Chapter 2

Theoretical Background
The purpose of this Chapter is to provide the necessary theoretical background applicable to
subsequent chapters in this thesis.

1) Fundamental Principles of Ultrasonic Test Process

Ultrasonic testing is a versatile nondestructive test (NDT) method that employs a transducer to
transmit high-frequency, short-pulse, inaudible sound waves through a test sample and receive them
back up at a receiver. Based on the reflection and propagation of these sound waves through the
medium flaws are detected and the material is characterized. It can also be used to measure thickness,
look for irregularities and monitor the condition of flaws, among other uses.

Ultrasonic non-destructive testing can be performed in several ways, the most common means use
piezoelectric transducer. This device converts acoustic energy into electrical energy and vice versa. It
essentially generates the acoustical energy that propagates through the test object. When the
propagating energy hits a discontinuity in the material, some of the acoustic energy reflects back for a
receiver to pick up. A flaw detector (a type of receiver) identifies these anomalies and displays them
on an oscilloscope, which displays the signals graphically and requires an expert for interpretation
[1,2].
The two types of transducer configuration used for testing includes:

 Based on the modes of transmission and reception of ultrasonic waves

 Pulse Echo Mode:
In this design, the acoustic energy is both emitted and received by the same transducer. Pulse-echo is
the most common mode, as it is simple to set up and more portable. If the propagating sound waves
makes contact with an imperfection in the test object, the waves reflect back to the probe. The probe
interprets the waves and show the corresponding signals propagating through the material. This
configuration is mostly preferred for large items where it is difficult to access the other side [2,3]
 Figure 1. Pulse Echo mode (without and with defect) [3].

 Through Transmission Mode:

With through-transmission, the object being inspected sits between a transmitter and a receiver. The
transmitter sends the waves through the object and toward the receiver. Any imperfection within the
object affects the sound signal reaching the receiver. This method is much suited for small test objects
with other side easily accessed [2].

Figure 2. Through Transmission Method from reference [4].

Another significant component of ultrasonic testing is the couplant gel, oil or water are some typical
couplant, and they all help to efficiently transmit the sound energy from the transducer to the object
being inspected. The couplant displaces the air between the transducer and the surface of test object in
order to allow for more efficient transmission of sound waves. The coupling of ultrasonic testing is
featured in a few varieties, primarily contact coupling and immersion coupling, Advantages of each
method is seen in different situations [1].

 Based on the means of coupling the probe device to the test sample
 Contact Coupling:
Ultrasonic contact coupling is a more portable option, great for large or object difficult to access. It
uses a pulse-echo transducer and a couplant, and the transducer requires contact with one surface of
the test object. They’re useful when only one side of the material can be reached. Plus, the transducer
can be moved to the test object most especially in the case of objects difficult or costly to transport.

Figure 3. Contact coupling Testing from reference [4].

 Immersion Coupling:
Another option is ultrasonic immersion coupling, which is more suitable for laboratory or factory
environment. This strategy is often used when the test object has a complex shape or curves. The
object is immersed in water, which serves as the couplant. The pulse-echo method is also applied here,
with a single probe transmitting and receiving the wave signals.
 Figure 4. Immersion Testing [4].

Some Industries that rely on ultrasonic testing are: Aerospace, automobile, medical, rail, construction,
oil and gas, metallurgy and manufacturing. Operators in these industries can use ultrasonic testing with
various materials, but some provide better results than others. Materials commonly used in ultrasonic
testing include metal, composites, plastic and ceramic. A high density improves the accuracy of the
image due to less attenuation, and other materials moderately dense are still well-suited to this
technology. Paper and wood are some of the few materials that ultrasonic testing is not suitable for,
along with coarse grains like cast iron that can reduce sound transmission.

Ultrasonic testing is widely preferred to other NDT method for some reasons, which includes; high
penetration power, high accuracy and sensitivity, safety, flexibility and adaptability, rapid preparation,
immediate results and automation [1,2]

2) Ultrasonic Wave Propagation

Sound waves propagate due to the compressibility of a medium. Based on the frequency, they can be
classified as (Fig. 1.1):

 Infrasound waves: below 20 Hz

 Acoustic waves: from 20 Hz to 20 kHz
 Ultrasonic waves: above 20 kHz
 Figure 5. Frequency ranges of sound

An ultrasonic wave is defined as acoustic sound waves with frequencies greater than 20kHz (the limit
of human hearing). An ultrasonic wave propagates by displacement of particles in the medium which
it is propagating through. There are many fields in our world today that have successfully applied
ultrasonic devices in practice, some of such are; medicine, non-destructive testing (NDT), flow
measurements and more. This section will provide some basic theory about ultrasonic wave
propagation.

 Wave Modes

According to the direction of particle oscillation with respect to the direction of propagation, the wave
type can be classified into longitudinal waves, shear waves, and surface waves. For a longitudinal
wave, the particles oscillate in the same direction of wave propagation , This occur in the medium due
to the propagation of energy through the atomic structure by means of a series of compressions and
expansion (rarefaction) motion, hence, longitudinal waves are sometimes called pressure waves. For a
shear waves, the particles oscillate perpendicular to the direction of propagation. Due to the nature of
their propagation the are sometime called transverse waves. These two wave modes are depicted in
Figure below. Longitudinal waves can travel through liquid, solid and gaseous medium as well,
whereas, shear waves require only an acoustically solid or a highly viscous liquid for effective
propagation, and therefore, they are not effectively propagated in materials such as liquids or gases.
When compared to pressure waves, shear waves are relatively weak. In fact, shear waves are usually
generated in medium using a fraction of the energy from longitudinal waves. The surface wave mode
typically exist at the surface of the medium as implied by the name or interface between to medium
. However, this work will not consider the surface waves mode.
[5,6]
 Figure 6. Depiction of longitudinal and Shear wave propagation [5,6]

 Effects of Interfaces and Discontinuities

When a propagating ultrasonic wave hits an interface between two media, components of the
longitudinal wave is partly reflected and partly transmitted. Consequently, both the reflected wave and
the transmitted wave are split into a longitudinal wave and a shear wave. The incident and reflected
wave are inclined by the same angle, according to the Snell’s law. This same law also applies to the
relationship between the resulting angles of transmitted waves and corresponding incident waves angle
formed with the normal to the surface, as shown in (Fig. 1.5). These angles depend on the values of the
impedances and transmission speed of the two media.

vP1 vS 1 vP2 v S2
= = = (1)
sin α 1 sin β 1 sin α 2 sin β 2
 Figure 7. Snell’s Law [6]

The v i are the propagation speeds of the P-waves (longitudinal) and S-type (shear) in the materials 1
and 2 respectively, whereas, α i and β i are the angles of inclination of the longitudinal and shear waves
formed with the normal to the interface between the two media respectively.

With a significant variation of acoustic impedance between two media, the intensity of reflected waves
increases and that of transmitted waves decreases. Since longitudinal waves travel more faster than
shear waves (with ratio of about 2:1), the angles of reflection and transmission β i are in general
smaller compared to those of α i, as shown in Fig. 1.5.

The fundamental relationship that governs the reflection of an ultrasonic wave at the interface between
two media of different acoustic impedance Z1 and Z 2 was formulated in the nineteenth century by
Poisson and the reflection coefficient is mathematically expressed as follows:

( Z ¿ ¿2−Z 1 )
R= ¿ (2)
(Z ¿ ¿ 2+ Z 1)¿
A similar relationship can be expressed in terms of the transmission coefficient T , as the two
coefficients sum up to equal 1:

2 Z2
T= (3)
( Z 2+ Z 1)
Tables of values of transmission speed, for water, acrylic plastic and steel materials are extracted from
the manuals [9].

 Attenuation

When ultrasonic waves propagate through a polycrystalline medium, they are subject to phenomena of
attenuation (loss of ultrasonic energy) caused by three different mechanism; spreading, grain scattering
and absorption. Thus, it is essential to put into account the effects of these energy losses due to
attenuation. Attenuation is defined by the combination of these effects. The spreading of cylindrical
ultrasonic beam in materials occurs due to the interference of the waves generated at various points on
the beam element within the near field region of the beam and expressed as:

−1
A D ∝ A 0∗R 2 (4)

Where A D is the amplitude of the ultrasonic waves due to spreading in the location around the near
field region, A0 is given as the amplitude of the initial ultrasonic pulse and R is the wave propagation
distance [7]. Attenuation due to grain scattering occurs when the microstructural orientation and sizes
of the material particles are not perfectly homogenous, this steers the incident wave in many different
directions, resulting in net loss of amplitude with increment in propagation distance in the direction of
propagation. This effect is commonly observed on the oscilloscope as high frequency noise around the
attenuated signal .
[8,5]

The attenuation due to absorption losses occurs as a result of mechanical energy converted to heat
energy (particle collision) during the wave propagation. Generally, the attenuation due to grain
scattering and absorption can be combined and expressed as [7,8,5]

−α ∙t
A R= A 0∗e (5)
Where A Ris the amplitude of the received ultrasonic waves from within the material, A0 is the
amplitude of the received ultrasonic waves from the outer surface of the material, α is the attenuation
coefficient, and t is the propagation time. The attenuation coefficient is the summation of components
of various attenuation mechanism, and is basically expressed as:

α =α S + α A (6)
Where α S is the attenuation coefficient due to scattering, and α A is the attenuation coefficient due to
absorption of ultrasonic waves in the medium. The attenuation coefficient due to absorption is very
small and is often neglected [7], Note that due to dissipative effect, elastic waves are strongly
attenuated.

 Effects of the Presence of Flaws

The features of the resulting reflected signal amplitude of a flaw depends on many possible influences
such as . size of the flaw such as (the Area), surface condition of the flaw (high roughness involves
[9,10]

scattering) orientation of the flaw relative to the sound beam ( 90° reflectors visible, 0° at practically
invisible) and the conformation (geometry contour). Other external factors are the features of the test
sample that influences the amplitude of the signal reflection are related to its degree of roughness.

The discontinuity is an obstacle to the propagation of ultrasonic waves in the material. Depending on
the ratio between the size of the beam produced by the transducer and the size of the flaw or
discontinuity, the diameter of the flat-bottomed holes used as simulated flaw for this thesis will be set
as fraction value of the diameter of the transducer used in accordance to the Distance-Gain-Size
Method (DGS).

Most ultrasonic testing is performed with pulses containing significant frequency components ranging
roughly from 1 to 20 mega-Hertz (MHz). The reason why ultrasonic testing range is limited to
approximately 1 MHz at the low frequency threshold is due to the difficulties encountered when
detecting flaws that are significantly smaller than a wavelength, for instance, detecting small flaws of
about 1 mm in length, will be difficult at frequencies much less than 1MHz. In contrast, ultrasonic
testing range at the high frequency range is limited to approximately 20 MHz for most structural
materials since, although the wavelengths are very small, material attenuation due to the scattering of
the waves by the grains of the material is high at such frequencies, limiting the effective penetration of
the ultrasonic waves. This work will consider a frequency testing range of about 2-5MHz .
[11]

3) Governing Equations
 Wave propagation equation for a fluid

The propagation equation for a fluid is basically derived by the relationship between the equation of
motion of an ideal compressible fluid in motion and the constitutive equation expressing pressure in
terms of the fluid motion displacement.

 The Constitutive Equations

The main purpose of the constitutive equation is to help express the differential equation of fluid
motion as function of a single variable, such as the pressure, for this reason, it is necessary to relate the
pressure, p, to the fluid motion {u }, through the material properties of the fluid. Thus, for an ideal
compressible fluid we have:

p=− λ÷.{u } (7)

i.e. the pressure is proportional to the divergence of the displacement vector, also known as the
dilatation of the fluid, where the proportionality constant, λ , is the bulk modulus of the fluid. The
minus sign indicates the decrement of the fluid volume. Due to the effect of the positive pressures.

 The Equations of Wave Propagation in fluid

By adopting the differential equation of motion derived for an arbitrary volume V of an ideal
compressible fluid in motion , we find that:
[11]

∂2 {u }
−∇ p ( x , t ) + f ( x , t )=ρ0 ∙ (x.t ) (8)
∂t 2

Where p ( x , t ) is the pressure in the fluid at any point, x , and time t , with ∇ being the vector gradient
operator, the vector quantities f is the body force (force/unit volume) of the fluid. ρ0 is the density at
the initial or undisturbed state of the fluid, (since all of the waves applicable in this work involve very
∂2 {u }
small displacements and velocities) and the acceleration a= , where {u } is the displacement
∂ t2
vector under these conditions. Taking the divergence of both sides of the above equation, we can apply
the constitutive relationship in equation(X) to obtain:

2 ρ 0 ∂2 p
∇ p+ f = ∙ 2 (9)
λ ∂t

where ∇ 2=∇ ∙ ∇ is the Laplacian operator and f =−∇ ∙ f is a scalar body force term. Equation (3.10) is
mainly the 3-D inhomogeneous wave equation for the pressure which can be rewritten as:
 2
2 1 ∂ p
∇ p− 2
∙ 2 + f =0 (10)
C f ∂t

Where the quantity C f =

√ λ
ρ0
is the speed of sound in the fluid.

Note that since we are modeling the fluid here as an ideal fluid (i.e. viscosity-free), the tangential
component of the velocities need not be continuous at the interface.

 Plane Waves in a Fluid

With respect to the general 3-D wave equation for the pressure, p, in a fluid given by Eq. (3.11). If by
assumption the disturbances traveling in the fluid is considered to vary only in one spatial dimension,
x , i.e. p= p (x , t) , then the wave equation reduces to:

2 2
∂ p 1 ∂ p
2
− 2 ∙ 2 =0 (11)
∂ x C ∂t
Whose general solutions can be obtained in the form

( Cx )+ g(t + Cx )
p=f t− (12)

where f and g are arbitrary functions and C=C f is the fluid wave speed. The solution corresponding
to f represents a wave traveling in the + x -direction while the solution corresponding to g represents a
wave traveling in the −x -direction. The two solutions are one-dimensional plane wave solutions, as
proven by determining the values at fixed time t 0, where the pressure is constant on the planes x=ct 0
and x=−ct 0 , respectively for f and g .
[11]

By applying Fourier transform relations to the solutions f t− ( Cx )∧g(t + Cx ), then:

+∞

( )
f t−
x
=
1
∫ F ( ω ) exp ⁡[iω(¿ Cx −t)]dω ¿
C 2 π −∞
(13)
 +∞

( )x
g t+ =
1
∫
C 2 π −∞
x
G ( ω ) exp ⁡[iω(¿− −t)]dω ¿
C
(14)

This equation shows that the effect of plane wave propagation through a distance d in the positive or
negative x -directions will introduce complex exponentials in their Fourier transforms given by
exp ⁡(± iωd), respectively.

The above equations indicate that an arbitrary plane wave traveling in one-dimension in the + x -
direction can be considered to be the superposition (over all frequencies) of a harmonic wave of the
form.

p= A exp[iω ( Cx −t )] (15)

Where A=
F ( ω)
2π ( )x
is its amplitude and ω −t is its phase. Alternate forms for the harmonic wave of
C
Eq. (4.6) are:

p= A exp [ ik ( x−Ct ) ] (16)

p= A exp [ 2 π i ( x−Ct )
λ ] (17)

ω 2π
Where k = is called the wave number and λ= is the wave length. Since ω is related to the
C k
frequency, f , in Hertz through ω=2 πf , so, f and λ can be directly related through the wave speed by:

fλ=C . (18)

 Wave propagation equation for a Linear Elastic solid

The propagation equations for an elastic wave in solid are fundamentally deduced from the
conservation equations of matter and momentum, complemented by the constitutive equation of the
propagation medium.

 The Constitutive Equations

In the case of a homogenous isotropic linear elastic solid considered in this work, the constitutive
equation, can be expressed by considering first, the influence of axial orthogonal transformations on
the stiffness tensor (required constant all through for isotropic solid).This condition is satisfied only on
a scalar or unit tensor δ ij (which is unaffected by any change in axes), thus, each component C ijkl of the
stiffness tensor can be expressed in terms of the components of the unit tensor while satisfying
symmetry condition [12]. The stiffness tensor takes the form:

C ijkl= λ δ ij δ kl + μ ¿ (19)
This shows that the properties of an isotropic solid are mainly defined by only two independent
constants: the Lamé constants λ and μ.

By substituting the obtained expression for stiffnesses into the generalized Hooke’s law, we obtain the
normal and tangential stresses which can be grouped into a single equation, as expressed below.

σ ij ( {u }.t )=λ δ ij ε kk ( {u },t ) +2 μ ε ij ({u },t ) (20)

Where ε represent the strain tensor.

Other couples of constants and parameters are applicable. In this work it is common and convenient to
introduce the Young’s modulus and Poisson’s ratio. The Young’s modulus E is the measure of
resistance of a bar to a traction along its axis Whereas, the Poisson’s ratio v compares the transverse
strain with the longitudinal strain, that is, along the direction of the applied force. These parameters are
expressed as functions of the Lamé constants by:

( 3 λ+2 μ ) λ
E=μ∙ ∧v = (21)
( λ+ μ) 2( λ+ μ)
The bulk modulus K and the shear modulus G are also applicable. The bulk modulus expresses the
material resistance to a change in volume under a hydrostatic compression. The shear modulus
indicates the material resistance to a change in shape. These parameters are given by:
 2μ E E
K= λ+ = ∧G=μ= (22)
3 3(1−2 v ) 2(1+ v)
In the case of a perfect fluid, since G=0, only one constant is needed:

the bulk modulus K= ρ c2 for a liquid, with c being the speed of sound, and the inverse, the
1
compressibility coefficient x= for a gas.
K

For reasons of mechanical stability, the parameters E , K and G are often required positive: a traction
F ∆l σ
σ= along the axis of a bar causes an elongation = >0 ,a compression p=−σ causes a
S l E
∆ V −p σ
decrease in volume = < 0 , a shear stress σ 12>0 creates a deformation ε 12= 12 >0 .
V K 2μ

 Elastic Wave Equation and Wave decoupling

The mechanical displacement {u } of this solid is the solution to the equation of motion also known as
the local equation of dynamics , which is an expressed form of linear momentum when taking into
[12]

account the conservation of mass.

d vi ∂ σ ij d {u}
ρ∙ =ρ F i+ ∨ρ ∙ =ρ { F }+{÷}[σ ] (23)
dt ∂ xj dt
Where v i is the velocity of the solid particle followed in its motion, ρ Fi is the force per unit volume
also known as force density which arises only from the internal stresses, σ ij , through which the
external medium acts upon the particle.

By ignoring the convection term in the initial particular derivative leading to the above equation of
motion, this is written as:

2 2
∂ u ∂ ui ∂ σ ij
ρ∙ 2
=ρ { F }+ { ¿ } [σ ]∨ρ ∙ 2 =ρ F i + (24)
∂t ∂t ∂x j
where ρ is the mass density of the solid and where the source term ρ {F } is a volume force. In the
absence of a source ({ F }={0} ) and considering the constitutive equation of an isotropic solid, the
local equation of dynamics takes the form of the Lamé-Navier equation:

∂ {u}
2
ρ∙ =( λ+ μ ) {grad } ( ¿ { u } ) + μΔ{u } (25)
∂t 2
This differential system couples the three components of displacement, thus, obtaining two
independent equations by transforming it becomes advantageous.

With the prior knowledge of the Laplacian of a vector the above equation of motion takes the form:

∂ {u}
2
ρ∙ =( λ+ μ ) {grad } ( ¿ {u })−μ rot (rot {u })
∂t 2 (26)

The Helmholtz decomposition is used to obtain decoupled equations for Longitudinal and Transverse
wave: this equation leads to two independent equations of propagation:

2
∂ {u L } 2
2
−V L { grad } ( ¿ {u L })={ 0 }
∂t
and
2 (27)
∂ {uT } 2
2
−V { rot } ( {rot }{uT }) ={0 }
T
∂t

where the corresponding velocities V L and V T are defined by the relations:

V L=
√ λ+2 μ
ρ
∧V T =
√
μ
ρ (28)

 Plane Waves in an Elastic Solid

The plane wave phenomena demonstrate the propagation of bulk waves in an elastic solid, this work
will fundamentally consider the one-dimensional plane wave to illustrate the wave propagation on 2-D
simulation domain.

In general, the Navier’s equations are not wave equations as found for a fluid media. However, in
some special case of 1-D motions, Navier’s equations are certainly simplified to ordinary wave
equations . For example, if we assume {u }=u (x , t)e x , i.e. the motion of the elastic medium is only
[11]

a function of x and t so that the only component of displacement, is in the x -direction, thus, Navier’s
equations become (for zero body forces):
 2 2
∂ u 1 ∂u
2
− 2 ∙ 2 =0
∂ x C1 ∂ t (29)

which is the ordinary homogeneous 1-D wave equation for P-waves, with solutions given by

(
u=f t−
x
C1 )
for disturbances traveling in the + x -direction. For this P-wave, the motion is entirely in

the direction of propagation, as shown in Fig. 4.4a.

Figure 8, (a) Propagation of a 1-D plane P-wave, (b) propagation of 1-D plane vertical-shear
(SV) and horizontal-shear (SH) waves

Whereas, when it is assumed that the displacement vector has components only in the y - or z -
directions respectively, i.e. {u }=u (x , t)e y , or {u }=u (x , t)e z , we find that Navier’s equation reduce to:

2 2
∂ v 1 ∂ v
2
− 2 ∙ 2 =0 (30)
∂ x C3 ∂ t
or

2 2
∂ w 1 ∂ w
2
− 2 ∙ 2 =0
∂ x C2 ∂ t (31)
with solutions given by v , w=f t− ( x
C2)for disturbances traveling in the + x -directions. These are

shear (S) waves that only differ by their displacement directions (or polarizations). In this case the v-
disturbances represent vertically polarized shear (SV) waves while the w- disturbances represent
horizontally polarized shear (SH) waves as shown in (Fig. 4.4b).

 Interface/Boundary Conditions

When dealing with problems that involves elastic solids, a variety of interface/boundary conditions
can be encountered. The solutions to the equation of motion (5) must satisfy boundary conditions,
which connect the mechanical quantities at the interface between two media. They differ in nature
depending on whether the surface of the solid is free, rigid immobile surface, welded Interface
between two elastic solids, perfect or smooth Interface between two elastic solids and Fluid-Solid
Interface [11,12]. This work will mainly focus on the case of a smooth or perfect fluid interface between
two elastic solid and that of a fluid-solid interface.

Figure 9. If the density of the force { p } exerted on the surface ∑ is zero, the traction vector
{T } is continuous over ∑ [12]
.
For the traction vector T I ( {l })=σ ij l j, we integrate equation (5) over a cylinder of height h , limited by
two neighboring parallel discs, placed on either side of the interface ∑ (Figure 1). Taking into account
an external force with a density per unit area pi, we obtain:

(ρ ¿ ¿ 1+ ρ 2) d vi
[ σ ij( 2) l j−σ ij( 1) l j ] d S+ ρ F i h d S+ pi d S= 2
hd S ∙
dt
¿ (32)

When h tends to zero, only the terms due to the surface forces remain:

( 2) (1 ) ( 2) (1 )
σ ij l j−σ ij l j=− pi , hence T i −T i =− pi (33)

 Smooth or Perfect Interface between Two Elastic Solids

If a thin fluid layer with interface of normal {l 1 } separates two elastic solids where the layer thickness
can be neglected and in the absence of any external force acting on the surface pi, then a smooth or
perfect contact implies that the normal component (oriented along {l }) of both the traction and
displacement (or velocity) vectors must be continuous whereas the tangential (shear) components of
the traction vector must vanish : These boundary conditions are formally written for two media (1)
[11,12]

and (2), as:

¿ (34)
This condition holds true regardless of the choice vector direction between materials. This work will
consider this boundary condition applicable to the case of a transducer mounted on a wedge which
then is placed in contact with an underlying solid (rail foot). Since in practical a fluid couplant exists
between the wedge and the solid, these smooth interface conditions are appropriate.

 Fluid-Solid Interface

In the case of a fluid (superscript 2) in contact with a solid (superscript 1), that is to say where the fluid
is medium one and the solid medium two, the interface conditions are very similar to that of the
smooth interface since in this case the normal traction and displacement (or velocity) also must be
continuous, and the shear stresses must also vanish . These conditions represented in fluid and
[11,12]

solid medium become:

{[σ ¿¿ 1] ( {x }, t ) . l=[σ ¿¿ 2] ( {x }, t ) . l ¿¿ {u1 } ( {x }, t ) . l={u2 } ( {x }, t ) . l at x ∈ ∑ (35)

Note that the normal traction [σ ¿¿ 1] ( {x }, t ) . l ¿ in the fluid medium can also be expressed in terms of
pressure in the fluid. This work will consider this boundary condition applicable in immersion
inspection method since the transducer is placed in a fluid and the sound beam it generates must pass
through a fluid-solid interface to inspect a solid medium. Note that the curly bracket notation in all
equations above indicates vectors whereas the square bracket notation indicates matrix.

4) Finite Element Analysis Method

 Application of FEM in PDEs

The language of mathematics is an essential means used to comprehensively understand and quantify
any physical phenomena, such as, wave propagation, structural or fluid behavior, thermal transport etc.
A vast majority of these processes are described with partial differential equations (PDEs) in space and
time dependent domain. However, due to the limitations of analytical methods in solving these PDEs
for most geometries and problems (especially complex forms), numerical techniques have been
developed over the last few decades and one of the most prominent today is the finite element method.
In this technique, an approximation of the equations is constructed, typically based on various forms of
discretization. These discretization methods approximate the PDEs with numerical model equations,
which can be solved using numerical methods. The solution to the numerical model equations are
consequently, an approximation of the real solution to the PDE’s. The finite element method (FEM) is
used to compute such approximations .
[13–16]

Regardless of the of the motivation for using finite element analysis, it is of utmost importance to
understand the various kinds of PDEs and their suitability for use with FEM. It is critical to note that
FEM is a tool and any tool is only as good as its user. One consequence of using a numerical
framework that is unsuitable for the type of PDE is that, such usage leads to solutions that are known
as “improperly posed.” This could mean that small changes in the domain parameters lead to large
oscillations in the solutions, or that the solutions exist only in a certain part of the domain or time,
which are not reliable. On a contrary, well-posed explications are defined as those where a unique
solution exists continuously for the defined data. Hence, for the sake of reliability, it is extremely
important to obtain well-posed solutions .
[16]

PDEs are mainly classified as elliptic, hyperbolic, and parabolic. When solving these differential
equations, boundary and/or initial conditions need to be provided, thus, they are commonly known as
initial boundary value problem (IBVP), the necessary inputs can be evaluated. Examples for PDEs in
each category include the Poisson equation (elliptic), Wave equation (hyperbolic), and Fourier law
(parabolic). The finite difference methods (FDM) and variational (or energy) methods are the two
main numerical approaches used to solve PDEs. FEM falls in the latter, the variational approaches are
fundamentally based on the principle of energy minimization .
[16,15]

Hyperbolic PDEs are commonly associated with discontinuous solutions. For instance, the wave
equation is a hyperbolic PDE. Due to the existence of discontinuities (or jumps) in solutions, the
original FEM technology (or Bubnov-Galerkin Method) was believed to be unsuitable for solving
hyperbolic PDEs. However, over the years, modifications have been developed to extend the
applicability of FEM technology. These compensating FEM technologies includes; Extended Finite
Element Method (XFEM), Generalized Finite Element Method (GFEM), Mixed Finite Element
Method, hp-Finite Element Method and Discontinuous Galerkin Finite Element Method (DG-FEM).
Amongst these methods, the Discontinuous Galerkin Finite Element Method (DG-FEM) has shown
significant promise for utilizing the idea of finite elements to solve hyperbolic equations, where
traditional finite element methods are limited .
[16,15]

 Principle of energy minimization

When a particular boundary condition is applied to a body, this can lead to several configurations but
yet in reality only a particular configuration is possible or achieved. Even when the simulation is
performed multiple times, same results prevail. This is governed by the principle of minimization of
energy. It states that when a boundary condition (like displacement or force) is applied, amongst the
numerous possible displacement configurations that the body conforms, only that configuration that
satisfies the equilibrium condition minimizes the total potential energy. Hence, the location with the
minimum total potential energy is selected. The principle of minimization of energy constitutes the
main backbone of finite element method [16]
.

Π=U e +V (40)
Where Π is the total potential energy, U e strain energy and V is the potential energy of external loads.

 Technical Overview of Finite Element Method

This section will provide a concise description of the mechanism of FEM. When we consider the case
of a simple 1-D problem to fundamentally depict the various stages involved in finite element analysis,
then the core terminologies describing FEM include;

 Weak Form

One of the initial steps in FEM is to identify the PDE associated with the physical phenomenon. When
the PDE (or differential form) is directly applied, it is commonly known as the strong form due to their
capability of providing the exact solutions however, they are mostly limited to simple cases, whereas,
the integral form is known as the weak form because they are only capable of solving by
approximation, which in turn negotiates accuracy but provides the advantage of solving even complex
problems with ease. Consider the simple differential equation as shown below. The equation is
multiplied by a trial function v (x) on both sides and integrated within the domain [0,1] .
[16]

2
d u (41)
2
=f (x)
dx

2
∫ dd xu2 ∙ v ( x )=∫ f (x )∙ v (x ) (42)

Now, by applying the integration by parts, the LHS of the above equation can be reduced to

|
1
du du dv (43)
∙ v (x ) −∫ ∙ =∫ f (x)∙ v (x)
dx 0 dx dx
As it can be seen from the new equation obtained, the order of continuity required for the unknown
function u(x ) is reduced by one. The earlier differential equation required u(x ) to be differentiable at
least twice whereas the integral equation requires it to be differentiable only once. The same is true for
multi-dimensional functions, but the derivatives are replaced by gradients and divergence. In
accordance to Riesz representation theorem u(x ) can be proved to have a unique solution for the
integral and also the differential form. Furthermore, if f (x) is smooth, it also ensures that u(x ) is
smooth .
[16]

 Discretization

Once the integral or weak form is set up, the next step is the discretization of the weak form. The
integral form needs to be solved numerically and hence the integration is transformed to a summation
that can be calculated numerically. In addition, one of the primary goals of discretization is also to
convert the integral form to a set of matrix equations that can be solved using well-known theories of
matrix algebra. The entire domain is divided into small pieces known as “elements” and the corner
point of each element is known as a “node” (the point of connection between elements). In order
words, from numerical treatment the model is discretized into a finite number of parameters, called
degrees of freedom (DOF). There are numerous of different styles of the finite elements. For a planar
model, the quadrilateral and triangular elements are the most commonly used, which can also be
converted into axisymmetric elements. . The unknown functional u(x ) are calculated at the nodal
[5,16]

points. Interpolation functions are defined for each element to interpolate, for values inside the
element, using nodal values. These interpolation functions are also often referred to as shape functions.
Thus. The unknown functional u(x ) can be reduced to:

nen
u ( x )=∑ N i ∙u i (44)
n=1

where nen is the number of nodes in the element, N i and ui are the interpolation function and
unknowns associated with node i, respectively. The weak form can now be reduced to the relation
between force and displacement acting on each node in a static system, given in the matrix form ;
[5,16,15]

{ F }=[ K ] {u } (45)
Where [ K ] is the global stiffness matrix, {u } is the vector depicting nodal displacements and { F } is the
vector depicting nodal acting forces. Generally, FEM over estimates the actual stiffness of the model,
which means an underestimation of displacements. The type of element used for discretization can
also influence the model stiffness, where for instance, the use of triangular elements increases the
stiffness of the model compared to rectangular elements. In this case, solution will in general be more
accurate in high stress regions, with the exception of high gradient regions. The analysis can be
assumed to be linear or nonlinear, in cases, where the computational time is reduced significantly with
the linear model. A linear model implies that the stiffness matrix [ K ] is assumed to be constant. This is
valid if no large deformations occur during the solution. Otherwise the model should be solved with a
nonlinear analysis, i.e. the stiffness matrix [ K ] will be calculated for every iteration during the
solution. For a dynamical system the equation of motion for the discretization becomes [5];

2
d u du
M 2
+C +[ K ]u=F (t) (46)
dt dt
2
d u du
where M is mass, C damping coefficients, 2 is the acceleration of mass, and is the velocity of
dt dt
mass. Note that whenever we apply external forces and calculate displacements, it is called
Displacement-based procedure whereas whenever we apply displacements and determine the forces, it
is called a Force-based procedure . Static analysis normally does not consider inertia mathematically
[15]

and the main difference between static and dynamic analysis is that in a static analysis, only the
stiffness matrix of the FEA model is solved. Whereas, in a dynamic analysis, in addition to the
stiffness matrix the solution to the mass matrix (and damping matrix, if not zero) is also determine .
[17]

 Solvers

Once the matrix equations have been established, the next step is to pass the equations to a solver in
order to solve the system of equations. Based on the type of problem, direct or iterative solvers are
generally used . A more detailed overview of the solvers and how they work, as well as tips for
[16]

suitable choice, are available to the reader here .

[18]

 The Discontinuous Galerkin Finite Element Method (DG-FEM)

In the conventional FE continuity of dependent variables and reciprocity of tractions across the
element interface remains constant, whereas, in the DGFE method, as the name implies, the continuity
of dependent variables across the element boundaries is not explicitly implemented. For this reason,
the number of unknown dependent variables is much compared to the conventional FE method. In the
DG method the governing differential equation is presented as two distinct differential equations of
smaller order than the original governing equation. Then the weak forms are determined for each of
the differential equations . In the weak form, numerical fluxes for the variables in the differential
[19]

equations are used. Various kinds of numerical fluxes are proposed in literature and used to merge the
variables across the inter-element boundaries [20–22]
. Details of the method and derivation steps are
illustrated below with a second-order differential equation. Consider once again the differential
equation within the domain [0,1] ;
[19]

2
d u
2
=f ( x ) ; 0 ≤ x ≤l (47)
dx

2
d u (48)
2
=f (x)
dx
In order to analyze this problem using the DG method a 2-element idealization is used in the following
with equal size elements of length h. The first step in this method is to rewrite Eq. (50) as two first
order differential equations:

du dτ
=τ (x) and =f (x ) (49)
dx dx
where τ (x) is a new variable. The next step is to develop weak forms for the two equations in Eq.
(51). This is accomplished by first writing the weight residual (WR) statements and then by integrating
by parts as;

∫ (¿ du
dx
– τ ( x ))∙ v ( x ) ∙ dx=0 ¿ (50)
0

and

∫ (¿ du
dx
– f ( x ))∙ w ( x ) ∙ dx=0 ¿ (51)
0

where v ( x ) and w ( x ) are the two weight functions.

In the DG method, Eqs. (52) are used to form the matrix equations between the nodal displacements
{u} and the independent variables τ ( x ) as;

[ B1 ] {u }−[ R ] { τ }={0 } and [ B2 ] {u }−[ k ] { τ }={ R } (52)

Where
 { u } ={u(11) u (21) u(12 ) u(12) } and { τ }={τ(1) (1 ) (2) (2 )
1 τ2 τ1 τ2 } (53)
In the DG method the variables { u } and { τ } are listed element by element; i.e. in the above Eq. (54)
where the subscript denotes the nodes and the superscript denote the element number. Note that [ R ] is
a symmetric matrix and has the same form as a ‘mass’ matrix in conventional FE. Also note that this
matrix is always invertible. Hence, the {τ} can be eliminated from Eqs. (75) and the final DG final
matrix can be expressed as [19]
;

−1
{ τ }=[ R ] [ B1 ] {u } (54)
And by substituting { τ } in the 2nd equation in Eq. (53) we obtain

−1
( [ B2 ] −[ k ][ R ] [ B1 ] ){u }={R } (55)
The above Eq. (56) is used to solve for the nodal displacements {u }. The LHS matrix, in the
parenthesis, is in general, not symmetric but is invertible. The nodal primary variables can be solved
using the above Eq. (76). The details of the development of the matrices will be ignored in this work.
In a simple sense, the DG method can be considered as a hybrid of the FE method and the finite
volume (FV) method, combining features of both numerical methods. Similar to the FE method, the
DG method uses shape functions within each element. And like the FV method also, the solution is
discontinuous between elements, and the elements interact by means of fluxes across the element
interfaces. The properties of the DG method make it an attractive option for simulations of certain
types of physical phenomena. This method can be a computationally efficient approach for solving
hyperbolic neutron transport and wave propagation equations, most especially with first-order terms.
In recent years, some FEM commercial software, such as COMSOL have introduced several interfaces
that use the DG method for applications in fluid dynamics, electromagnetics and acoustics .
[23]

In addition to these physics-specific interfaces, COMSOL users can implement custom DG models
using the Wave Form PDE interface that is found in the Mathematics >PDE Interfaces section of the
Select Physics and Add Physics windows. Compared to the FE method, the DG method can be more
efficient, accurate and numerically stable for systems that are described by hyperbolic PDEs. When
deciding whether to use the DG method for a particular problem, it may be of importance to consider
the following notes [23]
:

 The DG method is well-suited for managing sharp jumps in the solution variables (such as
shock waves or phase transitions).
  For large problems, the memory usage of DG models can be much less compared to FE
models.
 The solver time per timestep is usually much less for DG models compared to FE models.
 Timesteps for DG analyses normally need to be much smaller than for FE analyses with
implicit time-stepping, thus, the DG method commonly requires many more timesteps than the
FE method.
 Depending on the problem, DG models may require some tuning of the limiter, filter and
solver setting to enable reasonably large timesteps while maintaining numerical stability.

5) Equipment
 Theory (From Professor)

Chapter 3
Finite Element Simulation and Experimental Validation
1) Finite Element Method Procedure in COMSOL Multiphysics
(General Perspectives)
The use of FEM in the modelling and simulation of various physical phenomena (such as wave
propagation) related to assembled components by employing the commercial computational software
COMSOL Multiphysics is discussed in this section. The finite element analysis in COMSOL
Multiphysics is performed by following a procedural flow, as shown in the figure below.
 Figure 10. Fundamental steps to perform a finite element analysis in COMSOL Multiphysics.
The method can be divided into three basic steps, namely; pre-processing, solver, and post-processing
[25,13].
The pre-processing step enables the user to build the model. Since it contains all the information
regarding the FEM application for interconnect study, it is convenient to divide it into smaller sub-
steps. The first sub-step involves the creation of the interconnect geometry (1D, 2D, or 3D), to
represent the domain under study, and assign the material properties to the domain. Then, the virtual
physical environments of the problem under investigation are generated by assigning the underlying
physics (or Multiphysics), mathematical equations, and finite element formulation to the model.
Typically, the latter is embedded among the core of the commercial software. Subsequently, the
application of appropriate loadings, boundary and initial conditions to the domain under study, as well
as its discretization into finite elements, determines the matrix equation governing the model,

This step is followed by solving the set of algebraic equations, which provides the physics-related
nodal solutions of the model. COMSOL Multiphysics mainly employs two methods to obtain the
model solution, namely segregated step method and fully coupled method. The segregated step method
generates the set of algebraic equations for every physical model under consideration and calculates
the resulting solutions for each of those models by following a sequence specified by the user. The
fully coupled method generates the single set of algebraic equations for all the involved physical
models and implements them in a single iteration scheme which is repeated until convergence is
reached. Once the solutions of the problems are determined, the post-processing step enables the user
to evaluate the results of the finite element analysis by means of plotting and data exporting tools.

 The Pressure Acoustic Module

The Pressure Acoustics interfaces are used to model classical pressure acoustics. Various numerical
formulations of the governing equations exist. The finite element method (FEM) based physics
interfaces solve for the acoustic variations in pressure. The Helmholtz equation is solved in the
frequency domain and the scalar wave equation is solved in the time domain. Domain conditions exist
for modeling losses in a homogenized way in porous materials as well as in narrow regions. Sources
such as background fields are also available. Boundary conditions include sources, nonreflecting
radiation conditions, impedance conditions, periodic conditions, exterior field computation conditions,
as well as interior boundary conditions such as walls or perforated plates. A boundary element method
(BEM) based interface, Pressure Acoustics, Boundary Element, also exists for modeling in the
frequency domain. In the time domain, the Pressure Acoustics, Time Explicit interface, uses a
formulation based on the discontinuous Galerkin method (dG-FEM) and uses a time explicit solver.
The interface is well suited for modeling large transient linear acoustic problems containing many
wavelengths .
[26]

 The Elastic Wave Module

The Elastic Waves interfaces are used to compute the propagation of elastic waves in solids as well as
the coupled propagation of elastic and pressure waves in porous materials. Elastic wave problems can
be solved both in the frequency domain and in the time domain while the poroelastic wave
propagation is covered in the frequency domain. A dedicated interface, based on the discontinuous
Galerkin method (dG-FEM), that uses a time explicit solver also exists. The method is very memory
lean, It is suitable for transient linear elastic wave simulation with arbitrary time-dependent sources
and fields and propagation problems over large domains containing many wavelengths, thus applicable
in cases such as ultrasound propagation like nondestructive testing (NDT) or seismic wave
propagation in soil and rock. The interface exists in 2D (generalized plane strain) and 3D, and includes
absorbing layers that are used to configure effective nonreflecting boundary conditions (such as
sponge layers). The interface supports damping, as well as isotropic, orthotropic, and anisotropic solid
material formulations. The model solves the governing equations for a general linear elastic material in
a velocity-strain formulation and are valid for both isotropic and anisotropic material data. Dissipation
can be added to the model in the form of Rayleigh damping, consequently, leading to an additional
right-hand side terms to the equation of motion with a mass and stiffness damping term. The solid
domains can be coupled to fluid domains using one of the many built-in Multiphysics couplings such
as Pressure Acoustics. Predefined Multiphysics interfaces exist and are located in the Acoustic-
Structure Interaction folder. In order to maximize the advantages of the time-explicit formulation, the
use of so-called nonconforming meshes (meshes that are discontinuous at a boundary) is essential
when coupling domains with distinct properties, which is a natural and application of the properties of
the discontinuous elements. When meshing a model solved with the Elastic Waves, Time Explicit
interface, three notes that need to be put into consideration :
[27]

c min
 Resolve the smallest wavelength( λ min= ) with about 1.5 elements;
f max

Due to the fact that various waves in a linear elastic solid propagate at different speed and have a
distinct wavelength, the slowest wave c min has the shortest wavelength λ min, which is the length scale
defining the mesh size The default for the Elastic Waves interface (similar to other time-explicit
interfaces) is to use quartic (fourth-order) shape functions. This defines a Maximum element size as:

λmin c min
h max= =
1.5 1.5 ∙ f max

where f max is the maximal frequency component to be resolved in the propagating signal.
This implies that only about 1.5 mesh elements per wavelength is needed to resolve the wave as
quartic elements are used.
 Resolve geometry details and curved surfaces:
The computational mesh should, of course, also resolve geometric details and curved surfaces
adequately. Typically, with the dG method, curved surfaces are properly resolved by setting the
Curvature factor to a value between 0.3 and 0.4. (However, note that the internal feature of COMSOL
is such that it uses a higher-order mesh element that curve).

 Generate a mesh that avoids small elements, if possible.

The time-explicit time-stepping scheme used in this method sets a strict limit on the internal time step
the solver takes when stepping forward in time. The time step is controlled by the globally smallest
value of the so-called cell wave time scale. The cell wave time scale is defined as the ratio between the
local mesh size and the fastest wave propagating, at speed c max . This implies that a small mesh element
in a model will control the global time step the solver can take. In most applications, the pressure
waves propagate faster than the other type of waves, and hence in this work the default is, the pressure
wave speed (when advanced physics option is applied the value can be changed). We can visualize
the cell wave time scale by plotting the function ( elte . wtc). The value does not directly represent the
internal solver time step, but it enables the user to identify problems amongst the elements and control
the time step. In 2D, one option to generate a triangular mesh of uniform size is to first generate a
mapped mesh and then transform it into triangles.

2) Fluid-Solid Model setup and Corresponding Simulation

This part of the chapter will provide the initial experimental validation for the finite element
method applied to this work, in this section the immersion test simulation will be performed for
three different cases and the equivalent experimental validation will be performed for such cases.
The immersion test simulation in COMSOL is performed by pairing the pressure acoustic and
elastic wave interface in a Multiphysics manner. The main purpose of the immersion test is to
acquire some information necessary to complement the optimization case of the partial immersion
technique, bearing in mind that the actual simulation of the rail foot will ignore the thin layer of
water interface between the wedge and rail foot. In addition because this simulation is performed
on the frequency domain this will allow an appropriate comparison of FEM with that of the elastic
wave time domain interface.
 The models for the immersion test simulation is designed for a normal incidence setup without
flaw, with 10mm flaw normal, with 10mm flaw oriented at 30 degrees from normal incidence,
and a case of 5 degrees transducer incidence angle without flaw and with 10mm flaw. This
selection was done empirically and on the basis of tendency relative to a significant change.
The tables below shows the, material properties and test parameters for the immersion setup.

S/N Material Density(kg/m3) Speed of Sound Waves (m/s)

Longitudinal Shear
1 Water 1000 1480 -
2 Steel 7850 5840 3130
Table 1. material properties
 S/N Description Values
1 Diameter of Steel Specimen 63.5 (mm)
2 Depth of Steel Specimen 100 (mm)
3 Transducer Underwater Acoustic Path 100 (mm)
4 Diameter of defect 10 (mm)
5 Thickness of defect 0.1(mm)
6 Transducer Diameter 12.7 (mm)
7 Thickness of transducer element 3.5 (mm)
8 Thickness of Absorbing Layer 7 (mm)
9 Center Frequency 5 (MHz)
Table 2. Test parameters for immersion test.

 The main conditions considered are:

 An isotropic loss factor of zero for steel to enable sufficient wave energy visibility within the
domain, when plotting for acoustic Intensity.
 The poison ratio is considered to be that of a perfectly elastic material and it is used to find the
corresponding young’s modulus in relation to the longitudinal wave speed and density of steel.
The relation combines the equation (21) and (28) to determine the value of young’s modulus.
 To ignore transmission coefficient with minimal reflection or absolute transmission of waves,
acoustic impedance matching, plane wave boundary condition and low reflecting boundary
condition was applied.
 Model Geometry Domains, Materials, Mesh and Corresponding Physics
 Normal No Flaw Normal with Flaw Normal with 30degree flaw

5 degree with No Flaw 5 degree with Flaw 10 degree with no flaw

 Simulation Results

Normal No Flaw Normal with Flaw Normal with 30degree flaw

5degree with No Flaw 5degree with Flaw 10 degreewith No Flaw

 Plot the simulation results.

 Present corresponding experimental setup and results for each case.
 Discuss the correlation between the two results and conclude.to validate the FEM method
 Chapter 4
FEM Simulation on Rail foot and selection of optimal settings for experiment.
 Rail profile specifications
The numerical model will be designed in accordance to a standard 60kg/m international union of
railway (UIC) steel rail section that is supported on the foot with two fastener having a maximum size
of 2.75cm measured from the extreme edge of the rail foot and positioned on opposite sides of the rail
foot edge. A well labeled schematic representation of the rail with equivalent dimension is shown
below.
 R400 Centre Line

Figure. Standard 60kg/m UIC rail profile sketch and equivalent dimensions
 Rail model transformation and Simplification for computational suitability
Due to some realistic point of view and computational advantages, this work will consider some key
transformation and simplification of the original full scale rail model. In this work, the main rail
profile without a fastener will be replaced with the profile with a fastener only, this is so because for a
mobile detection system, the regions of rail foot covered with fasteners will certainly obstruct the
continuity of the process therefore, the design is done while considering such critical regions. In such
case this work select the rail foot slope that is a few distance away from the maximum size of the
fastener measured from the extreme edge of the foot and all the probes needed for full scale detection
will be placed in this region with their respective orientation to area of detection. The next step is to
simplify the model to the main area of interest in this work, which is the rail foot, to achieve this, a
major section from the rail head is removed, and only a portion of the rail web is attached to the rail
base, this is done for computational advantage. Finally, due to the symmetry of the desired rail foot
inspection system, in other words, the mobile inspection is performed in a duplicated form on both
sides of the rail foot simultaneously, only a halved section of the already simplified rail foot connected
to the web and fastener will be used.

 Test model and probe setup for different test region.

The test model for this work will be divided into three different cases, which depends on the location
of the flaws on the rail foot and corresponding probe orientation. At least two simulated flaws few
distance apart will be embedded around each region of focus, one of these flaws will be ensured to be
located at extreme position as possible from the probe, this is to allow the farthest possible detection
from the probe. Two flaws will be embedded towards the center of the rail foot, another set will be
placed directly under the slope away from the selected maximum fastener length from the rail foot
edge and finally another two sets of flaws will be placed toward the edge of the rail foot directly
beneath the fastener. The orientation of the probe is also configured in this manner respectively. Based
on the required tolerance from the fastener location the optimal geometric dimension of the wedge is
set as 19mm width, 16mm lateral length,8mm height for o degree incidence and 10mm height for the
angle beam orientation. With an exception of the normal incidence probe placed directly on the slope a
set of incidence angle will be selected empirically within a feasible range for the other orientation
towards the center and edge. The selected angles will be optimized, and the most suitable case will be
based on the extent of signal reflected or received (that is the amplitude of reflected signal). The size
of flaw is based on the flat bottom hole (FBH) flaw sizing theory, which sets the flaw size to a
fractional value of the transducer diameter, This work desires the flaw size to be as small as 2mm with
a tolerance of 1mm. For the sake of convenience when receiving the reflected signal the flaw types are
made to be circular holes positioned at the bottom of the rail foot, this is also designed with regards to
the corrosion pit located at the bottom of the foot.

 Initial input data and table values for mesh and Simulation

This model employs the Elastic Waves, Time physics interface that implements the time-explicit
discontinuous Garlerkin finite element method (dG-FEM). The source signal is given by a modulated
Gaussian pulse, controlled with two sets of central frequency of 2.25MHz and 2.5MHz.

Figure. Source signal at 2.25MHz

Figure. Source Signal at 2.5MHz

Tables for values of material Properties

 Models solved at a central frequency of 2.25MHz with equivalent mesh

Figure 1. Sample Models for 0 degrees Incidence without flaw

Figure 2. Sample Models for 0 degrees Incidence without flaw

S/N Optimal Geometric Dimensions for probe at 0 degrees Incidence

1. Wedge width (W) 19mm
 2. Wedge Height (H) 8mm
3. Lateral Length of Wedge (L) 15mm
4. Transducer Diameter (D) 14mm

 Corresponding simulation results for each case with the variation of frequency and angle of
incidence.
 Analyze and discuss the results to choose the best setting.

Chapter 5
Experimental setup and results for rail foot

 Provide detail explanation of equipment design and settings

 Provide the probe setting based on the optimal simulation settings
 Discuss how experiment is performed.
 Present the corresponding results to the best simulation results settings