Angle beam synthetic aperture focusing

in ultrasonics
f 2:!1
I / -,?'··- by

__.. 6c 7
/ --- Irfan M. Furniturewala
(l, /

A Thesis Submitted to the

Graduate Faculty in Partial Fulfillment of the

Requirements for the Degree of

MASTER OF SCIENCE
Department: Electrical and Computer Engineering
Major: Electrical Engineering

Approved: Member of the Committee:

Signatures have been redacted for privacy

Signatures have been redacted for privacy

Iowa St ate University

Ames, Iowa
1995

Copyright © Irfan M. Furniturewala, 1995. All rights reserved.

 11

TABLE OF CONTENTS

ACKNOWLEDGEMENTS . . . . . VI

CHAPTER 1. INTRODUCTION 1

NDE Techniques . 2

Problem Statement 3

CHAPTER 2. ULTRASONIC NONDESTRUCTIVE EVALUA-

TION . . . . 6
Generation and Propagation of Ultrasonic Waves 6

Comparison of Ultrasonic NDT with other Techniques 9

Ultrasonic Testing Techniques 10

Pitch-catch technique . 10

Pulse-echo technique 12

Other techniques . . 12

Modes of Ultrasonic Testing 14

Normal incidence . . . 14

Angle beam incidence . 14

Ultrasonic Imaging 16

A-Scan . 17

B-Scan . 18
 111

C-Scan . . . .. 20
Data Storage Issues . 21

CHAPTER 3. SAFT 25
Synthetic-aperture Fundamentals 25
Quantitative Interpretation of SAFT 27
Time domain back-propagation of wavefronts: SAFT 30
Qualitative Analysis of SAFT . . . . . . . 32

CHAPTER 4. ANGLE BEAM SAFT 37

Angle Beam Theory . . . . . . . . . . . . 37

Relevant Issues with Angle Beam Mode of Scanning . 39

Algorithm for Angle-Beam SAFT .. . . . . .. .. . 43
Mapping between Image Domain and Physical Domain 47
Mapping from image domain to physical domain . 48
Mapping from physical domain to the image domain . 49

CHAPTER 5. RESULTS AND DISCUSSIONS 51

Experimental Setup . 51

Results .. . 53
B-scan 53
B'-scan . 57

Multi depth B'-scan. 60

Discussion . 61
Conclusion . 64
Future Work 66
 IV

BIBLIOGRAPHY .. . .. . .. .. .. . 68
 v

LIST OF FIGURES

Figure 1.1: Typical nuclear pipe and weld geometry . . . . . . 4

Figure 2.1 : Normal pitch-catch technique of ultrasonic testing 11

Figure 2.2: Normal pulse-echo technique of ultrasonic testing . 13

Figure 2.3: Angle-beam pitch-catch technique of ultrasonic testing . 16

Figure 2.4: Angle-beam pulse-echo technique of ultrasonic testing 17

Figure 2.5: A-scan . . . . . . . . . . . . . . . . . . 19

Figure 2.6: Geometry for the flaw depth calculation 20

Figure 2. 7: B-scan imaging system 21

Figure 2.8: B-scan . . . . . . . . . 22

Figure 2.9: Raster scanning of the test specimen . 23

Figure 2.10: C-scan .. .. . . . . . . . . . . . . . 24

Figure 3.1: Scattering object and measuring surface . 31

Figure 3.2: Hyperbolic reflections from the same reflector in different A-

scans . . . . . . . . . . . . . . . . . . . 34

F igure 3.3: Constructive interference during SAFT 35

Figure 3.4: Destructive interference during SAFT 36

Figure 4.1 : Mode conversion of sound waves . . . 38

 Vl

Figure 4.2: Angle of incidence is greater than the first critical angle 40
Figure 4.3: Multiple sound paths in angle-beam pitch-catch mode of ul-

trasonic testing . . . . . . . . . . . 41

Figure 4.4: Periodic imaging of the same defect 43

Figure 4.5: B-scan with periodic imaging of a crack 44

Figure 4.6: Test specimen geometry for the angle beam SAFT algorithm 45

Figure 4.7: Metal path calculation geometry . . . . . . . . . 46

Figure 4.8: Mapping from image domain to physical domain 48

Figure 4.9: Mapping from physical domain to image domain 50

Figure 5.1: Experimental pipe geometry 54

Figure 5.2: B-scan imaging . . . . . . . 55

Figure 5.3: A B-scan image in which the defect can be seen clearly 56

Figure 5.4: After SAFT implementation on the B-scan . . . . . . . 57

Figure 5.5: A B-scan image in which the defect cannot be seen clearly . 58

Figure 5.6: After SAFT implementation on the B-scan 59

Figure 5.7: B '-scan imaging technique . . . . . . . . . 60

Figure 5.8: A B'-scan image in which the defect can be seen clearly 61

Figure 5.9: After SAFT implementation on the B'-scan . . . . . . . 62

Figure 5.10: A B'-scan image in which the defect cannot be seen clearly 63

Figure 5.11: After SAFT implementation on the B'-scan . . . . . . . . . 64

Figure 5.12: A B'-scan image in which the defect cannot be seen clearly 65

Figure 5.13: After implementing SAFT at 6 uniformly spaced depths in

the third leg of the B '-scan . . . . . . . . . . . . . . . . . . . 66

 Vll

ACKNOWLED GEMENTS

First and foremost, I wish to thank God, for giving me all the wonderful oppor-

tunities, and the best family and colleagues.

I would like to take this space to extend my deep-felt gratitude towards my

major professor, Dr. Lalita Udpa, for having the confidence and faith in me, and

for providing me with valuable suggestions and critique, throughout my master's

program. I also wish to offer my deepest thanks to my other committee members,

Dr. Satish Udpa and Dr. Soma Chaudhuri, for their encouragement and help. I

am thankful to the Electric Power Research Institute (EPRI), which was responsible

for sponsoring this research. Especially, I wish to thank Mr. Tom Taylor at EPRI,

for his complete guidance and support in all possible ways towards this research.

I acknowledge the support of Mr. Mike Elmo at EPRI, for providing me with his

P C_Image program, which served as a valuable tool during the entire research. I also
thank Mr. Kevin Levesque at LMT, Inc. and Mr. Mark Davis at Davis NDE, Inc. ,

for their time and suggestions on scanning techniques used in the industry. I express

my deep gratitude, towards my colleagues, Dr. Mani Mina and Mr. Wassef Masri,

who helped with the initial part of the research.

My parents, brothers and sister deserve a special acknowledgement for instilling

in me the confidence, and providing me with the support, that I needed to pursue
 Vlll

this program. I am extremely thankful to them for being the most wondeful family.

This space will not be complete without acknowledging the support of my dearest

wife, Asma, who has played a very important role in all my achievemnets, and has

made all the difference to my life. Last, but not the least, I thank my good friend,

Ashish Dixit, for extending his support in ways more than one.

Thank you to all of the above people for their enthusiasm, and confidence in me.
 1

CHAPTER 1. INTRODUCTION

Defects, big or small, can cause major disasters. Aircrafts have crashed, oil

pipelines have ruptured and nuclear reactors have failed because of undetected cracks.

In order to assess the safety of a part , it is important to establish (without damaging

it) whether it contains structural defects, and if so, to decide whether it can continue

to be used in operation. All this needs to be done without damaging the part.

Non-destructive evaluation (NDE) techniques are used extensively to accomplish this

inspection [1].

The aim of NDE is not to reject a product having a defect. This would imply

that finding smaller flaws would lead to rejecting more parts. On the contrary, the

aim of NDE is to characterize flaws and to evaluate their severity. Characterization

of flaws implies determining their size, shape, orientation, etc. By characterizing the

flaws, it is possible to quantify the safe limits for operation of the specimen being

tested.

Many nondestructive (NDT) methods have reached the stage of development

where they can be used by a semi-skilled operator following detailed procedural in-

structions, with safeguards built into the system. However, given the working con-

ditions and the tedious nature of evaluation, the accuracy of operator interpretation

can definitely be improved by enhancing the signals using signal and image processing
 2

techniques. With rapid advances made in the field of Very Large Scale Integration

(VLSI) technology, it is not inconceivable to have more sophisticated equipment to

pre-process the large volume of data generated in real-time. This increases both the

speed and the accuracy of evaluation of the operator, thus reducing inspection costs.

During the last few years, the potential economic pay-off of NDE has been recognized.

As a result there has been a tremendous emphasis on research towards improving all

aspects of nondestructive testing. A variety of NDE techniques have been developed

over the years to cater to a number of application problems. A brief summary of

some of these methods is given below.

NDE Techniques

Generally, defects are characterized into three main groups. The groups are:

1. Inherent defects (resulting from the manufacture of the raw material).

2. Fabrication defects (resulting from the conversion of the raw material to a

finished part).

3. Service induced defects (generated during the operation of a part) [2].

NDT techniques aim at detecting and characterizing all the above classes of defects.

Some of the most commonly employed nondestructive techniques are ultrasonic, eddy

current, magnetostatic, X-ray radiography and acoustic emission.

The ultrasonic testing method involves inducing high frequency (greater than

20 KHz) pulses into the specimen under test, and analyzing the reflected or scat-

tered energy for defect characterization. Eddy current testing involves the interac-

tion between an induced eddy current electromagnetic field and the test object. The
 3

corresponding change in the impedance of the probe coil constitutes the eddy cur-

rent signal, which contains information related to the condition of the specimen. In

magnetostatic NDE, the specimen, which is ferromagnetic, is magnetized by suitable

methods, and flaws at the surface or subsurface distort the magnetic field , causing

local flux leakage fields. In X-ray radiography, X-rays are passed through the test

specimen, and defects having a different density compared to the surrounding matter

are revealed by a localized difference in the blackness of the film image. In acous-

tic emission NDE, the sound emitted by solids subjected to a high level of stress is

analyzed for flaw characterization [3] [4].

Each of these NDT techniques have a wide range of applications and it is not

easy to compare the overall performance of one NDT technique over others.

Problem Statement

The problem addressed in this thesis is one module of the overall system to

automate flaw detection, location and classification in Boiling Water Reactors (BWR)

tubings. Stainless steel pipes of thickness 0.69 inches, and an outer diameter of

12 inches are inspected for cracks using ultrasonic nondestructive testing methods.

Cracks in these pipes, mainly develop close to the weld joining two pipes. Specifically,

the cracks develop in the heat affected zone (HAZ) of the pipes. The heat affected

zone extends axially up to 0.5 inches from the weld for thin pipes (thickness less than

0.5 inches) and upto 1.25 inches for thick pipes (thickness from 1.25 inches to 1.5

inches). A typical weld geometry is shown in Figure 1.1. The counterbore represents

the thinning of the pipes to ensure that pipes being welded are of the same thickness.

Most of the cracks are due to mechanical fatigue, thermal fatigue, or intergranu-
 4

Heat effected zone Weld crown

Pipe outer diameter

Pipe 1 Pipe 2

Pipe inner diameter

Counterbore

Flow

Figure 1.1: Typical nuclear pipe and weld geometry

lar stress corrosion. Each of these have a characteristic signature as far as the reflec-

tion and scattering of the induced signal is concerned. The pipes are tested for cracks

using ultrasonic NDE techniques. A typical setup comprises of an operator in com-

plete radiation-proof gear getting into t he reactor, scanning the pipes in an attempt

to characterize the reflected signal as being from counterbore, root-weld or crack.

Due to the high noise environment of the setup, it is quite probable, even for a highly

skilled operator, to misclassify the reflected signal. This may lead t~ catastrophic

results. It is, therefore, necessary for some preprocessing to be performed before the

operator interprets the signal. Synthetic aperture focusing technique (SAFT) is one

such widely used method to enhance the signal to noise ratio (SNR) of ultrasonic

and radar signals. Most of the research uptil now has concentrated on performing
 5

SAFT on data collected using the normal mode of incidence of the ultrasonic beam.

However, as will be discussed in later chapters, nuclear NDE uses the angle-beam

mode of inspection of test specimens. There has not been much research done in this

field. In this research, a SAFT algorithm has been developed for preprocessing the

ultrasonic data collected using the angle-beam technique. The algorithm has been

tested with ultrasonic signals from the industry, and initial results look promising,

thus enhancing detectability of defects. Algorithms related to flaw location have

also been developed in this research. These algorithms are valuable in locating and

characterising the defects , and other features of the specimen being evaluated.

The thesis is organized as follows. Chapter 2 provides information on the vari-

ous techniques and modes of ultrasonic NDE data collection and representation. A

discussion on the principles and formulation of normal SAFT is given in chapter 3.

Chapter 4 describes angle-beam SAFT in detail and presents the algorithm used in

t he research. A discussion of the results along with conclusions and areas for further

investigation are presented in Chapter 5.

 6

CHAPTER 2. ULTRASONIC NONDESTRUCTIVE EVALUATION

Ultrasound, as used in nondestructive testing, refers to sound waves slightly

above the audible range of the human ear, i.e. greater than 20 KHz. For years,

ultrasound has been used to cover all aspects of NDT research, ranging from the

detection of internal crack in materials to small sub-surface defects. These methods

have played a major role in the quality inspection of partially manufactured compo-

nents, as well as nondestructive testing of parts being used.

There are several modes of ultrasound propagation. The generation and propagation

of ultrasonic waves is considered in the next section [5].

Generation and Propagation of Ultrasonic Waves

Crystals of certain minerals and salts, when subjected to alternating electric

charges, expand and contract under the influence of these charges. Conversely, these

materials when subject to alternate compression and tension generate alternating

charges. This is the piezo electric effect. A large number of piezo-electric materials,

including man-made ceramics and polymers, have been used in the construction of

modern ultrasound transducers. However, the original material used was, and still

is, natural quartz. When a piezoelectric material is subject to an alternating voltage

across its thickness , vibrations, caused by contraction and expansion of the disc,
 7

gives rise to a compressional wave normal to the disc's surface. When the transducer

crystal vibrates at it 's natural frequency, wave generation is most efficient. Piezo

electric materials can be used for generating as well as detecting sound waves [5].

Ultrasonic waves are elastic waves, which can be transmitted through any medium.

However, a required criteria for the propagation of ultrasonic waves is that the

medium should be continuous for the energy to propagate freely. Any discontinuity,

such as internal voids, delaminations or cracks, will interfere with the transmission

of ultrasonic signals.

UT waves propagate in two dominant modes, namely longitudinal and shear. In

fluid medium, ultrasonic waves propagate in the longitudinal or compressional mode,

wherein the wave propagates in the direction of particle vibration [6]. However, in

solids, a shear wave component may arise, wherein the wave displacement is normal

to the direction of propagation. Elastic surface waves, called Rayleigh waves can also

occur.

The longitudinal mode of wave transmission is probably the most widely used in

ultrasonic testing. In most common materials, this wave has a short wave length in

comparison with the cross sectional area of the transducer. Consequently, the wave

energy is focused into a sharp beam with a small divergence. The main use of longitu-

dinal waves is for the detection and location of cracks that present a reasonably large

frontal area to the test surface. In fluids the longitudinal velocity 131 of ultrasound

waves is given by

(2.1)

where la is the adiabatic volume elasticity, and pis the density of the fluid. However,

for solids the expression for the longitudinal velocity is not as trivial. The longitudinal
 8

velocity ,,J 1 in the case of solids is given by

c(l - K)
(2.2)

where E is the Young's modulus of elasticity, K is the Poisson's ratio and p is the

density of the solid.

Shear waves have lower velocities than longitudinal waves. As a result, shear

waves have shorter wavelengths than longitudinal waves of the same frequency. This

shorter wavelength makes shear waves more sensitive to smaller defects, and conse-

quently they are more easily scattered within the specimen. Shear beams are mainly

used for inspection with the angle-beam technique. The shear wave velocity ,,Js of

propagation of ultrasound in solids is is given by

(2.3)

where ( is the modulus of rigidity of the material.

As sound propagates through a material it 1s attenuated with the distance of

propagation z [7], according to the equation

(2.4)

where I is the intensity at depth z and 10 is the intensity at the surface of the material

(z = 0). The attenuation coefficient, a, depends on the frequency, and increases with
an increase in the frequency. However, the frequency f, along with the velocity

of sound, 73, in a specific medium, determine the wavelength, A, of the ultrasonic

pulse. Note that the wavelength must be kept as short as possible to improve axial
 9

resolution [8]. These three variables are related by the following equation

(2.5)

It can be verified that higher the frequency, lower the wavelength, the better the

resolution. The trade-off is the high attenuation at higher frequencies.

Comparison of Ultrasonic NDT with other Techniques

Ultrasound inspection has several distinct advantages over other NDT meth-

ods [2]. Among the most significant are:

1. The inspection requires only one surface for inspection.

2. The ability of ultrasonic test waves to penetrate thick specimen.

3. Its ability to locate the exact position of a flaw.

4. The capability to detect minute flaws under appropriate conditions.

5. The compatibility of ultrasonic inspection for automated processing.

Ultrasonics, as do other NDT methods, has disadvantages , which require con-

sideration. Among these are:

1. The test object must be able to conduct sound without excessive losses.

2. Discontinuities just beneath the surface (dead zone) may not be detectable.

3. Need for qualified technical personnel to operate and interpret the information

resulting from an ultrasonic inspection.

 10

4. The geometry of the test piece may make ultrasonic inspection difficult or

impossible.

5. The necessity of a good contact for the sound to be transmitted into and re-

turned from the test piece.

6. The relatively small area of inspection due to the small area of the search

unit. This limitation makes the inspection of large surfaces dependent on using

multiple transducers and complicated mechanical scanning equipment [2].

U ltrasonic Testing Techniques

There are two methods for carrying out a UT inspection:

1. Pulse-echo

2. Pitch-catch

Both these methods rely upon the fact that inhomogeneities produce variations in

impedance of the surrounding media [4]. Each of these is detailed in the following

subsections.

Pitch-catch technique

In the pitch-catch or through-transmission mode of UT testing, two transducers

are applied on either side of the specimen under test as shown in Figure 2.1. This

in turn requires both surfaces of the test piece to be accessible, which may not be

possible in some cases.

 11

T
Front Wall Crystal
I I
I
I

I
r I
I

v ~
I
I I I
I
I Defect '
I I
I
\ Test Specimen
Back Wall Crystal
R

Figure 2.1: Normal pitch-catch technique of ultrasonic testing

One transducer injects (pitches) a sound pulse into the test piece, while the other

transducer receives (catches) the transmitted sound. Any inhomogeneity in the path

of the beam causes some of the ultrasonic energy to be absorbed or reflected, and the

eviden ce of the existence of the defect is shown as a reduction in the total amount of

the energy received at the receiving transducer, remote from the source of the energy.

In addition to flaw detection, transmission methods are used to determine grain size,

porosity, physical properties and wall thickness measurement.

A major difficulty encountered with t he pitch-catch technique, is that m case

of presence of bonding of materials in the test specimen, the acoustic transmission

properties of the materials vary [4]. The pulse-echo method is better suited for such
 12

applications, and is discussed in the following subsection.

Pulse-echo technique

The pulse-echo technique is the most popular UT testing method and is used

most often in industrial ultrasonic flaw detection systems. In the pulse-echo system,

an electric pulse is applied to the transmitter probe, which produces a short ultra-

sonic pulse which is propagated into the specimen through a couplant layer. The

same transducer is used as the receiver, since piezo-electric transducer generates me-

chanical vibrations under electrical stimuli and vice-versa. The transmitted beam is

reflected from the opposite surface or the defect back to the transducer as shown in

the Figure 2.2.

The pulse-echo technique depends upon the detection of the energy reflected back

to its source. For bonded structure, pulse-echo is a better option, since it is more

direct than the pitch-catch technique, as the actual reflection from the area under

examination is detected and evaluated. Unlike pitch-catch, pulse-echo technique

requires access to only one surface of the test specimen, which is an added benefit.

Other techniques

Besides these two main techniques, other methods used for ultrasonic testing are

resonance, acoustic imaging and frequency modulation [2].

The resonance technique utilizes the ability of a body to resonate due to ar-

tificially induced resonating frequencies. After the "harmonic frequency" has been

determined for the reference material, test pieces are subjected to the same frequency

or frequency spectrum. The lack of induced vibrations indicates a difference between

 13

T R
Front Wall Crystal
I I
I I
I
I

~
1 ~
r I

,;
I
I

Defect
"
I
~ Test Specimen
Back Wall

Figure 2.2: Normal pulse-echo technique of ultrasonic testing

the standard and the test piece. Generally, this method of ult rasonic testing is limited

to thin sections.

The acoustical imaging technique uses the visual indication of the sound energy

transmitted through a test piece. The methods of visualization include special elec-

tronic tubes, temperature sensitive compounds, lasers, and television apparatus to

"visualize" the test piece.

The frequency modulation method of testing employs one transducer whose fre-

quency is continually varying. Sound is constantly sent and received by the trans-

ducer. The ability of the transducer to receive a signal while it is emitting a signal is

due to the differences in frequencies of the signal received and the signal being sent.
 14

M odes of U ltrasonic Testing

An additional parameter that is usually varied in ultrasonic testing using either

the pulse-echo or pitch-catch mode, is the angle of incidence of the ultrasonic beam

into the test specimen. Based on this parameter, there are two modes of opera-

tion with each of the two techniques (pulse-echo and pitch-catch), namely normal

incidence and angle-beam incidence.

Normal incidence

In normal incidence ultrasonic testing, the ultrasonic beam is induced normal to

the major axis of the specimen being tested. It is most effective in flaw character-

ization where the flaws are mainly parallel to the axis of the unit under test. The

discussion in the earlier sections has implicitly assumed a normal incidence mode.

Figures 2.1 and 2.2 display normal incidence pitch-catch and pulse-echo modes of

testing respectively. Note that normal incidence testing uses longitudinal waves as

the sound waves used for flaw characterization.

Angle beam incidence

In order for a reflector to be displayed, t he axis of the sound beam must be per-

·pendicular to the reflector. Normal beam transducers are effective only for detecting

flaws that are parallel to the test surface. vVhen the largest face of the discontinuity

is expected to be at orientations other than parallel to the surface, angle-beam mode

is used for flaw detection.

In case of the nuclear pipe weld geometry illustrated in the Figure 1.1 , as stated

earlier, most of the cracks occur in the vicinity of the weld material, i.e. in the
 15

heat affected zone. A characteristic of the weld material is the granular structure,

resulting in scattering of the UT energy. Hence, it is not possible for a transducer to

be placed on the weld crown to detect cracks underneath using normally incidence

waves. Hence the technique used for ultrasonic testing of welded pipes is angle-beam

pulse-echo, which is a standard technique used in the industry.

Angle beams are produced in the test material using the principle of refraction.

Refraction is the bending of a sound beam when it passes through an interface be-

tween materials of different acoustic velocity. The relationship between the angle of

incidence and the angle of refraction is given by Snell's law, stated below

SinBl t91
(2.6)
SinB2 t92
where Bl is the angle of incidence, 82 is the angle of refraction, t91 is the velocity of

sound in the incident medium an t92 is the velocity of sound in the refracted medium.

The angle of refraction depends upon, both the incident angle and the ratio of acoustic

velocities through which the incident and refracted beams are travelling. As the angle

of the incident beam increases, the angle of the refracted beam also increases. At a

critical angle, the refracted beams are refracted go 0 , and are in effect, fully reflected

within the first material. This is called as the first critical angle, and only the shear

component of the refracted beam is transmitted to the second material. The second

critical angle is the incident angle that causes the refracted shear beam to be refracted
goo.

One of the most commonly used technique, for angle beam generation is the use

nf a suitable angle wedge, to launch the beam into the specimen, at the required

angle. Typical angles for angle-beam incidence is the 30° - 60° range.

Figure 2.3 shows angle-beam pitch-catch mode while Figure 2.4 shows angle-
 16

beam pulse-echo mode of ultrasonic testing.

Ultrasonic Imaging

The basic concepts underlying the imaging techniques using ultrasound are dis-

cussed in this section. The images produced by ultrasonic waves represent the internal

interaction with the mechanical properties of defects and other features in material,

and hence, serve as a powerful tool in modern NDT research [5]. This section is

intended to cover some of the imaging methodologies involved in the application of

defect characterization.

Front Wall , Wedge

\
I \
I \
I \
\ '\
\
\
\

'
\
\ '

, Defect,'
\

\
\

Test Specimen
Back Wall Wedge

Figure 2.3: Angle-beam pitch-catch technique of ultrasonic testing

 17

T
\ \

Front Wall
\ \

, , ',
I\
\

\ wedge
\
\
I \
\
I \
\
I \
\

,
I \

~
\
\
\ \

~
I

~
I
I
I Defect
Test Specimen
Back Wall

Figure 2.4: Angle-beam pulse-echo technique of ultrasonic testing

A-Scan

There are several ways in which received signals can be displayed in order to

extract the necessary information describing any particular medium, during a typical

ultrasonic test.

Ideally the display signal represents the reflectivity of the work piece, as a function

of time, where time is a measure of the depth into the specimen [9]. This display of

the reflectivity as a function of time is most commonly used and is referred to as the

A-scan signal.

A typical A-scan of a stainless steel pipe being scanned, using the angle-beam,

pulse-echo technique is shown in Figure 2.5. The largest pulse at the left-hand side
 18

corresponds to the front wall echo of the sample and the successive blips correspond

to the echoes emanating from defects. The amplitude of these echoes is generally

proportional to the size of the reflecting surface, but is also affected by the attenuation

effects introduced by both the traveled distance and the inherent composition of the

material itself. The A-scan is quite often used to determine the location of the

reflector, using the time of flight (TOF), and the velocity of sound in the medium.

For instance, consider the flaw in the sample whose A-scan is displayed in Fig-

ure 2.5. The A-scan has been obtained using angle-beam pulse-echo technology in

a stainless steel pipe. The angle of incidence B of the beam is 43° and the velocity

11 of sound in stainless steel is 0.11 inches/ µsecond. The corresponding geometry is

shown in the Figure 2.6. In Figure 2.5, the x-axis represents the time of flight, and

the y-axis represents the amplitude of the reflected signal. The time of flight (TOF)

corresponding to the maximium amplitude in the reflected signal is 29 µs.

The depth d of the flaw from the scanning surface can be calculated using

d = TOFv cos B (2.7)

For the given A-scan this value comes to 2.33 inches below the top surface which

is indeed the location of the flaw

B-Scan

The B-scan takes the same signals received in the A-scans and presents them in

a different pattern. It displays the reflectivity of a two dimensional slice through a

portion of the structure [10].

This format enables the recording of a defect location within a material. The
 25

CHAPTER 3. SAFT

Ultrasonic testing technique is used extensively because of its ability to pene-

trate thick specimen, to locate the exact location of a flaw, and to detect minute

flaws under appropriate conditions. However, there is excessive noise in the reflected

signals from grain boundaries and discontinuities in the geometry. Many signal pro-

cessing techniques have been developed over the years for enhancing the signal to

noise ratio of UT signals. This chapter describes the synthetic aperture focusing

technique (SAFT), which is used extensively in radar and ultrasonic NDE for image

enhancement.

Synthetic-aperture Fundamentals

Synthetic-aperture focusing refers to a process in which the focal properties of

a large aperture focused transducer are simulated, by a series of measurement made

using a small aperture transducer scanning over a large area. Synthetic-aperture

imaging can realize the theoretical resolution potential of ultrasonic waves, improve

the signal to noise ratio of the image, and take into account such complex material

interactions as irregular refractive interfaces and the complex angular scattering of

ultrasonic waves by a defect [11]. The synthetic-aperture method accomplishes these

tasks by extensive processing of the data from an ultrasonic transducer, in order to

 20

Transducer

Theta Fop
d * v·-.
-

Test Specimen

Figure 2.6: Geometry for the flaw depth calculation

abundantly, due to the intrinsic nature of the SAFT algorithm to work on a collection

of A-scans at a time.

C-Scan

The C-scan presentation is a plan view of the part under test, and a flaw is

indicated in a manner similar to an X-ray of the specimen. In a C-scan, corresponding

to a time slice of a region, the image pixel intensity is obtained by observing the peaks

in t he time slice of the scanned A-scans. The method of raster scanning of a sample

is illustrated in Figure 2.9. A typical C-scan of the same pipe, whose A-scan and

B-scan are illustrated in Figure 2.5 and 2.8 respectively, is illustrated in Figure 2.10.
 21

A-SCANS

Echo From Defect

'
''
Transducer
' I

I'
I '
I
I '

'
'
'
I
I

I
0 \
I
Defect ' \
'

Figure 2.7: B-scan imaging system

Accurate flaw depth cannot be determined from the C-scan presentation, although

the flaw sizing in the two dimension can be observed.

Data Storage Is sues

An important issue relevant to NDE in general and UT NDE in particular, is

the storage and processing of huge amounts of data. Currently, most of the data is

digitized and stored on computer-supported storage devices. These are mainly in t he

 22

Figure 2.8: B-scan

form of :floppy disks, hard drives and tape drives. A typical setup in which a 10" by

3" region of a pipe is scanned using a 2 MHz transducer, and then sampled at 25

mega-samples/second may lead to data files , as large as 100 Mbytes each. This slows

down the entire process of performing the subsequent signal processing on the data,

and hence it would be beneficial to the industry if real-time techniques are developed

using dedicated hardware resources for performing the signal processing in hardware,

instead of software.
 23

Top of Specimen

Transducer

Figure 2.9: Raster scanning of the test specimen

 24

10 20 30 40 50 60 70 80 90 100

Figure 2.10: C-scan

 25

CHAPTER 3. SAFT

Ultrasonic testing technique is used extensively because of its ability to pene-

trate thick specimen, to locate the exact location of a flaw, and to detect minute

flaws under appropriate conditions. However, there is excessive noise in the reflected

signals from grain boundaries and discontinuities in the geometry. Many signal pro-

cessing techniques have been developed over the years for enhancing the signal to

noise ratio of UT signals. This chapter describes the synthetic aperture focusing

technique (SAFT), which is used extensively in radar and ultrasonic NDE for image

enhancement.

Synthetic-aperture Fundamentals

Synthetic-aperture focusing refers to a process in which the focal properties of

a large aperture focused transducer are simulated, by a series of measurement made

using a small aperture transducer scanning over a large area. Synthetic-aperture

imaging can realize the theoretical resolution potential of ultrasonic waves, improve

the signal to noise ratio of the image, and take into account such complex material

interactions as irregular refractive interfaces and the complex angular scattering of

ultrasonic waves by a defect [11]. The synthetic-aperture method accomplishes these

tasks by extensive processing of the data from an ultrasonic transducer, in order to

 26

encode the resulting image information in such a manner, that the necessary human

interpretation can be done faster and more easily.

Ultrasonic images can contain, and effectively convey, an extremely large amount

of information. It is widely accepted among researchers that the ultrasonic pulse re-

turned from a test sample contains much more information about the interaction

between the ultrasound and the material, than is commonly extracted from con-

ventional displays. The more adept practitioners of the ultrasonic art can either

consciously or unconsciously make use of this wealth of information. The presence of

such adept operators is encouraging to those researches attempting to extract more

information from the ultrasonic signal, as it implies the existence of the data they

desire, but it is discouraging to those managers wishing to employ ultrasonic testing

on a routine basis, as it promises inconsistent results over time and with different

personnels. SAFT is able to enhance the images, so that it is possible for many more

people to reliably interpret the ultrasonic signals [11].

The synthetic-aperture technique utilizes the information in the ultrasonic data

to produce high quality images. The SAFT process presents the available data in

an image format that is much easier to interpret. The concept of synthetic-aperture

imaging is borrowed from radar (microwave) imaging techniques of the same name.

While there are significant differences between the ultrasonic and the microwave

implementation, there is a commonality in concepts and motivation, since the wave

physics of radar antenna and ultrasonic transducer is quite similar. The designers of

both, ultrasonic transducer and air borne radars , realized that there were practical

limits to their ability to fabricate the large antenna or transducers required to form

high resolution images. The solution in both fields is to move a small transducer
 27

in such a manner that the collected data can be later processed, to simulate the

desired large transducer. This processing, in effect, synthesizes the physical processes

occurring in a large, but unobtainable, aperture transducer and thus is referred to as

synthetic-aperture imaging

In both, ultrasonic and radar, the data are focussed, not by some physical mech-

anism employed during data acquisition, such as a curved transducer, but by sub-

sequent data processing. Since this data processing is done in software, it offers

considerable advantages in terms of fiexibilty and evaluation, as against varying the

focii of the transducers physically to image different parts.

Both ultrasound and radar synthetic-aperture techniques explicitly make use of

both the amplitude and the phase information contained in their respective signals.

Synthetic-aperture ultrasonic images are formed from RF wave forms, and the ultra-

sonic phase information manifested in the processed image through the mechanism of

constructive and destructive interference, plays an important role, in the subjective

improvement of the ultrasonic image [11 J.

Quantitative Interpretation of SAFT

Digital ultrasonic imaging can be viewed as collection of ultrasonic data followed

by processing of the data by an algorithm which represents the ultrasonic scattering

information in an image. This section presents the theoretical development of the

ultrasonic defect characterization and the resulting SAFT processing.

Quantitative interpretation of ultrasonic images requires careful investigation

and understanding of the inverse scattering scheme under consideration [12]. The
 27

in such a manner that the collected data can be later processed, to simulate the

desired large transducer. This processing, in effect, synthesizes the physical processes

occurring in a large, but unobtainable, aperture transducer and thus is referred to as

synthetic-aperture imaging

In both, ultrasonic and radar, the data are focussed, not by some physical mech-

anism employed during data acquisition, such as a curved transducer, but by sub-

sequent data processing. Since this data processing is done in software, it offers

considerable advantages in terms of fiexibilty and evaluation, as against varying the

focii of the transducers physically to image different parts.

Both ultrasound and radar synthetic-aperture techniques explicitly make use of

both the amplitude and the phase information contained in their respective signals.

Synthetic-aperture ultrasonic images are formed from RF wave forms, and the ultra-

sonic phase information manifested in the processed image through the mechanism of

constructive and destructive interference, plays an important role, in the subjective

improvement of the ultrasonic image [11 J.

Quantitative Interpretation of SAFT

Digital ultrasonic imaging can be viewed as collection of ultrasonic data followed

by processing of the data by an algorithm which represents the ultrasonic scattering

information in an image. This section presents the theoretical development of the

ultrasonic defect characterization and the resulting SAFT processing.

Quantitative interpretation of ultrasonic images requires careful investigation

and understanding of the inverse scattering scheme under consideration [12]. The
 29

increasingly difficult. One might wish to recover information by variation of the

transmitting transducer, Ro , so that information from all possible points is made

available to the receiving transducer.

Variation of R""'a, however does not extract all the information available, and

it would definitely be better if the receiving transducer, R, could be synthetically

increased, using the variations of R""'a. Synthetic aperture focusing technique is a

highly successful imaging scheme, which is based on the observation that there is a

scattering of amplitude travel-time wavefronts from the defects [13] [14]. The goal

of SAFT is to back-propagate these wavefronts to the scattering defect where they

came from. This is achieved implying the following assumptions:

1. The wavefronts are essentially treated as scalar wavefields; hence <fi becomes
-
<Pi and <Ps becomes <Ps·

2. The incident wavefield is supposed to emanate from a point source located at

Ra.

3. The defect is decomposed into single non-interacting scatterers, which act as

point sources after illumination with the incident wavefield.

4. The incident pulse is assumed to be 'nearly' infinitely broad-band, i.e. its time

structure is represented by F (t) , where F(t) denotes a very sharp nearly 8-like

impulse.

In the following sub-section we describe briefly how these assumptions lead to

the formulation of the imaging SAFT algorithms.

 28

important parameters of an ultrasonic scattering experiment are the mode of oper-

ation, be it either pulse-echo or pitch-catch, as well as the bandwidth of the trans-

mitted pulse. For imaging purposes, the scattered wavefield has to be observed at a

considerable number of spatial points within a synthetic aperture.

Let us denote the the position of the transmitting transducer by the spatial

vector R~ and that of the receiving transducer by R. Let us call the incident wave

field ii and the observable scattered field as</;~. Now</;~ is a function of (<h, 1(R'),
R~, R, t) where t is the time and 1( R') is a description of the geometry of the material
defect, which is defined as follows

1(R') 1 R' inside the defect (3.1)

0 -
R' outside the defect

The following relations have been developed assuming that the defect resides in a

linear, isotropic and homogeneous material. Now, assuming that ii is a short broad
band pulse, then data collection implies the observation and digital storage of ¢;~ as

a function of the receiving transducer R and the time t [12]. Consequently, an RF

data field is obtained, which supplies us with information to be exploited for imaging

purposes.

It is intuitively clear that confining the receiving transducer R to a spatially

limited region and restricting the frequency bandwidth of the incident pulse results

in a deterioration of the image quality, depending upon the complexity of the defect

geometry 1(R'). Stated in simple terms, this means that we would like the receiving

transducer (aperture) to be as large as possible. However, because of the manufac-

turing limitations on the size of the transducer, it is not feasible to increase the size

beyond a certain limit. Therefore, quantitative nondestructive evaluation becomes

 29

increasingly difficult. One might wish to recover information by variation of the

transmitting transducer, Ro , so that information from all possible points is made

available to the receiving transducer.

Variation of R""'a, however does not extract all the information available, and

it would definitely be better if the receiving transducer, R, could be synthetically

increased, using the variations of R""'a. Synthetic aperture focusing technique is a

highly successful imaging scheme, which is based on the observation that there is a

scattering of amplitude travel-time wavefronts from the defects [13] [14]. The goal

of SAFT is to back-propagate these wavefronts to the scattering defect where they

came from. This is achieved implying the following assumptions:

1. The wavefronts are essentially treated as scalar wavefields; hence <fi becomes
-
<Pi and <Ps becomes <Ps·

2. The incident wavefield is supposed to emanate from a point source located at

Ra.

3. The defect is decomposed into single non-interacting scatterers, which act as

point sources after illumination with the incident wavefield.

4. The incident pulse is assumed to be 'nearly' infinitely broad-band, i.e. its time

structure is represented by F (t) , where F(t) denotes a very sharp nearly 8-like

impulse.

In the following sub-section we describe briefly how these assumptions lead to

the formulation of the imaging SAFT algorithms.

 31

v(R')

R'
v

Figure 3.1: Scattering object and measuring surface

The 'inversion' of the Equation 3.6, i.e. the determination of 1(R') through the

knowledge of <l>s( R , t) can be achieved, recognizing the fact that every single point

scattered, for example Rs, contributes linearly to the scattered field [12] .

(3.7)

where R is varied along the measurement surface SM by some appropriate procedure.

Due to the sharpness of the impulse F(t), <P~·( R, t) is significantly non-zero only

for times

(3.8)

Intuitively, t s(R) corresponds to t he time when the incident wavefront from Ro

 30

Time domain back-propagation of wavefronts: SAFT

Figure 3.1 gives a geometrical illust ration of the scattering objects and the mea-

surement surface. Suppose the transmitting transducer is modeled as a scalar point

source Ro on a measurement surface SM,transmitting the impulse F (t) . The incident

field at the point R' is then given by

F(t _ IR'-Ro l)
ef>i(Ro, R' ,t) = IR'-Ro l (3.2)

where c is the wave velocity in the medium.

The material inhomogeneity represented by the defect, and its illumination by

the incident field, gives rise to an induced source ¢>B(Ro, t) which is responsible for

the scattered field . The entire defect volume V enclosed by 1(R' ) can be represented

by the scattered field equation

,;., (Rt)=
'f' s '
JJ{J</>B(R',IR't -- RI~) d3 R'
v
(3.3)

Now as per the third assumption, ¢>B is modeled as a superposition of indepen-

dent point sources, hence

¢>B(R', t) = 1(R' )</>i(Ro, R' , t) (3.4)

Therefore the scattered field becomes

,;., (R t ) =
'f's '
J J1(R' )¢;i(Ro,IR' R',
J oo
- oo - RI
t - ~) d R' 3
(3.5)

Using [3.2] for </>i,

<l>s (R, t ) -
_J Ji (R' )FIR'(t-_ R\\R' - Ro \
J oo
- oo
IR'~RI _ IR'~Rol )
dR
3 /
(3.6)
 32

reaches Rs and then to R.

A reconstruction of the defect geometry point Rs, given by 1(Rs) can therefore

be defined by integrating ¢~· (R, t) along the measurement surface SM, with respect

to R, for exactly the points defined by Equation 3.8. All contributions of the point

scatterer are summed up. It is intuitively clear that the point R' =/= Rs receives a

much smaller amplitude from ts(R) because the curve,

t'(R') = IR' - Roi+ IR' - RI (3.9)

c c

has only one point in common, namely (R' =Rs), with equation 3.8.

Accounting for every point scatterer within 1(R'), the defect can be recon-

structed as

1(R') = j fsM ef>s(R, t'(R'))IR' - RollR' - RI dD (3.10)

The integration is with respect to the surface elements on SM.

Qualitative Analysis of SAFT

As has been stated in earlier sections, SAFT refers to a process in which the

focal properties of a large-aperture focused transducer are synthesized by a series of

measurements, made using a small-aperture transducer, that scans a large area [15].

In other words, SAFT is the process of simulating a focused transducer.

SAFT has several advantages over physical focusing techniques. SAFT has a

great deal of flexibility in data collection and image formation [16]. If a test object

of irregular size and shape has to be inspected, the data can be processed in a way

that minimizes the error caused by refraction. To make a similar correction with a
 33

physical focus system would require a transducer with different focal properties to be

designed for every different part to be inspected [16].

In case of the nuclear pipe UT inspection technique, a small contact transducer

is positioned on the test specimen. It simulates a "point" transmitter-receiver, and

a broad divergent ultrasonic beam is launched into the object. As the transducer is

scanned over the surface of the object, a series of amplitude v /s time A-scans are

recorded for each position of the transducer. Every reflector produces a collection

of echoes in the A-scan records. The locus of each elementary reflector produces a

characteristic hyperbolic curve, in a set of linear positions, as shown in Figure 3.2.

The apex and curvature of the hyperbola are completely determined by the depth

o f the reflector within the test object. This deterministic relationship between the

echo location in the recorded A-scans and the location of the reflectors wit hin the

test object makes it possible to generate a processed image from the new data by the

coherent summation process [16].

If the scanning and surface geometries are known, it is possible to accurately

predict the shape of the locus of echoes for each point within the test object. The

process of coherent summation involves shifting the raw A-scans by the predicted

time-delay and summing of the shifted echo data. The time-compensated A-scans

are then averaged over a numLer of positions, which is the aperture size. The aperture

of the synthesized transducer is the number of A-scans that are coherently summed

in SAFT. For all processing purposes, the aperture can be treated as a window into

the collection of A-scans. The time-shifting is done with respect to the A-scan which

is at the center of the aperture ( COA) . Following the averaging, the resultant vector

is placed at the center of the chosen aperture.

 34

Figure 3.2: Hyperbolic reflections from the same reflector in different A-scans

If the point being imaged corresponds to a reflection from a defect, or in other

words, correlates with a locus of A-scan echoes, then all of the A-scans will add

in phase during the process of spatial averaging to produce a well-defined signal.

This is illustrated in Figure 3.3. However, if the point being imaged, corresponds

to noise, or in other words does not correlate with a locus of A-scan echoes, then

destructive interference of the A-scans will take place, and the spatial average will be

nearly zero. Figure 3.4 illustrates this destructive interference. Thus, there is both

an improvement in signal level and a reduction on noise level producing an image

with a high signal to noise ratio. It should be noted that the SAFT procedure is

performed at a specific depth site in the specimen under test.

 35

Figure 3.3: Constructive interference during SAFT

 36

I
I '
I
I

Figure 3.4: Destructive interference during SAFT

 37

CH APTER 4. ANGLE BEAM SAFT

Angle Beam Theory

In order for a reflector to be displayed, the axis of the sound beam must be

perpendicular to the reflector. Straight beam transducers are effective only for reflec-

tors that are parallel to the test surface. When the largest face of the discontinuity

is expected to be at an orientation, other than parallel to the surface, angle beam

transducers are used to aim the axis of the beam perpendicular to the discontinuities.

There are different ways of introducing angle beams into the test material, de-

pending on the coupling technique used. In order for the angle beams to be produced

in the test material, the transducer's beam axis must be aimed towards the test sur-

face at an angle. Contact testing employs angle wedges, immersion testing permits

continuously variable angulation of the transducer by means of an adjustable manip-

ulator assembly.

Angle beams are produced in the test material using the principles of refraction.

Refraction is the bending of a sound beam when it passes through an interface be-

tween materials of different acoustic velocity. The transducer is aimed so as to direct

the beam towards the test material interface at an angle.

When a sound beam is incident to an interface at an angle other than 90°, as

illustrated in Figure 4.1, a phenomenon called mode conversion occurs. In addition

 38

to the simple reflected beam, a portion of the incident beams energy converts at the

interface to a beam with a different mode of vibration, and reflects at an angle other

than the angle of incidence (or reflection).

When sound refracts, there is mode conversion, too. Thus, there is an additional

refracted beam which is a mode conversion of the incident beam.

Incident Beam : Mode converted beam

I
I

Reflected beam

Refracted longitudinal beam

Refracted shear beam

Figure 4.1: Mode conversion of sound waves

As the angle of the incident beam increases, the angle of the refracted beams

also increases. At a critical angle, the longitudinal refracted beams is refracted 90°,

and is in effect, fully reflected within the first material. This is called as the first

critical angle and only the shear component of the refracted beam is transmitted

to the second material as illustrated in Figure 4.2. The second critical angle is the
 39

incident angle that causes the refracted shear beam to be refracted 90°. For example,

with a plastic wedge producing angle beams in carbon steel, the first critical angle is

27.2° and the second critical angle is 55.8°.

In the experiments conducted, the data has been collected from stainless steel

pipes, using plastic wedges, so that the angle of refraction is 43°. The shear waves

dominate in this particular testing setup.

Relevant Issues with Angle Beam Mode of Scanning

Three of the most important issues which distinguish and complicate angle-beam

SAFT from normal-beam SAFT are discussed below.

1. Multiple Sound paths

In case of angle-beam pulse-echo mode of inspection, with finite depth of ma-

terial, multiple sound paths can contribute to the processed image.

There are three paths that can contribute reflection echoes in the data. These

paths are shown in Figure 4.3. One needs to be aware of the physics of the sit-

uation to avoid misinterpreting images or to use the multiple path information

to build greater accuracy into the final image. In Figure 4.3, the material is of

thickness T, and the defect of interest is located at x = 0 and at a depth D in

the material [16]

• The first path is the direct back-scattered path (P1 - P 1 ) from the sound

field entrance into the part to the defect and back to the receiver. Analysis

shows that the data along this path will be deterministic, that is, one

unique object depth corresponds to one unique hyperbolic curve. This

 40

Incident Beam Mode converted beam

I

Reflected beam

Refracted shear beam

Figure 4.2: Angle of incidence is greater than t he first critical angle

curve is defined by the equation

F(x, D) = Jn 2 + x2 ( 4.1)

• The second sound path is ( P 2 - P3 - P 1 ) from the sound field entrance

point of the part, to the back surface, to the defect and finally to the

receiver. The sound path is the sum of the length of the ray paths P 1, P 2

and P3 . The equation for the corresponding curve is

F(x, D, T) = j(D 2 + X 2 ) + j(2T - D) 2 +X2 (4.2)

This path is identical to the reverse sequence of (P1 - P3 - P 2 ). This path

contributes a very large return echo since it relies on the strong forward-
 40

Incident Beam Mode converted beam

I

Reflected beam

Refracted shear beam

Figure 4.2: Angle of incidence is greater than t he first critical angle

curve is defined by the equation

F(x, D) = Jn 2 + x2 ( 4.1)

• The second sound path is ( P 2 - P3 - P 1 ) from the sound field entrance

point of the part, to the back surface, to the defect and finally to the

receiver. The sound path is the sum of the length of the ray paths P 1, P 2

and P3 . The equation for the corresponding curve is

F(x, D, T) = j(D 2 + X 2 ) + j(2T - D) 2 +X2 (4.2)

This path is identical to the reverse sequence of (P1 - P3 - P 2 ). This path

contributes a very large return echo since it relies on the strong forward-
 41

0 \
\
\I
I

,,
\ I
X=O

6 1 T

Figure 4.3: Multiple sound paths in angle-beam pitch-cat ch mode of ultrasonic test-
mg

scattered signal rather than the weaker-back scattered signal [16].

• The third sound path to be considered is from the sound field entrance

point on the part, to the back surface, to the defect , back-scatter from the

defect to the back surface and return to the receiver or ( P2 - P3 - P3 - P2 ).

This path forms the mirror image of the defect and back relies on the
usually weak back scattered signal. The equation for the corresponding

curve is

F(x , D, T ) = V(2T - D)2 + X 2 (4.3)

 42

Depending on the defect type, orientation and other characteristic, one of the

three paths will dominate the recorded experimental data. The general-purpose

SAFT algorithms is capable of constructing all three of the propagation paths

for angle-beam data.

2. Periodicity

An interesting phenomenon involving angle-beam scanning is that the same flaw

is seen from two transducer position that are a full V-apart. This is illustrated

in Figure 4.4. A half-V corresponds to the time required for the beam to travel

an entire depth of the specimen. A full-Vis the summation of two half Vs. This

periodicity of the flaw in the A-scan records occurs since, during the angular

travel of the beam, it gets reflected from the front and the back wall surfaces.

Thus when the transducer is at position T1 , the beam path to the defect is

P1 - d - P1 . When the transducer is at T2 , which one full-V away from Ti, the

sound path is P3 - P2 - P1 - d - P1 - P2 - P3 . Note that for T2, because of

the longer path, the reflection in the A-scan occurs later. Also, the amplitude

of reflection from the defect in case of T2 is smaller, because of the higher

attenuation due to the longer path.

A typical B-scan in which this periodicity is exhibited is shown in the Figure 4.5.

The A-scans have been plotted vertically. As can be seen, the same defect is

seen multiply in the upper left hand side and the lower right hand side. The

difference in their time of flights and amplitudes can also be seen.

3. Ellipsoidal Incidence

An additional complication encountered while performing angle-beam SAFT is

 43

T2 Tl
Weld
p 3~

P2

v--~

Figure 4.4: Periodic imaging of the same defect

that the metal-path traveled from all points on the face of the transducer is

not the same. Thus the emerging beam is more in the shape of an ellipsoid, a

phenomenon not encountered in normal SAFT.

The angle-beam SAFT procedure has to account for all these issues and compli-

cations. The corresponding angle-beam SAFT algorithm developed, is discussed in

the following section.

Algorithm for Angle-Beam SAFT

The following algorithm outlines the major steps for performing angle-beam

SAFT for a single depth position. It can be iterated over multiple depth positions to
 44

200

400

600

800

1000

1200

1400

1600

20 40 60 80 100 120

Figure 4.5: B-scan with periodic imaging of a crack

get the SAFT processing on multiple positions.

1. Input SAFT processing parameters: depth d, aperture a, thickness T, velocity

of sound in specimen rJ , angle of incidence (), data sampling frequency S, total

number of A-scans numA, and number of data points numD in each A-scan.

The parameters can be better recognized in Figure 4.6.

2. Read the set of A-scans constituting the B-scans b-scans and the corresponding

transducer positions pos . Initialize the result matrix sa f t to 0.

 45

~ xdist ~

d theta

J : T
:
... .. .. .. .. .. ... .... ....... ... ... .... ... ..... ... · · ·· · ··· · ·· ··· ------ ---- -··· · · ·· -· -· -- ----

Figure 4.6: Test specimen geometry for the angle beam SAFT algorithm

3. Compute the lateral distance xdist and metal path mp to the depth.

xdist = d tan e (4.4)

mp = 2Jd2 + xdist 2 (4.5)

The factor of 2 is included to take care of both the forward and backward path

of the beam.

4. For i = 1: numA

For j =-a: a

Calculate the metal path mpj(j) to the point P on the path

of center of aperture (i) from the neighboring point (i +j )

as shown in Figure 4. 7.
 46

(i+j) :--- xdist

0
'
'
'
''
'

Figure 4.7: Metal path calculation geometry

mpj(j) = 2J(pos(i + j) - pos(i) + xdist) 2 + d2 (4.6)

Calculate the time shift ts(j) to be incorporated in the

A-scan at point (i + j)

ts(j) = mpj (j) - mp *S (4.7)

-a
Shift the Ph scan and add to the result .
For k = 1 : numD

Saft(i, k) = Saft(i , k) + bscan(i + j, k + ts(j)) (4.8)

 47

end

end

end

5. Average the result over the aperture

Saft(i, k) = Saft(i, k) (4.9)

(2*a+l)

This algorithm was implemented on a number of experimental B-scan images,

and the results are presented in the following chapter.

Mapping between Image Domain and Physical Domain

The final goal of any image processing technique is to represent the result in a

manner so that it is easy to interpret. It requires a fair amount of skill and practice to

be able to map a point from an A-scan, to its corresponding location in the geometry

of the specimen being inspected. Hence, mapping techniques, which will translate a

point in an A-scan at a particular transducer position, to its corresponding position

in the specimen, will greatly enhance visual interpretation. Preliminary mapping

techniques to translate a point in an A-scan to its location in the geometry was done.

The reverse mapping from the test geometry to the location in the A-scan was also

developed. Both these techniques are described in the following sub-sections.

 48

Mapping from image domain to physical domain

Mapping from the image domain to the physical domain implies that given the

axial (a), and the circumferential (c) position of the transducer, along with a specific

time of flight (to!), it is required to calculate the corresponding point ( x, y, z ), in the

geometry of the test specimen. If v is the velocity of sound in the medium, then the

metal path mp can be calculated as

mp=v*tof/ 2 (4.10)

This can better understood by referring to Figure 4.8 The axial, x, circumferential,

·theta
z
d
J.

Figure 4.8: Mapping from image domain to physical domain

 49

y, and depth, z, locati0"' of the point in the geometry can be calculated as

x = a - mp* sinB (4.11 )

y=c (4.12)

z = (m p* cosB )modulo( d) (4.13)

Here B is the angle of incidence of the beam into the specimen, d is the thickness of

the test specimen, and

Mapping from physical domain to the image domain

Mapping from the physical to the image domain implies that, given the axial

(x), circumferential (y), and depth (z) location of a point in the specimen geometry,
it is required to calculate the corresponding positions of the transducer (a, c), along

with the time of flight tof , of the point in the A-scan record. In this case, as shown

in the Figure 4.9, for each point in the geometry, there are two primary positions

(Tl and T2) that need to be considered. One of them (Tl) falls in the first half-V

(al, cl, to fl ), and the next (T2) in the second half-V (a2, c2, tof2). In the first half-
V, the location of the transducer and the corresponding time of flight is given by the

following equations.

al = x + z * tane (4.14)

cl= y (4.15)

tofl = 2 * z/(cosB * v) (4.16)

In the second half-V, the equations for the calculations are as follows.

a2 = x + (d - z)/tan(n"/2 - B) + d * tanB (4.17)

 50

a2---~

al

T2

Figure 4.9: Mapping from physical domain to image domain

c2 = y (4.18)

tof2 = 2 * (a2 - x)/(sin() * v) (4.19)

Thus, using these equations, it is easy to map from one domain to another, thus

enhancing visualization, and understanding. Following this, the next development

is to use these functions to map the entire image information into an appropriate

3-dimensional representation of the geometry being imaged, along with the flaw rep-

resentation.
 51

CHAPTER 5. RESULTS AND DISCUSSIONS

The algorithm for performing angle-beam SAFT on B-scan images and B'-scan

image was discussed in the previous chapter. In this chapter we present the results

of implementing this algorithm on B-scan, B'-scans at single depths and multiple

depths. This will be followed by the discussion of the results and finally the future

work that can be carried out in this field.

Experimental Setup

The data for the SAFT system was obtained through the Electric Power Research

Institute. In this section, a brief summary of the entire data collection scheme is given.

In a typical setup, the UT process starts with an operator entering the reactor with

his radiation-proof gear and the testing tools. Since most of the defects occur close

to the weld, he uses normal incidence and angle beam incidence methods to check

for reflections in the received echoes in those regions. On observing the reflection

from a possible flaw, he moves the transducer in the neighboring regions to test if

the reflection is from a defect or a counter-bore or the root of the weld.

All the above features produce distinct signatures in their corresponding reflec-

tions. For example, the reflections from a root weld are usually of a longer duration,

whereas the reflection from the counterbore and the defect are much sharper. Again,
 52

to distinguish between a counterbore and a crack, the operator moves the probe par-

allel to the counterbore. Since counterbore occurs throughout the pipe, reflections

from counterbore will be present for all transducer positions along the circumference,

whereas a crack can be characterized as a sharp reflection in a restricted neighbor-

hood.

Following the detection of the flaw , the location and sizing of the flaw are done

using the velocity of the wave, the thickness of the sample and the time of flight

corresponding to the reflection.

In case of the data used for our experiments, standard 43° refracted beam was

employed, with the same plastic wedges. All the pipes were stainless steel with a

thickness of 0.69" and 12" outer diameters. The pipes were scanned using transducer

probes centered at 1, 2.25, and 5 MHz. The analog received data was digitized

using 8-bit analog-to-digital converter. The reflected data was collected for 71.92µs

for each of the transducer positions. This analog data was digitized at 25 Mega-

samples/second to give A-scans of length 1798 samples each. A typical scan area was

3" X 10" with resolution of 0.025" (which is less than a quarter of a wavelength with
velocity {) = 0.11 inches/ µs and frequency f = 2.251\1 H z) and 0.1" along the axis
and circumference of the pipe, respectively. This led to individual data files which

contains

3 10
--X-Xl798
0.025 0.1

samples

Thus each of the files was close to 22 MBytes in size.

 53

Results

This section is divided into three parts, to present results of:

1. focusing at single depth for B-scan images

2. focusing at single depth for B'-scan images

3. focusing at multiple depths for B '-scan images

In all the sections, the test specimen is a stainless steel welded pipe of outer diameter

12" and thickness 0.69" . The ultrasonic data was collected by a scanning a 3" (axial)

by 10" (circumferential) region around the weld using angle-beam, pulse-echo tech-

nique. The center, frequency of the ultrasonic data was 2.25 MHz and the angle of

inspection was 43°. A plastic wedge was used to achieve this angle. The scanning

was done in a raster format along the axial direction. The axial resolution is 0.025"

and circumferential resolution is 0.1". The reflected analog data was collected for

71.92 µ s for each of the transducer positions. This was then 8-bit sampled to give

digital data of length 1798 for each of the A-scans. A wedge delay corresponding to

6.4 µs ( or 16 samples) has to be compensated for, in each of the A-scans.

The pipe has a crack very close to the root of the weld. The circumferential span

of t he crack is 1.618" and depth is 0. 33". The geometry of the pipe and the crack

being imaged is illustrated in Figure 5.1

B -scan

As has been explained in an earlier section, B-scan is a collection of A-scans.

In this case, the B-scan is in the axial direction, i.e. it has the higher resolution. A

typical B-scan imaging is illustrated in Figure 5.2.

 54

FLOW

12.00" - - - - -

0.69"

Pipe 1 Pipe 2
I

:-.o.so'~
I I
<·-----
0.12"
Figure 5.1: Experimental pipe geometry

The B-scan of the test specimen along a particular circumferential position is

shown in Figure 5.3. As can be seen, there is a high saturation region in the initial

part. This corresponds to the "big bang" or the reflection from the front wall-wedge

interface.

As the transducer, moves away from the weld, during the B-scan imaging, the

defect and the weld are seen at later positions in time or the time of flight (TOF) .

Note that, because of the periodicity in angle-beam imaging the same reflections

are periodically repeated at a large TOF for the transducer position further away

from this weld.

The transducer raster scans in the same directions as in B-scan imaging. As a
 55

Scanning direction

<····· ···
Weld-crown
< · · ·· · ···

Weld-root
Axial direction

Figure 5.2: B-scan imaging

result, the higher resolution of 0.025" is in this direction. Consequently, SAFT can

be applied over a larger aperture, since because of the physical proximity between

the transducer locations for the B-scan, the same reflectors can be seen over a larger

windows of A-scans. Figure 5.4 shows the result of SAFT performed on the B-scan

in Figure 5.3. The aperture chosen was 10 and the depth of the SAFT corresponds

to the third leg.

The third leg was chosen, because of two reasons:

1. Due to the high gain of the receiving amplifier in the scanning system, to be

able to detect the smallest flaws , t he data is highly saturated in the first leg of

the beam and hence t his leg could not be used.

 56

200

400

600

800

1000

1200

1400

1600

20 40 60 80 100 120

Figure 5.3: A B-scan image in which the defect can be seen clearly

2. There is a larger beam spread in the third leg and hence contributions from

the reflectors occur over a larger area, thus improving the ability of the SAFT

algorithms.

Another B-scan image in which the defect or the weld are not seen clearly is

illustrated in the Figure 5.5. The result of performing SAFT with an aperture size

of 10 is illustrated in Figure 5.6.

As can be seen, there is clearly an improvement in the resultant image after

performing the SAFT, as the fl.aw and the weld-reflections can be made out more
 57

200

400

600

800

1000

1200

1400

1600

20 40 60 80 100 120

Figure 5.4: After SAFT implementation on the B-scan

clearly.

B'-scan

The B-scan imaging is a collection of consecutive A-scans along the axial direc-

tion. The B'-scan, on the other hand, is a collection of consecutive A-scans along

the circumferential direction. This is illustrated the Figure 5. 7. Since the transducer

moves parallel to the weld and the crack in the corresponding B '-scan image, they

are "seen" at the same time of flight (TOF).

 58

200

400

600

800

1000

1200

1400

1600

20 40 60 80 100 120

Figure 5.5: A B-scan image in which the defect cannot be seen clearly

Note that the periodicity of the reflection is not observed in a B '-scan imaging,

since the transducer is being moved parallel to the weld.

Since the transducer scans along the lower resolution axis, SAFT is performed

over a small aperture size as compared to the B-scan case. Figure 5.8 shows a B '-scan

from the pipe, along a certain axial position.

Due to the reasons explained in the B-scan imaging the SAFT algorithm has

been implemented for the third leg of the beam. The B'-scan has been chosen such

that the defect lies in the third leg. The aperture size chosen is 3 because of the
 59

200

400

600

800

1000

1200

1400

1600

20 40 60 80 100 120

Figure 5.6: After SAFT implementation on the B-scan

lower resolution along the circumferential direction. Figure 5.9 shows the result of

the SAFT algorithm.

Figure 5.10 shows a comparatively degraded B'-scan image. Figure 5.11 illus-

trates SAFT with aperture size 3

As can be seen, there is an improvement in the detection ability after performing

the B'-scan SAFT.

 60

Weld root

Figure 5.7: B'-scan imaging technique

Multi depth B'-scan

During the multidepth implementation of the SAFT algorithm, the B'-scan im-

age of Figure 5.12 was selected. The SAFT was then performed at 6 uniformly spaced

depths in the third leg. The resultant images were then combined to give Figure 5.13.

As can be seen, in Figure5.13 the flaw profile comes up very clearly, due to focusing

at more than one depths. Thus, performing multidepth SAFT is definitely a better

t echnique than single depth focusing.

 61

200

400

600

800

1000

1200

1400

1600

10 20 30 40 50 60 70 80 90 100

Figure 5.8: A B'-scan image in which the defect can be seen clearly

Discussion

As can be seen from the results in the previous sections, SAFT is indeed a

valuable tool, to enhance ultrasonic testing techniques, to take care of the complex

scattering phenomenon.
T he depth of SAFT requires to be known apriori. Also the aperture size needs

to be directly related to the scanning resolution in the particular direction. If the

aperture size chosen is too large, besides the additional computational complexity,

there might be severe degradation of the resultant image. This is because the im-
 62

2 00

4 00

600

8 00

1000

1200

14 00

1600

10 20 30 40 50 60 70 80 90 100

Figure 5.9: After SAFT implementation on the B'-scan

plementation would average over A-scans in which there is no contribution from the

defect. This definitely needs to be avoided.

B'-scan imaging is better than B-scan imaging in this particular research. This is

because, the orientation of the cracks makes the B'-scan more perceptible, visually, as

it is easier to distinguish the weld and crack regions more clearly in the B'-scan image.

However, in this research, because of the higher resolution in the B-scan direction,

there is a larger area (or transducer positions) to which a reflector contributes.

Multidepth-focusing is a good option in this case, since the crack can be at any
 63

200

400

600

800

1000

1200

1400

1600

10 20 30 40 50 60 70 80 90 100

Figure 5.10: A B'-scan image in which the defect cannot be seen clearly

depth in the specimen. By attempting to focus at multiple depths, the amount of

apriori information required is reduced. Secondly, flaw profiles not envisaged, might

be seen in the same B'-scan after processing it via the SAFT algorithm.

The work reported in this thesis on the application of SAFT to angle-beam

pulse-echo ultrasonic testing technique clearly shows that this methodology is quite

powerful at achieving an improved resolution.

 64

200

400

600

800

1000

1200

1400

1600

10 20 30 40 50 60 70 80 90 100

Figure 5.11: After SAFT implementation on the B'-scan

Conclusion

Angle-beam pulse echo technique is the dominant mode of data collection in the

nuclear industry. The data obtained, however, is very noisy, and signal processing

techniques are needed to enhance the signal-to-noise ratio of this data, to improve

operator interpretation. Focus has been placed on SAFT as a technique to perform

this noise reduction and resolution improvement. SAFT is a valuable tool, for ul-

trasonic NDE because of the complex scattering phenomenon in the weld material.

Most of the research done earlier, has been in the field of normal SAFT. However,
 65

200

400

600

800

1000

1200

1400

1600

10 20 30 40 50 60 70 80 90 100

Figure 5.12: A B'-scan image in which the defect cannot be seen clearly

the algorithm for normal SAFT has to be enhanced, to incorporate issues unique

to the angle-beam mode of data collection. These include multiple sound paths of

the ultrasonic beam, and spatial periodicity of the defects. The angle-beam SAFT

algorithm was developed in this research.

Flaw location is very important in the nondestructive evaluation of a specimen.

However, it is also very difficult. In this research, algorithms have been developed to

aid in manual and computer flaw location.

From the results of performing SAFT on various B-scan and B'-scan images, it
 66

200

400

600

800

1000

1200

1400

1600

10 20 30 40 50 60 70 80 90 100

Figure 5.13: After implementing SAFT at 6 uniformly spaced depths in the third
leg of the B '-scan

can be concluded that SAFT is indeed a valuable tool for improvement of ultrasonic

images of nuclear tubings.

Future Work

Further development and refinement of the SAFT process is desirable, in order

to consider all possible factors. Proposals for additional studies may include the

following avenues:
 67

1. The algorithm should be developed independent of the apriori information re-

quired for the SAFT depth calculation. This could be done using a neural

network approach, which produces the data set to detect possible defects. The

neural network could inform the SAFT algorithm about the approximate loca-

tion of the defect.

2. The optimum aperture size for a given scanning geometry could be calculated

automatically. This can be done using the basic laws of physics to calculate the

beam spread in the test specimen using the sound wavelength, the ultrasonic

impedance of the test material and the scanning resolution in the particular

direction.

3. As a final suggestion, to optimize the inherent capabilities of SAFT, the pos-

sible expansion to real-time should be looked into, as a way to speed up the

computational time. Due to the fact that the SAFT-algorithm executes its

computation A-scan by A-scan, and the process is repeated for each of the

A-scan, all individual operat1 ns can be parallelized via the use of an array

processor. This may reduce the CPU-time by a large factor, which may be a

critical requirement in an extensive testing environment.

 68

BIBLIOGRAPHY

[1] Singh, A. (1980). Crack Depth Determination by Ultrasonic Frequency Analysis

Aided by Dynamic Photoelasticity. Ames, IA: Iowa State University.

[2] Gordon P. H. (1978). Introduction to Nondestructive Testing. Milwaukee, WI:

American Society for Quality Control.

[3] Halmshaw, R. (1987). Non-destructive Testing. London, UK: Edward Arnold

Publications.

[4] Fordham, P. (1968). Non-destructive Testing Techniques. London, UK: London

Business Publications Limited.

[5] Ghorayeb, S. (1992). Study of Beamforming Techniques for Ultrasound Imaging

in Nondestructive Testing. Ames, IA: Iowa State University.

[6] Kessler, L. W. (1989). Acoustic Microscopy Metals Handbook 17. Materials Park:
ASM International.

[7] Havlice, J. F. and Taenzer, J. C. (1979). Medical Ult rasound Imaging: An

Overview of Principles and Instrumentation. Proc. IEEE 67: 620-641.

[8] Haumschild, D. J. and Carlson , D. L. (1983). An Ultrasonic Bragg Scattering

Technique for the Quantitative Characterization of Marbling in Beef. Ultrasonics
21: 226-233.

[9] Macovski, A. (1983). Medical Imaging Systems. Englewood Cliffs, NJ: Prentice-
Hall.

[10] Woodcock, J.P. (1979). Ultrasonics: Medical Physics Handbook 1. Bristol, UK:
Adam Hilger.
 69

[11] Seydel, J. (1982). Research Techniques in Nondestructive Testing. London, UK:

Academic Press.

[12] Langenberg, K. J., Berger, M., Kreutter, T., Mayer, K. and Schmitz, V. (1986).
Synthetic Aperture Focusin Technique Signal Processing. NDT International 19:
177-188.

[13] Heyman, E. and Felsen, L.B. (1985) . A Wavefront Interpretation of the Singular-
ity Expansion Method. IEEE tranaction on Antenna Propagation 33: 706- 720.

[14] Langenberg, K. J., Aulenbacher, U., Bollig, G. , Fellinger, P., Weinfurter, J.

and Zanger, P. (1986). Pulsed Wave Scattering: Wavefronts and Resonances,
identification and Imaging. IEEE tranaction on Antenna Propagation 33: 312-
321.

[15] Furniturewala, I., Udpa, L. and Udpa, S. (1995). Angle Beam Synthetic Aper-
ture Focusing in Ultrasonics. Proceedings of the 4th Annual Midwest Electro-
Technology Conference: 100-103.

[16] Doctor, S. R., Hall, T. E. and Reid, L. D. (1986). SAFT - The Evolution of
a Signal Processing Technology for Ultrasonic Testing. NDT International 19:
163-167.