Ultrasonics 40 (2002) 110

www.elsevier.com/locate/ultras

Modeling for quantitative non-destructive evaluation

Jan D. Achenbach

Robert R. McCormick School of Engineering and Applied Science, Center for Quality Engineering and Failure Prevention, Northwestern University,
Evanston, IL 60208-3020, USA

Abstract
A quantitative approach to non-destructive evaluation (NDE) must be based on models of the measurement processes. A models
purpose is to predict, from rst principles, the measurement systems response to material properties and anomalies in a material or
structure. For the ultrasonic case a measurement model should include modeling of the generation, propagation and reception of
ultrasonic signals, and the ultrasonic interactions that generate the systems response function. A measurement model has many
benets, which are discussed in the paper. Three examples of the productive use of quantitative modeling in conjunction with
measured data are presented: the detection and sizing of fatigue cracks which emanate from weep holes in the risers of wing panels in
the interior of an aircraft wing by the use of ultrasound generated on the exterior surface of the wing, the determination of the elastic
constants of anisotropic thin lms deposited on a substrate, and the detection and sizing of surface-breaking cracks by the use of the
laser-source scanning technique for laser generated and detected ultrasound. 2002 Elsevier Science B.V. All rights reserved.
Keywords: Non-destructive; Ultrasonics; Gack detection; Acoustic microscopy; Thin lm; Scanning laser source

1. Introduction
A retrospective look at the eld of non-destructive
testing as it existed some thirty years ago shows a eld
that was much smaller than it is now, but one that was
already concerned with problems that are still being
considered today: detection and characterization of
cracks and cavities, disbonds and corrosion, and the
non-destructive determination of material properties,
among others. In those days the attention was primarily
directed towards suitable approaches for specic congurations and problems. In more recent years the
emphasis has shifted from the treatment of specic problems to general approaches. Also, experimental ingenuity, improved hardware and better signal processing
techniques have produced signicant progress. In addition, an important extension of the eld has been provided by the use of analytical techniques for the
development of measurement models. As discussed in
this paper such models are essential. Measured data
generally cannot be understood in a quantitative way
without a measurement model.
*

Corresponding author. Tel.: +1-847-491-5527; fax: +1-847-4915227.

E-mail address: achenbach@nwu.edu (J.D. Achenbach).

Thus, recent work is based on the premise that a

quantitative approach to non-destructive evaluation
(NDE) must be based on models of the measurement
processes of the various inspection techniques. A models principal purpose is to predict, from rst principles,
the measurement systems response to material properties and anomalies in a given material or structure, (e.g.
cracks, voids, distributed damage, corrosion, deviations
in material properties from specication, and others).
For the ultrasonic case, a measurement model should
include the conguration of probe and component being
inspected, as well as a description of the generation
propagation and reception of the interrogating ultrasonic signals. Detailed modeling of the ultrasonic interactions, which generate the measurement systems
response function, should also be included.
The availability of a measurement model has many
benets. First, numerical results, based on a reliable
model, are very helpful in the design and optimization of
ecient testing congurations. Second, a good model is
indispensable in the interpretation of experimental data
and in the recognition of characteristic signal features.
The relative ease of parametrical studies, based on a
measurement model, facilitates an assessment of probability of detection of anomalies. A measurement model
is a virtual requirement for the development of an

0041-624X/02/$ - see front matter 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 0 4 1 - 6 2 4 X ( 0 2 ) 0 0 0 8 3 - 5

J.D. Achenbach / Ultrasonics 40 (2002) 110

inverse technique based on quantitative data. Finally, a

measurement model whose accuracy has been tested by
comparison with experimental data provides a practical
way of generating a training set for a neural network or
a knowledge base for an expert system.
In this paper, three examples of the productive use of
quantitative modeling are discussed. The rst example
concerns the detection and sizing of fatigue cracks that
emanate from weep holes in the risers of lower wing
panels in the interior of the wing of a military aircraft,
the C-141, by the use of ultrasound generated on the
exterior surface of the wing. The second example concerns the determination of the elastic constants of anisotropic thin lms of the order of thickness of one
micrometer, deposited on a substrate. The third example
concerns a new scheme for the detection and sizing of
surface-breaking cracks by the use of laser generated
and detected ultrasound.

2. Fatigue cracks emanating from weep holes in C-141

aircraft
Weep holes are placed in the risers of inner wing
panels of C-141 aircraft in order to allow fuel to be
properly distributed during ight. Fig. 1 shows the location of weep holes along the riser of a wing panel.
Each C-141 inner wing contains over 700 weep holes
located either separately or in groups of two or three.

These holes gives rise to stress concentrations that can

lead to fatigue cracks. Two cases of fatigue cracks emanating from weep holes have been observed; top cracks
which propagate upward and bottom cracks which
propagate downward toward the wing panel surface.
Due to the location of top cracks relative to the external
surface of the wing, these cracks have been dicult to
detect by conventional ultrasonic techniques. Eddy
current inspection is currently carried out as a routine
depot maintenance procedure to determine the condition of the inner wing weep holes. The eddy current
inspection requires fuel tank entry by the inspection
technician, and is a costly and time-consuming process.
To minimize aircraft inspection costs and to reduce
potential damage to the interior of the wing, it is desirable to use an inspection process that can be performed from the exterior of the wing.
Prior work has shown the potential of a two element
ultrasonic transducer technique for the generation and
measurement of circumferential creeping waves around
the weep hole to determine the existence of top cracks
[1]. The use of both the pulseecho and pitchcatch responses was found to greatly increase the sensitivity of
the method. From these studies, a procedure was proposed to detect top cracks for the dry weep hole case
under laboratory conditions [1]. To further develop this
method and implement it as a robust protocol, the work
of Ref. [1] had to be supplemented by an accurate
measurement model, careful signal processing of the
measured traces, an automated technique for signal classication by the use of a neural network, and a probability of detection (POD) assessment. Here, we briey
describe this supplemental work, as well as the implementation of the automated procedure from the initial
proof of concept phase to in-eld application.
First, we consider the measurement model that was
developed. The signal produced by the sonifying transducer was t by a time-domain trace, and this trace was
subsequently converted to the frequency domain by an
exponential Fourier transform. For a range of frequencies the scattering of the incident signal by the weep hole
with crack conguration was determined by the boundary element method.
Fig. 2 shows the conguration of the weep hole and
the crack. The boundary of the weep hole with crack is
denoted by S. The outside region is V . In scattering
problems the total eld is usually written as the sum of
the incident eld and the scattered eld
in
sc
utot
i ui ui

Fig. 1. Weep holes in wing structure of C-141.

and

in
sc
stot
ij sij sij :

in
Here, uin
i and sij represent the displacement and stress
elds in the solid in the absence of the hole. Clearly, the
scattered eld satises homogeneous governing equations. For a point of observation, Q, not inside the hole
or on its boundary, an elastodynamic representation

J.D. Achenbach / Ultrasonics 40 (2002) 110

Fig. 3. Comparison of experimental and simulated pulseecho signals.

Fig. 2. Cylindrical hole with a radial crack.

integral may be used to express the displacement at xQ in

the form
Z h
i
Q
sc
G
Q
G
Q sc
usc
x

u
xs
x

x


u
x

x
s
x
k
i
ij;k
i;k
ij
S

 nj x dSx:

Here, the normal n points into the hole, and uGi;k x  xQ

is the displacement in the xi direction at the eld point x
due to a unit point load applied in the xk direction at
point xQ . The corresponding stress components
sGij;k x  xQ can be obtained from Hookes law. The
surface of the hole is free of surface tractions. Thus, with
tin tsc ttot , where ti sij nj the conditions on S become
in
tsc
n x; x tn x; x

and

in
tsc
t x; x tt x; x;

where the subscripts n and t dene normal and tangential components.

The integral representation given by Eq. (2) is not
immediately applicable to eld calculations of the displacements because the displacement components of the
scattered eld are not known. These displacement
components can, however, be computed by letting the
point xQ approach S, to yield a set of boundary integral
equations for the displacement components. These
boundary integral equations can be solved numerically
by the boundary element method. The numerical analysis requires careful attention, but the details will not be
discussed here. They can be found in Ref. [2]. Once the
displacement components on the boundary S have been
computed, the displacement at any eld point can be
obtained by the use of the representation integral given
by Eq. (2). The results in the frequency domain can be
converted back into the time domain by using the inverse Fourier transform. Fig. 3 shows a comparison
between simulated and measured traces. The agreement
is very good. Snapshots for contour plots of the displacements as time increases are shown in Fig. 4. The

wave motion propagating around the hole, its reection by the crack, and the spiral-like circumferential
waves generated in the surrounding solid, which propagate back towards the transducer, can clearly be observed.
Once the signals have been acquired, a gate is applied
beyond the specular response to sift out the circumferential signal. The reference specular pulseecho signal
was used to adjust the gate location for variation in hole
depth and diameter. The ratio of the peak-to-peak value
of the reected signal with respect to the peak-to-peak
value of the transmitted signal is used for classication
purposes. Due to hole size variation, an approach using
ratio values for multiple transducer locations and using
a neural network to classify this data set was chosen as
the most robust methodology. Concurrent development
of a C-scan based inspection technique was also pursued.
In order to understand the inspection procedures
and the possible impact of deviations from the set-up
parameters, a laboratory parametric investigation was
conducted (see Aldrin et al. [3] for details). In this investigation, the three factors, i.e., the transducer gain,
the scanner skew, and the timebase delay, which could
potentially impact the outcome of the inspection, were
systematically varied. An expert user carried out the
laboratory runs and applied the algorithms presented in
Ref. [3] to make a call.
POD curves for the laboratory runs of the inspection procedure are shown in Fig. 5. These inspections
involved 280 specimens, some without aws, some with
notches, others with actual cracks. For a top crack of
length 0.020 in., the POD was obtained as 90% with a
false call rate of 0.1%. For the C-scan results, the corresponding numbers were 0.045 in. for a POD of 90%,
with a false call rate of 1.4%.
To determine the reliability of the inspection process
when implemented in a eld environment, an experiment was designed to address the issues of inspector
experience and training, the inspection environment,
and normal operating conditions [3]. Four inspectors at
Warner Robins Air Force Base were chosen to inspect

J.D. Achenbach / Ultrasonics 40 (2002) 110

Fig. 4. Snapshots of the total displacement eld for an incident in-plane shear pulse on a 1/4 in. diameter weep hole with a 1.8 mm (0.070 in.) crack,
for four time steps, s 0,1,2,3,4,5,6,7,8; s t  t0 ct =r.

3. Line-focus acoustic microscopy

Fig. 5. POD results using neural network assisted (NN) and C-scan
(CS) classication approaches for top crack (Top) and bottom crack
detection.

the panels without knowledge of the aw placements.

Three of these inspectors had prior experience with the
Ultra Image IV (or similar) equipment and had signicant experience in the eld testing environment. One
inspector was familiar with the operation of the Ultra
Image IV, but had limited eld testing experience. All
four inspectors received training on the procedures, and
all were asked to perform the inspection using both the
C-scan and the neural net algorithm procedures. The
results are reported in Ref. [3].

Thin lm coatings are widely used in surface engineering to extend the life or enhance the performance of
components. For example, transition metal carbides and
nitrides of the IV and VI group of the periodic table
have extremely high melting points and extreme hardness, excellent high-temperature strength, and good corrosion resistance. They are widely used for thin lm
coatings to produce wear-resistant surfaces, to provide
corrosion protection against harsh environments, and
for other applications [4].
Basically, a coated component, here referred to as a
lm/substrate system, is a layered structure, with a single
or multiple-layered thin lm deposited on a substrate.
Film/substrate systems are usually prepared by deposition techniques, such as physical vapor deposition and
chemical vapor deposition. The mechanical properties
of the thin lm, and the adhesion between coating layers
and between the coating and the substrate, determine its
functional characteristics. These properties are very
sensitive to a number of factors, which are determined
by the deposition technique and by processing parameters. The thin-lm properties may be quite dierent from
bulk properties. Therefore, there is a need for a non-

J.D. Achenbach / Ultrasonics 40 (2002) 110

destructive technique to characterize these thin lm

coatings.
Quantitative line-focus acoustic microscopy is a
suitable technique for non-destructive characterization
of thin lms. The technique is based on the measurement
of the so-called V z curve, which is a recording of the
voltage output of the transducer when the distance between the lens and the specimen is decreased. Fig. 6
shows the conguration. Oscillations appear in the V z
curves due to interference between the rays specularly
reected by the specimen and the rays associated with
propagation of a leaky SAW on the specimen. The
phase velocity of a leaky surface wave, cLSAW can be
determined from the period, Dz, of the periodic oscillations, using a well-known relationship based on ray
considerations [5,6].
"

2 #1=2
p
cf
cLSAW cf 1  1 
 cf f Dz;
4
2f Dz
where cf is the wave velocity in the coupling uid, and f
is the operating frequency. Eq. (4) implies that the
measurable Dz must satisfy
cf
 1:
2f Dz

The period Dz can be accurately extracted from the

measured V z curve by a spectrum analysis, as discussed in detail in Ref. [6]. Therefore, by measuring V z

Fig. 6. Line-focus acoustic probe and typical V z curve.

curves for a specimen at dierent frequencies or at the

same frequency, but dierent wave propagation directions, the dispersion curves or the directional variations
of the leaky surface wave velocity can be measured.
By numerically simulating the V z measurement
procedure, a V z measurement model can be developed
to calculate theoretical V z curves for a layered specimen of arbitrary conguration for use with a line-focus
acoustic microscope. In this paper, we use the approach
proposed by Achenbach et al. for synthesizing V z
curves [7]. In that approach, the output of the line-focus
acoustic probe is expressed as a Fourier integral over the
product of the characteristic functions of the acoustic
lens, L1 kx and L2 kx , and the reectance function of
the uid-loaded specimen Rkx , as expressed by
Z 1
V z
L1 kx L2 kx Rkx exp2ikz z dkx ;
6
1

where kx ; kz denes the wave vector,

q
x
x
kz kf2  kx2 ; kf ; kx
:
cf
cLSAW
As shown by Eq. (6), the theoretical measurement
model for the V z curve contains the reectance function of the lm/substrate system as a principal component. Well-known theoretical methods are available to
determine reectance functions (see for example, Ref.
[8]).
By using the theoretical reectance function model
for parametrical studies, the phase velocities of the
possible SAW modes and their corresponding mode
reection coecients can be predicted for an arbitrary
lm/substrate system. Therefore the reectance function
can be used as a predictive tool to estimate in which
frequency range a wave mode may be picked up by
acoustic microscopy. By applying the V z measurement
model, the phase velocities can be calculated for a lm/
substrate system at dierent frequencies. For trial values
of the unknown elastic constants of the thin lm, these
theoretically calculated phase velocities have been used
in an iterative manner together with phase velocities obtained from V z measurements to determine the elastic
constants of thin lms by minimization of the deviation
between theoretical and experimental results.
Fig. 7 shows the conguration of a layered medium
with N layers on a substrate. A time-harmonic plane
wave is incident from a coupling uid, at arbitrary incident angle h. The response of the layered medium to
the incident time-harmonic plane wave can be expressed
in terms of the reectance function. This reectance
function carries information on the layered medium,
including its mechanical properties, the thickness of
each layer, and the interface conditions. A general theoretical model has been developed which quantitatively
relates the reectance function to these characteristic
properties of the layered specimen (see Ref. [8]). It has

J.D. Achenbach / Ultrasonics 40 (2002) 110

mode. By virtue of the relations between the incident

angle h and the phase velocity c shown in Eq. (9), the
reectance function can be used to obtain the phase
velocities of the wave modes for the layered system at
a xed frequency. The reectance function undergoes
a 2p phase shift and its real part undergoes a
jump, decreasing its magnitude as the incident angle
h passes a critical angle hcr , which corresponds to a
wave mode. Using this criterion to determine the critical
angles, the phase velocity of a wave mode can be calculated as
c

Fig. 7. N-layered lm on a substrate.

been shown that the reectance function is of the general

form
Rh; f

A  bB
;
A bB

where
iq x2
b f ;
af k x


T
A  31
T41

7

T33 
;
T43 


T
B  21
T41


T23 
:
T43 

Here qf and cf are the density and the velocity of the

coupling uid, and f and x are the frequency and the
angular frequency of the incident wave with incident
angle h, and kx is the tangential component of the wave
vector, where
q
kx kf sin h; kf cf kx c x; kz kf2  kx2 ;
q
cf
2
c
; af c=cf  1
9
sin h
and Tij are the elements of the global transfer matrix [T].
For the numerical calculations, the delta-matrix reformulation is introduced to avoid the precision problem
(see Ref. [9]).
The reectance function given by Eq. (7) can be used
for wave mode identication. The vanishing of its denominator, i.e.,
A bB 0

10

is exactly the characteristic equation of the layered medium. Solutions for the velocity and the amplitude of the
possible wave modes can be obtained by numerically
solving this equation. Setting b 0 takes out the eects
of the uid loading, and thus the equation A 0 can be
used to identify the free wave modes.
For a given frequency, the reectance function is a
function of the incident angle, h. The reectance function shows distinct behavior when the incident angle
equals the critical angle associated with a possible wave

cf
:
sin hcr

11

As discussed by Guo et al. [10], the magnitude of the

reectance function at the critical angle hcr , indicates
how strongly the wave mode will be reected. For
convenience, that magnitude will be referred to as the
mode reection coecient. A strongly reected SAW
mode with a large mode reection coecient can be
picked up easily by acoustic microscopy. The results
show that the theoretical mode reection coecient of a
true surface wave mode is unity and that of a pseudoSAW mode that radiates energy into the substrate is
somewhere between 0 and 1. A low mode reection coecient indicates that the wave mode is weakly reected
and will be more dicult to be picked up by the LFAM
for generation of a V z curve.
Theoretical calculations and experimental observations show that the LFAM cannot pick up any SAW
mode with a mode reection coecient smaller than
around 0.15. A strongly reected SAW mode can be
easily picked up by the LFAM and its velocity can be
extracted with high accuracy. The V z measurement
will be less precise for a weakly reected surface wave
mode. There is a signicant decrease in the number and
magnitude of the oscillations of the V z curve for a
weakly reected SAW mode. Comparisons of theoretical results for the phase velocity for a slow-on-fast system calculated both directly from the reectance
function and from the V z curve via Eq. (4), presented
by Guo et al. [10], illustrate the signicance of an appropriate selection of the frequency range for modes
with a suciently large value of the reectance function.
For a fast-on-slow system the material constants have
been determined from the experimental V z results by
the iterative optimization method mentioned earlier in
this section. After determination of the material constants the corresponding phase velocities were recalculated to verify the procedure. Results are shown in Fig.
8. The best results are obtained in the frequency range
dened fd < 0:3, i.e. before the cut-o of the true SAW
mode. In that range the magnitude of the reectance
function is unity. Based on the excellent agreement
displayed in Fig. 8 it may be expected that the material

J.D. Achenbach / Ultrasonics 40 (2002) 110

Fig. 9. Leaky SAW phase velocities calculated using the V z model,

using known material constants of the thin lms and the determined
interface parameters KL and KT (indicated by lines), plotted together
with the measured values (symbols).

Fig. 8. Phase velocities of the SAWs calculated using the V z model

with the determined material constants (indicated by lines) plotted
together with the measured data (symbols) for TiN thin lms of different thicknesses on a M2 substrate.

constants have been obtained with good accuracy. A

detailed discussion of the accuracy can be found in the
paper by Guo et al. [10].
This section is completed with some comments on the
eects of non-perfect interfaces. In the theoretical approach an imperfect interface is replaced by a layer of
extensional and shear springs. Such a layer of springs
can easily be included in the method of calculation of
the reectance function that was earlier presented in this
section. In Ref. [11] it has been attempted to correlate
calculated values of the spring constants with results
obtained by a destructive scratch test for strength determinations. Specimens with dierent KL and KT interface conditions were prepared by using dierent lengths
of time for etching of the substrate surface. For lm
deposition by the reactive sputtering technique etching of
the substrate for dierent time periods has an observable eect on the stiness constants KL and KT of the
interface, as shown in Fig. 9. The determined constants
KL and KT show that for 10 min etching (specimen 1)
they are about 2.7 times the values for the case of 20 min
etching (specimen 2). The scratch test results indicate
that the adhesion of specimen 1 is much better than that
for specimen 2. These limited experimental results show
better adhesion for a higher stiness interface.
However, it is important to note that there is no
theory which shows that non-destructive techniques can
directly measure interface strength. At the present time,
only destructive techniques, such as the scratch test and

the pullout test, can provide us with a measure of the

interface strength. In Ref. [11] the LFAM at high frequency was used to measure the interface stiness. For
most of the lm/substrate systems studied in Ref. [11]
the interface stiness parameters seem to have a strong
correlation with the destructively measured interface
strength, such as the critical normal force of a scratch
test. Therefore, measured interface stiness parameters
may provide an indication of the bond quality between a
lm and a substrate.

4. Scanning laser source technique

Surface breaking cracks in a structure can be ultrasonically detected using Rayleigh or Lamb waves. As
discussed in Section 2, conventional ultrasonic aw detection methodologies require the generation of an ultrasonic wave packet that travels through a structure
and interacts with existing aws within the structure.
Either reected echoes or transmitted signals may be
monitored in the pulseecho or pitchcatch mode of operation. In the rst method, the presence of an unexpected reected signal is a possible indicator of a aw,
whereas in the second method the transmitted signal
may be signicantly attenuated by the presence of an
intervening aw.
Pulseecho and pitchcatch techniques can also be
used with laser ultrasonics where a high-power pulsed
laser is used to generate ultrasound thermoelastically (see
Ref. [12]). Laser-based ultrasonic techniques provide a
number of advantages over conventional ultrasonic
methods, such as higher spatial resolution, non-contact
generation and detection of ultrasonic waves, use of ber
optics, narrowband and broadband generation, absolute
measurements, and ability to operate on curved and
rough surfaces and in hard-to-access locations. On the
receiving side, surface ultrasonic waves can be detected
using piezo-electric (PZT) or EMAT transducers, or

J.D. Achenbach / Ultrasonics 40 (2002) 110

optical interferometers in a completely laser-based ultrasonic system.

In both the pitchcatch and pulseecho methods, the
source is expected to generate a well-established ultrasonic wave, which then interacts with existing aws. The
limitations on the size of aws that can be detected using
pulseecho or pitchcatch methods are determined by
the ultrasonic reectivity or transmittance of the aws
for the particular wavelength used, and by the sensitivity
of the ultrasonic detector. The reectivity and transmittance of Rayleigh waves as a function of defect geometry,
and the change in frequency spectra of ultrasonic waves
upon transmission or reection by a defect, have been
investigated. As might be expected, small aws give rise
to weak reections and small changes in the amplitude of
transmitted signals. These small variations are often too
weak to be detected with existing laser detectors. In this
paper we briey discuss an alternate approach for ultrasonic detection of small surface-breaking cracks using
laser-based techniquesthe scanning laser source (SLS)
technique. This approach, discussed in detail in Refs.
[13,14], does not monitor the interaction of a wellestablished ultrasonic surface wave with a aw, but rather monitors the changes in the generated ultrasonic
signal as the laser source passes over a defect. Changes in
amplitude and frequency of the generated ultrasound are
observed which result from the changed constraints
under which the ultrasound is generated over uniform
versus defective surface areas. These changes are quite
readily detectable using existing laser detectors even for
very small aws.
The SLS approach has been experimentally veried
by testing of an aluminum specimen with a surfacebreaking fatigue crack of 4 mm length and 50 lm width.
A broad-band heterodyne interferometer [15] with 115
MHz bandwidth has been used as an ultrasonic detector. The SLS was formed by focusing of the irradiation
of a pulsed Nd:YAG laser (pulse duration 10 ns, energy
5 mJ) into a line of 5 mm length and 0.4 mm width. The
generation of ultrasound was done in the thermoelastic
regime. The distance between the source and detector
was kept at 40 mm (see Fig. 10).
A plot of ultrasonic amplitude of the generated signal
versus the SLS position as it was scanned over the crack
is shown in Fig. 11. A characteristic signature of the
crack can be seen on this plot as a specic variation of
the ultrasonic amplitude. The following aspects of this
signature should be noted.
(1) In the absence of a crack or when the source is far
from reaching the crack, the amplitude of the generated
ultrasonic direct signal is stable (see zone I in Fig. 11),
and the signal is of sucient amplitude above the noise
oor to be unambiguously picked up by the laser detector, as shown in Fig. 12a. Note that a weak reection
from the crack is barely visible amidst the noise in
Fig. 12a.

Fig. 10. Principle of the SLS method with laser detection.

Fig. 11. Typical characteristic signatures of ultrasonic amplitude vs.

SLS location as the source is scanned over a defect: (I) far ahead, (II)
close to, and (III) behind the defect.

(2) As the source approaches the crack, the amplitude

of the detected signal signicantly increases (zone II in
Fig. 11). This increase (from a level that was already
suciently above the noise oor) is more readily detectable with a laser interferometer than any weak echoes from the crack (see Fig. 12b). We attribute this
increase in signal amplitude to interference of the incident wave with the wave reected by the crack.
(3) As the source is very close to, or right above the
defect, a steep decay is observed, which is attributed to
the changes in the conditions for generation of ultrasound, due to the presence of the crack.
(4) Subsequently, as the source moves behind the
crack, the amplitude remains low due to attenuation of
the signal generated by the crack (see zone III in Fig.
11). An example of this attenuated signal is shown in
Fig. 12c. Note that when the crack depth is smaller than
the wavelength of the generated ultrasound, a signicant
portion of the ultrasound can pass underneath the crack

J.D. Achenbach / Ultrasonics 40 (2002) 110

conductivity, s is the relaxation time, ce is the heat capacity at constant strain, W is the heat source, b is the
thermoelastic coupling constant: b 3k 2laT , aT is
the linear thermal expansion coecient, and a is the
parameter of thermodynamic temperature.
A possible expression for the energy deposited by
laser illumination is

 3

2 2r2 =R2
8t 2t2 =g2
vz
G
W E1  Ri ve
e
e
;
RG
g4
14
where E is the laser pulse energy, Ri is the surface reectivity, RG is the Gaussian beam radius, g is the laser
pulse risetime, and v is the extinction coecient.
For the two-dimensional case, where the elds are
independent of x2 , solutions to Eqs. (12)(14) can be
obtained by the use of the exponential Fourier transform with respect to x1 and the Laplace transform with
respect to time. The inverse transforms can be obtained
by numerical techniques. The resulting solutions represent elds generated by a laser-line source illuminating a
homogeneous elastic half-space. Fig. 13 shows the vertical displacement at position x1 9:1 mm as a function of time. The corresponding stress component rx is
shown for x1 5:1 mm in Fig. 14. The symbols in Figs.
13 and 14 represent the elds generated by a dipole
represented by Dd0 x1 dx3 H t where D is the magnitude of the dipole. As pointed out by earlier authors
([12,16]) intuitively the actions of a local injection of
heat and the application of a dipole should be expected
to produce equivalent elds. A local temperature increase produces an expansion of material right under the
area of laser illumination. This expansion is prevented in
the x2 direction because of the two-dimensional nature
of the problem, but it is free in the x3 -direction because

Fig. 12. Representative ultrasonic time-domain signals detected by the

heterodyne interferometer at a xed location when the laser source is:
(I) far ahead, (II) close to, and (III) behind the defect.

when the source is behind it. Theoretical models for the

generation of ultrasound by laser illumination have been
presented by a number of authors. See the book by
Scruby and Drain [12] and the review article by
Hutchins [16]. The equations governing the coupled
temperature and displacement elds are
kr2 T ce sT ce T_ T0 br  u_  W ;

12

lr2 u k lrr  u q
u brT arT_ ;

13

where T is the absolute temperature, T0 is the ambient

temperature, u is the displacement, q is the mass density,
k, l are the Lame elastic constants, k is the thermal

Fig. 13. Normal displacement, u, generated by laser line-source (5 mJ).

10

J.D. Achenbach / Ultrasonics 40 (2002) 110

tration under Contract #DTFA 03-98-D-0008 through

the Air Transportation Center of Excellence in Airworthiness Assurance. The author wishes to acknowledge the assistance of John Aldrin, Zhiqi Guo and Irene
Arias.

References

Fig. 14. Stress components rx generated by laser line-source (5 mJ).

of the free surface at x3 0, and it is resisted by the

material in the x1 direction. The stresses due to expansion in the x1 -direction are equivalent to the application
of a dipole Dd0 x1 . For convenience, the dipole is applied at x3 0: From the governing Eqs. (12)(14) or
from a separate consideration of an element of the
material at the surface it can be shown that the magnitude of the dipole should be
D

2l3k 2l
aT DT
k 2l

15

Figs. 13 and 14 show that the elds obtained from the

dipole equivalency argument display excellent agreement with the curves obtained from the full formulation.
The expression for the dipole given by Eq. (15) shows
that D depends linearly on the coecient of linear
thermal expansion. Since a damaged region or a region
containing a crack has a smaller coecient of thermal
expansion than an undamaged region, we can conclude
right away that the presence of the laser source very
close to, or right above the crack will produce a smaller
ultrasonic signal than for the source over an undamaged
area, in accordance with the experimental result shown
in Fig. 11. A quantitative treatment of the laser-sourcescanning method, including interaction eects just ahead
of the crack, is currently being completed (see Arias,
Murray and Achenbach [17]).

Acknowledgements
The material presented in Sections 24 represents
results obtained in the course of research funded by
three sponsors: Section 2Aging Aircraft Program
Oce of Wright Patterson Air Force Base; Section
3Oce of Naval Research under Contract N0001489-J-1362 and Section 4Federal Aviation Adminis-

[1] P.B. Nagy, M. Blodgett, M. Golis, Weep hole inspection by

circumferential creeping waves, NDT&E International 27 (3)
(1994) 131.
[2] J.C. Aldrin, Models and classication procedures for ultrasonic
inspection of holes for fatigue cracks, Ph.D. Dissertation, Northwestern University, June 2001.
[3] J.C. Aldrin, J.D. Achenbach, G.A. Andrew, C. Pan, B. Grills,
R.T. Mullis, F.W. Spencer, M. Golis, Case study for the
implementation of an automated ultrasonic technique to detect
fatigue cracks in aircraft weep holes, Materials Evaluation 59 (11)
(2001) 13131319.
[4] L.E. Toth, Transition Metal Carbides and Nitrides 7, Academic
Press, New York, 1971.
[5] R.D. Weglein, A model for predicting acoustic material signature,
Appl. Phys. Lett. 34 (1979) 179181.
[6] J. Kushibiki, N. Chubachi, Material characterization by line-focus
beam acoustic microscope, IEEE Trans. Sonics Ultrason. SU32
(1985) 189212.
[7] J.D. Achenbach, J.O. Kim, Y.-C. Lee, Measuring thin-lm elastic
constants by line-focus acoustic microscopy, in: A. Briggs (Ed.),
Advances in Acoustic Microscopy I, Plenum Press, New York,
1995, pp. 153208.
[8] A.H. Nayfeh, D.E. Chimenti, Reection of nite acoustic beams
from loaded and stiened half-spaces, J. Acoust. Soc. Am. 75
(1984) 13601368.
[9] E.C. Pestel, F.A. Leckie, Matrix methods in elasto-mechanics,
McGraw-Hill Book Company, New York, 1963.
[10] Z. Guo, J.D. Achenbach, A. Madan, K. Martin, M.E. Graham,
Integration of modeling and acoustic microscopy measurements for thin lms, J. Acoust. Soc. Am. 107 (5) (2000) 2462
2471.
[11] Z. Guo, J.D. Achenbach, A. Madan, K. Martin, M.E. Graham,
Modeling and acoustic microscopy measurements for evaluation
of the adhesion between a lm and substrate, Thin Solid Films 394
(2001) 189201.
[12] C.B. Scruby, L.E. Drain, Laser ultrasonics methods and applications, Adam Hilger, New York, 1990.
[13] A.K. Kromine, P.A. Fomitchov, S. Krishnaswamy, J.D. Achenbach, Scanning laser source technique and its application to
turbine disk inspection, in: D.O. Thompson, D.E. Chimenti
(Eds.), Review of Progress in Quantitative Non-destructive
Evaluation, Kluwer Academic/Plenum Publishers, New York,
1999, pp. 381386.
[14] A.K. Kromine, P.A. Fomitchov, S. Krishnaswamy, J.D. Achenbach, Laser detection of surface-breaking discontinuities: Scanning laser source technique, Mater. Eval. 58 (2) (2000) 173177.
[15] B.R. Pouet, R. Ing, S. Krishnaswamy, D. Royer, Heterodyne
interferometer with two-wave mixing in photorefractive crystals
for ultrasound detection on rough surfaces, Appl. Phys. Lett. 69
(1996) 37823784.
[16] D.A. Hutchins, Ultrasonic generation by pulsed lasers, in: W.P.
Mason, R.N. Thurston (Eds.), Physical Acoustics XVIII, Academic Press, San Diego, 1988.
[17] I. Arias, T.W. Murray, J.D. Achenbach, Interaction of ultrasound
generated by a scanning laser source with a surface-breaking
crack, in preparation.