportional to
the distance covered; such losses are usually associated with divergence of the beam.
Attenuation of a medium is usually limited to phenomenon that cause loss proportional to
the propagation distance, expressed as
L =a d, where d is the propagation distance and
a is the attenuation coefficient.
Ultrasonic
attenuation
is the rate of decay
of mechanical
energy
as a wave
propagates through a material. There are two major classes of attenuation mechanisms
considered important for ultrasonic materials characterization, which can be expressed as:
& = & absorption
(3.4)
+a scattering’
46
First,
absorption
conduction,
acoustic
converts
energy
to
absorbed
elastic hysteresis, etc. The
heat
viscosity,
via
energy
relaxation,
heat
is irreversibly lost from the
acoustic field and is dissipated in the medium. Second, scattering converts the energy of
the coherent, collimated beam into incoherent, divergent waves as a result of interaction
with inhomogeneities in the material. The scattered energy is not necessarily lost as at
least part of it can be also picked up by the same transducer used to receive the coherent
wave as backscatter. Hence, scattering does not only reduce the amplitude of the coherent
signal, but also gives rise to an incoherent
material
noise which
further limits the
detectability of defects in the attenuated signal. On the other hand, the same signals that
block our view to the development
of low-level
(fatigue) damage,
also allow us to
characterize materials, based on ultrasonic backscattering.
Attenuation measurements were conducted in the same titanium sample set that
velocity
was
measured,
as
described
in
the
previous
measurements were taken in accordance with the ASTM
section.
The
attenuation
standard practice, (E664) [23].
For these measurements a conventional immersion ultrasonic system was used to avoid
difficulties in coupling breakdown suffered by contact ultrasonics. One of the main
problems encountered was due to the fact the attenuation is spatially distributed in the
titanium
alloy
samples
due
to
the
highly
inhomogeneous
nature
of the
material.
Therefore, attenuation measurements were spatially averaged over approximately 1" x 1"
on each sample. This was done by averaging signals while the transducer is being scanned
over the sample. Approximately 2 k waveforms were averaged over a 1" x 1" area on
each of the samples and repeated several times to ensure consistent results. The basic
measurement
involves digitally averaging waveforms
while the transducer is moving,
resulting in a spatial average. Each of the acquired spatially averaged signals correspond
to multiple round trip responses inside the material, as illustrated in Figure 3.10. The
transducer is pulsed via broadband excitation and then is a receiver for a relatively long
period (about 1 ms) during which time the wave propagates back and forth multiple times
through the 1.5" thick titanium samples. As the wave travels inside the material, the
amplitude of the coherently reflected waves continuously diminish due to grain scattering,
beam
reflection,
non
slightly
and
divergence,
parallel
plane
surfaces.
beam
The
divergence problem is one of diffraction from an aperture and has been solved for
longitudinal waves from a circular transducer of a given radius which radiates waves into
different fluids [24, 25]. These solutions work well for wave propagation in
focused
transducer
excitation
pulse
amplitude |
specular
reflection
water path
rn
Ae
‘ ee
“
ae
backwall echoes
re
'
7
15*round
'
>
ql
t
'
“‘sample'
Figure 3.10
mclope ~ exp(-otx)
3"
time
trip echo
Illustration of the ultrasonic wave propagating back and forth inside the
‘material and the associated waveform showing the multiple coherent
echoes.
48
homogeneous, isotropic materials. The problem of beam divergence has also been solved
for anisotropic media [26, 27] along axes of three, four, and six fold symmetry. The
corrections required to compensate for diffraction can be experimentally measured for a
given transducer or can be analytically calculated for models.
The simplest and most
often used model for ultrasonic transducers.is a circular piston radiator. The acoustic
pressure field of an unfocused circular radiator is illustrated in Figure 3.11. There is a
distinct difference in acoustic behavior between the near-field and far-field regions. The
transition between these two regions is defined as:
where a
is the aperture and 4 is the wavelength. For these attenuation measurements all
the sample thickness’ were the same and all measurements were taken in the far-field so
beam corrections were not used.
for the purpose
of
that the quantities we measure
for
While the attenuation is a valuable parameter to measure
materials characterization, we should recognize
attenuation are composed
of many
complex
interacting phenomenon
and not easily
separated. Moreover, the attenuation is integrated over the entire sound path inside the
material so the measurement is not well suited for locating anomalous indications or for
resolving local structural variations. Attenuation measurements are also very dependent
on the frequency and the coupling between the probe and the object.
nfat
far-field zone
near-field
Figure 3.11
The acoustic field of a circular piston source showing
variations at discrete distances from the transducer.
the
pressure
The data from the attenuation experiments demonstrate that the microstructural
differences between the samples are quite modest, with the exception of the coarse beta
annealed sample, which clearly has the highest attenuation. In Figure 3.12, a plot of the
frequency
dependent
attenuation losses (see also Table
3.5 - Ultrasonic Attenuation,
Appendix A). Its not surprising sample CBA is by far the highest attenuating sample in
both axial and transverse orientations with its very large prior beta grain structure and the
presence of large randomly oriented a+
colonies, which tend to strongly scatter the
incident energy. However, one unfamiliar with the nature of attenuation in Ti-6Al-4V
may be surprised to learn the AR(0) sample has the next highest apparent attenuation of
the lot. This is especially interesting after realizing that in AR(0) the wave is traveling
parallel to the aligned macroscopic colonies. Despite the fact that the waves are traveling
parallel to the alignment direction, the attenuation is approximately double that of the
transverse sample, AR(90). One of the reasons for this unusual attenuation phenomenon
has to do with the phase sensitivity of the receiver and the fact that the phase of the
wavefront is much more nonuniform in the axial direction. Part of this attenuation
anomaly is thought to result from localized extremes in velocity that perturb the
25 4
~N
oO
i
l
—* DA(90)
—® RA(90)
—4— FBA(90)
—e CBA(90)
—*- AR(90)
j
—
—
~
nN
N
Oo
!
Radial Attenuation
faa
as)
\/
~”
N
nN
2
8
10
12
14
frequency (MHz)
a)
aa)
Mo)
—o— DA(0)
— RA(0)
—2- FBA(0)
2 CBA(0)
—*— AR(0)
1
—
SN
N
Oo
l
Axial Attenuation
iv)
Ny
—
So
\
~~
nN
WN
2
b)
Figure 3.12
10
12
14
frequency (MHz)
Frequency dependent attenuation losses for the titanium alloy samples in
the (a) radial direction, and (b) axial direction.
51
wavefront. The phase distortion effect on attenuation, which is really not related to the
loss of energy commonly
associated with attenuation, will be discussed in Chapter 5
where a system to map microscopic phase variations has been developed.
Samples DA(0) and RA(0) have about the same attenuation followed by the group
consisting of FBA(0), AR(90), and FBA(90), respectively. Similar to the attenuation of
AR(0),
samples
DA(0)
and RA(O)
have
somewhat
unusual behavior as the optically
apparent microstructure in either sample is quite highly refined with equiaxed grains. The
explanation for the relatively high apparent attenuation in DA(0) and RA(0) is due to the
minor effect the annealing has on the attenuation. Remember, DA(0) was held in the a+B
phase field for only one hour at 937 °C and RA(0) was held at only 900 °C for four hours,
which, despite the fact that the microstructures clearly change, evidently does not perturb
the elastic properties of the structures in a very meaningful way. It is also clear from the
metallography results that the banded macrocolonies are leftover from the mill anneal in
both DA(0)
and RA(0)
in spite of the highly
refined microscopic
appearance
with
equiaxed primary alpha grains averaging about 20 1m in diameter.
The group consisting of samples FBA(0), AR(90), FBA(90) is closely separated in
the attenuation chart, with FBA(0) being only slightly higher than the other two. Samples
FBA(0) and FBA(90) are not distinguishable in terms of the optical microscopy between
the two orientations and, being beta annealed, as some of the texture has been removed
due to recrystallization. The ultrasonic properties of sample FBA are slightly directionally
dependent.
Sample
AR
has
strong
directional
52
dependencies
and
quite
differently
appearing microstructures between the two orientations of interest. Samples DA(90) and
RA(90) also had very similar velocity and obviously are alike in the sense that the data is
in both cases taken from the transverse orientation.
In general, the transverse orientations tend to be less attenuative than the axial
orientations. What is probably most intriguing about these lower attenuating samples,
including AR(90), is the fact that the highest scattering direction coincides with lower
attenuation.
For
example,
sample
AR(90)
significantly
has a
higher
degree
of
microstructural scatter (~12 dB) than AR(0). This is because in AR(90) the waves are
propagating perpendicular to the aligned colonies of the mill annealed microstructure.
Propagation of sound waved in this transverse direction is optimal for the generation of
backscattered energy and yet, AR(90) is one of the lower attenuative samples. This is
valuable
and
information
underscores
the
importance
of understanding
peculiarities
associated with titanium alloys. To further illustrate this point, Figure 3.13 shows two
different oscilloscope traces taken at random from the transverse and axial direction of
the mill annealed microstructure using identical data collection parameters (i.e., gain,
damping, energy, focal distance, etc.). This figure demonstrates that the attenuation is not
necessarily
related
to
the
microstructure
scatter
in
Ti-6Al-4V,
which
is unusual
considering the vast majority of materials have their attenuation primarily dictated by
scattering. The amplitude of incoherent microstructural scattering is also very important
from the standpoint of defect detection. Materials with higher scattering are generally
more
difficult in terms of detecting and characterizing tiny defects, as the flaws are
masked by the scattering. The problem of evaluating the as-received titanium alloy from
AR(90)
microstructure
scatter
AR(0)
frontwall
echoes
Figure 3.13.
backwall
echoes
Oscilloscope signal traces showing the difference in amplitude
microstructure scatter from the mill annealed samples.
the transverse direction is closely related to the billet NDE
of the
qualification issue. Large
titanium billets are also examined by probing from the transverse direction, hence the
qualification of billets is hindered by the presence of very strong microstructure scatter,
making it difficult to discriminate between benign scattering versus genuine defects.
3.4 Ultrasonic Grain Scattering
Polycrystalline materials consist of individual grains of the constituent material,
which
compactly
mold
together
forming
a structure.
The
grains
are
generally
a
distribution of shapes and sizes filling the space within the boundaries of the medium
with uniquely
titanium
oriented
crystalline
alloys, the grains
structures.
In inhomogeneous
often tend to combine
into large
materials
colony
like some
structures
with
preferential orientation existing at the macroscopic level. Hence it is possible to have a
54
material that is inhomogeneous
on more
than one dimensional
scale due to a wide
distribution of grain sizes or the presence of macroscopic grain colonies. The grains may
state or by recrystallization during materials
form by crystallization from the molten
processing. For different grain morphologies, a single grain may be composed of a single
crystal; it may have two or more phases breaking it up, or the structure may consist of
both single grains and heterogeneous grains, as is typically the case for Ti-6Al-4V. Each
individual grain can be assigned a set of axes corresponding to the principal crystal
directions of the major constituent phase. The crystal orientation is generally different
from grain to grain, except in samples containing a preferred orientation. The geometry of
the grains may be flattened, elongated, or basically spherical, which are termed equiaxed.
The simplest polycrystalline materials are equiaxed and homogeneous (narrow grain size
distribution),
with
inclusions.
Simple
and
no
preferred
microstructures
are
nicely
single
phase
or
orientation,
colonies,
voids,
for
ultrasonic
materials
suited
characterization techniques to determine such parameters as grain size, stress state, and
elastic moduli [31, 32, 33]. More complicated structures tend to be difficult to evaluate in
terms of deriving quantitative material characteristics from the ultrasonic response. This
difficulty is mainly due to multiple scattering and other interference effects.
Scattering measurements, both forward and backward, have been conducted on a
number of different titanium microstructures. This section is geared towards developing
an understanding of ultrasonic scattering from a complicated polycrystalline material. The
main goal for this section is to describe the scattering for the various titanium samples
and discuss the mechanisms responsible for originating the scatter. One of the approaches
55
used
to determine
material
characteristics
is to collect
data
from
different
sample
orientations. By measuring the scattering response from different directional orientations,
we may determine if the signals are consistent with the presence of structural alignment,
indicating mechanical anisotropy, based on the scattering coefficients. This section will
also feature the importance
of signal processing techniques which are crucial to the
interpretation of the data. The frequency dependence of scatter is another area where
useful information can be collected, but this is generally true for grain size measurement
in simple microstructures only. For most materials, grain scattering primarily influences
the attenuation. However for titanium alloys, this is not always the case due to factors
related to materials processing and the resulting complex microstructure and development
of texture. The anisotropy of titanium, especially in the highly textured mill annealed
condition, is an important factor that influences not only the mechanical load response,
but also the scattering and attenuation.
Figure 3.14 shows the narrow-band and broadband experimental configurations
for scattering measurements. For simple microstructures, the narrow-band experimental
approach is typically used to acquire information regarding the nominal dimensions of
scatterers in the material. This approach is taken to determine the frequency dependence
of attenuation and scattering [34], which can be used to assess grain size in single phase
materials with narrowly distributed equiaxed
however,
a
broadband
experimental
grain sizes. In complex
approach
is
usually
taken
as
microstructures
the
frequency
dependence does not provide much insight into characteristics of the nominal scatterer.
The broadband approach allows the qualitative determination of different microstructure
Zz
digital oscilloscope
computer
controlled
scanner
X
.
pre-amplifier
receiver
output
hd
2
9
t
J
2
9
C=—)
function generator
°
rf power amplifier
°
in
©
out
&
+55 dB
trigger output
sine wave burst
~ 15 cycles
Narrow-Band Experiment
computer
Zz
digital oscilloscope
controlled
transmitter / receiver
Ur
eee
sync
output
Ad
|!
$F-
Broadband Experiment
Figure 3.14
different experimental
of the two
Schematic
microstructure characterization using backscattering.
used
set-ups
for
types, but generally provides little in terms of quantitative material characteristics. Both
the narrow-band and broadband experiments involve measurements of the backscattered
signals.
The
main
difference
between
narrow
band
versus
configurations is the excitation source. For the narrow-band
broad band
experimental
experiment, a sinusoidal
burst of approximately 20 cycles of the desired harmonic frequency is amplified and used
to excite the transducer.
This results in the excitation of a narrow
range
of signal
frequencies. For the broadband experiment, a spike (impulse) excitation is used, which
results in the generation of a wide range of frequencies in the transmitted and received
signals.
57
Many problems can arise in experimentally determining the frequency dependence
of attenuation and
scattering in polycrystalline
metals.
For example,
every time the
frequency is changed, the amount of energy imparted to the material also changes, hence
gain-compensation should used in order to maintain consistent experimental conditions in
terms of the incident energy. Generally, ultrasonic transducers are broadband enough that
gain-compensation requirements are negligible so long as we use frequencies near the
of the spectrum.
middle
arising
problems
due
to
translates into a problem
gain compensation
The
frequency
where
is easily manageable
beam
dependent
divergence.
Beam
compared
to
divergence
frequency variations result in a different interaction
volume every time the frequency is changed. In other words, lower frequencies tend to
have higher beam divergence and increasing the frequency results in a more collimated
beam. Hence, every increase in frequency results in a smaller and smaller interaction
volume of material in which the waves interact with the medium. In practical terms what
this means is that experimentally it is difficult to observe the predicted power of four
frequency dependence
in the Rayleigh
scattering regime (power of two in stochastic
regime) unless sophisticated diffraction corrections are implemented. This correction is
not
so
difficult
for
simple
microstructures,
but
is quite
in the
more
in polycrystalline
alloys
challenging
complicated microstructures seen in titanium alloys [24, 26, 27].
An
important
technique
used
to evaluate
scattering
involves spatial averaging in either the time or frequency domain [35, 36]. We know, for
example, that from location to location on the samples the ultrasonic backscatter response
changes. These changes are demonstrated in Figure 3.15 showing backscattering signals,
58
using a focused transducer, from five randomly chosen locations on the AR(90) sample.
Each
has
position
a different backscatter
response
because
the
structure
within the
interaction volume changes. Backscatter signals are highly influenced by slight changes in
the experimental set-up. For example, by changing the incidence angle of the transducer
\
amplitude
fitter
time
b)
vonateoonnny
m
ot mrelprartan
bo
d)
>
oN A
>)
iN .
rectified rf signals a,b,c,d,e
10K rectified rf signals
transducer focused on surface
\——— transducer focused ~ 0.5" inside sample
‘\
Figure 3.15
Spatial averaging of backscattered longitudinal waves in the time domain,
using broadband excitation with a transducer having a center frequency of
approximately 8.5 MHz.
59
of a degree
a fraction
by only
results
in an entirely different backscattered
signal.
Likewise, a position change of approximately one grain diameter or one wavelength will
suffice to drastically change the appearance of the signal. These signal disturbances in Ti6Al-4V could be due to many factors, such as grain-to-grain elastic property variations, In
ultrasonic materials characterization, polycrystalline metals generate an incoherent and
highly divergent field that appears as "grain noise" in the detected ultrasonic signals. It is
important to realize that these signals do not necessarily correspond to any particular
physical feature inside the material structure, but rather, are generated due to interference
and multiple scatter effects of the sound waves interacting with a volume of material.
Multiple scattering is especially important for titanium alloys as they tend to scatter
ultrasonic energy on more than one dimensional scale consistent with the formation of
microscopic features (i.e. grains) as well as macroscopic features (i.e. grain colonies).
When
produces
a sound wave
its own
filled with tiny objects, each object
traverses a volume
array of scattered waves.
These
scattered waves reinforce in some
directions and interfere in others, and the wave incident on each scatterer is affected by
the presence of other scatterers. These interactions gives rise to coherent, incoherent, and
multiple
scattering.
The
coherently
scattered
signals
contribute
additively
to
the
amplitude of the transmitted wave. By subtraction of the coherent response, the remaining
energy is due to incoherent and multiple scatter. The incoherent scatter is apparent as low
amplitude signals (40 to 60 dB below the coherent echoes) occurring between the two
surface reflections from the front and back walls of the sample [37]. This incoherently
scattered energy can be easily measured with a receiver which has the appropriate signal
60
amplification and filtering capabilities.
For backscattering, the data collection scheme starts with the sample being placed
in the immersion tank on a tilt stage and then leveled to approximately
4/10
or+5 pm
flatness over about one square inch on the sample. The leveling is achieved by repeatedly
scanning
the transducer
over the sample
and adjusting the tilt stage
so that as the
transducer moves, there is no change in the water path length. If the water path varies as
the transducer scans, the oscilloscope signal trace correspondingly varies in time. Once
the sample is level, the time drift disappears. Sample leveling is extremely critical if
accurate results are to be acquired. Next, the nominal coherent transducer response must
be acquired by collecting a global spatially averaged signal. The transducer response is
to every signal trace in the spatially averaged scan and generally
the signal common
requires
performed
15
k to 20
k signal
samples
for adequate
convergence.
The
averaging
is
as the transducer is scanned over approximately a 1" x 1" region on the
sample. The transducer response must be re-acquired for each new material sample and
the
experimental
conditions
must
not
change
from
sample
averaging measurement would be extremely time consuming
to
sample.
The
spatial
and tedious without the
advent of computer controlled scanning and an on-line digitization capability provided by
a digitizing oscilloscope. With the transducer response acquired, the final data acquisition
steps can be performed, resulting in the collection of signals representative of the average
backscattered energy for each sample. A new spatial average is started, essentially
repeating the process used to acquire the coherent response, except and for each collected
rf signal, the coherent response is subtracted out on-line. What is remaining after the
response
coherent
incoherent
is
subtracted
at each
signals are then squared
and a
point
is the
incoherent
scattering.
These
of all of these signals
spatial summation
provides the input for the last step in the backscatter energy acquisition process. The last
step is to take the summed average of the squared incoherent signals and then take the
square root. The signal resulting from this last step is directly related to the backscatter
energy response of the sample as it is an rms average of all the material responses.
PART II: The New Forward Scattering Method
3.5 Ultrasonic Forward Scattering
Forward
scattering
measurements
provide
information
on
the
macroscopic
structural composition of the alloy. Unfortunately, this research area is relatively new in
terms of information documented in the literature. This is a little surprising, considering
the advantages this type of measurement
has over conventional backscattering.
First,
forward scattering requires a transmission experiment so the amplitude and divergence of
the scatter are both viable sources of information. Secondly, the data evaluation is open to
conventional
C-scanning, rather than the more
complicated
spatial averaging used in
backscattering. Finally, unlike backscattering, the time dependency of the scatter can be
mapped in forward experiments.
Figure 3.16 shows the schematic of the forward scattering experiment. A 0.25"
contact transducer (transmitter) is mechanically mounted to one surface of the sample. On
the opposite surface, a second transducer (receiver) is aligned at normal incidence to the
transmitter. This position becomes the center of the C-scan to map the amplitude and
divergence of the incoherent forward scatter. The transmitter launches a wave through the
material
receiver.
the
towards
This
coherent
wave
at the
is detected
receiver
and
immediately following its arrival begins the forward scattered energy, which is actually
the backscatter from the internally reflected coherent wave.
Peak-detecting electronic
gates are used to sample the forward scattered data, which forms the image in the Cscans. By raster scanning the receiver the scattered field can be mapped to assess the
influence of texture and microstructure on the amplitude and divergence of the scatter.
Unfortunately, this mapping procedure can not be used in backscattering experiments,
since the transmitter and receiver are one and the same.
.
.
scanning receiver
first coherent
transmission
sample / transmitter
reflection
th
7
incoherent
forwar
gate positions
scatter
<t |
reflected
—_
coherent waves
ot
Ci
a
FORWARD
SCATTER
mounted”
transmitter
Oscilloscope
controlled
scanner
transmitter / receiver
tt
syne outpu
water tank
Figure 3.16
Schematic of the forward scattering experiment.
63
Spatial
forward
and
directional
and backward
measurements
averaging
directions
the Ti-6Al-4V
been
have
samples.
The
conducted
amplitude
in the
modulated
incoherent scattering signals are specially suited to evaluate materials for microstructural
variations.
Monitoring
of the
orientation
dependence
and
spatial
variation
of the
amplitude modulation in scattered signals provides an indication of the shape of the
nominal material inhomogeniety.
For randomly oriented grain structures the scattering of
sound waves from material inhomogeneities generally occurs in all directions. Randomly
oriented materials with coarse grain structure generally have higher attenuation because
the anisotropy of individual grains scatter away the energy of the propagating waves
faster than in the equivalent fine grained material.
In Ti-6Al-4V, the material inhomogeneity can take on the form of colonies or
clusters of similarly oriented grains, which often appear as a result of wrought processing
operations like forging. Figure 3.17 shows an example of a backscattered signal with and
without high-gain amplification, including the front and back surface echoes. The normal
incident
spatially averaged
backscatter
energy
data for the titanium
samples
in the
transverse and axial orientations is shown in Figure 3.18 (see also Table 3.6 — Ultrasonic
Backscattering, Appendix A). These signals were acquired using broadband excitation
with a 10 MHz
ultrasonic transducer focused at the surface of each sample.
Surface
focusing allows for the evaluation of grain scattering without the influence of focusing
affects inside the material, since the propagating sound waves continuously diverge from
the entry surface. The original cylindrically shaped samples were prepared such that data
could be collected from both the axial and transverse directions through flat and parallel
64
I
Broadband excitation
10 MHz longitudinal wave
amplifier: + 19 dB
rn
|
backwall echo
frontwall echo
Broadband excitation
10 MHz longitudinal wave
amplifier: + 59 dB
backscatter
Figure 3.17
Backscattered signals from sample AR(90) showing the effect
increasing the amplification by 40 dB to reveal the backscatter signals.
of
entry surfaces to allow direct comparison of scattering from the samples. The data was
collected in both orientations using the same transducer, under the same scan parameters
and hardware scan settings.
At normal incidence, clearly the highest backscattering occurs in the transverse
orientation of the mill annealed sample, AR(90).
highly aligned nature
of the inhomogeneity
This is not too surprising given the
in the axial direction of the sample.
In
contrast, the axial direction of the mill annealed sample (AR(0)) demonstrates the lowest
backscattering of longitudinal waves, along with a group of other samples which all have
65
Figure 3.18
i
°
‘p
amplitude (volts)
o
ro
o
a
—™ AR(90)
— DA(90)
—*— RA(90)
—®— FBA(90)
—+- CBA(90)
—— CBA(0)
1+ FBA(0)
—t— AR(0)
—— DA(0)
Spatially averaged backscattered signals from the axial and transverse
orientations, using flat entry surfaces with beam focused on front surface.
nearly identical backscattering characteristics. With the exception of the beta annealed
RA(90) also clearly show evidence of the highly aligned macroscopic features similar to
sample AR(90).
Sample DA(90)
appears to have backscatter coefficients only slightly
samples (i.e., FBA(0) and CBA(0)) all of the samples, in the axial direction, had virtually
the same negligible backscatter response. Hence, the mill annealed sample has both the
lowest and highest scattering conditions depending
on which
orientation the data is
collected from. The scattering is highly anisotropic, with the optimal generation of scatter
occurring the transverse orientation.
Samples
DA(90)
and RA(90)
also clearly show
evidence of the highly aligned macroscopic features similar to sample AR(90). Sample
DA(90) appears to have backscatter coefficients only slightly stronger than RA(90). The
metallographic results show the DA and RA heat treatments have more equiaxed grains
as compared to the heavily elongated grains and colonies of the mill annealed condition.
However, both DA and RA
samples clearly retain a significant portion of the original
texture, and there clearly exists evidence of banding on a macroscopic scale.
direction, the lowest backscatter energy
For data collection in the transverse
comes
from sample
CBA(90)
with sample FBA(90)
These
beta annealed
samples
are more
structures
as
indicated
by
the
scattering only slightly stronger.
closely
representative
and
attenuation
velocity
of randomly
results.
oriented
Moreover,
the
backscattering data shows that in sample CBA the response is not strongly influenced by
orientation affects. The beta annealing heat treatments have significantly transformed the
microstructures
due
to the
dwell
time
above
the beta transus
resulting
in a more
thoroughly recrystallized structure where most of the original texture is overwritten by the
more random orientation of the new grain structure. The reason for the stronger scatter in
FBA(90) is due to the presence of retained texture. Hence, sample FBA is simple more
anisotropic than sample CBA.
In the axial orientation, the two beta annealed samples clearly backscatter the most, with sample CBA(0) the strongest, followed by sample FBA(0). This scattering is
consistent with the idea that these samples
generate more
randomly
directed scatter.
Samples AR(0), DA(0), and RA(0) all backscatter energy the same low intensity manner,
with virtually 100%
overlap of the signals in the axial orientation.
67
Surprisingly, this
minimal backscattering characteristic is true for the entire mill annealed sample set (i.e.,
samples AR(0),
except AR(90).
AR(30), AR(45), and AR(60))
Overall however, this
should not be taken to imply that these samples scatter any less than AR(90). Rather, only
the
backscattering
energy
180°
(directed
from
the
wave
propagation)
shows
this
minimum characteristic. At the other than orthogonal orientations, the scatter is simply
directed away from the backward direction.
Figure 3.19 shows results from the mill annealed samples in the axial and radial
directions over a 2" x 2" area. For the axial direction the forward
scatter is highly
divergent and has a characteristically conic symmetric for all directions in the plane of the
scan.
Moreover,
images.
this divergence
can be clearly observed
In contrast, the mill annealed
sample
C-scan
in the time-lapsed
in the radial direction has a distinctly
different pattern to the forward scatter. In this case, the initial divergence (i.e. gate 1) is
much more highly concentrated and the image sequence shows a dramatically different
scattering pattern as compared to the axial case. The scatter divergence emerges as a fan
beam,
rather than a cone
beam
as observed
in the axial case.
The
for the
reason
contraction in the axial direction (indicated by the white arrow in gate 2) is that the
elongated grains and grain colonies tend to act as a diffraction grating. Hence, for uniform
and
equiaxed
grains,
as
viewed
from
the
axial
direction,
the
scatter
diverges
symmetrically as opposed to the elongated grain colonies, as viewed from the radial
direction, which causes the divergence to be much more one-dimensional. In terms of
amplitude, it is of great significance to note that the direction of wave propagation is of
little or no consequence as the distribution of amplitude values is virtually the same
68
Sample AR(0)
(axial direction
is out of page)
Sample AR(90)
(axial direction
indicated by
white arrow)
gate 1
Figure 3.19
gate 2
gate 3
Time-lapsed C-scan images mapping the amplitude and divergence of
forward scattering in mill annealed Ti-6Al-4V,
showing clearly
distinguishable differences between the axial versus radial directions.
between the axial and radial directions, only the shape of the divergence pattern changes.
Figure 3.20 shows the amplitude and divergence of forward scatter for the coarse
beta annealed (CBA) sample over a 2" x 2" area. This sample was held for approximately
six hours above the beta transus (~1000 °C) providing ample time for the material to fully
recrystallize, thereby removing virtually all of the initially present texture. As such, this
sample
has
large
randomly
oriented
grain
colonies
and
upon
viewing
the
metallographically prepared surfaces under magnification, it is not possible to distinguish
the axial direction from the radial one. The disappearance of the macroscopic anisotropy
and random
grain orientation is also apparent in the forward scattering results. These
results demonstrate
the lack of distinguishing
features between
the axial and radial
directions for a random structure, as opposed to the highly anisotropic mill annealed
69
Sample CBA(0)
Sample CBA(90)
gate 2
Figure 3.20
gate 3
Time-lapsed C-scan images mapping the amplitude and divergence of
forward
scattering in coarse beta annealed Ti-6Al-4V,
showing
indistinguishable differences between the axial versus radial directions.
structure. Additionally, the amplitude distributions are virtually the same between axial
and radial directions of the CBA samples, similar to that of the AR samples.
Based
on the backward
and forward
scattering results it is possible to more
thoroughly describe the scattering cross-sections. This is particularly important in terms
of understanding
attenuation
results
the
in
attenuation
results
mill
annealed
the
for the
mill
structure
annealed
are
unusual
microstructure.
The
the
high
due
to
backscattering in the radial direction, but low attenuation than that of the axial direction.
By constructing the notional ultrasonic cross-sections, we can better understand why this
unusual attenuation behavior exists. Figure 3.21 shows these notional scattering crosssections for the axial and radial directions. The total scattering for the axial direction in
the mill annealed structure is larger than that of the radial direction. Therefore, despite the
70
stronger backscattering in the radial direction, the axial direction has significantly higher
attenuation.
WIA
shear
setetetetetete longitudinal
back
random structure
Figure 3.21
Notional scattering cross-sections in the axial and radial directions for
both anisotropic and random structures, which were generated based on
the backward and forward scattering results.
Chapter 3, References:
1. Materials Property Handbook on Titanium Alloys, eds., R. Boyer, G. Welsch, and E.
W. Collings, (ASM International, 1994) pp. 483-500.
2. Titanium, A technical guide, ed., M. J. Donachie, Jr., (ASM International, 1988) p. 44.
3. S. L. Semiatin et. al., “Hot working of titanium alloys — An overview,” Advances in
Titanium Alloy Processing, eds., 1. Weiss, et. al., Proceedings from the 125" TMS
Meeting held in Anaheim, California, 5-8 February, 1996, (TMS, 1997) pp. 4-26.
4. Metals Handbook (ASM, 1973) pp. 45-47.
5. D. Eylon, "Faceted fracture in beta annealed titanium alloys," Met. Trans. A Vol. 10,
pp. 311-317 (1979).
6. D. Eylon and J. A. Hall, “Fatigue behavior of beta processed titanium alloy IMI 685,”
Met. Trans A, Vol. 8, pp. 981-990 (1977).
7. S. Hirsekorn, “The Dependence of Ultrasonic Propagation in Textured Polycrystals,”
Journal of the Acoustical Society of America, T9(5), (1986).
8. C. M. Sayers, “Angular Dependent Ultrasonic Wave Velocities in Aggregates of
Hexagonal Crystals,” Ultrasonics, Vol. 24, pp. 289-291 (1986).
9. C. M. Sayers, “Ultrasonic Velocities in Anisotropic Polycrystalline Aggregates,”
Journal of Applied Physics D, 15, p. 2157, (1982).
10. E. S. Fischer and C. J. Renken, "Single-Crystal Elastic Moduli and HCP - BCC
Transformation in Ti, Zr, and HF," Physical Review, 135(2A), pp. A482-A494, (1964).
11. A. Migliori, et. al., “Resonant Ultrasonic Spectroscopic Techniques for Measurement
of the Elastic Moduli of Solids,” Physica B, Vol. 183, pp. 1-24, (1993).
12. L. Adler, “Ultrasonic Spectral Analysis for Nondestructive Evaluation,” (Plenum
Press, 1981), p. 21-25.
13. A. F. Brown, “Materials Testing by Ultrasonic Spectroscopy,” Ultrasonics,
September, pp. 202-210, (1973).
14. W. M. Visscher, et. al., “On the Normal Modes of Free Vibration on Inhomogeneous
and Anisotropic Elastic Objects,” Journal of the Acoustical Society of America, 90(4),
Part 1, pp..2154-2162, (1991).
15. ASTM Standard E 494-75, “Standard Practice for Measuring Ultrasonic Velocity in
Materials,” ASTM (1985).
16. E. P. Papadakis, “ The Measurement of Ultrasonic Velocity,” Ultrasonic
Measurement Methods, Physical Acoustics , Vol. XIX, Eds. R. N. Thurston and A. D.
Pierce, (Academic Press, New York, 1990) p. 91.
17. F. Nye, "Physical Properties of Crystals - Their Representation by Tensors and
Matrices,” J. (Clarendon Press, Oxford, 1985) p. 143.
18. In Proceedings on Advances in the Science and Technology of Titanium Alloy
Processing, eds., I. Weiss et. al., (TMS Publication, Aneheim California, 1996) p. 16.
19. W. F. Hosford, "The Mechanics of Crystals and Textured Polycrystals," (Oxford
University Press, New York, 1993), p. 16.
20. P. C. Waterman and L. J. Teuntonico, "Ultrasonic Double Refraction in Single
Crystals," Journal of Applied Physics, 28(2), pp. 266-270 (1957).
21. R. Truell, L. J. Teuntonico, and P. W. Levy, "Detection of Directional Neutron
Damage in Silicon by Means of Ultrasonic Double Refraction Measurements," Physical
Review, 105(6), pp. 1723-1729 (1957).
,
72
22. Ultrasonic Testing of Materials, eds.,
J. Krautkramer and H. Krautkramer, (Springer-
Verlag, NY, 1990), pp. 108-115.
23. ASTM Standard E 664-93, “Measurement of the Apparent Attenuation of
Longitudinal Ultrasonic Waves by Immersion Method,” ASTM (1989).
24. H. Seki, A. Granato, R. Truell, "Diffraction Effects in the Ultrasonic Field of a Piston
Source and Their Importance in the Accurate Measurement of Attenuation," , Journal of
the Acoustical Society of America, 28(2), pp. 230-238, (1956).
25. K. Tjadens, "Longitudinal Wave Ultrasonic Absorption in Aluminum Alloys,"
Acoustica, Vol. 11, pp. 127-336, (1961).
26. E. P. Papadakis, "Diffraction of Ultrasound Radiating into an Elastically Anisotropic
Medium," Journal of the Acoustical Society of America, 36(3), pp. 414-422, (1964).
27. E. P. Papadakis, "Ultrasonic Diffraction Loss and Phase Change in Anisotropic
Materials," Journal of the Acoustical Society of America, 40(7) pp. 836-876, (1966).
28. M. P. Blodgett, P. B. Nagy, “Anisotropic Grain Noise in Eddy Current Inspection of
Noncubic Polycrystalline Metals,” Appl. Phys. Let., 72(9), pp. 1045-1047 (1998).
29. M. P. Blodgett, W. Hassan, P. B. Nagy, “Theoretical and Experimental Investigations
of Lateral Resolution in Eddy Current Imaging,” Materials Evaluation, May, 2000.
30. Microscopy Handbooks 12, A. Briggs, “An introduction to scanning acoustic
microscopy,” (Oxford University Press, 1985) pp. 1-15.
31. E. P. Papadakis, “Ultrasonic Attenuation due to Grain Scattering in Polycrystalline
Metals,” Journal of the Acoustical Society of America, Vol. 37, p. 711 (1965).
32. J. Szilard, and G. Scruton, “Revealing the Grain Structure of Metals by Ultrasonics,”
Ultrasonics, May, p. 114, (1973).
33. J. H. Rose, “Ultrasonic Backscatter from Microstructure,” in Review of Progress in
ONDE, Vol. 11, eds D. O. Thompson, D. E. Chimenti (Plenum, New York 1992) p. 1677.
34. R. L. Smith, “The Effect of Grain Size Distribution on the Frequency Dependence of
the Ultrasonic Attenuation in Polycrystalline Materials,” Ultrasonics, September, p. 211,
(1982).
35. J. H. Rose, “Theory of Ultrasonic Backscatter From Multiphase Polycrystalline
Solids,” in Review of Progress in ONDE, Vol. 12, eds D. O. Thompson, D. E. Chimenti
(Plenum, New York 1993) p. 1719.
36. K. Goebels, “Structure Analysis by Scattered Ultrasonic Radiation,” Research
Techniques in Nondestructive Testing, Vol. 4, edited by R. S. Sharpe, (Academic Press,
New York, 1980), Vol. IV, pp. 87-157.
37. F. J. Margetan, R. B. Thompson, and I. Yalda-Mooshabad “Modeling Ultrasonic
Microstructural Noise in Titanium Alloys,” Review of Progress in Quantitative
Nondestructive Evaluation, Vol. 12B, eds. D. O. Thompson and D. E. Chimenti,
(Plenum, New York, 1993), p. 1753.
73
CHAPTER IV
SUBSURFACE MICROSTRUCTURE ANOMALY DETECTION IN BULK
TITANIUM ALLOY USING ULTRASONIC BACKSCATTERING
4.1
Introduction
In this chapter, the main objective is to demonstrate the capabilities of ultrasonic
inspection for the purpose of identifying and characterizing anomalous conditions in 2.5"
diameter cylindrical bar samples. Implanted microstructure anomalies were diffusion
bonded in the axial and radial orientations to simulate processing remnant cast structure
at the center of the forged bar. Primarily, conventional B-scanning [1, 2, 3] is used as a
foundation
for this chapter.
B-scanning
of the backscattered
ultrasound
a
provides
realistic representation of the statistically distributed ultrasonic data, unlike C-scanning,
which generally provides only the peak-detected signals, thereby giving a distribution of
data skewed towards the large set values. B-scans were collected from a number
of
different regions on each of the diffusion bonded samples to examine the extent to which
the anomalous microstructure inserts could be detected and characterized. This approach
was also taken in order to make
an assessment of the optimal scan configuration to
qualify the mechanical suitability forged Ti-6Al-4V bar stock for further processing.
74
Other
backscattering,
scan
configurations
including
forward
were
investigated
scattering
and
besides
side
conventional
scattering,
to
ultrasonic
determine
the
effectiveness of these nonconventional approaches to identify anomalous microstructure.
Unfortunately, the results clearly demonstrate that there is no advantage to using forward
or side scattering as an inspection tool over ultrasonic backscattering. In fact, forward and
side scattering are complicated by the requirement for a pitch-catch (i.e. two transducer)
scan configuration, which is difficult to keep aligned. Moreover, the main goal of the
inspection is to determine the origin of anomalous indications inside of the alloy, and the
side scattering approach is lacking in this regard due to the uncertainty in the source of
the
anomalous
backscattering,
signals.
Forward
but is complicated
scattering
provides
by alignment
essentially
constraints
the
same
data
as
and appears to offer no
definitive advance in detection capabilities. Hence, no significant effort was made to
develop new NDE techniques for this section, due to the need for more comprehensive
data based on existing ultrasonic (backscattering) approaches.
The forward scattering
results revealed no clear advantage over backscattering, but they could be significantly
improved through the use of more sophisticated articulation the transmitter and receiver
to
map
the
backscattering,
forward
which
scatter
as
a function
has been thoroughly
of
incidence
researched
angle.
throughout
Unlike
ultrasonic
the last couple
of
decades both experimentally and theoretically [4, 5, 6, 7, 8, 9, 10, 11], forward scattering
remains an open for further research. Useful information could be further acquired
through the collection of data at multiple frequencies to provide more quantitative detail
into the physical aspects of material inhomogeneities. Unfortunately, these developments
would require an investment of resources that goes beyond the scope of the current
75
research. However, the samples developed for this dissertation and the research area will
remain an open opportunity for future efforts.
4.2 Sample Development
sets of samples have been developed
Two
characterization of subsurface
to investigate the detection and
in the titanium bar stock. Additionally,
anomalies
a
preliminary sample set was constructed of 1" diameter Ti-6A1-4V bar, to assess whether
or not the diffusion bonding parameters (i.e., 950 °C, 15 ksi, 4 hours) were adequate to
join the metal. These 1" diameter samples were axially bonded and were available for
destructive
metallographic
examination.
After
the
initial
trial
bond
samples
were
generated, two additional sample sets were constructed so that anomalous microstructure
detection could be investigated from both the axial and radial directions.
For data collection in the axial direction, sample coupons of the 2.5" diameter bar
stock were rough cut into disks of approximately 1" thick, as shown in Figure 4.1. Five
different samples
CBA/CBA,
of this orientation were to be made,
AR/FBA/AR,
and AR/CBA/AR
including AR/AR,
AR/CBA,
(see Table 4.1 — Axially Bonded Samples,
Appendix A).
Following the preparation of the roughly cut disks, some of the disks were heat
treated to arrive at the fine and coarse beta annealed condition. Of course, heat treating in
air results in a thick scale on the samples which is easily removed, but worse, the exterior
76
a
>.
—“—
a
bonded sample
axial bond coupons
Figure 4.1
Schematic for the axial diffusion bonded Ti-6A1-4V samples.
material of the annealed samples becomes case hardened due to the chemical reaction that
takes place between the room air and the Ti alloy. Hence, prior to joining, this alpha case
must be removed if the coupons are to be appropriately joined. The case removal was
achieved by a combination grinding and lathe machine work. The axially bonded samples
were further prepared by generating flat mating surfaces on the lathe and then lapping the
surfaces samples to approximately 1 1m rms roughness. The next step is to thoroughly
clean the samples to remove
any traces of contamination, like fingerprints, grease or
grime. Finally, the samples were electron beam welded and placed into the hot isostatic
press
(HIP)
where
the combination
of time,
temperature,
and pressure
results
in a
diffusion bond. The actual samples are shown in Figure 4.2.
For data collection in the radial direction, which is the important direction form an
industrial billet qualification standpoint, an additional sample set was constructed. These
samples have three different diameters of inserts with three different microstructures,
77
which are diffusion bonded to the interiors of the bored cylinders. Figure 4.3 shows the
schematic for the radially bonded samples. Insert sizes of 0.2", 0.4", and 0.9" were
AR/AR
Figure 4.2
CBA/CBA
AR/CBA
The five axially diffusion bonded samples.
Hp
po
‘ : leckedesh
—-
Figure 4.3
AR/CBA/AR
AR/FBA/AR
:
ptt
L
: :
——_—
radial bonds
; 4 wd: : |
radial inserts:
0.2",
0.4",
Schematic for the radial diffusion bonded Ti-6A1-4V samples.
78
5
0.9"
generated with the AR, FBA,
and CBA
microstructures. These inserts were first heat
treated, then machined on the lathe to snugly fit into the appropriate bore of the 2.5"
diameter cylinders. Following the final machining and cleaning steps, the inserts were
press fitted into the cylinders
and then electron beam
welded
on the ends prior to
placement in the HIP chamber. The actual samples are shown in Figure 4.4 (see Table 4.2
— Radially Bonded
Samples, Appendix A). Figure 4.5 shows a metallographic results
from the destructive evaluation of the 1" diameter AR/AR
Clearly,
this
sample
has
some
defects
bondline
which
diffusion bonded
may
have ‘been
sample.
due
to
contamination, but as a result of finding defects the bonding parameters were modified to:
950 °C, 30 ksi, 4 hours (i. e. the HIP pressure was doubled) to ensure better consolidation
across the interface. The HIP process clearly softens the elongated grains and grain
Figure 4.4
The nine radially diffusion bonded samples.
79
Cae
original
axial direction
Figure 4.5
Metallographic results comparing the original as-received microstructure
with that of the HIP'ed sample. Both images were taken at 200x from the
same orientation of the bar.
colonies present in the original mill annealed structure, resulting in more equiaxed grains
with a distribution of grain sizes centered at approximately 30 zm and no evidence of
macrocolonies. Figure 4.6 shows a comparison of the metallographic results from a alpha
case contaminated bondline versus a clean bond interface. In the case of the contaminated
bond, the alpha case hardened surface that results from heat treating in lab air was not
thoroughly ground off prior to diffusion bonding. This sample was the motivation for
extensive grinding of the beta heat treated sample surfaces that followed to assure there
was no contamination at the interfaces.
4.3 Detection of Anomalous Microstructure
The internal structure of the diffusion bonded samples was examined using a 10
MHz,
3 in. focal length ultrasonic transducer in the pulse-echo mode. For the five
80
samples bonded in the axial configuration, B-scans were taken from several locations
through the flat entry surface on the ends of each of the samples, which were ground and
polished
for flat and parallel opposing
surfaces.
In contrast, for the radially bonded
samples, the B-scans were taken through the curved entry surface on the sides of each of
the samples. B-scanning is a means of visualizing the actual ultrasonic response of the
structure in the form of an image.
The B-scan is taken as a line of data with each successive interval representing a
separate, but adjacent, data point. Figure 4.7 shows a B-scan image as it would look
without grayscale coding. The scan is simply a series of closely separated ultrasonic
signals, which are all stacked on top of one another to represent a cross-section of the
material inspected. One of the disadvantages of the B-scanning approach is that it is
possible to generate huge volumes of data that can quickly overwhelm the computational
resource needed to process the data. To get around this problem, only a few randomly
chosen
data files will be shown
from
each
of the samples.
However,
if one
were
interested in reconstructing the scattering response over a significant volume of material,
extensive computer memory and processing power would be essential. Generally, B-scans
are encoded according to a color scale as depicted in Figure 4.8. Also shown is a line plot
taken directly out of the B-scan to demonstrate the relationship between the B-scan
image, the ultrasonic waveform, and the color scale.
In all of the B-scans collected, the gain on the receiver was turned up to + 52 dB
in order to observe the grain scattering signals. The high gain saturates the front and back
81
Peers:
alpha case contaminated bond
axial direction
Metallographic results from dissimilar microstructures samples (i.e.
AR/CBA) comparing an alpha contaminated bond with a clean bond. Both
images were taken at 100x from the same orientation of the bar.
Figure 4.6
f
rn
AS ad
oe
length (inches)
.
Vy Y DLS
ne
15)
4
\ fer
sn
as.
Sven
NY [ON
Np ot.
i
"y
5
PPP
avy
VPN
F
PAL NINAi
pnt
.
INI POOL
*
.
a NVA
Nees LIBS
LANL LNGect
“ny
AVN,
PAA Noe,
i
tN f Mi
om
LOLA
o
NL
Figure 4.7
A
vd. Sy YC
backscattered
Smt
LALO
ultrasonic
B-scan
coding.
82
n
EPALIVS
LLG, PRL
/
ny
NES POD
PS SIONON
DLN
NN.
wa
ae
tinct a oN
pl
Ot
LBL RL ONLEL VDL
INS SAAS AN LN
shown
.
NG vars SPOON SOI NON
without
OP
IIL
I \ et
NLSO J
et ff
OLIN
t
LE Loren
the
r
advent
NS
of color
B-scan
85.61
.
line plot
backscattered grain noise
8917
1123
Pixels
0
color scale
just beyond the
center
front wall echo
~10ps
Figure 4.8
An illustration showing the relationship between the B-scan image, the
backscattered waveform, and the associated color scale.
wall echoes on the analog-to-digital converter used to collect the data, so there is really no
useful information from either of these locations. As such, the B-scan gates from which
the rf data is collected begin slightly after the saturated front wall echo and extend
approximately 11 1s into the material, which corresponds to more than half way through
the thickness
of the 2.5" diameter
samples,
as shown
in Figure
4.9. Based
on this
approach, the backscattered rf signals can be used as a means of differentiating between
the different microstructures. The B-scans are essential to the process of discriminating
between the different microstructures because the statistical distributions of the amplitude
data between the AR, FBA, and CBA microstructures overlap extensively. The primary
83
feature that allows us to detect the anomalous microstructure is the correlation length [1,
2], which is a measure of the degree of association in a series of signals and changes
between the three microstructures chosen to manufacture the bonded samples. A B-scan
represents such a series. The autocorrelation of a two dimensional image (the B-scan)
provides a measure of length over which the signals in a series are closely matched in
terms of the amplitudes and phases of the scattering centers.
In the axially bonded sample set, only one scan configuration was used, as shown
in Figure 4.10.a. In this case, the samples were scanned through the flat entry surfaces on
either end of the samples. In the radially bonded sample set, two different configurations
were used to investigate the optimal scan plan for detecting anomalous microstructure, as
a+
1.5+
|
interface signal from
it
-0.5+
-1+
(+ 52 dB gain)
jil
ost
0 {|
0.9" CBA implant
!
ihWa HN Mh. bali A Ah at
FOV RUN TORR
| all 4"I
10
TTT
UCTYYIE
12
ets
time (us)
ere
14
YTetene ie
saidy
16
rediarerrdiid
18
Wz
B -scan gate
20
h iN
[b,
a
back wall echc
TOT
front wall echo
2
Figure 4.9
An ultrasonic waveform which was collected through the curved surface
on the side of the radially bonded AR/CBA/0.9" sample.
84
=
scan line
=
~~
axial scan
NTT
To
eee
LA
radial
scan
axially bonded sample
radially bonded sample
(a)
(b)
Figure 4.10
The B-scan
samples.
configurations
used
for the two
sets of diffusion bonded
shown in Figure 4.10.b. In this case B-scans were collected through the curved entry
surface in both the axial and radial directions. For both sample sets (axially and radially
bonded) data was collected at two different focal lengths such that the effect of focusing
on the surface versus focusing below the surface could be assessed.
4.4 Results and Discussion
Since the receiver was set to nearly full gain, it is important to remove from each
B-scan the part of the signal associated with ringing of the transducer. To do this, the first
step is to collect a spatial average in the time domain using approximately 10 k signals.
85
This coherent signal is then subtracted out of each B-scan such that all that remains is the
backscatter response of the material. Figure 4.11 shows an example of a raw data B-scan
compared with one that has the reference signal subtracted out. This reference subtraction
is a simple processing routine which can be performed in any spreadsheet software, but is
necessary in order that the material response data is not confused with the response of the
ultrasonic transducer. The remaining images throughout this section will all have had this
simple processing
step performed
on the data. Additionally,
the B-scan
image
data
histograms have been equalized to provide the optimal contrast. The remainder of this
section is broken up into two different parts, which address the axially bonded samples
and the radially bonded samples.
4.4.1
Axially Bonded Samples
All of these samples were bonded with adjacently cut coupons in a manner that
maintained the original alignment to minimize the evolution of a bondline interface and
any associated ultrasonic reflection. Figures 4.12(a) shows the axially bonded AR/AR
sample, Figure 4.12(b) and 4.12(c) shows the AR/CBA sample and Figure 4.12(d) shows
the CBA/CBA sample. The AR/CBA and CBA/AR samples are the same, except flipped
over to observe differences in scattering due to the initial microstructure encountered
first. In the case of sample CBA/AR, there is no evident bondline indication from the Bscans. Moreover, there is no indication of a change in character of the scattering between
the two different microstructures in CBA/AR, unlike the case when the sound enters the
AR microstructure first in AR/CBA. Each of the surface focused, 10 MHz B-scans from
Fig. 4.12 are 11.3 ps in length, which corresponds to a one way path length of
86
i
;
P;
Be
it
+&
>é
a
ce
go
3
477.59
spatially averaged
reference
73.82
0
Pixels
509
raw - reference
cuss
Figure 4.11
=.
An illustration showing how the transducer reference signal is removed
from the B-scan, leaving primarily only the material response.
approximately 1.35". In each case, the B-scans started at approximately 1.5 ps into the
sample to avoid the meaningless saturated front wall echo signals and extended through
the thickness of the sample such that the bond interface is approximately in the center of
each of the scans. The bond interfaces in Fig. 4.12 are indicated by discontinuities in the
B-scan data, which are consistent with the predicted arrival time from the reflections.
From the backscattering results of Chapter 3, it became clear that the axial direction of
the mill annealed structure yields the least in terms of backscattered ultrasonic energy.
87
The diffusion bonded AR/AR sample is true to form in this regard, even though the HIP
processing changes the microstructure from one of elongated grains and grain colonies to
Figure 4.12
~ AR/CBA
CBA/AR
(b)
(c)
(d)
Enhanced ultrasonic B-scans using equalized histograms from the axially
bonded sample series. AR/AR shows a clearly evident bondline signal.
AR/BA clearly demonstrates the typical change in the scattering
characteristics between two different microstructures. No indications of
any kind are present in BA/AR. BA/BA has a strong bondline signal and
clearly did not properly form a good diffusion bond.
88
one with essentially equiaxed grains and no apparent grain colonies. While a bondline
signal is clearly evident in the AR/AR sample, the structure noise on either side of the
interface has essentially the same character, both spatially and temporally. In contrast, the
axially bonded AR/CBA sample has a distinct change in the character of the signal noise
before and after the bond interface. This change in the character of the signal noise is not
represented by the statistical distribution of the amplitude data, but by the correlation
length
of the ultrasonic
backscatter.
A
of the histograms
comparison
nearly
shows
completely overlapping signal distributions between the two distinctly different regions of
the AR/CBA
sample. However, a comparison of the correlation lengths of the same
regions demonstrates a dramatic change, which is representative of the change in the
degree of randomness (of the crystallographic grain structure) between the two scattering
fields originating from the two different microstructures.
Additional B-scan images were collected from the AR/AR
subsurface,
rather
defocusing
provides
B-scan
surface
focusing.
These
significantly
stronger
ultrasonic
than,
and AR/CBA
with
showed
that
the
signals,
thereby
images
backscattering
allowing for easier observation of signal variations from the joined samples. However,
this increased scatter intensity is not necessarily helpful in terms of defect detection
because both the scattering and the defect indications, from imperfections at the bond
interfaces, are equally intensified, without a net improvement to the signal to noise ratio.
Image
enhancements,
performed
to these
like histogram
images
to bring
equalization
and
out more
detail
microstructures.
89
edge
detection,
in signal
can be easily
differences
between
The
images
from
Figure
show
4.13
B-scans
from
the
AR/FBA/AR
and
Figure 4.13
Hci
Ver.
ica ae
at
AR/CBA/AR sandwich bonded samples. Both images cover approximately 19 us of
Enhanced ultrasonic B-scans from the AR / X/ AR sample series. Both the
FBA and CBA implant images show indistinguishable characteristics, with
clear indications of scattering changes between the joined microstructures.
90
data in the interior of each sample, which corresponds to a one way path of approximately
2.4", allowing both bond interfaces of the sandwich samples to be captured. The initial
microstructure transitions are clearly visible, based on the scattering distinction between
the AR and FBA/CBA bond interface. However, the second transitions (back to the AR
microstructure) are undetectable, similar to the problem observed with the CBA/AR
sample. The scattering characteristics do not reverse back to those typified by the AR
structure upon encountering the second bond interface. Hence, there is really no way of
distinguishing the CBA
structure from the AR
or FBA
structure after the second
interfaces. While the signal differences are clearly distinct before and after the first bond
interfaces, there is no distinction ‘at the second
These
interface.
interface
secondary
reflections do not stand above the background level of the scattering, but are partially
visible due of the coherency of the ultrasonic reflection from the interface. This signal
coherency
at the
interface
second
the
make
bonds
marginally
but
visible,
without
knowing in advance that the structure is bonded, the detection would be quite difficult if
not impossible. It is also surprising that there is very little distinction between the
scattering fields from the FBA
sample
versus the CBA
sample.
Without
using any
sophisticated signal processing, one would have great difficulty in differentiating between
the
FBA
and
examination,
CBA
there
structures
are
subtle
sandwiched FBA and CBA
based
the
on
characteristic
B-scan
differences
data.
scan
between
data.
the
Upon
closer
axially
bonded
inserts. However, these distinctions are only realized upon
viewing the scattering correlation lengths between the two data sets. Basically, the CBA
correlation length was shown to be slightly longer than that of the FBA sample, which is
probably due to the larger prior beta grain size in the CBA sample insert.
91
4.4.2 Radially Bonded Samples
In this section, 10 MHz B-scan results are presented from data collected through
the curved surface on the sides of the 2.5" diameter radially bonded
samples.
Three
different sets of samples were generated, which have different sizes of diffusion bonded
inserts (i. e., 0.9", 0.4", and 0.2") and different microstructures (i.e., AR, FBA, and CBA).
B-scans were taken from both the axial and radial configurations with surface focusing.
Figure 4.14 shows axial B-scans from the three AR/AR/X samples, where X is 0.9", 0.5",
or 0.2". In these images, the bond interfaces for the 0.9" and 0.4" inserts are marginally
visible in the backscattered data as indicated by arrowheads. However, for the 0.2" insert,
there is no evidence of any kind of indication. Moreover, beyond the indications from the
interfaces, there is no evidence of a change in the character of the scattering results.
Hence, we can conclude from this data that these samples are all well bonded and there
exists no overwhelming indication of any bonding problems.
In Figure 4.15 the same sample (i.e. sample AR/AR/X) was scanned, except this
time in the radial direction, using surface focusing at 10 MHz. In this case, the bondlines
are clearly detectable for the 0.9' and 0.5" inserts, but again there is no clear indication of
a bondline reflection or the presence on an anomaly for the 0.2" insert. Somewhat
surprisingly, defocusing fails to clarify the presence of bondline signals. The defocusing
appears
to brings
out stronger
signals, but these
same
signals mask
over the bond
reflections that were evident with surface focusing. Hence, radial scanning with surface
focusing appears to highlight bond imperfections, but with internal focusing the detection
of bondlines is sub-optimal.
92
Bey.
~AR/AR/0 Q/a
Figure 4.14
Basst Pcesecis
/ AR/0.5/ a
AR/AR/0.9/a
B-scans in the axial direction through the curved surface on the side of the
AR/AR/X bonded cylinders focused at surface at 10 MHz. Enhanced
ultrasonic B-scans from the AR / AR / X / x sample series, where X
designates the diameter of the implanted anomaly, and x indicates the scan
orientation (a - axial).
ee,
AR/ 0.2
Figure 4.15
Enhanced B-scans from the AR / AR / X / x sample series, where X
designates the diameter of the implanted anomaly, and x indicates the scan
orientation (r - radial). No indications of diffusion bondlines or anomalies
are evident for the 0.2" diameter implant, but bondlines are clearly evident
in both of the samples with
indicated with arrowheads.
94
larger diameter
implanted
anomalies,
as
is presented for the AR/FBA/X
In Figure 4.16 the same data as in AR/AR/X
sample set, where X is 0.9", 0.4", and 0.2". In this case, regardless of the focusing, the
axial imaging
approach demonstrates
no clear indications of bondline reflections for
either the 0.9" or 0.4" anomaly inserts. On the other hand, there are clearly obvious
changes in the character of scattering data for both of the larger implanted anomalies.
Unfortunately, AR/FBA/0.2"
has a severe bondline defects, which makes it difficult to
comment on the detectability due to uncertainty regarding how much acoustic energy is
propagating
through
the
imperfection.
Moreover,
the
ultrasonic
ringing
from
the
unbonded interface causes difficulty in determining the origin (in time) of the reflection.
Further inspection of the 0.2" sample showed lack of bonding over virtually the entire
implant.
Defocusing
appears
to
have
no
advantage
in
detectability
for
bondline
imperfections or microstructure variations.
In Figure 4.17, the B-scans were collected in the radial direction. Again, it appears
that scanning in this orientation tends to preferentially bring out the bondline signals, but
really does not provide any greater insight into the scattering perturbation caused by the
microstructure anomalies. Moreover, scanning in this orientation with defocusing appears
to wash out all indications of the bondline, except for the 0.2" anomaly, which obviously
did not fully join, as indicated by the strong echoes. This apparent inability of defocused
radial
scanning
to
highlight
bond
anomalies
is unfortunate,
considering
a similar
approach is a standard practice for large (~10" diameter) billet qualification examinations.
The defocusing technique clearly suffers from beam disturbances, tending to mask over
the grain scattering , which is the basis of the microstructure characterization.
95
ER
RR
x,
AR/FBA/2/a
Figure 4.16
Enhanced ultrasonic B-scans from the AR / FBA / X / x sample series,
where X designates the diameter of the implanted anomaly, and x indicates
the scan orientation (a - axial). A strong bondline appears from the 0.2"
implant, which indicates the bond did not properly form. The nature of
scattering clearly changes in both of the images from the bigger diameter
implanted samples, but neither show evidence of bondline signals.
. 96
OESAR/EBA/.9/
Figure 4.17
Enhanced ultrasonic B-scans from the AR / FBA / X / x sample series,
where X designates the diameter of the implanted anomaly, and x indicates
the scan orientation (r - radial). A strong bondline appears from the 0.2"
implant, which indicates the bond did not properly form. A bondline signal
is evident from the 0.4" implant, but only marginally so for the 0.9"
implant. No characteristic scattering changes are evident in any of these
images.
In Figure 4.18 data is presented from the AR/CBA/X sample set, where X is 0.9",
0.4", and 0.2". The axially collected B-scans again demonstrate marginal detectability of
the 0.9" and 0.4" microstructure anomalies, but fail to show any indications from the 0.2"
anomalies. Clearly, all the samples from this set fused in such a way that the bondlines
are virtually invisible. The coarse microstructure of the CBA
inserts present a less
obvious target in terms of detectability compared to the FBA
targets, based on these
backscattering results. In Figure 4.19, the radial scanning preferentially delineates the
bondlines, at the expense
of characteristic variations associated with microstructure
scattering.
The autocorrelation length, which is expressed as:
(4.1)
A(t) = [f(OF(t-aat
provides a measure of the similarity
between functions (or waveforms in this case). The
correlation lengths of coarse and fine beta annealed inserts and the mill annealed inserts
have been calculated from the radially bonded samples (surface focusing), as shown in
Figure 4.20. This parameter appears to be the only reliable indicator that allows key
differences in the scattering to be described. The capability to distinguish between the
different imbedded
microstructure
statistical amplitude distributions
microstructure
anomalies
tend
overlap
to
has the longest correlation
is important,
especially considering the
significantly.
The
length due to the aligned
98
mill
annealed
features
in the
structure, while the correlation lengths are much shorter for the beta annealed samples
due to recrystallization.
AR/CBA/2/a
Figure 4.18
Meaapne
AR/CBA/4/a.—AR/CBA/9/a
Enhanced ultrasonic B-scans from the AR / CBA / X / x sample series,
where X designates the diameter of the implanted anomaly, and x indicates
the scan orientation (a - axial). No indications of diffusion bondlines or
anomalies are evident for the 0.2" diameter implant, but the characteristics
of the scatter appears to change in both of the samples with larger diameter
implanted anomalies, as indicated with arrowheads.
99
ee
AR/CBA/.4/r
Enhanced ultrasonic B-scans from the AR / CBA / X / x sample series,
where X designates the diameter of the implanted anomaly, and x indicates
the scan orientation (r - radial). Bondlines are marginally detectable for the
0.2: implant. The bondline from the 0.4" implant is clearly detectable, as is
that of the 0.9". No characteristic changes in the scattering are evident in
any of these samples.
100
- + = coarse beta annealed
i\
———~ fine beta annealed
4
. 0.168"
——
mill annealed
60 5
40 |
J
0.064"
pre
0.047"
207")
0
.10
.20
30
40
50
.60
length (inches)
Figure 4.20
Correlation lengths of the radially bonded samples in the axial direction.
4.5 Summary
Diffusion
bonded
samples
have
been
constructed
to
simulate
incomplete
processing anomalies in the Ti-6Al-4V alloy. These samples were bonded in both the
axial and radial orientations to allow an assessment to be made regarding the detection of
anomalous
microstructure.
For
the
samples
bonded
in
the
radial
orientation,
an
assessment was made of the optimal scan configuration for the detection of simulated
processing anomalies, but for the axially bonded samples only one scan configuration was
used. A 10 MHz,
3" focused, 0.5" aperture ultrasonic transducer was used to conduct
these experiments. B-scans were taken at two different focal lengths (i.e. at the surface
and defocused 1" from the surface) to ascertain whether or not it is advantageous to focus
below the surface. The ultrasonic results from the axially bonded samples clearly show it
is possible to detect anomalous microstructure via B-scanning in the axial direction. For
the radially bonded samples, the results indicate the 0.9" and 0.4" diameter microstructure
101
inserts are marginally detectable when scanned in the axial direction and more difficult to
detect when
scanned
in the radial direction.
Radial
scanning
tends to preferentially
resolve the diffusion bondlines at the expense of scattering information related to the
microstructure
anomalies.
On the other hand,
axial scanning
tends to bring out the
characteristic differences in scattering between the microstructure inserts and the base
metal.
Probably
the most
significant finding
is that the 0.2"
undetectable in all cases, except the AR/FBA/0.2"
diameter
inserts were
sample, which clearly did not fully
bond during the HIP process. The correlation length appears to be a reliable parameter to
characterize the microstructure differences in the implanted anomalies, but the statistical
amplitude distributions overlap significantly.
102
Chapter 4, References:
1. Fundamentals of Ultrasonic Nondestructive Evaluation - A Modeling Approach, L. W.
Schmerr, Jr., (Plenum Press, NY, 1998).
2. Ultrasonic Testing of Materials, eds., J. Krautkramer and H. Krautkramer, (SpringerVerlag, NY, 1990).
3. Ultrasonic Waves in Solid Media, J. L. Rose, (Cambridge University Press, 1999).
4. E. P. Papadakis, “Ultrasonic Attenuation due to Grain Scattering in Polycrystalline
Metals,” Journal of the Acoustical Society of America, Vol. 37, p. 711 (1965).
5. E. P. Papadakis, “Ultrasonic Attenuation Caused by Scattering in Polycrystalline
Media,” Physical Acoustics Principles and Methods, Vol IV, Part B, Ed. Mason, W. P.,
(Academic Press, New York, 1968) p. 269.
6. F. J. Margetan, R. B. Thompson, and I. Yalda-Mooshabad “Modelling Ultrasonic
Microstructural Noise in Titanium Alloys,” Review of Progress in Quantitative
Nondestructive Evaluation, Vol. 12B, eds. D. O. Thompson and D. E. Chimenti,
(Plenum, New York, 1993), p. 1753.
7. B. Fay, “Theoretical Considerations of Ultrasonic Backscatter,” Acoustica 28, p. 354
(1973).
8. C. B. Guo, P. Holler, and K. Goebbels, "Scattering of Ultrasonic Waves in Anisotropic
Polycrystalline Metals," Acoustica 59, pp. 112-120 (1985).
9. S. Hirsekorn, “The Scattering of Ultrasonic Waves in Polycrystalline Materials with
Texture,” Journal of the Acoustical Society of America, 77(3), (1985).
10. J. H. Rose, “Ultrasonic Backscattering from Polycrystalline Aggregates Using TimeDomain Linear Response Theory,” Review of Progress in QNDE, Vol 11, eds D. O.
Thompson, D. E. Chimenti (Plenum, New York 1992) p. 1677.
11. J. H. Rose, “Ultrasonic Backscatter from Microstructure,” in Review of Progress in
ONDE, Vol 11, eds D. O. Thompson, D. E. Chimenti (Plenum, New York 1992) p. 1677.
103
CHAPTER V
THE INFLUENCE OF TEXTURE AND PHASE DISTORTION ON
ULTRASONIC ATTENUATION IN Ti-6Al1-4V
5.1 Introduction
In this chapter,
a methodology
to characterize
microstructure
and
study the
influence of texture and phase scattering on ultrasonic attenuation in Ti-6Al-4V
was
was
also
demonstrated.
developed
and
developed
to map
the phase
A
high
resolution
experimental
of ultrasonic
and magnitude
waves
capability
transmitted
in the
titanium solid. The advancement presented in this work is provided by laser detection of
the ultrasonic energy over a microscopic aperture of approximately 50 um. The system is
built around a computer controlled scanner and a confocal Fabry-Perot interferometer,
which uses a diode pumped Nd:YAG
laser as a light source. Wave propagation in the
axial and radial directions of a 2.5" diameter bar of highly textured (mill annealed) Ti-
6Al-4V was investigated in this study.
The
work
was
motivated
by
the
observation
of
unusually
high
apparent
attenuation in the axial direction of the as-received bar, thought to be mainly associated
with phase distortion rather than actual energy loss. The current phase mapping results,
104
using
a focused
show
spot,
laser
high
relatively
wavefront
distortion
and
more
nonuniform distribution of the transmitted energy in the axial direction. The contribution
to attenuation associated with phase cancellation loss was also investigated. These
measurements
show the laser detected attenuation to be substantially lower than the
piezoelectrically measured attenuation. However, even the relative phase insensitivity of
focused laser detection approach clearly indicates the attenuation to be strongest in the
axial direction. This work demonstrates the orientation dependence of attenuation stems
from scattering effects associated with texturing and the elongated macroscopic grain
structure in the mill annealed Ti-6A1-4V bar generated during processing, which may also
affect
diffraction
and
divergence.
beam
This
chapter
provides
a
foundation
for
nondestructive evaluation to observe worked microstructure, indicated by structure bands,
and may provide a process control method for alloy forging and ingot coarseness.
For Ti-6AI1-4V, the manufacturing processes used to fabricate stock materials (e.
g. bar, billet, plate) tend to impart a preferred crystallographic orientation due to the
restricted nature of mechanical
evidence
of macroscopic
slip [1], leaving the material with easily measurable
anisotropy.
For example,
5 MHz
shear wave
birefringence
measurements conducted on 1.5"-thick sections of 2.5"-diameter mill annealed Ti-6A1-4V
bar demonstrate only about 1% variation in velocity for wave propagation in the axial
direction, while in the radial direction the difference in velocity between the slow and fast
pure shear modes is nearly 5%. In addition, compared to the axial direction, the 10 MHz
longitudinal
wave
velocity
is approximately
2%
faster in the radial
direction.
The
observed velocity anisotropy is consistent with the development of a texture in which the
oriented
randomly
basal
normals
of the
alpha
phase
(hexagonal
symmetry),
which
constitutes more than 90% of the alloy by volume, tend to preferentially rotate to the
transverse direction of the bar during the course of processing.
Ultrasonic measurements, using piezoelectric detection, show significantly higher
attenuation in the axial direction as compared with the radial one. Similar results have
been described for austenitic stainless steels in which the material develops a columnar
grain structure due to solidification, rather than hot-working as is the case for titanium
alloys [2, 3]. Typically, attenuation is considered to be based on energy decay of the
propagating ultrasonic waves, which is proportional to length and mainly affected by
absorption
and
scattering.
However,
for
some
polycrystalline
titanium
alloys
and
austenitic stainless steels comprised of primarily hexagonal symmetry, phase distortion of
the propagating wavefront and the resulting signal diminution at the receiver has been
speculated as another source of loss. This phase cancellation loss at the receiver, rather
than actual energy-based losses, is thought to be partly responsible for the attenuation
behavior observed in the axial direction of mill annealed Ti-6AI-4V.
Recently,
Panetta
et al.
demonstrated
that
fluctuations
in the
amplitude
of
backwall echoes observed as the ultrasonic beam is scanned over different samples of
titanium alloys are not due to energy losses [4]. Their work supports the idea that phase
distortion of the transmitted beam contributes to the higher attenuation values in the axial
direction of the billet. Using a small diameter "point" receiver, they also revealed that the
scattering and absorption as deduced from the transmitted energy are indeed anisotropic
106
in Ti-6Al-4V billets, but are smaller and isotropic in Ti-5Al-2Sn-2Zr-4Mo-4Cr (Ti-17)
[5].
was
finding
This
both
fact that
the
despite
observed
showed
alloys
distinct
characteristics of macroscopic anisotropy in the two orientations studied. Similar phase
mapping studies have been undertaken using small aperture hydrophone receivers [6] and
array transducers
[7]. In this chapter,
we
use
focused
detection
laser spot
thereby
minimizing the phase sensitivity, allowing high resolution imaging of structure-induced
wavefront distortion.
Spatial uniformity of the transmitted ultrasonic beam's pressure distribution has
been shown to play an important role in the measurement of absorption and scattering [8].
For
phase-sensitive
receivers,
impinging
the aperture
electrical
signal.
any
fluctuations
in the
are instantaneously averaged,
Conversely,
the
phase-insensitivity
phase
of the
pressure
field
resulting in cancellation of the
offered
an
by
acoustoelectric
receiver, constructed from a single crystal of the piezoelectric semiconductor cadmium
[9], demonstrates
sulfide
more
reliable estimates
of quantitative
intensity dependent
parameters like attenuation, at the expense of reduced resolution and sensitivity.
Medical ultrasonic characterization of tissue samples is another area where the
development of phase distortion in the propagating wavefront has important technological
ramifications. Diagnostic analysis of tissue inhomogeneities generally involves imaging
provided by phased array ultrasonic systems. Tissue-induced wavefront distortion can
severely
degrade
the
quality
of medical
ultrasonic
107
imaging
and
often
correction
techniques are needed to account for phase aberrations to generate higher quality images
(10, 11].
A
transmitted
focused
laser
incident
beam
on
[12]. Besides
acoustic radiation
a suitable
surface
the obvious
can
advantages
used
be
to
detect
offered by laser
interferometry (e. g. noncontacting, elevated temperature usage, complex geometries) for
detecting ultrasonic waves, diffraction limited apertures of sub-wavelength dimensions
can be easily achieved via focusing of the incident laser beam. In contrast to other
approaches used to restrict the aperture, such as applying a pinhole [13] or needlelike
stylus tip [14] to achieve high resolution, laser interferometry offers the best sensitivity
when the laser is concentrated to the smallest spot size [15]. However, in some highly
backscattering materials like that of the transverse direction in mill annealed Ti-6Al-4V,
the grain noise level significantly increases as the spot diameter of the laser decreases
[16], which would obviously have adverse affects on the detection of small defects.
For laser detection, as the incident beam is scanned over the optically reflective
surface of the titanium sample
undergoing
surface displacements,
the reflections are
collected and passed to the interferometer. The surface displacements generated by the
propagating ultrasonic waves cause the reflected laser beam to be frequency modulated,
resulting in a Doppler
shift of the carrier frequency
interferometer
demodulates
corresponding
to the
the
acoustic
propagating in the sample.
reflected
light
particle
velocity
[17]. The confocal
to reconstruct
(or
pressure)
Fabry-Perot
a time-domain
signal
elastic
waves
of the
The
main
goal of this chapter
is to explore
the use
of laser interferometric
detection to better understand the effects of macroscopic texture and phase perturbation
on the attenuation of longitudinal waves in the mill annealed Ti-6A1-4V alloy. To begin,
we will review the basic ultrasonic properties as a function of orientation in the mill
annealed Ti-6Al-4V bar, measured with standard piezoelectric ultrasonic detection. We
then discuss the operation of the laser interferometric detection experiment, detailing the
differences
in hardware
components
used
to map
the phase
and
magnitude
of the
transmitted ultrasonic waves. Next follows a presentation of the experimental results and
discussion, demonstrating the structure-induced phase jitter in the axial direction to be
significantly higher and more nonuniform than that of the radial direction. Finally, in
order to assess the attenuation loss which develops due to phase cancellation, we compare
attenuation measurements
performed
with a conventional
wide-aperture
piezoelectric
receiver versus laser-based detection, using a microscopic spot size.
5.2 Ultrasonic Properties of Mill Annealed Ti-6A1-4V
Samples of the mill annealed Ti-6A1-4V alloy were cut into 1.5"-thick sections
with flat, plane parallel surfaces such that the angle between surface normals and the axis
of the bar were at 0° (axial), 30°, 45°, 60°, and 90° (radial). Ultrasonic measurements of
longitudinal and shear wave velocity, attenuation, and scattering (forward and backward)
were conducted on the samples. The microstructures of the two key sample orientations
are shown in Figure 5.1. The axial orientation of the Ti-6Al-4V bar shows a relatively
equiaxed and random grain structure, while that of the transverse shows highly aligned
colonies of elongated grains.
The longitudinal ultrasonic velocity was measured using the standard pulse-echo-
overlap technique at 10 MHz, consistent with guidelines specified in the literature for
measuring ultrasonic velocity [18, 19]. The longitudinal wave velocity results are shown
in Figure 5.2. The data clearly shows that the wave velocity is highest in the radial
direction of the bar (see Table 5.1 - Orientation Dependence of Longitudinal Velocity,
Appendix A). A difference of about 2% is observed in the average velocity between the
axial and radial directions. The longitudinal data also shows a linear increase in velocity
as the propagation direction changes from axial to radial. The velocity is associated with
mechanical stiffness, therefore the material is stiffer in the radial direction.
< ER
hes
as
OWE
Figure 5.1
ad i
The as-received (mill annealed) condition of Ti-6AI-4V bar stock from
surfaces of the (a) axial (0°) and (b) transverse (90°) samples. Both
microstructures are shown at 200x.
110
longitudinal velocity (m/s)
30
45
60
orientation angle (degrees)
Longitudinal wave velocity as a function of orientation for 1.5"-thick mill
annealed Ti-6AI-4V samples.
Figure 5.2
Measurements of the shear wave velocity provide more details on the macroscopic
texture
of the
propagation
structure
direction.
as these
Shear waves
waves
can
for the
same
to the surface of a sample
will
be polarized
incident normal
differently
generally split into two transverse waves vibrating at right angles to each other and
propagating at different velocities, due to birefringence [20, 21]. This effect is even more
dramatic in single crystals than in anisotropic polycrystalline metals due to the stronger
directional dependence of velocity. The shear wave velocity data of Figure 5.3 shows
significant velocity variations between the differently oriented titanium alloy samples. For
each sample, several measurements were taken at two different shear polarization angles
consistent with the slow and fast pure shear modes. We can see from the average
3350 +
Cislow mode
3300 +
ca
ES
ba
60
| Pd
90
v4oS
Loe)
}
ch
3200 aah
oS
Oo
|
1
—
3150 +
ww
shear velocity (m/s)
(fast mode
3050 + |x
3000
ba
45
Tite,
30
°
orientation angle (degrees)
Figure 5.3
Shear wave velocity
annealed Ti-6Al-4V.
as a function
of orientation
for
1.5"-thick
mill
shear velocity data that the primary factor affecting the velocity is the presence of a
preferred crystallographic orientation. The most dramatic evidence of this influence is
illustrated in the 90° sample, with the difference between the slow and fast pure modes of
4.8%. We
also see a steady increase in the birefringence delta (difference in velocity
between the slow and fast pure modes) as the wave propagation direction goes from axial
to radial, with an axial birefringence
delta of only about
1%.
This small degree
of
mechanical anisotropy is evident in the axial direction because the texture is slightly offaxis and varies a little for different locations along the bar. Hence, the shear velocity is
minimally dependent on the polarization direction for propagation in the axial direction;
however, for propagation in the radial direction, the shear polarization has a major
influence
on the wave
velocity
(see Table
112
5.2
- Orientation
Dependence
of Shear
Velocity, Appendix A). Wave propagation in the radial direction always gives rise to the
fast pure
mode
when
the particle
displacements
are perpendicular
to the elongated
macroscopic inhomogeneities, while the slow pure mode always occurs when the particle
displacements are parallel to the aligned inhomogeneities. Clearly, the particle oscillation
of the fast pure shear mode is consistent with the preferred orientation of hexagonal basal
normals, i. e. in the principal direction. This finding also appears to make sense in light of
the longitudinal velocity results, showing faster speeds for waves traveling perpendicular
to the axially aligned macroscopic structural features.
Attenuation measurements
were also conducted on the same
set of 1.5"-thick
samples. The ultrasonic attenuation coefficient is broadly defined as the rate of decay of
the measured signal as a wave propagates through a material. The apparent attenuation
can be significantly different depending on how the signal is measured (e. g., phasesensitive versus phase-insensitive measurements). Attenuation loss is defined as the ratio
of two amplitudes and expressed in logarithmic units, neper or decibel. In some cases, the
loss can occur locally as a result of interaction with a material discontinuity. Similarly,
losses could originate from reflection and transmission at an interface, internal voids or
inclusions, scattering at a rough surface, or from microstructural changes. There are other
losses occurring over a given distance as the wave propagates in the medium, which are
not necessarily proportional to the travel distance. Such losses are usually associated with
beam divergence. The primary loss mechanisms investigated in this chapter are those due
to the presence of texture and orientation.
There are two major attenuation mechanisms considered important for ultrasonic
materials characterization. The first is absorption, which converts acoustic energy to heat
via viscosity, relaxation, heat conduction, elastic hysteresis, ete. The absorbed energy is
irreversibly lost from the acoustic field and is dissipated in the medium. While absorption
is difficult to separate from scattering in terms of a practical measurement, all the obvious
indicators seem to point to the fact that the absorption is virtually the same, regardless of
orientation in the mill annealed alloy. The second important attenuation mechanism is
scattering, which converts the energy of the coherent beam into incoherent, divergent
waves as a result of interaction with inhomogeneities in the material. The scattered energy
is not necessarily lost as part of it can be recovered by the same transducer used to
transmit, as backscatter. Scattering not only reduces the amplitude of the coherent signal,
but also gives rise to an incoherent material noise which further limits the detectability of
defects. Fortunately, the same incoherent noise that hinders our view to the development
damage
of low-level
or defects also provides
a source
of information for ultrasonic
characterization of grain structure [22, 23, 24].
The
attenuation
was
measured
underwater
using
a standard
10
MHz,
0.5"-
diameter unfocused transducer, driven with narrow-band tone-burst excitation to allow
easy adjustment of the frequency. The orientation dependence of the attenuation is shown
in Figure
5.4
as a function
of frequency
(see
Table
5.3
- Orientation
Dependent
Attenuation Loss, Appendix A). For these measurements, no effort was made to account
for beam divergence losses since all the samples were of the same thickness and because
we were mainly interested in making a simple comparison of relative attenuation
114
uO deg
16,
x30 deg
14-
.45 deg
_
124
960 deg
3~
191
*90 deg
7
E
:
ZB
>
6+
t
4t
:
ff
;
i
=
&
&
§
&
| |5
i
Hf
4
2+
0
5
t
t
t
t
6
7
8
9
10
t
t
11
12
—
13
frequency (MHz)
Figure 5.4
Orientation dependence of the average attenuation loss versus frequency
for 1.5"-thick mill annealed Ti-6A1-4V samples.
between the various sample orientations. All the immersion attenuation measurements
were done with pulse-echo in the far-field of the transducer.
The distance between the normally incident transducer and the front surface of
each specimen was set to 3". A function generator and power rf amplifier were used to
excite the transducer at a given frequency.
For each frequency, the attenuation was
measured by taking the Fast Fourier Transform (FFT) of the signal from the first backwall
echo and subtracting the FFT of the second backwall echo. The difference spectrum was
then evaluated at the excitation frequency (the role of the FFT processing was to average
over the length of the signal, rather than to provide spectral information). Because the
115
¢
signal amplitudes tend to have some modest spatial variations in the mill annealed alloy,
ten different locations on each
were repeated for approximately
the measurements
sample. The excitation frequency was then changed and the measurement repeated for a
range
of frequencies
between
7 and
calculated
We
12 MHz.
the impedance
losses
solid / water interface using a
contributing to the attenuation based on the Ti-6A1-4V
density of 4.4 grams/cc and a velocity of 6180 m/s, yielding a value for the reflection
coefficient, R, of .9 or R* ~ —2 dB. This loss was accounted for in the results.
The 0° sample (axial direction) clearly has the highest attenuation, followed by the
30°, and 45°
samples, respectively. The
60°
and 90°
samples completely
overlap in
attenuation, considering the scatter in the data. The error bars of Figure 5.4 are based on +
one standard deviation from the mean, representing the data scatter. The 45°
sample
showed very little scatter in the data (< + 0.3 dB), but the 0° and 30° samples showed
significant scatter ( up to about + 2 dB) with increasing scatter at the higher frequencies.
Moderate data scatter, on average of about + 1 dB, was observed in both the 60 and 90
The piezoelectrically measured
degree
samples.
clearly
orientation
dependent.
What
be
attributed
to
attenuation
can
we
losses
would
coherent
"plane wave"
like to know
from
phase
measurements
were
incurred
attenuation is
much
of this
perturbations
of the
conducted
in the
is how
transmitted signals.
Forward
and backward
scattering
also
titanium alloy samples for the axial and radial directions. Backscatter measurements were
done at normal incidence, using a 10 MHz, .5"-diameter transducer with a 3" focal length.
116
The transducer was focused on the front surface and the backscatter was measured in
based
pulse-echo
on
a spatial
average
over
approximately
1"
a
x
1"
This
area.
measurement is relatively simple using a digital oscilloscope. The first step is to be sure
the sample is parallel to the scan plane of the transducer such that the signal does not drift
in time during scanning. Next, we remove the coherent component by taking a global
spatial average in the rf and subtracting this averaged signal from each independent rf
backscatter signal. What is left after subtracting the coherent part is squared and spatially
averaged. Once the summed average runs its course, the final step is to take the square
root of the spatially averaged
backscatter response.
in Figure
returned as a function of time is shown
Dependence
Appendix
of Backscattering,
A).
The
average
5.5 (see Table
Clearly,
the
significant backscatter is from the radial direction in which
backscatter
signals
5.4 - Orientation
with
any
the incident waves
are
only
sample
traveling normal to the elongated grains and grain colonies. In each of the other samples,
the
bulk
of the
scattered
energy
is directed
away
from
the
backscatter
path.
The
backscatter in samples cut at 0, 30, 45, and 60 degrees to the bar axis is negligible, but the
overall scatter is not. For example, Figure 5.6 shows schematically what is happening for
scattering in the 0 and 30 degree samples. Here, the forward and side scattered acoustic
energy is equally, if not more important, to the backscatted energy for the purpose of
measuring attenuation. Using a large-aperture phase-sensitive transducer, the effect of
scattering is basically all the same insofar as the attenuation is concerned as part of the
energy
is removed
from
the
coherent
beam,
scattering is detected or not.
117
regardless
of whether
the
incoherent
average backscatter (a. u.)
sample (degrees)
time (us)
Average backscattering intensity as a function
orientations of mill annealed Ti-6Al1-4V.
Figure 5.5
of time
in
different
Forward scattering measurements were performed in through-transmission using a
10 MHz, 0.25" diameter contact transducer, mechanically coupled to the backside of the
sample to transmit the acoustic energy.
A 10 MHz, 0.5"-diameter focused transducer was
placed over the opposite surface to receive the transmitted scatter. Using this approach,
the forward
directed
incoherent
scatter can be observed
between
the first coherent
transmission and the first subsequent coherent echo. By scanning the receiver over the
forward
scattered
field,
we
can
assess
the
magnitude
and
divergence
of forward
scattering. A simple way to view forward scatter is by taking a C-scan, gating on the
amplitude modulated incoherent noise. Figure 5.7 shows the forward scattering C-scan
results, using a 1 ps gate width, after averaging several different images with slightly
different gate placements. Probably the most notable aspect of these results is the overall
axial
ieee
transducer
:
axial
transducer
direction
a direction
/
J
1
|
|
;
|
:
WW
|
f /
iwi
L\A
Al.
,
wave
direction
primary
scattering
direction
Y
y
30 degree sample
0 degree sample
higher
/
ge
forward scatter .
Figure 5.6
2
Schematic diagrams showing the effect of the macroscopic structure on
scattering for the 0 and 30 degree samples.
divergence
of forward
scatter in the axial direction.
Divergence
is inversely
proportional to average grain size [22, 25], that is why wave propagation in the radial
direction results in essentially the same divergence width in the horizontal direction (from
Figure 5.7.b) as seen for the axial direction. The forward scattering is more divergent in
the axial direction, but the amplitude of scatter signals in the radial direction is observed
to be at least as strong to those of the axial direction. Intuitively, the combination of
strong
backscatter
and relatively
strong
forward
scatter in the radial
direction may
logically lead to the assumption that the radial attenuation should be higher, but this is
clearly not the case. For radially directed wave propagation, due to the normal orientation
of the elongated grain colonies, essentially all scattering is concentrated in the backward
119
Wie
jaxial direction
(b)
.
(a)
C-scans of the forward scatter in the axial (a) and radial (b) directions,
taken over a 1" x 1" area at 10 MHz with a 0.5"-diameter, 3"-focal length
receiver, focused on the surface.
Figure 5.7
and forward directions. Since the scattering is very weak in all but these two specific
directions, the coherent wave attenuation is actually very low. Clearly, the total scattering
cross-section must be considered to assess the orientation dependence of the coherent
wave attenuation.
This is an area in need of further investigation, especially the forward scattering, if
we are to fully understand the higher attenuation losses in the axial direction of mill
annealed Ti-6AI-4V. In the next section, we move to the laser-based ultrasonic detection
experiment
to
discuss
magnitude
and
phase
disturbances on attenuation.
120
mapping
and
the
influence
of phase
5.3 Laser Detection of Ultrasonic Phase Distortion
A system was developed to image the phase
and magnitude
of transmitted
ultrasonic waves using a computer-controlled scanner in conjunction with a Fabry-Perot
interferometer. Figure 5.8 shows the schematic diagrams of the experimental systems.
The systems include a unique combination of analog and digital components, allowing
the measurements to be made without extensive signal processing. The phase mapping
system is configured to reconstruct the shape of the transmitted wavefront, including
distortions which develop as the waves propagate. Several different sample thicknesses,
including 0.25", 0.5", 1.0", and 1.5", were examined in the axial and radial directions.
Unfortunately, phase mapping of the thinner samples (< 0.5") revealed some inherent
near-field unevenness that we decided to exclude from this chapter.
A 0.25"-diameter
10 MHz
contact transducer was used to transmit the acoustic
energy, which has a near-field transition length (NV = a’/ A) of about 16 mm (0.63"). This
transducer was mechanically mounted to the backside each of the inspected samples to
maintain constant acoustic coupling. The transmitter generates a tone-burst, derived from
the combination of a function generator (HP 3314), capable of delivering a sine-wave
burst at up to 20 MHz,
and a power rf amplifier to boost the tone-burst signal by
approximately 50 dB. The power amplifier output excites the transmitter which launches
a burst of energy into the sample. The sample, with the attached transmitter, is fixed to a
bracket on the x, z scanner and the laser is held stationary. The transmitted signal is
detected on the front-side of the sample by a laser beam which passes the Doppler shifted
carrier signal to a 100-MHz-bandwidth Fabry-Perot interferometer for demodulation and
Nd:YAG Laser
10X microscope
objective
Transmitter
X, Z Scanner
on vertical stand
ly,
4, feyeS
O
Fiber-optic
cable Nw
Computer
:
Collecting
Fabry-Perot
lens
Interferometer
Function
Receiver
generator
(5072 PR)
(HP 3314)
RF amplifier
(DG535)
bd
|
—
trigger in
[
“|
t
:
= =
Analog oscilloscope
(aN
gate out
signal in
coerge
Ane.
Digital oscilloscope}
9310 AM)
Signal averager
(SR 280)
Avg signal
Delay generator
eet
Digital oscillosco
Function generator
(HP 3314)
-
~_t {|
(9310 AM)
A
rt
j—
Nd: YAG Laser
10X microscope
objective
Transmitter
“ey,
Fiber-optic
cable Nw
Fabry-Perot
Interferometer
(5072 PR)
O
Computer
Collecting
lens
unction
RF amplifier
generator
CHP 3318)
RE
Signal averager
Del
elay
generator
avg signal
(SR 280)
YE
(DG535)
out
triggerin
TR
Digital oscilloscope
X, Z Scanner
on vertical stand
A
Ona
Functi
.
Receiver
Sg
id
-
GO
. be
bf
a
RPL] |] signatin — sateout
i
ANB
i
Vertical signal out
Digital oscillosco
|e | fee] ES
9310 AM) | | Oscilloscope / Spectrum analyzer
Gate
signal marked trigger]
mow
it
J
in
[———"_
| }}
b)
Figure 5.8
Schematic diagrams of the experimental set-ups used to map phase (a) and
magnitide (b).
reconstruction of the time-domain response. The laser beam is focused on the surface of
the sample through a 10x microscope objective to make the spot size microscopic.
The focused laser spot size on the surface was estimated, assuming a Gaussian
beam, based on the following equation:
(5.1)
d(z) = dy)[1+ (442/ adj)? }”
is the diameter of the focused laser spot and zis the distance from the waist of
where d
the laser beam [26]. The laser has a wavelength (1) of 532 nm and a divergence angle
(0) of about 2.2 mrad. The output beam is assumed to be Gaussian with an initial beam
diameter (d;) of 320 um
and a waist diameter (dy) of approximately
@=44/ nd) ~2.2 mrad). Using a value of 16.9 mm
307 um
(from
for the focal length of the 10x.
microscope objective and assuming the objective was focused to within about 1 mm of
the actual focal length, we estimate the spot size of the focused laser light to be from 25
to 50 um.
The phase disturbances are measured by a precision time-to-voltage converter,
assembled
from
general-purpose
laboratory
instruments.
The
titanium
samples
were
finish polished with 9 pm diamond paste to yield nicely reflecting surfaces. The laser
light reflects off of the shiny titanium alloy surface undergoing displacements and is
gathered by a collecting lens which focuses the light onto a 400 .m-diameter fiber-optic
123
cable. This cable passes the light to the interferometer which is operated in reflection
mode. The time-domain signal output from the interferometer is then routed to a receiver
(Panametrics 5072 PR) to control the gain and then to an analog oscilloscope which has a
trigger
delayed
output
and
pre-triggering
capabilities.
Pre-triggering
allows
local
triggering on the waveform of interest. The delayed trigger output, which is dictated by
the pre-trigger, is used to synchronize the boxcar averager with the phase variation of the
interferometer output signal. A single cycle triangle wave is input to the boxcar averager
to discriminate the phase variations. With this experimental set-up, any movement
in
arrival time (phase) of the interferometer signal simultaneously moves the gate on the
boxcar averager and changes
its corresponding voltage input according to the linear
portion of the triangle wave. The boxcar gate is generally centered on the middle zeroof the triangle
crossing
wave.
With
this
approach,
variations
in the phase
of the
interferometer signal, detected as the laser is scanned over the specimen, correspond to a
linear change in voltage.
To increase the signal to noise ratio in the images, we usually decrease the scanner
speed to about 10 s per line (200 points per line) and increase the number of signal
averages
performed
on the boxcar.
The
sensitivity
of the phase
imaging
system
is
adjusted by manipulating the slope and amplitude of the triangle wave input to the boxcar
averager. Obviously, higher frequency triangle waves yield steeper slopes and therefore
provide higher sensitivity. Hence, a variation in phase as indicated by movement of the
boxcar gate, can have dramatic or slight corresponding voltage change, determined by the
characteristics of the triangle wave.
124
For magnitude imaging, the interferometer output signal is hardware gated with a
stepless gate on the signal of interest and then introduced to a spectrum analyzer. The
spectrum analyzer is tuned via a local oscillator to the appropriate frequency (generally
the same frequency as the tone-burst signal) and the instrument's
3 MHz
bandwidth
determines the length and shape of the signal output. This output is a bell shaped curve
whose amplitude corresponds that of the detected interferometer signal over the
3 MHz
frequency range. Again, the boxcar averager is used to sample and hold the peak of the
bell curve, except under normal triggering synchronization. The averaged output from the
boxcar, which is proportional to the strength of the gated signal in the narrow frequency
band centered around the chosen nominal measuring frequency, is fed into the analogdigital converter for sampling and digitization.
The
computer
controlled
scanner
always
generates
200
x 200
array images,
regardless of the scan size (from 0.1" x 0.1" up to 2" x 2") that are digitized with 16 bit
resolution. The C++ computer program we use to generate the images was written do
some modest signal averaging internally, but we also generally over-sample extensively
on the boxcar averager (100 or more measurements per pixel) and reduce the speed to
increase the signal-to-noise ratio of the phase and magnitude pictures. While this scanner
is modest in terms of the sophistication of the analog to digital conversion, it is versatile,
allowing data to be collected from a number of different experimental configurations.
5.4 Results and Discussion
The objective of phase imaging is to map the shape of the transmitted ultrasonic
wavefront. The transmitter is mechanically fixed on the backside of the titanium alloy
sample. The waves are launched from the transmitter and eventually begin to diverge
after the near-field / far-field transition zone. As the waves diverge, the shape of the
wavefront
becomes
curved
due
to diffraction
and
beam
spreading.
This
wavefront
curvature is especially apparent in points laterally displaced from the direct line-of-sight
transmission axis. The phase lag increases for further off-axis locations on the sample, as
shown
schematically
in Figure
5.9.
The
phase
mapping
system
accommodates
the
sometimes very large phase variations over a given area simply by restarting the grayscale
after every 2x phase shift. Hence, at every 2m phase transition a fringe is apparent. The
development of fringes is consistent with the discontinuous presentation of phase, folding
back after every 2n increment. It is possible to develop an algorithm to reconstruct the
waveform in the absence of fringes, but this would not change the information held in the
image,
only reshape
it. Moreover,
in this paper
we
are mainly
interested
in high-
resolution imaging of phase disturbances in the directly transmitted lowest order fringe.
Figure 5.10 shows some examples of phase and magnitude imaging. These images
were taken at 9.7 MHz through 0.5"-thick samples, cut to allow access to the axial
(images a and b) and radial directions (images c and d) in the mill annealed material over
a 1" x 1" area centered on the transmission axis. Both phase images were taken under
identical scan parameter settings and reveal multiple 21 phase transitions due to the
curvature of the propagating wavefront.
Darker
shades of gray in the phase
images
represent a phase lag. In this case (0.5"-thick samples) the first transmitted wave burst
arrives at the detector prior to the near / far-field transition zone. As such, we see a highly
126
fundamental
fringe (all data within a 27
phase field)
Lo
| subsequent
NRA ff
rw nna’
4 WWMM
4 iA LALS-
shifts
J.
A
AA
—~S
=
UV
position
aml
:
:
Y
MIATA
’
VA
VW
A
AA
MAA
TNT NT NT
MAA
Av A
A
A
A
ae
—
LI
fA
>
time
LJ
from oscilloscope
mounted
transmitter
(b)
(a)
Figure 5.9
Schematic diagrams showing (a) three regions on the specimen over which
the phase map is repeatedly folded back by 2x and (b) the variation in
arrival time at different positions across the wavefront.
collimated wavefront that dramatically lags in phase outside of the center fringe, thus the
large number of fringes. Also evident is the near-field effect of phase lag in the centrally
transmitted zone of the lowest order fringe. The corresponding magnitude images (Figure
5.10.b and 5.10.d) which were both taken with the same scan settings, are shown next to
the phase images. Here, lighter shades indicate higher amplitudes.
Figure 5.11 shows a similar effect to the wavefront distortion we are interested in
observing in the mill annealed titanium samples. In this case, we mapped the phase
through a 0.2"-thick Ti-6Al-4V
beta annealed (~1040°
sample, which was vacuum encapsulated in glass and
C) for an exaggerated period, then slowly cooled to yield very
127
coarse lamellar «a + B grain colonies. In some places these colonies extended through the
thickness
of the
sample.
Here,
we
used
a 0.5"-diameter
transmitter to increase the
interaction volume in hopes of encountering a grain colony with significantly different
(b)
(d)
Figure 5.10
Phase (a) and corresponding magnitude (b) images taken at 9.7 MHz,
through 0.5"-thick samples of the mill annealed Ti-6A1-4V from the axial
direction, covering a 1" x 1" area. Same images are shown for the radial
direction (c), (d).
128
(b)
(a)
Figure.5.11
Phase maps taken through a 0.2"-thick Ti-6Al-4V sample which had been
heat treated to generate a coarse widmanstitten microstructure with very
large lamellar o + B colonies. Phase images taken at (a) 7.5 MHz and the
same region at (b) 9.7 MHz.
wave propagation characteristics to the surrounding structure. The upper middle part of
the center fringe shows the desired effect. Phase images were taken at two different
frequencies (7.5 and 9.7 MHz) using the same settings as those from Figure 5.10. Clearly,
in Figure 5.11 there is a region in the sample that lags significantly in phase from the
surrounding
material.
At 9.7 MHz,
the
shorter
wavelength
causes
the
grayscale
to
reinitiate in the center of the anomalous colony, indicating the arrival time delays by more
than a full cycle at that frequency. Similarly, Figure 5.12 shows the phase and magnitude
images for a 1.5"-thick mill annealed Ti-6AI-4V sample in the axial direction (a, b) and
radial direction (c, d) using the 0.25"-diameter transmitter at 9.7 MHz. In this case, the
first transmitted waves are in the far field of the transmitter. These phase images are
representative of the ultrasonic wavefront distortion in the axial and radial directions
129
through 1.5". In the axial direction, the phase perturbation was severe enough to cause the
central fringe to be significantly distorted from the generally observed concentric ring
pattern, as seen in the radial direction. Again, the corresponding magnitude images are
shown next to the phase images.
Figure 5.12
Phase (a) and corresponding magnitude (b) images over a 1" x 1" area
through a 1.5"-thick sample of mill annealed Ti-6A]-4V in the axial direction at 9.7 MHz. Same are shown for the radial direction (c), (d).
130
So far, we have shown representative data from 0.5"-thick and 1.5"-thick mill
annealed samples, with neither case being ideal for the generation of phase images due to
the presence of near-field artifacts or too much phase variation. Figure 5.13, a and b
shows the raw phase images from a 1.0"-thick sample in the axial and radial orientations,
respectively, using the 0.25" transmitter at 9.7 MHz.
In this case, we optimized the
sensitivity to see the best contrast in the center fringe of the image. The scan settings were
identical for both the axial and radial samples. Each of these scans were taken over a 0.4"
x 0.4" area with a 0.002" stepping increment. To allow easier manipulation of the data,
we stripped the first 27 phase transition from the phase images. We also filtered the data
to remove low spatial frequency phase variations associated with beam divergence. The
resulting high-pass filtered images are shown in Figure 5.13, c and d.
From
the
statistical
distribution
of the high-pass
filtered data, we
measured
approximately 30% higher phase variation in the axial direction, as compared with the
radial direction, for a 130 x 150 pixel array covering the center portion of data in each
the magnitude
of these high-frequency
image.
While
modest,
only up to about 20 degrees,
nonuniform in the axial direction.
the phase
wavefront
distortions
is rather
scatter is clearly higher and more
The patterns generated by the high-frequency phase
variations are also quite different between the two orientations. In the axial direction the
phase contrast appears as more or less random speckles, while in the radial direction the
contrast reveals long vertical striations. These patterns are both consistent with the
corresponding macroscopic grain structures.
Figure 5.13
Phase images taken through 1"-thick samples of the mill annealed Ti-6Al4V covering a 0.4" x 0.4" area using the 0.25"-diameter transmitter at 9.7
MHz. Raw data from the axial direction (a) after removing the first 2x
foldback, same for the radial direction (b). High-pass filtered data from
axial (c) and radial (d) directions.
Finally in addition to texture, we wanted to better understand the influence of
phase distortion on attenuation. Laser-based attenuation results were collected similar to
the approach used in the pulse-echo immersion experiment, where we simply measured
132
the loss in dB
between
the first and
second
echoes.
backwall
The
only
significant
difference for the laser detection approach is that we needed to account for the impedance
mismatch loss due to mechanical mounting of the transmitter to the sample. This loss is
due to the relatively low reflection coefficient at the interface between the specimen and
the coupled contact transmitter, compared to the specimen / water interface. Here, part of
the energy is lost due to wave propagation back into the transmitter.
This loss was
measured to be approximately 6 dB. Otherwise, except for the fact that the measurement
was performed in through-transmission, the data collection routine was the same as the
immersion attenuation measurements. Figure 5.14 shows the outcome of these attenuation
a O deg, 0.5" aperture
0 90 deg, 0.5" aperture
average loss (dB)
e 0 deg, laser spot
o 90 deg, laser spot
9
10
11
frequency (MHz)
Figure 5.14
Comparison of attenuation measurements from a 0.5"-diameter unfocused
immersion transducer in pulse-echo, and from laser interferometric
detection using a spot size of approximately 50 pm in throughtransmission.
measurements, directly comparing to data from the immersion experiment (0.5"-aperture
transmitter / receiver) with the laser experiment (approximately 50 jm- spot size). The
error bars represent the scatter in the data points collected, derived from + one standard
deviation
from the mean.
The
laser measurements
consistently
show
a reduction
in
attenuation across the range of frequencies examined for the axial direction, with only
slight overlap in the scatter. In contrast, the laser detected attenuation results from the
radial direction are virtually indistinguishable from the immersion measured attenuation
results, with
extensive
data overlapping.
Clearly,
phase
perturbation
accounts
for a
significant part of the attenuation in the axial direction and has no measurable effect in
the radial
direction
(see
also
Table
5.5
- Phase-Sensitive
Versus
Phase-Insensitive
Attenuation, Appendix A).
5.5 Summary
The influence of crystallographic texture and the resulting phase perturbation on
attenuation was experimentally investigated for the axial and transverse directions of a
mill annealed Ti-6AI-4V bar. The orientation dependence of longitudinal and shear wave
velocities indicate the material is highly textured and this was also verified with analytical
x-ray
results.
While
the
structural
features
including
grains,
grain boundaries,
and
macroscopic grain colonies all tend to be aligned in the axial direction of the bar, the
principal direction of the hexagonal crystallographic lattice (i.e. the basal normal) for the
grains preferentially lie in the transverse plane of the bar. The phase mapping results
revealed approximately 30% higher scatter, measured over the center fringe in the axial
direction, relative to the transverse direction. The higher degree of phase scattering in the
axial direction of the bar is due to the presence of elongated macroscopic inhomogeneities
aligned in the same direction. These extended colonies of similarly oriented grains in the
axial direction are thought to essentially behave
as single crystals, resulting in local
disturbances in the arrival time of the propagating wavefront that extend beyond the finite
aperture of the transmitter. While there also exists phase scatter in the radial direction, it
is significantly smaller and more uniform than that of the axial direction. Measurements
also revealed
substantially lower
attenuation values
for the laser experiment
due its
relative phase-insensitivity, provided by the microscopic footprint of the focused laser
detector.
While
laser
detection
measurements
demonstrate
immersion results, the attenuation is still clearly dominant
lower
attenuation
than
the
in the axial direction, as
compared with the transverse direction. These measurements demonstrate that the wider
apertures generally used by conventional transducers clearly suffer phase cancellation
losses, resulting in significantly higher attenuation results. The attenuation in this material
is unusual considering that the backscatter is 2 to 3 times stronger in the radial direction
than in the axial one; and the forward scatter is at least as strong, although less divergent,
than in the axial direction. This is because elongated grains and grain colonies present
larger total scattering cross-section (larger effective size) along their axis compared to
normal to them, but backscattering and forward scattering coefficients are actually higher
in the normal direction. Therefore, despite the stronger backscatter in the radial direction,
the axial direction has higher attenuation. Finally, this method
development made
possible to image the microstructure by using the laser detection technology.
135
it
Chapter 5, References:
1. J. C. Williams and E. A. Starke, Jr., "The Role of Thermomechanical Processing in
Tailoring the Properties of Aluminum and Titanium Alloys," Deformation, Processing,
and Structure, G. Krauss, ed. (ASM International, 1984) pp. 306-314.
2. A. Juva and M. Haarvisto, "On the Effects of Microstructure on the Attenuation of
Ultrasonic Waves in Austenitic Stainless Steels," The British Journal of Nondestructive
Testing, 19(6), pp. 293-297 (1977).
3. B. L. Baikie, A. R. Wagg, M. J. Whittle, and D. Yapp, "Ultrasonic Inspection of
Austenitic Welds,” Journal of the British Nuclear Energy Society, 15(3), pp. 257-261
(1976).
4. P. D. Panetta, F. J. Margetan, I. Yalda, and R. B. Thompson, "Ultrasonic Attenuation
Measurements in Jet-engine Titanium Alloys," Review of Progress in Quantitative
Nondestructive Evaluation, Vol. 15B, Thompson and Chimenti, eds., (Plenum Press, NY,
1996) pp. 1525-1532.
5. P. D. Panetta, F. J. Margetan, I. Yalda, and R. B. Thompson, "Observation and
Interpretation of Microstructurally Induced Fluctuations of Back-surface Signals and
Ultrasonic Attenuation in Titanium Alloys," Review of Progress in Quantitative
Nondestructive Evaluation, Vol. 16B, Eds., D. O. Thompson and D. E. Chimenti,
(Plenum Press, NY, 1997) pp. 1547-1554.
6. T. Seldis and C. Pecorari, "Scattering-induced Attenuation of an Ultrasonic Beam in
Austenitic Steel," Journal of the Acoustical Society of America (submitted for
publication).
7.M. R. Hollard and J. G. Miller, "Phase-insensitive and Phase-sensitive Quantitative
Imaging of Scattered Ultrasound Using a Two-dimensional Pseudo-array," Ultrasonics
Symp. Proc. IEEE Cat. No. 88CH2578-3, pp. 815-819 (1988).
8. L. J. Busse and J. G. Miller, "Response Characteristics of a Finite Aperture, Phase
Insensitive Ultrasonic Receiver Based Upon the Acoustoelectric Effect," Journal of the
Acoustical Society of America, 70(5), pp. 1370-1376 (1981).
9. L. J. Busse and J. G. Miller, "Detection of Spatially Nonuniform Ultrasonic Radiation
with Phase-sensitive (piezoelectric) and Phase-insensitive (acoustoelectric) Receivers,"
Journal of the Acoustical Society of America, 70(5), pp. 1377-1386 (1981).
10. D. Liu and R. C. Waag, "Estimation and Correction of Ultrasonic Wavefront
Distortion Using Pulse-echo Data," Ultrasonics Symp. Proc. EEE Cat. No. 0-7803-36151/96, pp. 1391-1394 (1996).
11. J. Lwin and W. D. O'Brien, Jr., "Tissue-induced Ultrasonic Wavefront Distortion,”
Ultrasonics Symp. Proc. YEEE Cat. No. 0-7803-3615-1/96, pp. 1415-1418 (1996).
12. J. P. Monchalin, "Optical Detection of Ultrasound," JEEE Transactions on
Ultrasonics, Ferroelectrics, and Frequency Control, 33, pp. 485-499 (1986).
13. S. Hirsekorn and W. Amold, "High Resolution Materials Characterization by
Conventional Near-field Acoustic Microscopy," Ultrasonics, 36, pp. 491-498, 1998.
14. B. T. Khuri-Yakub, S. Akamine, B. Hadimoglu, H. Yamada, and C. F. Quate, "Nearfield Acoustic Microscopy," Scanning Microscopy Instrumentation (SPIE, Bellingham,
1992) No. 1556, pp. 30-39.
15. J. P. Monchalin, R. Heon, P. Bouchard, and C. Padioleau, "Broadband Optical
Detection of Ultrasound by Sideband Stripping with a Confocal Fabry-Perot," Applied
Physics Letters, 55(16) pp. 1612-1614 (1989).
16. W. Hassan and P. B. Nagy, "Experimental Investigation of the Grain Noise in
Interferometric Detection of Ultrasonic Waves," Journal of NDE (submitted for
publication).
17. M. Hercher, "The Spherical Mirror Fabry- -Perot Interferometer," Applied Optics, 7(5),
pp. 951-966 (1968).
18. ASTM Standard E 494-75, “Standard Practice for Measuring Ultrasonic Velocity in
Materials,” ASTM (1985).
19. E. P. Papadakis, “ The Measurement of Ultrasonic Velocity,” Ultrasonic
Measurement Methods, Physical Acoustics , Vol XIX, Eds. R. N. Thurston and A. D.
Pierce, (Academic Press, New York, 1990) p. 91.
20. P. C. Waterman and L. J. Teuntonico, "Ultrasonic Double Refraction in Single
Crystals," Journal of Applied Physics, 28(2), pp. 266-270 (1957).
21. R. Truell, L. J. Teuntonico, and P. W. Levy, "Detection of Directional Neutron
Damage in Silicon by Means of Ultrasonic Double Refraction Measurements," Physical
Review, 105(6), pp. 1723-1729 (1957).
22. K. Goebbels, "Structure Analysis by Scattered Ultrasonic Radiation," Research
Techniques in Nondestructive Testing, Sharpe, ed. (Academic, New York, 1980) Vol. IV,
pp. 87-157.
23. H. Willems and K. Goebbels, "Characterization of Microstructure by Backscattered
Ultrasonic Waves," Metal Science, 15, pp. 549-553 (1981).
24. C. B. Guo, P. Hiller, and K. Goebbels, "Scattering of Ultrasonic Waves in
Anisotropic Polycrystalline Metals," Acoustica, 59(2), pp. 112- 120 (1985).
25. K. Goebbels, "Evaluation of the Structure of Steels by Ultrasonic Scattering,"
Materials Testing. 77(7), p. 231-233 (1975).
26. H. W. Koglenik and T. Li, "Laser Beams and Resonators," Applied Optics, 5, pp.
1550-1555 (1966).
CHAPTER VI
DEVELOPMENT OF AN EDDY CURRENT MATERIALS CHARACTERIZATION
METHOD FOR TITANIUM ALLOYS
6.1 Introduction
This chapter discusses the role electrical anisotropy plays in the structural integrity
of polycrystalline
assessment
titanium
alloys
from
the
standpoint
of fatigue
crack
detection and the related issue of microstructural noise. In eddy current inspection of
noncubic
crystallographic
classes of polycrystalline
metals the electric anisotropy
of
individual grains produces an inherent microstructural variation or noise that is very
similar to the well-known acoustic noise produced by the elastic anisotropy of both cubic
and noncubic materials in ultrasonic characterization.
The
presented
results demonstrate
the electrical
that although
grain noise
is
detrimental in eddy current nondestructive testing for small flaws, it can be also exploited
for characterization of the microstructure in noncubic polycrystalline materials such as
titanium
alloys
in
the
same
way
acoustic
grain
noise
characterization of the microstructure in different materials.
138
is
used
for
ultrasonic
Elastic anisotropy of single crystals plays an important role in ultrasonic materials
characterization of polycrystalline materials. Microscopically homogeneous but randomly
oriented
individual
medium
which
grains
produces
make
up
incoherent
isotropic
but inhomogeneous
scattering commonly
called "grain noise."
a macroscopically
wave
While acoustic grain noise has an obvious adverse, often prohibitive, effect on ultrasonic
flaw detection, [1, 2] it can be also exploited for ultrasonic characterization of the grain
structure [3, 4, 5, 6]. Electric anisotropy exhibited by specific types of crystallographic
classes can play a very similar role in electromagnetic testing of polycrystalline metals.
All physical properties relating two first-order tensor quantities are characterized by
second-order tensors, the directivity of which can be represented by a symmetric ellipsoid
(7, 8]. Such properties include electrical and thermal conductivity, thermoelectricity, diaand paramagnetism, and dielectricity. In the most common
degenerates
into a sphere
and these properties
become
cubic system, the ellipsoid
fully isotropic.
However,
in
noncubic materials the same physical properties are inherently anisotropic. In contrast,
elastic material properties relate two second-order tensor quantities they are characterized
by fourth-order tensors [9]. As a result, from an elastic point of view, cubic crystals are
also anisotropic just like other crystallographic classes.
In nondestructive
materials
characterization
electrical
conductivity
is usually
measured by the non-contacting eddy current method. Neighbor was the first to extend
the eddy current method to electrically anisotropic materials and showed theoretically that
one can obtain the full conductance tensor from such measurements. Special eddy current
coil configurations that allow the simultaneous measurement of electrical conductivity in
two principal directions have been developed for texture assessment in plates [10, 11].
Just like in the case of elastic anisotropy, the source of electrical anisotropy can be either
(i) intrinsic crystallographic anisotropy in single crystals and textured polycrystals or (ii)
structural anisotropy caused by oriented reinforcement in composite materials. The latter
can be exploited for eddy current assessment of constituent volume fractions in metal
matrix composites [12, 13]. Grain boundary contributions to the electrical resistivity [14]
can cause
additional
electrical anisotropy
materials with elongated
in polycrystalline
grains aligned in preferred orientation due to thermal or mechanical treatment.
It should be emphasized
that, in contrast with elastic properties, the electric
conductivity is completely isotropic in cubic crystals which constitute the overwhelming
majority
of polycrystalline
therefore,
metals;
the
role
of intrinsic
crystallographic
anisotropy in eddy current testing has not been investigated in detail. However,
less
common materials of hexagonal symmetry can exhibit strong electrical anisotropy with
significant difference in conductivity between the basal plane and normal to it. Titanium
is one of the few structural metals
applications,
which
preferentially
of practical
crystallizes
importance,
in hexagonal
especially in aerospace
symmetry
and
therefore
exhibits strong electrical anisotropy. There are two areas where this electrical anisotropy
becomes very relevant from the point of view of mechanical fatigue in titanium alloys.
First, eddy current inspection is probably the most commonly used nondestructive testing
technique for fatigue crack detection in airframe structures and engine components and
electrical grain noise presents the same
problem
in eddy
current crack detection
as
acoustic grain noise does in ultrasonic flaw detection. Second, proper microstructure is
140
absolutely essential for assuring good
grain
electrical
noise
exploited
be
can
fatigue tolerance in the material
[15] and the
of the
characterization
for nondestructive
microstructure by eddy current inspection in the same way as acoustic grain noise is used
in ultrasonic characterization of the microstructure.
Another goal of this chapter is to investigate the feasibility of exploiting the
eddy
unique
current
grain
noise
the
for
alloys
in titanium
observed
purposes
of
nondestructive materials characterization. The main achievements of this effort are (i) the
strong
electrical
grain
demonstrated,
(ii) the
physical
very
noise
in
titanium
alloys
for
responsible
mechanism
been
has
this
experimentally
contrast
been
has
theoretically explained, (iii) analytical, finite element, and experimental methods were
used
to
investigate
lateral
that assure
characterization
were
the best
developed
eddy
of
resolution
procedures
optimization
materials
the
imaging
and
current
microscopy,
resolution
verified.
It is
and
(iv)
for microstructural
shown
that
electric
anisotropy exhibited by noncubic crystallographic classes of materials can play a very
similar role in electromagnetic materials characterization of polycrystalline metals to that
of elastic anisotropy in ultrasonic materials characterization. Titanium is one of the few
~ structural metals of practical importance,
especially in aerospace
applications, which
preferentially crystallizes in hexagonal symmetry and therefore exhibits strong electrical
anisotropy. At the same time, the titanium alloy microstructures of interest tend to form a
rather coarse, locally textured microstructure featuring large colonies of hexagonal alpha
grains of similar orientation. The fracture and fatigue resistance of this material is
strongly affected by the type of microstructure, hence there is a continued need for new
141
nondestructive evaluation techniques that are capable of both imaging and quantitatively
characterizing microstructure. It was found that the lateral resolution of eddy current
imaging is ultimately limited by the probe-coil geometry and dimensions, but both the
inspection frequency and the phase angle can be used to optimize the resolution, to some
degree, at the expense of sensitivity. Although eddy current imaging is still in its infancy,
a direct comparison of 5-MHz eddy current and 40-MHz acoustic microscopic images of
the
same
coarse-grained
Ti-6Al-4V
indicated that the same
sample
features
can be
observed by both methods at approximately the same resolution level. This work also
shows
experimentally
that
eddy
microscopy
current
can
be
enhanced
via
a high-
resolution, small diameter probe-coil which delivers a unique materials characterization
tool well suited for the evaluation of Ti alloys.
Eddy
current
imaging
has
been
used
for years
in flaw
detection,
corrosion
mapping, and other nondestructive testing and materials characterization applications to
increase the amount of information obtained by conventional point-by-point inspection
and enhance its clarity [16, 17, 18, 19, 20, 21, 22, 23]. Another emerging technology
capable of mapping electrical conductivity distributions at considerably better resolution
though with inherently less sensitivity is near-field microwave imaging. Ash and Nichols
were probably the first to use near-field scanning for electromagnetic imaging in 1972
[24]. Using 3-cm microwave radiation, they achieved a resolution of
direction and
2/20
for two-dimensional
4/60
in one
objects. Since the early use of microwave
radiation, several techniques have been developed for near-field imaging. Among
the
various versions developed during the last few years, the transmission-line resonator
technique appears to be the most promising for high-resolution near-field microwave
inspection of surface conductivity [25, 26, 27]. Eddy current inspection can be readily
adapted for imaging via automated scanning since it is noncontacting in nature and less
sensitive to surface topography than near-field microwave
imaging.
One
of the most
important advantages of eddy current imaging over ultrasonic imaging is that there is no
grain noise in cubic materials which constitute the overwhelming majority of structural
However,
metals.
electrical
that noncubic
it was
recently
found
that
presents
a serious
anisotropy
limitation
materials
for
flaw
exhibit
substantial
detection
[28].
In
particular, in many titanium alloys microstructure scatter generates noise that can be very
severe since often large colonies of similarly oriented alpha grains form a coarse, textured
structure that effectively hides small defects.
As an example, Figure 6.1 shows the eddy current images of small fatigue cracks
of approximately 0.025 inch in length in 2024 aluminum and Ti-6Al-4V specimens at 2
MHz. A 0.060"-diameter coil was used to scan a 0.5" x 0.5" area over the surface of the
specimens. The crack is somewhat smaller than the inside diameter of the coil, therefore
it produces a characteristic double-image. Two dark spots appear at the two opposite ends
of the crack when it intersects the path of the eddy currents. The diameter of the dark
spots is approximately 0.040" and is determined by the size of the probe as discussed
later. It is quite obvious from these results that even this less than fully resolved flaw is
readily detected in aluminum
since the background is quite uniform, i.e., there is no
microstructural noise. In comparison, a similar crack is barely detectable in titanium
because of the presence of strong grain noise caused by the unique electrical anisotropy of
Ti-6Al-4V, 0.026-in.-crack
Al 2024, 0.025-in. crack
Eddy current images of small fatigue cracks in 2024 aluminum and Ti6AI-4V specimens (0.5" x 0.5", 2 MHz, 0.060"-diameter coil).
Figure 6.1
hexagonal crystallites. The physical origin of this unique crystallographic noise and its
possible role in quantitative analysis of the grain structure in titanium alloys was recently
shown using single crystals of cubic Al and Cu, and hexagonal
Cd
[28]. This letter
experimentally demonstrates the dependence of electrical conductivity on the relationship
between
the
surface
normal
direction
and
the
principle
direction
in
noncubic
crystallographic classes of materials.
Based on our current understanding, we know (i) that in coarse-grained Ti-6Al4V, the detectability of small fatigue cracks by eddy current inspection is ultimately
limited by grain noise and (ii) that this grain noise can be exploited for the purpose of
characterizing the microstructure by high-resolution eddy current imaging. Because of the
small penetration depth of eddy current inspection, only the grain structure at the surface
can be inspected. The electrical anisotropy of noncubic materials like titanium and its
144
alloys can be also exploited to assess texture
at the surface
and, by direct contact
resistivity measurements, even in the interior of the material.
The grain noise observed in eddy current images is due to the variation of the
average conductivity between different crystallographic planes. In spite of the difference
in the origin of contrast, namely electrical versus mechanical properties of the material,
the eddy current contrast mechanism is quite similar to the contrast produced in spherical
acoustic microscopy
and therefore
[29]. Both techniques are sensitive to crystallographic orientation
produce
images
on which
large colonies
orientation show up as essentially homogeneous
of smaller
grains
of similar
domains. This occurs in spite of the
partitioned appearance of individual grains of a macroscopic colony as featured in an
optical micrograph of metallurgically prepared samples which, due to chemical etching
(as in Fig.
6.2.a),
produces
a contrast
regardless
whether
there
is a difference
in
orientation between neighboring features or not. Figure 6.2 illustrates that essentially the
same macroscopic inhomogeneity of the microstructure can be observed in coarse-grained
polycrystalline Ti-6Al-4V
via eddy current imaging
and acoustic microscopy.
These
1"x 1" images were scanned from a specially heat treated sample to bring about a high
degree of grain consolidation to the structure. The large lamellar colonies are composed
of alternating plates of the alpha and beta phases. The principal direction in each of these
large colonies is essentially uniform, thereby forcing the colony to behave as if it were a
single crystal. The main reason for generating a sample with such exaggerated structure is
to determine if like features could be observed with both eddy current and acoustic
scanning techniques. Clearly, these like features do simultaneously appear in the images
145
|
@
f%
Figure
oie
fs
c) 40 MH zZ acoustic m. icrog raph
Comparison of (a) optical, (b) eddy current, and (c) acoustic microscopic
images of a coarse-grained Ti-6Al-4V sample (1" x 1") from nearly the
same area on the sample.
Figure 6.2
of
b) 5 MHz eddy current
)
/
a) optical image
ese
due
6.2,
to
the
relative
crystallographic
orientation
the
of
various
microstructural features encountered during scanning.
6.2 Eddy Current Experiments
Single crystal materials generally behave anisotropically in response to a given
stimulus such as heat, electricity, or force due the symmetry conditions of the atomic
lattice
structure.
In
contrast,
polycrystalline
materials
tend
to
behave
essentially
effect associated with the presence
of a random
crystallographic orientation distribution of the individually homogeneous
grains which
isotropically due to the averaging
constitute the solid. This paper demonstrates that eddy current evaluation is capable of
resolving the crystallographically related directional dependence of electrical conductivity
in noncubic polycrystalline metals. In coarse-grained structural alloys of polycrystalline
titanium
the
spatial
variation
in
electrical
conductivity
significantly
reduces
flaw
detectability. On the other hand, the physical mechanism responsible for reduced flaw
146
delectability provides a source of information about the microstructural make-up of the
material, and therefore can be used for materials characterization.
Eddy
current
testing
is
the
most
common
electromagnetic
nondestructive
evaluation method and is widely used in the aerospace industry. Small diameter coils
combined with a computer controlled scanning mechanism can be readily used for eddy
current imaging. The coil impedance is determined by the resistivity of the specimen as
measured by the eddy current, which runs parallel to the surface in a concentric circle
with the coil. In this way, an eddy current probe measures the average resistivity in a
given plane rather than in a given direction. As the probe is moved along the surface, it
measures the local average resistivity along the path of the eddy current in the plane of
the surface. The resistivity is integrated over the entire probe circumference in the eddy
current path, resulting in grain contrast that is proportional to the average resistivity
between the different crystallographic planes. This contrast is similar to the mechanical
contrast produced by spherical acoustic microscopy, which is determined by the variation
of the average surface wave velocity between different crystallographic planes [29].
For a hexagonal crystal like pure titanium and its most common alloys the axial
symmetry
around the principal
direction (the hexagonal
axis) allows
the directional
dependence of the electrical resistivity to be described over the entire space by two
orthogonal axes and the directivity can be represented as an ellipsoid:
(6.1)
pg) = p,cos’¢ +p sin’¢,
147
in cubic materials the electrical resistivity is fully isotropic due to the balanced symmetry
of the lattice structure, i.e., the resistivity becomes a single scalar value and the ellipsoid
describing
measurements
eddy
current
to
a sphere.
For
of electrical resistivity in a hexagonally
symmetric
single crystal, the
its directional
dependence
degenerates
average surface resistivity can be expressed from Eq. (6.1) as:
pO) = 1/2[p, sin’ 0+ p (it cos’),
where
0
denotes the inclination angle between the basal plane and the surface of the
specimen. For example, in pure titanium
the resistivity is approximately 6%
P 1 = 48 pQcem
and
p= 45.35 wOcm,
lower in the basal plane than normal
1..,
to it (30).
Because of the above described averaging effect of eddy current inspection, the actual
grain contrast is expected to be 50% lower in eddy current inspection. In titanium, the
average resistivity is approximately 3%
lower when the basal plane is parallel to the
surface than when it is normal to it.
In order to assess the feasibility of eddy current materials characterization and
flaw detection in structural alloys of noncubic symmetry, we carried out two sets of
experiments. First, we used an eddy current probe to measure the directional variation of
the electrical conductivity in pure single crystals of aluminum, copper, and cadmium; the
former two materials consist of a cubically symmetric crystallographic lattice, the latter
one consists of a hexagonally symmetric lattice (unfortunately, titanium single crystals
be
cannot
grown
to
sizes
enough
large
eddy
accurate
for
current
conductivity
measurements). Second, we used an eddy current scanner to map the electrical grain noise
in Ti-6Al-4V titanium alloy specimens of different microstructures.
The Al, Cu, and Cd single crystals used in this study were of random orientation.
Each specimen was a solid cylinder of approximately 2" length and 0.5" diameter, large
enough to section into multiple test samples of varying surface orientation. Eddy current
resistivity measurements were taken on the various single crystal sample sets using a
Nortec 19e eddy current instrument and a 0.060"-diameter probe at 2 MHz. If the eddy
current probe is energized in air, away from the conducting sample, the instrument will
register the baseline
impedance
of the
coil.
As
the probe
is moved
closer
to the
conductive sample, the impedance indication will begin to change and becomes strongest
when the probe is directly on the sample. The variation in the eddy current indication
related to the spacing between the coil and the sample is known as "lift-off." When the
probe is in contact with the sample, the coil impedance
indicated on the instrument
represents the electrical resistivity. Moreover, the fact that the lift-off curve approaches
the resistivity curve at an angle allows the separation of lift-off from resistivity by proper
adjustment of the phase angle on the instrument [31]. For each set of samples, the phase
angle was set to isolate lift-off to the horizontal direction, the sensitivity was adjusted,
and the instrument was nulled. The vertical output from the eddy current instrument,
corresponding to the average electrical resistivity, was captured on a digital oscilloscope.
With this automated approach, it was possible to statistically analyze the population of
average
surface
resistivity
values
corresponding
149
to
approximately
500
individual
measurements
(see
sample
for each
Electrical
- Normalized
6.1
Table
Resistivity,
Appendix A).
The measured data are shown in Figures 6.3.a through ¢ as histograms of the
probability distributions of the surface resistivity for various
surface crystallographic
orientations in the three single crystals. For each set, only three surfaces showing the most
extreme differences in average resistivity are displayed. It should be noted that these
values in average
electrical resistivity are subject to a variety of small experimental
an
errors, including thermal drift from the instrument or sample, probe alignment and
associated probe rocking effect, inevitable thickness and edge effects, etc., hence the
variability in the data. These factors were considered during the data collection and
efforts
were
taken
to
minimize
their
The
affects.
data
from
6.3.c
Figure
clearly
of the electrical resistivity in cadmium
demonstrates the crystallographic dependence
representing noncubic materials, as opposed to the lack of separation demonstrated by
cubic copper and aluminum in Figures 6.3.a and 6.3.b. In the cadmium crystal the values
of electrical resistivity are
py = 83 pQcm
and
p=
6.8 wQcm,
a
relatively large
difference of approximately 22% between the basal plane and the normal to it. Due to the
averaging effect, the most extreme resistivity separation which could be expected in Cd
by eddy current measurement
present
in the randomly
is approximately
cut Cd
samples
was
11%. The average resistivity variation
clearly measurable
with
a maximum
variation of approximately 3% in resistivity. Considering that we did not necessarily find
the principal planes of maximum separation, the measured variation is reasonable.
Probability Density
a) Aluminum
Probability Density
b) Copper
2
n
Swo
Qa
2
6
Fe
2
°
a
Probability Densi
=
Normalized Surface Resistivity [%]
d)
Figure 6.3
Electrical resistivity probability distributions for three single crystal
surface orientations in a) aluminum, b) copper, and c) cadmium and d) on
the surface of polycrystalline Ti-6V-4V (solid lines are best fitting
Gaussian distributions).
Because
of our particular
interest
in nondestructive
testing
of high-strength
titanium alloys by eddy current methods, a special attempt was made to obtain the same
type of data from a pure alpha (hexagonal) phase Ti single crystal. However, due to the
inherently small size of the available Ti single crystals, it was not possible to collect data
actually representative of the material's electrical resistivity due to edge affects, which
tend to diminish the accuracy of the measurements. Moreover, in titanium, the maximum
difference in average resistivity is expected to be only about 3%, i.e., only one fourth of
the corresponding variation in Cd. Nevertheless, based on the results from the Cd crystal
sample, the evidence of electrical anisotropy in noncubic crystalline materials is clearly
further
supported.
To
distribution
of the
demonstrate
surface
this
point,
Figure
resistivity for a Ti-6Al-4V
6.3.d
shows
polycrystalline
the
probability
specimen.
As
expected, there is a significantly wider variation in the resistivity from point to point than
on single crystals, which will be shown later to be caused by the relatively coarse grain
structure.
Structural alloys of titanium are comprised of microscopically anisotropic grains
of random
order,
which
macroscopically
behave
isotropically.
However,
often
the
materials fabrication process results in both small- and large-scale structures which lack
the degree of randomness in the crystallites’ orientation required to allow the behavior to
be fully isotropic. These materials are said to contain texture, which is generally imparted
to the material via plastic deformation, like forging. Texture results, for example, in the
alignment of like crystallographic slip planes parallel to the rolling plane, while certain
slip directions tend to align in the direction of rolling or wire drawing. The development
of this preferred orientation also tends to align microstructural features like inclusions,
second phase particles, or grain boundaries and the texture affects are often observed on a
large scale relative to the individually homogeneous grains, often spanning several inches
or more. In some polycrystalline titanium alloys, certain microstructural conditions give
rise to a highly localized form of crystallographic microtexture
causing
fractures to
preferentially occur along certain weak erystallographic directions [15].
The macroscopic inhomogeneity of the microstructure in polycrystalline Ti-6AlIAV can be observed via eddy current imaging as shown in Figure 6.4, which correspond
1" area on the samples. The specimen shown in Figure 6.4.a contains a gross
to 1" x
microstructural anomaly clearly visible on the right side of the eddy current image. Figure
6.4.b shows a typical billet microstructure with texture related features in the horizontal
direction while Figure 6.4.c shows a large grained sample and Figure 6.4.d shows an
equiaxed
annealed
beta
microstructure.
In short,
some
interesting
parallels
can
be
observed between the reported electromagnetic approach and conventional ultrasonic
evaluation methods.
can be used to exploit the fact that in
Ultrasonic techniques
polycrystalline materials, grain to grain differences in crystallographic orientation and the
presence of grain boundaries provide a source for scattering of ultrasonic energy. The
presence of texture and additional phases of material also play an important role in the
ultrasonic response of the material and the scatter provides a source of data which can be
used to characterize the microstructural features. In ultrasonic flaw detection, the acoustic
grain
noise
is
clearly
detrimental
due
to
reduced
detection
threshold.
Likewise,
electromagnetic inspection techniques benefit from the fact that noncubic systems exhibit
d)
Scanned eddy current images of different Ti-6Al-4V microstructures; a)
sample containing a severe microstructure anomaly (right side, middle); b)
the billet microstructure showing texture related features in the horizontal
direction; c) a large grained sample; and d) equiaxed beta annealed
microstructure (dimension 1" x 1").
electrically
anisotropic
properties,
characterization,
for microstructural
allowing
and
suffer from the fact that the electrical scatter originating from varying local resistivity
raises the noise
floor, thereby
reducing
The
detectability.
flaw
electrical
anisotropy
observed with eddy currents in noncubic metals is therefore analogous to the elastic
anisotropy
observed
nondestructive
with ultrasonic
of polycrystalline
evaluation
anisotropy
of noncubic
knowledge,
the
techniques
crystals
significant
role
titanium
known
is a well
played
and has
by the
strong
alloys.
physical
microscopic
implications
for the
the
electrical
Although
fact, to the best of our
electrical
anisotropy
of
individual grains in the macroscopic eddy current response of the polycrystalline material
has never been pointed out or investigated in any depth.
6.3 Resolution of Eddy Current Imaging
In contrast to other more
conventional
applications
of eddy current imaging,
microstructure characterization crucially depends on achieving truly microscopic imaging
resolutions. The lack of published theoretical and experimental investigations on the
factors affecting the lateral resolution of eddy current inspection prompted us to initiate
this research
effort aimed
at laying
down
the
groundwork
for developing
a high-
resolution eddy current microscope capable of resolving the fine details of the textured
microstructure in titanium alloys. Numerous efforts have been made in the past to study
the eddy current distribution of probe-coils by analytical, numerical, and experimental
means. In spite of the general nature of these methods, most of the published results were
focused mainly on the axial penetration depth of the probe, which is undoubtedly the
primary consideration in the overwhelming majority of NDT
155
applications. Much less is
known about the radial penetration depth or lateral resolution of eddy current inspection
except that it is essentially governed by the geometry and dimensions of the probe-coil. It
is usually assumed that the eddy currents induced in the material are essentially mirroring
the excitation
in the probe-coil,
current running
therefore
the
lateral
extent
of the
inspected region under the coil is more or less independent of frequency. However, there
is clearly a radial spreading of the eddy current distribution as the frequency is lowered,
which will adversely affect the lateral resolution of the eddy current inspection by small
coils used in eddy current microscopy. This is because in eddy current microscopy, such
as microstructural characterization of polycrystalline titanium alloys, we are trying to
optimize the inspection parameters for maximum
lateral resolution in the near-surface
region of the specimen, therefore very small probes are used at very high frequencies. By
comparison, in conventional eddy current imaging, such as flaw detection or corrosion
assessment, a relatively large penetration depth is desirable which inevitably requires a
large probe diameter [32, 33].
All the necessary analytical [34, 35] and numerical [36, 37] tools for investigating
the radial penetration depth or lateral resolution of eddy current microscopy are readily
available
in
the
literature.
It
has
been
previously
demonstrated
that
on
simple
axisymmetric configurations these techniques provide essentially identical solutions for
the axial penetration depth of the eddy current [38], as such they are also expected to
work equally well for radial penetration depth estimations. In our finite element (FE)
calculations
we
used
a commercially
available
software
capable
of simulating
the
magnetic field and eddy current density distributions for axisymmetric configurations
156
Figures 6.5 and 6.6 show the magnetic
[39]. As an example,
field and eddy current
distributions, respectively, produced by a small pancake coil in titanium at four different
frequencies (air-core, inside diameter 1 mm, outside diameter 1.5 mm, height 2 mm).
most
The
convenient,
albeit somewhat
over-simplified
measure
of frequency
relative to the conductivity of the material and the size of the probe is the so-called
standard penetration depth [40]
OHO
@
where
(6.3)
:
5 = | 2
is the angular frequency, and
specimen,
of
the
w=
4n10-7Vs/Am
conducting half-space
respectively
and
6
(in
and
our
o
finite
are the permeability and conductivity
element
simulations
we
took
A/ Vm). For a plane wave incident on a
o = 214x 10°
actually gives the 1/e skin depth of the exponentially decaying
eddy current density, but for any finite-sized probe-coil it is only a useful parameter
which
happens
to be the upper limit for the axial penetration depth.
Whenever
the
standard penetration depth is very large with respect to the dimensions of the coil the
magnetic field is essentially unaffected by the flow of eddy currents in the specimen and
can be approximated
by the magnetic
field produced by the coil far away
from the
conducting half-space. In this frequency range (up to about 10 kHz or 6 ~ 3.4mm in our
case) the eddy current distribution is also independent of frequency while its absolute
density proportionally increases with frequency. Whenever the standard penetration depth
157
1mm
<——_>|
Figure 6.5
Magnetic field distribution produced by a small pancake coil in titanium at
four different frequencies.
1 mm
<—_>|
Figure 6.6
~° Eddy current distribution produced by a small pancake coil in titanium at
four different frequencies.
is very small with respect to the dimensions of the coil the magnetic field is essentially
eliminated by the flow of eddy currents in the specimen below a certain skin depth which
approaches the standard penetration depth. In this frequency range (above approximately
1 MHz or & ~ 034mm _ in our case) the eddy current distribution is also limited to this
shallow layer determined by the standard penetration depth. Figure 6.7 shows the axial
penetration depth versus frequency curve for a 1-mm-inside diameter pancake coil in
titanium. The symbols represent the numerical results calculated by finite element (FE)
simulation, the solid line represents the general trend of the FE data, and the dashed line
is the plane wave asymptote calculated from the standard penetration depth according to
Eq. (6.3). The axial penetration depth was calculated from the eddy current intensity
directly under the coil at its middle
dpiddle = (douter + Ginner)/2. This figure well
demonstrates the substantial difference between the true axial penetration depth of a
finite-diameter coil and the standard penetration depth at low frequencies, which was first
pointed out by Mott [32] and later further investigated by Stucky and Lord [38].
Our main interest in this study is a similar investigation of the radial penetration
depth of eddy currents generated by a finite-diameter probe-coil, which can be readily
done by analyzing the same set of FE data*. Figure 6.8 shows the radial penetration
versus
frequency
for a 1-mm-diameter
pancake
coil
in titanium.
The
solid
circles
represent the numerical results calculated by finite element (FE) simulation and the solid
line illustrates the general trend of the FE data. In addition, the empty circles represent‘the
analytical
results
calculated by Dodd
and Deed's
method
[34]
and the dashed
line
illustrates the general trend of the analytical data. The radial penetration depth was
160
measured from the axis of the coil to the point on the surface of the conducting half-space
where the eddy current density dropped to 1/e relative to the maximum, which is directly
under the middle of the coil. In spite of a minor numerical discrepancy between the FE
simulation and the analytical results the agreement is quite acceptable. Both methods
indicate
that
the
penetration
radial
approaches
low-frequency
a
asymptote
of
approximately 1.8 mm. This value is clearly sensitive to the shape of the coil (both insideto-outside diameter and diameter-to-height ratios) but it is roughly equal to the outer coil
10000
standard penetration
depth
—®—
=
finite element
=.
[a
G~~
QO.
oO
A
c
io)
eer
iss}
5
oO
q
oO
[a
is
10
0.00001
v
Lf
LU
q
0.0001.
0.001
0.01
0.1
Frequency [MHz]
Figure 6.7
Axial penetration depth versus frequency for a 1-mm-diameter pancake
coil in titanium. The symbols represent the numerical results calculated by
finite element (FE) simulation, the solid line represents the general trend
of the FE data, and the dashed line is the plane wave asymptote calculated
from the standard penetration depth according to Eq. (6.3).
161
e
veeee $ wen eeeee
finite element
---Q--
analytical
ip
Radial Penetration
[mm]
18
——@®—_
1.2 4
0.8
0.00001
t
T
LJ
T
J
v
0.0001
0.001
0.01
0.1
1
10
100
Frequency [MHz]
Figure 6.8
Radial penetration p versus frequency for a 1-mm-diameter pancake coil
in titanium. The solid circles represent the numerical results calculated by
finite element (FE) simulation, the solid line illustrates the general trend of
the FE data, the empty circles represent the analytical results calculated by
Dodd and Deed's method, and the dashed line illustrates the general trend
of the analytical data.
diameter. At high frequencies, the radial effective outer diameter of the coil is slightly
larger than the outer diameter of the coil, ie., dog > doyter. In order to better illustrate
this relationship, Figure 6.9 shows the corrected radial penetration Peg,
= P — deg 12
versus frequency curve for the same 1-mm-diameter pancake coil in titanium compared to
the standard penetration depth, which is of course independent of the size and shape of
the coil. The effective diameter was found simply by best fitting the high-frequency
162
asymptotic behavior as predicted by the FE results to the standard penetration depth. In
case
considered
here,
turned
diameter
effective
the
out
to
be
the
particular
dey
~ 1.79mm, i.e., approximately 19 % higher than the outer diameter of the coil. This
value is clearly sensitive to the shape of the coil but we can still conclude that over a wide
frequency range the radial penetration can be more or less accurately approximated
=
10000
Stes
‘
)
.
3
q
=
1000
>
100
-
—@®—
standard penetration
depth
finite element
e
_@
s
©
5
Ay
“s
3fa
a
2
3)
oO
q
fo)
O
10
0.00001
T
+
T
J
LU
v
0.0001
0.001
0.01
0.1
]
10
100
Frequency [MHz]
Figure 6.9
Corrected radial penetration
Pcorr = P — Feg /2
(dog = 119 douter)
versus frequency for a 1-mm-diameter pancake coil in titanium. The solid
circles represent the numerical results calculated by FE simulation, the
solid line illustrates the general trend of the FE data, and the dashed line is
the standard penetration depth calculated from Eq. (6.3).
163
as the sum of the standard penetration depth, which depends on the material properties
and frequency, and the effective coil diameter, which depends on the size and, to some
degree, on the shape of the coil.
6.4 Results and Discussion
The
now
until
results
indicate
in eddy
that
current
the
microscopy
lateral
resolution is essentially determined by the probe diameter although it slightly improves
with
frequency.
increasing
commercially
These
predictions
Nortec
3-mm-diameter
available
were
experimentally
pencil-probe
of
tested by using
500
kHz-1
a
MHz
frequency range. The nominal diameter actually indicates the case only and the inner and
2.5
outer diameters of the small coil were measured to be approximately 1.25 mm and
the
mm, respectively. The height of the coil could not be established without damaging
probe. It was also verified that the coil had a ferrite core, but according to our finite
element
simulations
such
a core
has
only a
relatively
weak
effect
on
the
lateral
distribution of the eddy current in the specimen. In order to measure the lateral resolution
of this probe we first used a polycrystalline Ti-6Al-4V specimen of very coarse grain
structure, zoomed in on the interface between the two largest neighboring colonies, and
the
determined the 10% to 90% width of the transition range by analyzing the contrast of
eddy current micrograph. However, it was found that since the grain boundary is not
necessarily normal to the surface of the specimen the measured lateral resolution is also
affected by the axial penetration depth. In order to eliminate this artifact, we prepared a
special specimen made of two flat and well polished Ti-6A1-4V halves of very fine grain
structure, which
were
tightly compressed
and held together by a bolt. The
164
reduced
electrical conductivity of the resulting imperfect interface between the halves produced a
dark or bright stripe on the eddy current micrograph depending on the rotation angle used.
The width of this stripe was used as a measure of the lateral resolution of the eddy current
probe
at the given inspection frequency.
Since the contrast profile was
found to be
approximately Gaussian in most cases, the accuracy and repeatability of the measurement
was
further increased
by first approximating
the measured
profile with a normal
distribution and then defining the lateral resolution as the standard deviation (1/e halfwidth) of the best fitting Gaussian curve. The measurement was carried out over a very
wide frequency range from
15 kHz to 10 MHz
far in excess of the 500 kHz-1
MHz
nominal frequency range of the probe. In order to facilitate reliable measurements at both
very low and very high frequencies where the sensitivity of the probe was marginal at
did not use the Nortec
best we
investigation.
19e eddy
current scope
Instead, we used a low-impedance
as we
did in our previous
(5 (2) bridge and an SR
530 low-
frequency analog lock-in amplifier up to 100 kHz and a high-impedance (50 Q) bridge
and an SR 844 high-frequency digital lock-in amplifier above 100 kHz.
As we will demonstrate later, the lateral resolution is also affected by the phase
angle of the detection (rotation angle of the impedance plane) therefore we had to assure
that
this
magnitude.
angle
remains
consistent
as the
frequency
changes
over
three
orders
of
One possibility is to use a changing reference angle, for example, always
adjust first the lift-off to be horizontal and then measure the vertical component, as we
did during imaging experiments. Figure 6.10 shows the experimental impedance diagram
at 2 MHz
in titanium. The rotation angle (-156° reference phase angle on the lock-in
amplifier) was chosen so that the lift-off curve is horizontal and the instrument was
nulled when the probe was above the titanium specimen but far away from the interface.
As the probe approached the interface (without changing the lift-off distance from the
surface of the specimen) the impedance point moved downward and to the right, which
corresponds to an apparent decrease of conductivity and increase of lift-off distance. The
of measuring only the vertical change
advantage
at horizontal lift-off is that surface
alignment, curvature, and topography will not affect the contrast. However, this mode of
operation would make the change of frequency rather cumbersome therefore we used a
similar, but much simpler approach which is not available on commercial eddy current
scopes but can be done very simply when a lock-in amplifier is used. We applied phaseinsensitive absolute value measurement to quantify the imbalance of the bridge. Since the
differential output signal of the bridge is mainly due to the presence of eddy currents in
the conductive half-space (lift-off curve) and is only weakly perturbed by the imperfect
interface (interface curve), measuring the magnitude of the differential output signal of
the bridge is essentially the same as measuring the component of the interface curve that
is parallel to the lift-off curve without repeated re-adjustment of the reference phase.
inspection
experimentally
6.11
shows
the
frequency
curve
for the previously
Figure
determined
described
lateral
commercial
resolution
versus
pencil-probe
in
titanium (see Table 6.2 — Frequency Dependent Lateral Resolution, Appendix A). The
solid line represents only the trend of the theoretical predictions based on our previously
shown
finite
element
simulations.
Direct
comparison
of these
predictions
and
the
experimental data is not feasible partly because of the lack of accuracy in the geometrical
166
50
n
fan)
Vertical Amplitude [mV]
Oo
j
* lift-off
interface
“100 4eeeee-eeee feceeeeeeeeee leveeeeeeeeees a
ee
-150
100
50
0
-50
150
Horizontal Amplitude [mV]
Figure 6.10
| Experimental impedance diagram in titanium at 2 MHz. The rotation angle
was chosen so that the lift-off curve is horizontal.
dimensions and material properties (core) of the probe-coil used in our calculations but
mainly because of the conceptual differences in how the radial penetration was calculated
and
the
lateral
resolution
was
measured.
The
calculations
require
an axisymmetric
geometry therefore the flat boundary or interface cannot be incorporated into the model
therefore the calculated radial penetration is not expected to be quantitatively identical to
the experimentally measured lateral resolution. However, both parameters are related to
the lateral spread of the eddy current distribution in the conducting specimen therefore
167
1800
-
1600
_
=
1400
——
adjusted FE prediction
@
low-frequency lock-in
O
high-frequency lock-in
Ei 1200 4
i=
-=
=
9
%
1000
‘S5
600 4
2%
-
800 4
3
400
-
200
-
—
0
10
1
0.1
0.01
Frequency [MHz]
Figure 6.11
| Experimentally determined lateral resolution versus inspection frequency
by a commercial pencil-probe in titanium. The adjusted FE prediction for
the radial penetration is plotted to indicate the trend of the data.
they are expected to exhibit very similar frequency dependence. In order to bring out this
similarity, the calculated radial penetration was shifted upward by an arbitrary amount of
180
pm.
With
this
adjustment,
the
qualitative
agreement
between
the
theoretical
predictions and the experimental data is quite good except maybe at very high frequencies
where the experimentally determined lateral resolution does not follow the still dropping
radial penetration. Over all, the effect of inspection frequency on the lateral resolution is
rather modest; over the studied three orders of magnitudes the resolution changes barely
by a factor of two, which confirms our previous analytical conclusion that the principal
168
parameter is the diameter of the probe-coil with only minor improvement produced by
increasing the frequency.
The effect of frequency on lateral resolution is demonstrated in Figure 6.12. In
a
this case we used a commercially available Uniwest 1180-b eddy current probe with
coil diameter of approximately 0.020" and an extended, sharpened ferrite core. This
to
probe, while having a slightly reduced frequency range and sensitivity in comparison
the 0.060" Nortec probe, has the obvious advantage of a smaller coil diameter, thereby
providing better lateral scanning resolution.
The rather small improvement produced by increasing the inspection frequency
prompted us to look at the other parameter that can be most easily adjusted to optimize
the lateral resolution of eddy current microscopy, namely the rotation angle. Assuming
a) 300 kHz
Figure 6.12
b) 1 MHz
c) 3 MHz
Eddy current images (0.5" x 0.5") taken at three different frequencies to
demonstrate the effect on lateral resolution. These images were scanned
from an extremely large grained polycrystalline titanium alloy.
169
that the specimen is flat therefore lift-off effects can be eliminated by careful parallel
alignment of the specimen with respect to the scanning plane, the rotation angle can be
freely chosen to achieve the best combination of resolution and sensitivity. Figure 6.13
shows the impedance diagrams and resolution profiles at three different rotational angles
relative to the horizontal
lift-off angle at 2 MHz
in titanium.
Solid lines show the
measured data and dotted lines are best fitting Gaussian distributions used to approximate
the measured contrast profiles. The thereby obtained standard deviation represents the
experimentally determined lateral resolution. The rotation angle is measured from the
reference phase where the lift-off direction is horizontal. In this case (Figure 6.13.a),
which was used in all of our earlier shown eddy current images because of its complete
cancellation of lift-off effects and was also shown in Figure 6.13, the lateral resolution is
about 790 pm. The apparent curvature of the interface line on the impedance diagram
allows us to further increase the resolution by simply increasing the rotation angle. For
example at 120° (Figure 6.13.b), the rotated impedance locus first moves upwards as the
probe approaches the interface then turns downwards as the probe gets closer to it. As a
result, the lateral resolution as calculated from the Gaussian fit reduces to 510 pm at the
expense
of slightly reduced and somewhat
distorted contrast. The contrast distortion
produces weak bright bands on both sides of the otherwise dark interface region. We will
show
later that even better lateral resolution can be achieved by further rotating the
impedance diagram up to about 140°, but at that point the reduction and distortion in
contrast both become unacceptable for imaging purposes. At even higher rotational angles
the contrast becomes positive as the two bright bands gradually overtake the dark stripe at
the center. In this range the contrast is both low and badly distorted. At 160° (Figure
a) 0° from lift-off, 790 pm standard deviation
Ss
;
ob
:
§
eso
::
3,o
of
iB
;
,
oO
3
NN
>
>
s
:
3
2
'
2
}
Oo
.
oO
>
wl
7
=
'
=
4
6
>
Distance [1 mm/div]
Horizontal Voltage [a. u.]
Vertical Voltage [a. u.]
Vertical Voltage [a. u.]
b) 120° from lift-off, 510 pm standard deviation
Distance [1 mm/div]
Horizontal Voltage [a. u.]
c) 160° from lift-off, 1,020 um standard deviation
3
2,
8
i)
>
:
3
.
rossrsrs sss
:
Sp
8
i)
>
—
: an
oO
3
3
2
=
2
=
>
>
o
o
Distance [1 mm/div]
Horizontal Voltage [a. u.]
Figure 6.13
1|
Impedance diagrams and resolution profiles at three different rotational
angles relative to the horizontal lift-off angle in titanium at 2 MHz.
171
6.13.c), the dark center line completely
disappears
and the contrast profile becomes
Gaussian again, however the lateral resolution is very poor at about 1,020 pm.
Similar measurements were carried out at different frequencies to establish the
relationship between the phase dependence of the lateral resolution and the previously
considered frequency dependence. Figure 6.14 shows the experimentally measured lateral
resolution versus rotation angle curves for three different frequencies (see Table 6.3 —
Phase
Angle
Dependent
Lateral
Resolution,
Appendix
A).
At
frequency,
each
the
rotational angle was measured from the reference angle necessary to assure that the liftoff curve is horizontal. Clearly, all three curves reveal essentially the same dependence on
the rotational
increasing
angle
frequency.
as well
as a slight improvement
It should
be remembered
in the lateral resolution
that the previously
presented
with
lateral
resolution versus frequency curve (see Figure 6.11) was measured in a phase-insensitive
way and represents a medium value corresponding to a rotational angle around, but not
exactly, 90°. As mentioned before, measuring the absolute value of the relatively large
output signal of the bridge yields only that component of the small interface signal which
is parallel to the main signal. Assuming that the only imbalance in the bridge is due to the
eddy current load of the measuring
coil with respect to the unloaded
reference coil
a
positioned in the shielded case, the absolute value measurement would correspond to
rotation angle of 90°. However, the inevitable lack of symmetry between the measuring
and reference coils and possible differences on the order of a few tenth of a percent
between the driver resistances cause the bridge to be slightly out of balance even without
magnetic loads on the measuring coil, therefore the lift-off signal is not exactly in phase
172
with the output signal of the bridge. In addition, the lift-off curve is slightly curved and
the rotation angle is measured not from its average direction but from the slope at its end
as shown in the blown-up picture of Figure 6.10.
It should be emphasized again that although Figure 6.14 seems to indicate that the
lateral resolution can be as low as 300-350 jum, but the accompanying contrast distortion
renders this optimum impractical for imaging purposes. However, considering the rather
oo
=
&.
c
Sgvar)
a
°
N
oO
fa
Sspo
2
is]
—
60
Rotation Angle [deg]
Figure 6.14
| Experimentally measured lateral resolution versus rotation angle curves
for three different frequencies in titanium.
of
modest improvements produced by increasing the inspection frequency, the benefits
the
this much simpler approach should not be underestimated. Figure 6.14 indicates that
the
lateral resolution can be improved by approximately a factor of two via optimizing
rotational angle with respect to the routinely used horizontal lift-off. As an example
Figure
shows
6.15
the
eddy
current
c-scan
images
of a coarse-grained
Ti-6A1-4V
evident
specimen at 5 MHz and two different rotational angles (0.5" x 0.5"). It is quite
eddy
that the higher rotation angle resulted in a significant resolution improvement. The
Figure
current and acoustic micrographs of the same specimen were previously shown on
6.15
6.2 at a two-times lower magnification. The large bright grain at the center of Figure
can be clearly recognized
inverted).
The
optimized
by its characteristic
5-MHz
eddy
shape
in Figure
current micrograph
6.2 (the contrast is
of Figure
6.15.b
exhibits
essentially the same fine details as the 40 MHz acoustic micrograph shown in Figure
b) 125° from horizontal lift-off
a) at horizontal lift-off
Figure 6.15
| Eddy current c-scan images of a coarse-grained Ti-6A1-4V specimen at 5
MHz and two different rotational angles (0.5" x 0.5").
174
6.2.c.
Even
be
could
better resolution
at higher
achieved
although
frequencies
the
decreasing sensitivity makes it increasingly difficult to visualize the rather weak (1-2%)
conductivity
variations
grains.
between
Further
in resolution
improvements
without
sacrificing the sensitivity must come from further reducing the coil size.
6.5 Summary
The main goal of this effort was to investigate the feasibility of exploiting the
unique
eddy
current
noise
grain
observed
in titanium
alloys
the
for
purposes
of
nondestructive materials characterization. It was shown that electric anisotropy exhibited
by
noncubic
crystallographic
electromagnetic
materials
classes
of materials
can
play
a very
similar
role
in
characterization of polycrystalline metals to that of elastic
anisotropy in ultrasonic materials characterization. Titanium is one of the few structural
metals of practical importance, especially in aerospace applications, which preferentially
crystallizes in hexagonal symmetry and therefore exhibits strong electrical anisotropy. At
the same time, titanium tends to form a rather coarse, locally textured microstructure
characterized by the presence
of large colonies
of hexagonal
alpha grains
featuring
similar orientation. The fracture resistance of the material is strongly affected by this
microstructure therefore there is a continued need for new nondestructive evaluation
techniques
that
are
capable
of both
imaging
and
quantitatively
characterizing
this
microstructure. During the previous year, we already demonstrated that the strong grain
noise in titanium alloys is caused by the unusual electrical anisotropy of hexagonal alpha
grains and theoretically explained the physical mechanism responsible for this contrast. In
this year, we further developed our eddy current imaging technique with special emphasis
175
on
improving
the
imaging
Higher
resolution.
resolution
current
eddy
micrographs
allowed us to determine the types of microstructure that most seriously hinder fatigue
crack detection and to demonstrate that eddy current imaging can be used to detect locally
textured microstructures that adversely affect fatigue resistance.
In pursuit of even better resolution we used analytical, finite element simulation,
and experimental methods to investigate the lateral resolution of eddy current microscopy
and
developed
optimization
procedures
that
assure
the best
imaging
resolution
for
microstructural materials characterization. It was found that the lateral resolution of eddy
current imaging is ultimately limited by the probe-coil dimension, but both the inspection
frequency and the phase angle can be used to optimize the resolution, to some degree, at
the expense of sensitivity. Generally, the higher the inspection frequency, the higher the
imaging resolution, but the improvement was found to be rather modest. Although eddy
current imaging is still in its infancy, a direct comparison of 5-MHz eddy current and 40MHz
acoustic
microscopic
images
of the
same
coarse-grained
Ti-6Al-4V
sample
indicated that the same features can be observed by both methods at approximately the
same resolution level. It is expected that in the future eddy current microscopy can be
further enhanced by the introduction of special high-resolution, microscopic probe-coils
which will make it a truly unique materials characterization tool especially well suited for
microstructural evaluation of titanium alloys.
* TI would like to express my sincere thanks to Dr. Waled Hassan (United Technologies
Research Center) for his expert support with the analytical calculations and finite element
results.
Chapter 6, References:
1. J. H. Rose, "Ultrasonic backscatter from microstructure,” in Rev. Progr. Quant.
Nondestr. Eval. Vol. 11B (Plenum, New York, 1992) pp. 1677-1684.
2. F. J. Margetan, R. B. Thompson, and I. Yalda-Mooshabad, "Backscattered
microstructural noise in ultrasonic toneburst inspections," J. Nondestr. Eval. 13, 111-136
(1994).
3. K. Goebbels, "Structure analysis by ultrasonic radiation," in Research Techniques in
Nondestructive Testing Vol. TV (Academic, New York, 1980) pp. 87-157.
4. A. Hecht, R. Thiel, E. Neumann, and E. Mundry, "Nondestructive determination of
grain size in austenitic sheet by ultrasonic backscattering,” Mat. Eval. 39, 934-938 (1981).
5. H. Willems and K. Goebbels, "Characterization of microstructure by backscattered
ultrasonic waves," Met. Sci. 15, 549-553 (1981).
6. C. B. Guo, P. Holler, and K. Goebbels, "Scattering of ultrasonic waves in anisotropic
polycrystalline metals," Acustica 59, 112-120 (1985).
7. W. A. Wooster, A Textbook on Crystal Physics (University Press, Cambridge, 193 8).
8. J. F. Nye, Physical Properties of Crystals, Their Representation by Tensors and
Matrices (Clarendon Press, Oxford, 1985).
9. J. E. Neighbor, "Eddy current method for measuring anisotropic resistivity," J. Appl.
Phys. 40, 3078-3080 (1969).
10. V. M. Tatarnikov, "Use of probes for measuring the electrical conductance of
anisotropic plates," Meas. Tech. 13, 877-881 (1970).
11. V. IL. Gordienko and V. G. Rybachuk, "Device for measuring the anisotropy of
electrical conductivity," Meas. Tech. 32, 371-374 (1989).
12. R. Pitchumani, P. K. Liaw, D. C. Yao, D. K. Hsue, and H. Jeong, "Eddy current
technique for the measurement of constituent volume fractions in a three-phase metalmatrix composite," J. Comp. Mat. 28, 1742-1769 (1994).
13. R. Pitchumani, P. K. Liaw, D. C. Yao, D. K. Hsue, and H. Jeong, "Theoretical models
for the anisotropic conductivities of two-phase and three-phase metal-matrix composites,"
Acta. Metal. Mat. 43, 3045-3059 (1995).
14. P. L. Rossitier, The Electrical Resistivity of Metals and Alloys (Cambridge University
Press, Cambridge, 1987) pp. 185-195.
15. D. Eylon, "Faceted fracture in beta annealed titanium alloys," Met. Trans. A 10, 311317 (1979).
16. “Device Eases Skin Crack Detection,” Aviat. Week Space Tech. Vol. 131, p. 137
(1989).
17. “Eddy Current Imaging Device Offers Prospect of Simple Crack Detection,” Aviat.
Week Space Tech. Vol. 132, p. 96 (1990).
18. N. C. Haywood and J. R. Bowler, “Eddy Current Imaging of Buried Cracks by
Inverting Field Data,” JEEE Trans. Magn. Vol. 28, p. 1336 (1992).
19. T. W. Guettinger, K. Grotz, and H. Wezel, “Eddy Current Imaging,” Mat. Eval. Vol.
51, p. 444 (1993).
20. G. L. Fitzpatrick, D. K. Thome, and R. L. Skaugset, “Magneto-optic/Eddy Current
Imaging of Aging Aircraft; a new NDI technique,” Mat. Eval. Vol. 51, p. 1402 (1993).
177
21. J. H. Hippler, H. Ermert, and L. von Bernus, “Broadband Holography Applied to
Eddy Current Imaging by Using Signals with Multiplied Phases,” J. Nondestr. Eval. Vol.
12, p. 153 (1993).
22. M. Gramz, “Eddy Current Imaging - A Theoretical Approach,” Compel. Vol. 13, p.
19 (1994).
23. M. W. Kirby, “Experience with Eddy Current Imaging Detection for the Detection of
Cracks within and under the Clad in a Pressure Vessel,” Mat. Eval. Vol. 54, p. 30 (1996).
24. E. A. Ash and G. Nichols, “Super-resolution Aperture Scanning Microscope,” Nature
237, pp. 510-512 (1972).
25. C. P. Viahacos, R. C. Black, S. M. Anlage, A. Amar, and F. C. Wellstood, “Near-field
Scanning Microwave Microscope with 100 pm Resolution,” Appl. Phys. Lett. Vol. 69, pp.
3272-3274 (1996).
26. D. E. Steinhauer, C. P. Vlahacos, S. K. Dutta, F. C. Wellstood, and S. M. Anlage,
“Surface Resistance Imaging with Scanning Near-field Microwave Microscope,” Appl.
Phys. Lett. Vol. 71, pp. 1736-1738 (1997).
27. D. E. Steinhauer, C. P. Vlahacos, S. K. Dutta, B. J. Feenstra, and F. C. Wellstood,
“Quantitative Imaging of Sheet Resistance with a Scanning Near-field Microwave
Microscope,” Appl. Phys. Lett. Vol. 72, pp. 861-863 (1998).
28. M. P. Blodgett, P. B. Nagy, “Anisotropic Grain Noise in Eddy Current Inspection of
Noncubic Polycrystalline Metals,” Appl. Phys. Let., 72(9), pp. 1045-1047 (1998).
29. A. Briggs, An Introduction to Scanning Acoustic Microscopy (Oxford University
Press, Oxford, 1985). pp. 35-41.
30. G. T. Meaden, Electrical Resistance of Metals (Plenum, New York, 1965) p. 31.
31. H. L. Libby, Introduction to Electromagnetic Nondestructive Test Methods (WileyInterscience, New York, 1971).
32. Z. Mottl, “The Quantitative Relations Between True and Standard Depth of
Penetration for Air-cored Probe Coils in Eddy Current Testing,” NDT Intern. Vol. 23, pp.
11-18 (1990).
33. F. Thollon, B. Lebrun, N. Burais, and Y. Jayet, “Numerical and Experimental Study
of Eddy Current Probes in NDT of Structures with Deep Flaws” NDT&E Intern. Vol. 28,
pp. 97-102 (1995).
34. C. V. Dodd and W. E. Deeds, “Analytical Solutions to Eddy Current Probe-coil
Problems,” J. Appl. Phys. Vol. 39, pp. 2829-2838 (1968).
35. C. V. Dodd, “The Use of Computer Modelling for Eddy Current Testing,” in
Research Techniques in NDT Vol. 3, Ch. 13 (Academic Press, London, 1977).
36. R. Palanisamy and W. Lord, “Prediction of Eddy Current Probe Signal Trajectories,”
IEEE. Trans. Magn. Vol. 16, pp. 1083-1085 (1980).
37. R. Becker, K. Betzold, K. D. Boness, R. Collins, C. C. Holt, and J. Simkin, “The
Modeling of Electrical Current NDT Methods: Its Application to Weld Testing (Part 1),”
British. J. NDT Vol. 28, pp. 286-294 (1986).
38. P. Stucky and W. Lord, “Skin Depth Considerations in Eddy Current NDT,” in Rev.
Progr. Quant. Nondestr. Eval. Vol. 11A (Plenum, New York, 1992) pp. 299-306.
39. PC-OPERA for Windows, Vector Fields, Inc. (Aurora, Illinois, 1998).
40. D. J. Hagemaier, “Eddy Current Standard Depth of Penetration,” Mat. Eval. Vol. 43,
pp. 1438-1442 (1985).
CHAPTER VII
CONCLUDING REMARKS
7.1 Summary
effort was
A comprehensive
undertaken to develop
evaluation
nondestructive
methods to characterize microstructure and material inhomogeneities in Ti-6Al-4V. The
influence of texture, grain structure, and crystallographic orientation on ultrasonic wave
propagation, electrical conductance, and microstructure anomaly detection in heat treated
investigated.
samples
was
received
material
types
microstructure
establish
a
and
baseline
Samples
were
generated
additional
samples
were
(generated
via heat
treatments
for
this
study.
these studies in the as-
generated to make
Joined
in five
of the
different
common
as-received
material)
samples
of
microstructure
the
to
same
composition, but dissimilar microstructures were also generated by diffusion bonding to
simulate processing anomalies in the axial and radial orientations.
The metallographic results indicate the presence of strong macroscopic texture in
the as-received alloy, exhibiting alignment of the grains, grain boundaries, and grain
colonies, which
is a key factor in assessing flaw (anomaly)
detectability due to the
generation of microstructural background noise. The heat treated samples showed the
179
these
diminution of the macroscopically aligned grain features in the microstructure, but
samples maintain most of the macroscopic texture, except in the coarse beta annealed
the two
samples. All of the samples from this study were axisymmetric and evaluated in
a
principal directions (i.e. axial and radial). Only the beta annealed samples demonstrated
also
reduction in the texture, due to recrystallization. Ultrasonic velocity measurements
the basal
indicated the presence of a strong preferred crystallographic orientation in which
by
normals preferentially rotate to the transverse direction. This finding is supported
analytical x-ray results, using Shultz back-reflection.
The orientation dependence of ultrasonic scattering and attenuation in the textured
studied
bar material was also investigated. Conventional ultrasonic backscattering was
tering
and a forward scattering method was developed to assess texture. The backscat
is not
results indicate defect detection in the radial direction of the as-received bar stock
in the
favorable due to the strongest presence of ultrasonic grain noise. Backscattering
is
axial direction is negligible; however, the ultrasonic attenuation in the radial direction
current
significantly lower than in the axial direction, a behavior not predicted from
t
theoretical scattering models. Typically, the direction of highest attenuation is consisten
broken
with the direction of highest backscattering energy. This common rule is clearly
reasoning
for the case of the Ti-6Al-4V forged bar material studied in this work. The
of
behind the unusual properties was demonstrated to be related to the forward scatter
be
ultrasonic energy in the elongated microstructure. This work shows that scatter must
considered from multiple directions, due to multiple scattering, in order to relate scatter
with attenuation.
as
measured
significantly misleading
due
to the phase-sensitivity
receivers.
This
result was
verified
based
on the
shown
was
means,
conventional
using
attenuation,
The
of the wide-aperture
to
be
ultrasonic
of a laser ultrasonic
development
detection method, which removes the phase-sensitivity of the receiver by focusing the
light down to a 50 pm spot, a subwavelength aperture. The laser detection system allows
mapping of the phase and amplitude of waves transmitted in a solid. This high resolution
mapping capability revealed for the first time, images of microstructural features based on
phase distortion of the propagating waves.
An eddy current materials characterization technique was also developed to map
electrical
conductance
variations
from
resulting
the
grain
anisotropy
and hexagonal
symmetryof the Ti-6AIl-4V alloy. This research demonstrated for the first time, an eddy
current materials characterization capability geared specifically towards titanium alloys to
map
electrical properties of the macroscopic
grain structure. An
optimization of the
electrical conductivity mapping technique demonstrated a factor of two increase in spatial
resolution by increasing the AC
frequency.
Further improvements
were
achieved by
appropriately rotating the phase angle; however, this research clearly shows the primary
factor influencing the resolution is the diameter of the probe. This research has lead to
new research in development of nearfield scanning microwave microscopy, which is
promising
for further
improving
the
spatial
Similarly, while the electrical mapping
microstructure
characterization,
it has
resolution
technique was
lead to new
magnetoresistance sensors for deeper inspection depths.
of eddy
current
microscopy.
geared towards near surface
interests
in the development
of
Finally, an investigation was
conducted
to detect and characterize anomalous
microstructure, using diffusion bonded samples containing microstructure inserts of three
different sizes (i.e., 0.9", 0.4" and 0.2"). The results demonstrate that the anomalous
microstructure inserts can be detected via ultrasonic backscattering in all but the smallest
dimension.
The
correlation
parameter in making
length of the grain scattering was
an assessment
shown
of microstructural uniformity.
to be the key
Unfortunately, the
scattering signals lack uniqueness, are frequency dependent, and are difficult to interpret
in terms of actual physical features within the material, due to signal interference and
multiple scatter. Hence, these ultrasonic backscattering results are qualitative, and lack
the specifics to allow a comprehensive description of the microstructure characteristics
(e.g., grain size, shape, anisotropy, etc.).
7.2. Conclusions
General
1) Background ultrasonic and eddy current measurements have revealed unusual
elastic and electrical properties in the Ti-6Al-4V forged bar stock. What is unusual is the
simultaneous high attenuation and low backscattering in the axial direction of the forged
bar stock, due to forward scattering; and the fluxuating local electrical conductivity, due
to the hexagonal
symmetry
of the alpha phase These measurements
also showed the
material contains significant deformation and crystallographic textue, and this fact was
verified using x-ray diffraction.
182
Ultrasonic
2)
backscattering
results
measurement
show
0.9”
the
and
0.4”
subsurface anomalies are detectable, but the 0.2” anomaly inserts are undetectable. These
measurements demonstrated the difficulty involved with finding small localized interior
microstructure anomalies in forged titanium alloys, especially in the noisy (i.e radial)
direction.
3) Three new NDE methods have been developed and demonstrated during this
dissertation to characterize titanium alloys. These methods are: i) forward scattering for
qualitative
assessment
assessment
and
evaluation
significantly influences
ii) ultrasonic
of texture,
of
distortions
attenuation;
phase
mapping
the
propagating
in
for grain
and ili) electrical anisotropy,
alignment
wavefront,
which
using eddy current
evaluation, for mapping of macroscopic grain structure.
4) The new NDE methods represent a sound experimental foundation for anomaly
detection and characterization in titanium alloy products.
5) This work should be continued to further enhance the NDE
capabilities for
titanium alloy billet, bar, and plate.
Ultrasonic Phase Mapping
1)
Conventional
ultrasonic
measurements
anisotropy.
183
show
evidence
of
strong
elastic
2) Phase mapping allows texture-induced wafefront disturbances to be visualized.
3) The sample orientation (i.e. described by the angle between the surface normal
and the bar axis) has a primary effect on ultrasonic properties (e.g. attenuation, scattering,
and velocity).
4) Phase cancellation is mainly evident in the axial direction of the Ti-6Al-4V
forged bar stock; the transverse direction has much smaller degree of distortion.
5) Phase cancellation at the receiver (in conventional wide-aperture transducers)
significantly influences attenuation and can generate misleading results for Ti alloys.
Eddy Current Microscopy
1) Eddy current measurements revealed strong electrical anisotropy in the Ti-6Al4V alloy due to the hexagonal nature of the alpha phase. This study revealed the difficulty
in detecting small flaws in titanium alloys due to strong background noise.
2) Good agreement was demonstrated between the analytical, finite element, and
experimental results on the eddy current lateral resolution investigation.
3) The lateral resolution is primarily driven by the eddy current probe diameter.
An increase in the AC frequency can also improve the resolution by about a factor of two
and phase angle can improve the resolution, but these are both secondary effects.
7.3 Comparison of Results for Microstructure Anomaly Characterization
Multiple NDE
conventional
ultrasonic
NDE
techniques
and backscatter)
and some
newly
developed
including forward scattering, laser interferometric phase mapping,
and eddy
These
dissertation.
(measurements
methods,
approaches have been investigated throughout the course of this
of velocity,
current microscopy
Conventional
approaches
attenuation,
(see Table
ultrasonic
include
7.1 — Comparison
approaches
provide
of NDE
a baseline
Results,
capability
to
Appendix
A).
evaluate
the
influence of texture and microstructure on the elastic properties. Despite the fact that
these techniques are not new, they are invaluable and provide unique insight into the
structural
make-up
of
polycrystalline
Specifically,
metals.
ultrasonic
velocity
measurements (longitudinal and shear wave velocity) provide a means to evaluate texture
in a quantitative manner. For example, the forged bar investigated in this dissertation
clearly contained a significant degree of mechanical texture and the ultrasonic velocity
measurements supported this fact, which was later confirmed via X-ray measurements
(Shultz back reflection) showing the basal normals of the forged bar tend to preferentially
rotate to the transverse direction during fabrication.
Conventional ultrasonic attenuation and backscattering measurements were also
taken to provide information on the micro- and macroscopic structural makeup of the
alloy.
The
attenuation
measurements
provide
an
indication
of the
impedance
to
propagating waves as a function of length, but do not specify what material features are
to
responsible for the decay in signal amplitude. A number of factors have been shown
affect attenuation in this dissertation, including phase sensitivity of the receiver, surface
185
texture
interactions,
the grain
orientation,
and
grain
size, and
shape.
Unfortunately,
attenuation is a bulk measurement and all of these interactions combine unseperably in
what we call attenuation. Ultrasonic backscattering measurements are generally not more
specific in terms of their behavior (compared to attenuation) and are also difficult to
interpret due
to competing
and
effects related to elastic property
crystallographic
variations, grain and colony boundaries, frequency dependence, and multiple scatter. In
titanium alloys, this problem is even more complicated by second phase particles and
density
their associated
differences
constituent,
relative to the primary phase
which
generates another source of scattering. Moreover, conventional scattering theory (based
on the Rayleigh,
stochastic, and geometric
scattering regimes)
is not geared for the
complexity of most titanium alloys. However, recent research demonstrates that by taking
the two-point elastic correlation, as measured by crystallographic mapping of the grain
a much
improved
prediction
structure,
provides
complex
microstructures, backscatter measurements
characteristic,
similar
to
attenuation.
While
these
for
backscattering.
provide
Generally,
for
only an overall response
measurements
(attenuation
and
backscatter) are not specific, they are useful for comparison purposes to observe relative
differences between effects of microstructure, texture, and manufacturing process. The
forward
scattering
results
described
in
this
dissertation
add
another
useful
bulk
measurement tool for microstructure characterization. This tool is particularly useful for
characterization of macroscopic texture, based on the amplitude and divergence of the
forward scatter. The results indicate this tool is useful for separating textured versus
random structures.
186
Phase mapping, using laser interferometric ultrasonic detection, was demonstrated
to be useful to characterize the influence of macroscopic orientation and the resulting
phase perturbation on attenuation in the axial and transverse directions of a mill annealed
Ti-6Al-4V bar. These phase mapping results revealed approximately 30% higher phase
scatter, measured over the center fringe in the axial direction, relative to the transverse
direction. The higher degree of phase scattering in the axial direction of the bar is due to
the presence of elongated macroscopic inhomogeneities aligned in the same direction.
These extended colonies of similarly oriented grains in the axial direction result in local
disturbances in the arrival time of the propagating wavefront that extend beyond the finite
aperture of the transmitter. While there also exists phase scatter in the radial direction, it
is significantly smaller and more uniform than that of the axial direction. Measurements
also revealed
substantially lower
attenuation values
for the laser experiment
due its
relative phase-insensitivity, provided by the microscopic footprint of the focused laser
detector.
While the laser interferometric measurements demonstrate lower attenuation than
the immersion results, the attenuation is still clearly dominant in the axial direction, as
compared with the transverse direction. These measurements demonstrate that the wider
apertures generally used by conventional transducers clearly suffer phase cancellation
losses, resulting in significantly higher attenuation results. The attenuation in this material
is unusual considering that the backscatter is 2 to 3 times stronger in the radial direction
than in the axial one; and the forward scatter is at least as strong, although less divergent,
than in the axial direction. This is because elongated grains and grain colonies present
187
larger total scattering cross-section (larger effective size) along their axis compared to
normal to them, but backscattering and forward scattering coefficients are actually higher
in the normal direction. Therefore, despite the stronger backscatter in the radial direction,
the axial direction has higher attenuation. This method made it possible to image the
and
microstructure by using the laser detection technology. The macroscopic orientation
directionality can be clearly identified and possibly quantified. Work is now in progress
to apply the same techniques to other Ti-6Al-4V microstructures with varying degrees of
structural directionality, to confirm this finding.
Finally,
was
microscopy
current
eddy
demonstrated
as
the
result
of
an
investigation on the feasibility of exploiting the unique eddy current grain noise observed
in titanium alloys for the purposes of nondestructive materials characterization. It was
shown that electric anisotropy exhibited by noncubic crystallographic classes of materials
can
play
a
very
similar
role
in
electromagnetic
materials
characterization
of
polycrystalline metals to that of elastic anisotropy in ultrasonic materials characterization.
Titanium is one of the few structural metals of practical importance, especially in
aerospace
applications,
which
preferentially
crystallizes
in hexagonal
symmetry
and
a
therefore exhibits strong electrical anisotropy. At the same time, titanium tends to form
rather coarse,
locally textured microstructure
characterized
by the presence
of large
colonies of hexagonal alpha grains featuring similar orientation. The fracture resistance of
the material is strongly affected by this microstructure therefore there is a continued need
for new nondestructive evaluation techniques that are capable of both imaging and
quantitatively characterizing this microstructure.
188
In
pursuit
of
better
resolution
analytical,
finite
element
simulation,
and
experimental methods were used to investigate the lateral resolution of eddy current
best
imaging
resolution
for
microstructural
that
assure
the
characterization.
It was
found that the lateral resolution of eddy
microscopy
materials
current imaging
is
ultimately limited by the probe-coil dimension, but both the inspection frequency and the
phase angle can be used to optimize the resolution, to some degree, at the expense of
sensitivity.
Generally,
the
higher
the
inspection
frequency,
the
higher
the
imaging
resolution, but the improvement was found to be rather modest. Although eddy current
imaging is still in its infancy, a direct comparison of 5-MHz eddy current and 40-MHz
acoustic microscopic images of the same coarse-grained Ti-6Al-4V sample indicated that
the same features can be observed by both methods at approximately the same resolution
level. It is expected that in the future eddy current microscopy can be further enhanced by
the introduction of special high-resolution, microscopic probe-coils which will make it a
truly unique materials characterization tool especially well suited for microstructural
evaluation of titanium alloys.
7.4
7.4.1
Future Nondestructive Materials Characterization Research
Application and Development of Array Transducers
One area that will be beneficial for future study is the development and use of
phased array transducers. The main problem of detecting anomalous scatterers in bulk
material is that only one incidence angle is generally investigated (i.e. normal incidence).
189
Phased
arrays
enable
the capability to probe
the material
for obscure
(nonspecular)
scatterers from a multitude of incidence angles, thereby increasing the probability of
detection. Unfortunately, these array systems have a high capital equipment investment
barrier that makes their application difficult for research. However, with the increased
need to find and characterize smaller and smaller internal flaws, especially in fracturecritical engine
equipment.
components,
Phased
this approach
array systems
is viable,
could provide
despite the large investment
a tremendous
in
increase in inspection
throughput for critical aircraft airframe and engine components. Moreover, the electronic
beam steering and dynamic focusing offered by phased arrays makes their use a potential
for improved billet inspection in the titanium industry.
7.4.2 Geometry Insensitive Techniques
One of the key difficulties in the nondestructive inspection of aircraft components
is due
to their
geometric
complexity.
Some
recent
developments
in thermoelectric
materials evaluation have demonstrated the capability to discriminate between varying
degrees of damage
(simulated by varying degrees of plastic deformation) without the
strong sensitivity to shape or surface character that some practical methods, like eddy
current inspection, demonstrate. The technique is based on the measurement of magnetic
field variations resulting from thermal currents in a conductor imposed by an external
thermal gradient (i.e., magnetic sensing of the Seebeck effect). While little research has
been done to describe the details of what may be possible in this area, the preliminary
research suggests this could be an extremely valuable NDE method. One application that
would represent a breakthrough is the demonstration of a capability to nondestructively
190
would
a
play
role
strong
in
assessment capability
A fatigue damage
provide local assessments of fatigue damage.
better
developing
prediction
life
cycling
mechanical
of thermal and
tools
and
general
on fracture critical
understanding
of the effects
components.
While this capability would likely be geared towards more macroscopic
measurements,
on the order of 0.25" spatial resolution, the benefits of observing the
progress
of
damage
without
destructive
microscopic
initiation
and
techniques
would
premature
to judge
as the
fundamental
experimental
foundation
still needs to be built. Clearly,
it will take
some
unparalleled.
be
at this
point
The
implications
of this
capability
and
examination
are a
little
theoretical
time to lay down
the
groundwork, make some investments in instrumentation, and develop practical methods.
7.4.3 Residual Stress Gradient Measurement
This research
is directed towards
the
capability to nondestructively
measure
residual stress depth gradients, which is a goal for improved life prediction practices.
NDE
research is needed to enable the capability to measure
approximately 400 ym
stress depth profiles to
in shot peened materials and deeper in laser shock peened
materials. Currently residual stress gradients can be measured, but only destructively by
surface material and measuring the stress by x-ray diffraction.
successively removing
Scanned x-ray diffraction measurement
measurements
over
a
small
area,
but
capabilities also exist, which provide stress
very
slow.
Ultrasonic
stress
measurement
capabilities also exist, based on the acoustoelastic effect, but these techniques generally
only provide an evaluation of the bulk material, similar to an attenuation measurement.
Local stress measurements using scanning acoustic microscopy (i.e., V(z) measurements
191
of the Rayleigh wave or longitudinal surface skimming wave) have been demonstrated to
be stress sensitive, but further research is necessary to describe the limits and prove the
approach. Similar to velocity, electrical measurements are marginally sensitive to stress
and cold working.
Researchers
are currently investigating the potential of frequency
dependent local acoustic velocity measurements and eddy current stress measurements.
Unfortunately, the surface roughness occurring as a result of shot / laser shock peening is
clearly a troublesome hurdle to these approaches. As such, other approaches should be
simultaneously investigated to overcome the obvious limitations of the ultrasonics and
Alternate approaches may
electrical approaches.
include:
alternating current potential
difference, thermal mapping, thermoelectric, and energy dependent x-ray. It's not obvious
that we will ever be able to achieve a nondestructive stress depth profiling capability, so
this is clearly a high risk research area, but the payoff for improved life prediction / life
extension requires us to do this research.
7.4.4 Forward Scattering Using Dual-Transducer Articulation
Finally, preliminary research results from this thesis project have demonstrated
the some
capabilities of ultrasonic forward
scattering to characterize bulk material
inhomogeneities. However, further research is needed to detail some of the fundamental
aspects of forward scattering, such as experimentally describing the scattering crosssections in Ti-6Al-4V.
The
majority of past efforts have
concentrated
on ultrasonic
backscattering, which is a well investigated area, due to the vast amounts of literature
currently
available.
More
details
in forward
research in ultrasonic materials characterization.
scattering
is an opportunity
to further
APPENDIX A
- Data Tables
Sample Description
Nomenclature
As-received (0 degrees relative to the axial
AR(0)
direction of the forged bar)
As-received (30 degrees relative to the
axial direction of the forged bar)
AR(30)
As-received (37 degrees relative to the
axial direction of the forged bar)
AR(37)
As-received (45 degrees relative to the
axial direction of the forged bar)
AR(45)
As-received (60 degrees relative to the
axial direction of the forged bar)
AR(60)
As-received (90 degrees relative to the
axial direction of the forged bar)
AR(90)
Duplex anneal (axial direction)
DA(0)
Duplex anneal (transverse direction)
DA(90)
Recrystallization anneal (axial direction)
RA(0)
Recrystallization anneal (transverse
direction)
RA(90)
Fine beta annealed (axial direction)
FBA(0)
Fine beta annealed (transverse direction)
FBA(90)
Coarse beta annealed (axial direction)
CBA(0)
Coarse beta annealed (transverse direction)
CBA(90)
Table 3.2
Ti-6AI-4V sample nomenclature for ultrasonic measurements. Sample
descriptions are based on the microstructure and the orientation of wave
propagation.
Longitudinal Wave Velocity
Table 3.3
Sample
Velocity (m/s)
Difference (%)
AR(90)
6248.5
AR(0)
6131.3
DA(90)
DA(0)
6230.8
6110.1
1.93
RA(90)
6234.7
1.70
RA(0)
6128.3
FBA(90
(90)
FBA(0)
CBA(90)
6200.8
6150.8
6197.4
CBA(0)
6173.9
187
0.80
038
Microstructural effects on longitudinal wave velocity, based on data from
Figure 3.9. All samples are approximately 1.5” thick.
Shear Wave Velocity
Sample
Table 3.4
| fast mode (m/s)| slow mode (m/s)|__
fast / slow
AR(90)
AR(0)
3319.57
3163.79
3164.39
3140.27
1.04903
1.00749
DA(90)
DA(0)
RA(90)
RA(0)
3281.54
3158.39
1.03899
3140.01
3273.65
3153.03
3119.29
3143.27
3133.84
FBA(90)
3204.37
3165.33
1.00664
1.04148
1.00612
1.01235
FBA(0)
CBA(0)
CBA(90)
3181.85
3178.00
*
3158.34
3168.00
*
1.00744
1.00315
*
Microstructural effects on shear wave velocity, based on data from Figure
3.12. All samples are approximately 1.5" thick.
194
to get reproducible results)
attenuative
too
* Data not available (sample was
[je Ul rojoRZ & Ape] st UoNeSedoid saeM oY} JO UOTRIUSO
"Wao }daoxa ‘soydures
“sATENusye 4svo] oY} SEM (06) VC pue satjenusye
ysour oy) ApIeaTD sem (06 PUe O HIG) VAD sduIeg “LTE aMBrJ Jo sy[nsor UO paseq SI BEC] “UOHRIOXS JsINq
-quo} Suisn ‘yIBusT YEdioyem ,,L"] & YIM “AaJoweIP ,,¢°Q pesnoogun “ZHY| OT & Sew Joonpsuel] “gP ul sa0yoo
L’81
S61
L1é
TOr
II
col
vil
6 Tl
vl
v8
Lg
$6
6
Lgl
LI
ss
Ls
cL
L’6
yuasyoo JUanbasqns u99M49q SSO] [BUSIS UO paseg ‘sesso] UOTENUS}e yuopuodap Aouanbey pue TemyonysoIoly|
ysl
(o)va
(06) Vda
(ova
(06) UV (06)VdO
(o)VdoO
(qp) uonenuapy stuoseni
S'€ BQRL
‘uoTOOIIp (sa1d0p
0) jeIxe oy} UeY}) Joneosyoeg JosUONs ATUBOIJTUSIS SayeIsUOWSp UOTAIIP (sa1Zap 06) aSISAsUeI} 9} ‘AT[eIBUAyH
‘Iayyeosyorq [RINJONsOIOIUN OU AjyenyA syussaidos YOTYM “AW €TE0'O 1% SUIZ9q P[OYSEAY} SSIoU [BOLO9[9 aU}
OT & Sem Joonpsuell
JY} SION ‘YISUaNs Zuployeosyorq JO Japio Burpusosep Ut sojdures YIM ‘ZE"¢ SINSIJ UO paseq SI eye] “UOTEWOX
asindur Zuisn ‘ysucy yyedioyem ,,¢ & YIM sd"JINS oY} UO posndoy ‘JooWeIp ,.S°0 ‘ZH
‘SULIOJOARM (NOT ATayeuTxosdde Burdesoae Alyereds soye (sv) auI Jo UOTOUNY ev se BUTIOWeOSyoeQ OTUOSeNI/)
sTeteo0
STETEO0
STETEO0
STETEO0
STETEOO
STETEO0
STETEO'O
SIETEOO
STETEO0 809700
STETEO0 LOEvr0'0
STEIEO0 £tplv00
cIETEO0
CZIE90'O
vL8S0'°0
Lv9S7l'0
8L6101°0
90L790°0
6S1$90'°0
9PPTLO'0
889200
605680°0
E1E8S0°0
9TT6170
CCPOPT 0:
STELTI'0
eTclolo
(06) Va
STI6b7 0 ILbLee0 COPETH 0
S8STI810 9758070 86EP0E0
6r76~L0°0 vS€1800
6608L0°0 767880°0
vE8680°0 6LSS60°0
TEzeOr oO vIPror'o
ILestloO 9096110
LOEVET 0 COPOST'O
9989910 ST6IL1°0
COL6P70
906LST°0
CSPI9T'O
91SL07'0
9°€ BIQRL
STETEOO
STETEO 0
STETEO0
eL6l110
STSSETO
8786910
(06) Va
pySL90T'0
STETEO0
STETEO0
€S6£800
CLZOLT'0
(06)VdO
618870°0 chIELo'O
Lysvso0 8086200
€67090'°0 8L7760'0
107890°0
(O6)UV
pS9OSET 0
SsssIl oO
STETEO'0
886010
(o)vao
T7€8S0'0
7EPI9N'O
L0S690°0
L6SS€0'0
679£0'0
69LS€0°0
v06rE0°0
6S9£0°0
elespo0'o
(o)Vas
(A Ul) SuLsyeOsWoe_ [BINjonsOIOI] posesoay Ayjeneds
copa
ope]
mM]
869990'0
819€0'0
pL8ev00
(UV
+l
9¢8LL0°0
L6EC80°0
v6Svr0'0
(o)va
an]
(06) Va
(ova
NI
Axially Bonded Samples
Table 4.1
Sample
Thickness (in.)
AR/ AR
1.79
AR/CBA
1.82
CBA/ CBA
2.33
AR/FBA/ AR
2.94
AR/CBA/ AR
2.65
Diffusion bonded samples oriented such that the ultrasound propagates in
the axial direction.
Radially Bonded Samples
Table 4.2
Sample
Insert Thickness (in.)
AR/ AR
AR/ AR
AR/ AR
AR/FBA
AR/FBA
AR/FBA
0.2
0.5
0.9
0.2
0.4
0.9
AR/ CBA
AR/ CBA
AR/ CBA
0.2
0.4
0.9
Diffusion bonded samples oriented such that the ultrasound propagates in
the transverse direction. Samples were all 2.5" in diameter, with varying
sizes of implants and implant microstructures.
Orientation Dependence of Longitudinal Velocity
Table 5.1
Sample
Velocity (m/s)
AR(0)
6131.32
AR(30)
6159.37
AR@G7)
6175.64
AR(45)
6193.10
AR(60)
6215.44
AR(90)
6248.52
Difference (%)
0.46
0.28
0.53
Orientation dependence of longitudinal wave velocity, based on data from
Figure 5.2. All samples are approximately 1.5" thick. Note, the difference
between the two extremes in orientation is approximately 2%. Scatter in
the data is + 0.5%.
Orientation Dependence of Shear Velocity
Sample
| fast mode (m/s)} slow mode (m/s)} __ fast / slow
3163.79
3217.53
3231.50
3243.28
3293.06
3319.57
AR(0)
AR(30)
AR(37)
AR(45)
AR(60)
AR(90)
Table 5.2
3140.27
3166.77
3173.00
3171.62
3161.31
3164.39
1.007
1.016
1.018
1.022
1.041
1.049
Orientation dependence of shear wave velocity, based on data from Figure
5.3. All
samples
are approximately
1.5" thick. Note,
the difference
between the fast versus slow modes is less than 1% in the axial (0 degree)
direction and approximately 5% in the radial (90 degree) direction. This
data supports the conclusion that the material is highly textured, with basal
normals preferentially oriented in the transverse direction. Scatter in the
data is + 0.5%.
Orientation Dependent Attenuation Loss (dB)
sample
freq ty
Table 5.3
| AR(O) | ARGO) | ARGS)
| AR(60) | AR(90)
8.10
5.58
4.25
3.38
3.28
8.96
6.33
4.65
3.52
3.44
10.19
11.87
6.80
7.52
4.82
3.76
3.79
5.35
3.98
4.18
13.38
15.28
8.36
| 9.56
5.80
6.60
4.42
4.94
5.25
6.07
Orientation dependent attenuation losses, based on signal loss between
subsequent coherent echoes in dB. Transducer was a 10 MHz, unfocused
0.5" diameter, with a 1.7" waterpath length. Data is based on results of
Figure 5.4.
Orientation Dependence of Backscattering (mV)
AR(O)
|AR(30) | AR(45) | AR(60) | AR(90)
| 0.0448
| 0.0387
0.4162
0.0358 | 0.0357 | 0.0360
| 0.0372
0.2988
0.0353 | 0.0358
| 0.0348
| 0.0333
0.2508
| 0.0313
| 0.0313
| 0.0331
0.2047
0.0313 | 0.0313
| 0.0313
| 0.0320
0.1611
0.0313
| 0.0313
| 0.0313
| 0.0318
0.1561
0.0313
| 0.0313
| 0.0313
{0.0319
0.1358
0.0313 | 0.0313 | 0.0313
| 0.0308
0.1191
{| 0.0313
| 0.0310
0.1067
Oo
[& [dy
npn lL
Lolr
0.0445 | 0.0442
Table 5.4
0.0313
0.0313
| 0.0313
Orientation dependence of ultrasonic backscattering as a function of time
after spatially averaging approximately 2K waveforms
at each position.
Transducer was a 10 MHz, 0.5" diameter, focused on the surface with a 3"
waterpath length. Data is based on Figure 5.5. Note that the electrical noise
threshold begins at 0.0313 mV.
199
Phase-Sensitive (PS) -vs- Phase-Insensitive (PI) Attenuation (dB)
sample
freq (MHz
Table 5.5
| AR(90)PS | AR(90)PI
4.65
4.02
4.61
4.05
4.46
4.10
AR(O)PS
8.86
9.68
10.56
| AR(O)PI
7.4
8.48
9.04
12.3
10.6
4.47
4.98
13.7
11.48
5.26
5.7
15.4
12.72
6.37
5.91
Phase-sensitive (0.5" aperture, piezoelectric) versus phase-insensitive (50
uum, laser interferometric) measurements of attenuation in dB, based on
Figure
5.14.
Note,
a significant
contribution
of phase
sensitivity
to
attenuation is only a factor in the axial (0 degree) direction; however,
sample orientation is clearly the primary factor affecting attenuation in this
material.
Normalized Electrical Resistivity (%)
Al
Cu
Cd
Ti-6AI1-4V
0.1
-0.05
| -1.7.
| data ranges
Surface 2 | 0.15
-0.1
0.025
from
15 to 15
Surface 3 | 0.2
0.055
1.6
Surface 1]
Table 6.1
Average electrical resistivity for three single crystal surface orientations,
which were arbitrarily determined, and from the surface of polycrystalline
Data
is based
on Figure
6.3. All data was
calibrated using
different
aluminum
alloys as the
Ti-6Al-4V.
measured
and
Note,
the
standard.
differences in resistivity for the cubic crystals (Al and Cu) are well within
the scatter of the measurement. However, the hexagonally symmetric Cd
crystal has orientation dependent resistivity variations that extend well
beyond the data scatter.
201
Frequency Dependent Lateral Resolution
Frequency (MHz)
Resolution (um)
0.01
1300
0.1
1175
900
725
1.0
10.0
Table 6.2
Frequency
dependent
lateral resolution, demonstrating
approximately
a
factor of two improvement in resolution over three orders of magnitude in
frequency. Data is based on Figure 6.11.
Phase Angle Dependent Lateral Resolution
Table 6.3
Phase Angle
Resolution
(degrees)
0
30
60
1 MHz
(um)
|2MHz | 3 MHz
876
736
660
|743 | 685
1655 _ | 615
|607 | 582
90
599
1558.5 | 541
496
1419
884
_|472.6 | 471
|1348
[1183
|746 | 687
Phase dependence of lateral resolution for three different frequencies over
180 degrees. Note, the cycle repeats for every 1 phase angle increment.
This
chart demonstrates
that, similar to frequency,
phase
can lead to
approximately a factor of slightly better than two in resolution, although at
the expense of lift-off insensitivity. Data is based on Figure 6.14.
Comparison of NDE Results
Usage
NDE Approach
.
Velocity
Advantages
Disadvantages
Provides information on
Accuracy of +,- 0.5% can be
This ia a bulk measurement
elastic property variations
routinely achieved and velocity
and does not allow for
and texture resulting from
is directly related to stiffness.
.
Attenuation
Backscatter
Forward Scatter”
Can potentially provide
Measurement is simple and
information on grain size
not time consuming.
Phase Mapping”
source of the loss and not to be
compare materials.
used for locating anomalies.
Mainly used to characterize
A single transducer is used as
Measurement is complicated
grain structure in simple, cubic
transmitter and receiver.
materials, can be used as a
materials discriminator.
different materials.
Anomaly detection possible.
and not unique.
Used to qualitatively assess
Data is in the form of an image
the structure of metals.
Allows insight into the
No real advance for anomaly
detection and experimental
A good way to discriminate
effective grain size and
orientation effects
Not suited to production environ.
Phase-insensitive due to the
narrow 50 micron aperture.
Allows imaging of the wavefront,
microstructure and phase
distortion.
Experimental system is complex
and instrumentation intensive.
Approach is for bulk wave
evaluation of structures.
Not for anomaly detection.
Laser detection allows high
resolution imaging of the
transmitted acoustic field.
Experimental system is complex
and instrumentation intensive.
Approach is for bulk wave
Offers a unique way to describe
the shape of a propagating
ultrasonic wavefront and map
m'structures phase distortion.
Laser Based
:
:
Amplitude
Mapping
A bulk measurement tool
not specific in terms of the
Allows for an easy way to
and grain morphology.
Allows for an basis to compare
between textured-vs-random mat.
Laser Based
localization of anomalies
except via surface waves.
processing.
Used to map the strength of
ultrasonic waves propagating
.* | through a solid.
e
'
”
and time consuming.
| Data is frequency dependent
set-up is complicated.
evaluation of structures.
Not for anomaly detection.
ready Curr ent
icroscopy
Used to image local electrical
conductivity variations in
noncubic materials.
Near surface detection of
microstructure anomalies.
Noncontacting, computer
Surface and near surface only.
controlled scanning of materials. | Resolution is primarily driven
Resolution is at the physical limit. | by the coil diameter.
Eddy current materials evaluation | Further probe developments are
required to yield higher resolution
is in its infancy.
* New method, developed under this dissertation project.
Table 7.1
Comparison of NDE results.
203