Engineering Fracture Mechanics 145 (2015) 43–53

Contents lists available at ScienceDirect

Engineering Fracture Mechanics

journal homepage: www.elsevier.com/locate/engfracmech

Analysis of strength and wave velocity for micro-damaged

elastic media
M.A. Sumbatyan a,⇑, M. Brigante b
a
Southern Federal University, Institute of Mathematics, Mechanics and Computer Science, Milchakova Street 8a, Rostov-on-Don 344090, Russia
b
Department of Structures for Engineering and Architecture, University of Naples – Federico II, Via Claudio 21, 84125 Napoli, Italy

a r t i c l e i n f o a b s t r a c t

Article history: A new macroscopic theoretical approach is proposed for strength analysis of damaged elas-
Received 11 February 2015 tic bodies. Two types of defects are considered: cracks and round voids. The method
Received in revised form 30 June 2015 applied is founded upon the Boundary Integral Equations (BIEs) method. The problem is
Accepted 3 July 2015
studied for both anti-plane and in-plane mode of deformation. We give respective dia-
Available online 10 July 2015
grams for calculated physical quantities. The results obtained in frames of the BIE method
are compared with an alternative technique based on a propagation of ultrasonic impulses
Keywords:
through the damaged media. There is evaluated a correlation between strength of the
Defects
Boundary elements
micro-damaged media and the wave speed of the of ultrasonic (US) impulse.
Strength Ó 2015 Published by Elsevier Ltd.
Micro-damaged media
Ultrasonic evaluation

1. Introduction

It is well known in the engineering practice that elastic materials with arrays of micro-defects possess absolutely different
mechanical properties when compared with homogeneous materials. Strength and fracture analysis of such elastic bodies
with defects is an important problem in various fields of engineering science, in mechanics of concretes, composites, soils,
ceramics, powder materials, etc. [1–4].
Various techniques are usually used to evaluate mechanical behavior of damaged materials. From the practical point of
view, standard non-destructive techniques like ultrasonic evaluation may be applied to this problem [5,6]. However, appli-
cation of such classical methods, efficient for metals, plastic materials, some kinds of composite materials, is faced with sig-
nificant obstacles when operating with strongly damaged media. The nature of the principal difficulties in the application of
ultrasonic techniques to strongly damaged media is connected with the fact that there is no reliable model, which can pre-
dict qualitative and quantitative correlation between the ultrasonic wave speed and the damage level (i.e. the number of
flaws per unit volume) of the medium. In mechanics of concretes, there has been a lot of heated debate among researchers
about the matter whether a change of the through-transmission wave velocity can correctly predict the change of mechan-
ical properties of the material. In spite of sharp criticism [7], the evaluation of strength properties of concrete on the basis of
the ultrasonic velocity in the through transmission technique is nowadays a conventional classical method of analysis.
Chronologically, first approaches to evaluate mechanical properties of micro-damaged media were founded on stochastic
theories [8,9]. However, a more advanced approach has been developed based upon analytical, numerical and experimental
investigation of the damaged media modeled as normal elastic bodies with a lot of randomly distributed defects [10]. Some

⇑ Corresponding author.
E-mail addresses: sumbat@math.rsu.ru (M.A. Sumbatyan), brigante@unina.it (M. Brigante).

http://dx.doi.org/10.1016/j.engfracmech.2015.07.005
0013-7944/Ó 2015 Published by Elsevier Ltd.
44 M.A. Sumbatyan, M. Brigante / Engineering Fracture Mechanics 145 (2015) 43–53

Nomenclature

u displacement vector
T vector of the internal stress in the fractured medium
k; l elastic moduli
n unit normal to the elementary area inside the medium
s unit tangential vector on the elementary area inside the medium
sij stress tensor
x; y Cartesian coordinates
e semi-length of the elementary crack
U Green’s function
U ij fundamental displacement tensor
Rij fundamental stress tensor
kN maximum stress intensity coefficient at the tip of a crack in the cracked medium with N cracks
Ts mean time delay of Ultrasonic impulse
vN through-transmitted wave speed in the cracked medium

deterministic methods have also been applied by many authors to this subject of investigation [11–17]. Alternative methods
are required to study systems of regularly distributed defects, like cracks in [18], where a good survey can be found too. Let
us note that the methods proposed to both chaotic and regularly distributed arrays of flaws allowed the authors cited above
to establish important qualitative and quantitative properties of the micro-damaged media. We only note in this connection
that one of the most powerful approaches here is the application of the Boundary Integral Equations (BIEs) method, applied
to both single [19] and multiple [10,14–16] defects.
In the present paper we further develop this promising direction of investigation, to apply it to the deterministic strength
analysis of the damaged elastic solids loaded by a static uniform outer load. The problem is studied for both anti-plane and
in-plane mode of deformation. The method is founded upon a numerical treatment of the system of BIEs holding over faces
of all flaws forming a huge array. The defects are simulated as either cracks or round voids. In the case of cracked media the
BIE itself is allied to a special treatment of the Boundary Element Methods (BEM) in the form of ‘‘displacement discontinuity
method’’, where a square-root singularity of stress is not explicitly extracted from the structure of the solution but this is
obtained numerically.
We give respective diagrams on results of the computations. The results obtained in frames of BIE are compared with an
alternative method based on a through transmission technique with the use of ultrasonic impulses. There is evaluated a cor-
relation between strength of the micro-damaged media and the through-transmission velocity of the ultrasonic (US)
impulse.
It should also be noted that another approach to evaluate internal structure and related crack geometries is connected
with the so-called ‘‘inverse’’ identification problems. Some our recent works with co-authors are devoted to this topic
[20–22], where the reader can also find some other helpful references.

2. Basic equations of the elasticity theory for anti-plane and in-plane modes of deformation

The outer load s0 applied to the solid is modeled as a uniform stress distributed over a pair of the opposite straight lines
y ¼  h taken somewhere far from the domain containing the considered set of defects. In the case of the so-called
‘‘anti-plane’’ (or ‘‘shear-stress’’, SH) problem, this applied stress is directed perpendicularly to the considered plane ðx; yÞ,
so that the mode of deformation is identical for all cross-sections z ¼ const [23]. Then in a fixed rectangular Cartesian coor-
dinate system Oxyz the non-trivial components of the stress tensor and the governing equation are

@uz ðx; yÞ @uz ðx; yÞ @ 2 uz ðx; yÞ @ 2 uz ðx; yÞ

sxz ðx; yÞ ¼ l ; syz ðx; yÞ ¼ l ; þ ¼ 0; ð2:1Þ
@x @y @x2 @y2
where l is the shear elastic modulus, and u  ¼ f0; 0; uz ðx; yÞg is the displacement vector. The internal stress vector
T ¼ fT x ; T y ; T z g on arbitrary elemental area with the unit normal n  ¼ fnx ; ny ; 0g in the considered case of anti-plane deforma-
tion has the only non-trivial component:
 
@uz @uz @uz
T x ¼ T y ¼ 0; T z ¼ sxz nx þ syz ny ¼ l nx þ ny ¼ l : ð2:2Þ
@x @y @n
Then the boundary conditions over a set of cracks ‘ located in a certain domain inside the medium should be satisfied over
the faces of all defects which are free of load. Let us represent the full solution as a sum of the stress caused by the applied
load s0 in the undamaged medium and the perturbed one:
 M.A. Sumbatyan, M. Brigante / Engineering Fracture Mechanics 145 (2015) 43–53 45

so
uz ðx; yÞ ¼ uoz ðx; yÞ þ wðx; yÞ; uoz ðx; yÞ ¼ y; T z ðx; yÞ ¼ s0 ny þ sz ðx; yÞ: ð2:3Þ
l
It is obvious that the latter is defined as a solution with the specific boundary condition:
 
@uz  @w s0
T z j‘ ¼ 0;  sz j‘ ¼ s0 ny ; ) ¼ 0;  ¼  ny : ð2:4Þ
@n ‘ @n ‘ l
In the case of in-plane problem the equations of equilibrium are given by the following system of partial differential
equations
8 2
< @ u2x þ c2 @ 2 ux @2 u
@x @y2
þ ð1  c2 Þ @x@yy ¼ 0; l
c2 ¼ ; ðc2 < 1Þ; ð2:5Þ
: @ 2 uy þ c 2 @ 2 uy
þ ð1 
2
ux
c2 Þ @@x@y ¼ 0; k þ 2l
@y2 @x2

where l and k are elastic constants, and fux ðx; yÞ; uy ðx; yÞ; 0g denote the components of the displacement vector u
 . The non-
trivial components of the stress tensor are given by the following relations

rxx @ux @uy rxy @ux @uy ryy @uy @ux

¼ þ ð1  2c2 Þ ; ¼ þ ; ¼ þ ð1  2c2 Þ : ð2:6Þ
k þ 2l @x @y l @y @x k þ 2l @y @x

Let us represent the displacement field as the one connected with the outer load plus the perturbed one:

ux ðx; yÞ ¼ u0x ðx; yÞ þ uðx; yÞ; u0x ðx; yÞ ¼ 0;

r0 y ð2:7Þ
uy ðx; yÞ ¼ u0y ðx; yÞ þ v ðx; yÞ; u0y ðx; yÞ ¼ :
k þ 2l

It is assumed in Eq. (2.7) that the outer load in the in-plane problem is again modeled as a uniform normal stress r0 dis-
tributed over a pair of straight lines y ¼  h far from the cracks. Let us write out the normal and the tangential components
of the perturbed stress vector:

T n ¼ rxx n2x þ 2rxy nx ny þ ryy n2y ; T t ¼ rxx nx sx þ rxy ðnx sy þ ny sx Þ þ ryy ny sy ; ð2:8Þ

where n and t designate the unit normal and the tangential vector on a certain elemental area, respectively. Then the bound-
ary condition over the traction-free faces of the cracks can be rewritten in the following form:

T n j‘ ¼ r0 n2y ; T t j‘ ¼ r0 ny t y : ð2:9Þ

3. Strength analysis of the multi-cracked elastic medium

In the forthcoming investigation of the cracked media we follow the standard so-called ‘‘displacement discontinuity
method’’ [24]. Its specific feature is that this does not take explicitly the classical square-root singularity of the stress, which
takes place in a small vicinity of all cracks’ tips in the cracked medium. Instead the singularity is obtained as a result of the
application of an appropriate numerical treatment. In this sense the method under consideration is allied to the standard
BEMs, see for example [25]. Our published work [26] shows that such an approach permits not only a correct calculation
of the stress field at arbitrary point inside the medium but also a precise calculation of the stress intensity factor, which
is the basic quantity in the strength analysis [27].
Let us begin with a single elementary crack of small length e located horizontally in the chosen Cartesian coordinate sys-
tem ðx; yÞ, see Fig. 1. To be more specific, Fig. 1 is related to the in-plane problem, but for the simpler anti-plane case the
geometry is the same, with the only difference that the nontrivial component T z of the load applied to the crack’s faces is
directed orthogonally to plane ðx; yÞ. If the length of the elementary crack is small, then the load may be set constant over
such a short interval ðe=2; e=2Þ.

Fig. 1. An elementary crack in the elastic medium.

46 M.A. Sumbatyan, M. Brigante / Engineering Fracture Mechanics 145 (2015) 43–53

In the anti-plane case the problem for the elementary crack can simply be solved by the classical Fourier transform along
x-axis. By introducing new unknown function gðxÞ, the opening displacement between the crack faces:

gðxÞ; x 2 crack;
wþ ðx; 0Þ  w ðx; 0Þ ¼ ð3:1Þ
0; x R crack;
the stress field at arbitrary point in the medium can be expressed as follows:

syz ðx; yÞ eg y2 þ ðe=2Þ2  x2

¼ ;
l 2p ½y2 þ ðx þ e=2Þ2 ½y2 þ ðx  e=2Þ2 
ð3:2Þ
sxz ðx; yÞ eg xy
¼ :
l p ½y2 þ ðx þ e=2Þ2 ½y2 þ ðx  e=2Þ2 
To investigate the strength properties of the micro-cracked elastic medium, the problem is simulated by a deterministic
multi-crack geometry, see Fig. 2 where the geometry, to be more specific, is again shown for the in-plane case. In the con-
sidered anti-plane problem the outer load T z ¼ s0 is tangential, in contrast to the normal stress T y ¼ r0 shown in the figure.
Let us consider the pair of elemental displacement discontinuities which are defined, respectively, by the data fxi ; yi ; n  i ; t i g
 
and fxj ; yj ; nj ; t j g in the Cartesian coordinate system ðx; yÞ. Here the first two quantities designate coordinates of the central
point, and the unit vectors t; n  are directed along the elementary crack and along the direction of its normal, respectively, see
Fig. 3. Let us study the contribution of j-th element to the tangential stress at i-th element. For this aim we couple with i-th
crack the local Cartesian axes ðx0 ; y0 Þ, and with j-th point the local axes ðx00 ; y00 Þ. It is obvious from Fig. 3 that if the
radius-vector of point i in the coordinate system ðx00 ; y00 Þ is denoted as r 00i ¼ fx00i ; y00i g, then
   
r00i ¼ r j  r i ; x00i ¼ r 00i  t j ¼ r j  r i  tj ; y00i ¼ r 00i  n
 j ¼ rj  ri  n
j : ð3:3Þ
Therefore, in the coordinate system ðx00 ; y00 Þ one obtains, see Eq. (3.2):

s00yz ðxi ; yi Þ e ½y00i 2 þ ðe=2Þ2  x002 i g j

¼ ;
l 2p ½y002
i þ ðx 00
i  e=2Þ 2
½y 002
i þ ðx00i þ e=2Þ2 
ð3:4Þ
s00xz ðx; yÞ e x00i y00i g j
¼ :
l p ½y002 2
i þ ðxi  e=2Þ ½yi þ ðxi þ e=2Þ 
00 002 00 2

Then in the coordinate system ðx00 ; y00 Þ the full stress vector on i-th elemental area is
T i ¼ f0; 0; s00xz n00i x þ s00yz n00i y g; ) siz ¼ s00xz n00i x þ s00yz n00i y ; ð3:5Þ

Fig. 2. A randomly generated system of equal cracks of length 2 mm in a 5 cm  5 cm domain.

 M.A. Sumbatyan, M. Brigante / Engineering Fracture Mechanics 145 (2015) 43–53 47

Fig. 3. The mutual influence of i-th and j-th elementary displacement discontinuities.

where siz is the shear stress on this area and quantities s00xz ; s00yz are defined by expressions (3.4). Further,
n  i  t j ; n
 00i ¼ fn i  n
 j g; )  i  tj ;
n00i x ¼ n n00i y ¼ n
i  n
j : ð3:6Þ

Now, the substitution of (3.6) into (3.5) gives

siz ¼ s00xz ni  tj þ s00yz ni  nj ; ð3:7Þ

The remaining local variables present in Eq. (3.4) can obviously be determined from Eq. (3.3) as follows:
x00i ¼ ðxj  xi Þt j x þ ðyj  yi Þt j y ; y00i ¼ ðxj  xi Þnj x þ ðyj  yi Þnj y : ð3:8Þ

Expressions (3.7) + (3.4) + (3.8) define the influence of the displacement discontinuity g j located at j-th point to the shear
stress siz acting on the elemental area at i-th point, and all quantities are now defined in the basic coordinate system. If
the total number of elements is I, then the complete contribution of all elements to siz is a superposition of elemental expres-
sions taken in the form (3.7). This can be written symbolically as follows:
X
I
siz ¼ K ij g j ; i ¼ 1; . . . ; I: ð3:9Þ
j¼1

with some known quantities K ij . By taking into account boundary condition (2.4), the problem is finally reduced to the fol-
lowing system of linear algebraic equations (LAS):
X
I
K ij g j ¼  s0 ni y ; i ¼ 1; 2; . . . ; I: ð3:10Þ
j¼1

So long as this system is numerically solved, the value of the stress at arbitrary point inside the medium can directly be
calculated from the expression coinciding with Eq. (3.9) and applied at the interior point concerned. This permits calculation
of the stress intensity factor, which is the basic quantity in the strength analysis [27]. It is interesting to note that the pro-
posed algorithm does not give explicitly the square-root singularity of the stress around cracks’ tips, which defines the stress
intensity factor – the key parameter in the present study. However, an appropriate algorithm to calculate the stress intensity
factor has been proposed in [26], just for such a numerical collocation technique. This is discussed in more detail in Section 5.
In the case of in-plane problem, for the same elementary crack of length e, the two unknown functions can be introduced
for the normal and the tangential opening displacement of the crack faces:
ðuþ  u Þðx; 0Þ ¼ g t ðxÞ; ðv þ  v  Þðx; 0Þ ¼ g n ðxÞ; jxj 6 e=2: ð3:11Þ
The solution to this simple in-plane problem for the elementary crack can again be directly obtained by applying Fourier
transform along x-axis, or by some other approaches [24]. In particular, the formulas analogous to anti-plane Eqs. (3.4), have
the following form:
* ( ) ( )+
pr00xx ðxi ;yi Þ t 1 1 y002 ðe=2x00 Þ2 y002 ðe=2þx00 Þ2
 ¼ g j 2y00  y00 
lð1c Þ 2 2
y002 þðe=2x00 Þ y002 þðe=2þx00 Þ 2 2
½y002 þðe=2x00 Þ2  ½y002 þðe=2þx00 Þ2 
2

8 9
*( ) > >+
e=2x 00
e=2þx 00 < e=2x 00
e=2þx 00 =
þg nj þ 2y002 h þ
i2 h i2 > ; ð3:12aÞ
y002 þðe=2x00 Þ2 y002 þðe=2þx00 Þ2 >
: y002 þðe=2x00 Þ2 y002 þ ðe=2þx00 Þ2 ;

*( ) ( )+
pr00xy ðxi ;yi Þ t e=2  x00 e=2 þ x00 e=2  x00 e=2 þ x00
 ¼ gj þ  2y002 þ
lð1  c2 Þ 002 00 2
y þ ðe=2  x Þ y þ ðe=2 þ x Þ
002 00 2 2
½y002 þ ðe=2  x00 Þ  ½y002 þ ðe=2 þ x00 Þ2 
2 2

( )
y002  ðe=2  x00 Þ2 y002  ðe=2 þ x00 Þ2
þ yg nj 2
 2
; ð3:12bÞ
½y002 þ ðe=2  x00 Þ2  ½y002 þ ðe=2 þ x00 Þ2 
48 M.A. Sumbatyan, M. Brigante / Engineering Fracture Mechanics 145 (2015) 43–53

( ) *( )
pr00yy ðxi ; yi Þ 00 t y002  ðe=2  x00 Þ2 y002  ðe=2 þ x00 Þ2 e=2  x00 e=2 þ x00
 ¼ y gj  þ g nj þ
lð1  c Þ 2
½y002 þ ðe=2  x00 Þ2 
2
½y002 þ ðe=2 þ x00 Þ2 
2
y002 þ ðe=2  x00 Þ 2
y002 þ ðe=2 þ x00 Þ2
( )+
e=2  x00 e=2 þ x00
þ2y002 2
þ 2
: ð3:12cÞ
½y002 þ ðe=2  x00 Þ2  ½y002 þ ðe=2 þ x00 Þ2 

Then expressions (2.8) imply

T ni ðxi ; yi Þ ¼ r00xx ðxi ; yi Þ n002 00 00 00 00 002

ix þ 2rxy ðxi ; yi Þ nix niy þ ryy ðxi ; yi Þ niy ;

T ti ðxi ; yi Þ ¼ r00xx ðxi ; yi Þ n00ix t 00ix þ r00xy ðxi ; yi Þ ðn00ix t 00iy þ n00iy t 00ix Þ ð3:13Þ

þr 00 00 00
yy ðxi ; yi Þniy t iy ; n00ix  i  t j ;
¼n n00iy i  n
¼n j ; t00ix ¼ t i  t j ; t 00iy ¼ t i  n
j :

One thus obtains a closed form (3.13) + (3.12) + (3.8) for the influence of the displacement discontinuities ðg nj ; g tj Þ, defined
over a small linear area located at j-th point with coordinates ðxj ; yj Þ, with the unit normal vector n  j ¼ fnjx ; njy g and the unit
tangent vector t j ¼ ft jx ; tjy g, to the normal and tangential stresses, T ni ðxi ; yi Þ; T ti ðxi ; yi Þ acting on the area element at i-th point,
with the unit normal vector n  i ¼ fnix ; niy g and the unit tangent vector t i ¼ ft ix ; tiy g.
If again the total number of elements is I then, due to linearity of the problem, the complete contribution of all small
boundary elements to T ni ðxi ; yi Þ; T ti ðxi ; yi Þ is a superposition of elemental expressions (3.13) that can be written symbolically
as

X
I X
I X
I X
I
T ni ðxi ; yi Þ ¼ K nn n
ij g j þ K nt t
ij g j ; T ti ðxi ; yi Þ ¼ K tn n
ij g j þ K ttij g tj ; ð3:14Þ
j¼1 j¼1 j¼1 j¼1

for any element i ¼ 1; . . . ; I. These relations contain 2I unknown quantities g nj ; g tj ; i ¼ 1; . . . ; I. Obviously, by using boundary
conditions (2.9) applied for i-th crack, i ¼ 1; . . . ; I, one can obtain the closed-form 2I  2I linear algebraic system regarding
these unknown quantities
8
>
> PI nn
XI
>
> j¼1 K ij g nj þ
> K nt
ij g j ¼  r0 niy ;
t 2
i ¼ 1; . . . I;
< j¼1
ð3:15Þ
>
> PI X
I
>
> tn n
K ttij g tj ¼  r0 niy t iy ; i ¼ 1; . . . I:
: j¼1 K ij g j þ
>
j¼1

Analogously to what is discussed for the anti-plane case in the paragraph after Eq. (3.10), if system (3.15) is numerically
solved, the stress tensor at any point and any elemental area inside the medium can directly be calculated from Eq. (3.14).
This also provides correct calculation of the intensity factor at any crack tip in the multi-cracked medium, by using the
approach developed in [26]. The so-calculated intensity factors are used in the energetic strength criterion of Irwin [27],
in the following way. The indicated strength criterion in the in-plane problem is connected with the combination
K 2I þ K 2II , where K I is the intensity factor for the normal stress and K II , the intensity factor for the tangential stress, respec-
tively. In the anti-plane problem above there is only one (tangential) stress intensity factor.

4. Strength analysis of the elastic medium with an array of round defects

In this section we study mechanical properties of the elastic medium with an arbitrary number of relatively small round
voids of the same radius, randomly distributed in a specimen, see Fig. 4.
In the anti-plane case the classical potential theory for Laplace Eq. (2.1) [25], under assumption (2.3), the following rela-
tion is valid:
Z
@U
wðx; yÞ ¼ uz ðn; gÞ d‘; ð4:1Þ
L @n

where ðx; yÞ is taken at arbitrary position out of any void. Here

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ln r
U¼ ; r ¼ jrj; r ¼ fn  x; g  yg; r¼ ðn  xÞ2 þ ðg  yÞ2 ð4:2Þ
2p
S
is Green’s function for the Laplacian. Besides, L ¼ Nm¼1 ‘m is the union of all ‘m ; m ¼ 1; 2; . . . ; N, where ‘m is the boundary of
m-th void, N is the number of voids. In Eq. (4.1) both unit normal vector n  and elemental arc length d‘ are applied at point
ðn; gÞ, not ðx; yÞ.
 M.A. Sumbatyan, M. Brigante / Engineering Fracture Mechanics 145 (2015) 43–53 49

Fig. 4. A randomly generated system of equal round flaws of radius 1 mm in a 5 cm  5 cm domain.

Taking into account that wðX; YÞ ¼ uz ðX; YÞ  uoz ðX; YÞ and the representation for uoz in (2.3), by using well-known proper-
ties of the double-layer potential, with the limit ðx; yÞ ! ðX; YÞ 2 L, Eq. (4.1) leads to the basic Boundary Integral Equation
(BIE) of the considered problem:
Z
gðX; YÞ 1 nn ðn  XÞ þ ng ðg  YÞ s0
þ gðn; gÞ d‘ ¼ Y; ðX; YÞ 2 L: ð4:3Þ
2 2p L ðn  XÞ2 þ ðg  YÞ2 l
This equation is a Fredholm integral equation of the second kind holding over the set of all boundaries of the considered
voids, and its kernel is continuous for all values of ðn; g; X; YÞ if the boundary of the voids is smooth [25], the property valid
for the round defects.
If function gðn; gÞ is defined from Eq. (4.3), the displacement at any point ðx; yÞ in the medium can be defined by using
formula (4.1), in the following form:
Z
s0 1 nn ðn  xÞ þ ng ðg  yÞ
uz ðx; yÞ ¼ y gðn; gÞ d‘: ð4:4Þ
l 2p L ðn  xÞ2 þ ðg  yÞ2
Then the value of the internal stresses at the same point ðx; yÞ is given by the constitutive law:
Z
@uz nn ½ðn  xÞ2 þ ðg  yÞ2   2½nn ðn  xÞ þ ng ðg  yÞðn  xÞ
sxz ¼ l ¼ 2
gðn; gÞd‘;
@x L 2p½ðn  xÞ2 þ ðg  yÞ2 
Z ð4:5Þ
@uz ng ½ðn  xÞ2 þ ðg  yÞ2   2½nn ðn  xÞ þ ng ðg  yÞðg  yÞ
syz ¼ l ¼ s0 þ 2
gðn; gÞd‘:
@y L 2p½ðn  xÞ2 þ ðg  yÞ2 
In the in-plane problem the general idea is the same. We give here only the basic formulas [25]. The following represen-
tations for the components of the perturbed displacement vector are valid, founded on Betti’s integral theorem (principle of
reciprocity of work):
Z Z

ux ðx; yÞ ¼ U xx T x ðn; gÞ þ U yx T y ðn; gÞ dl  Rxx ux ðn; gÞ þ Ryx uy ðn; gÞ dl;

ZL ZL ð4:6Þ

uy ðx; yÞ ¼ U xy T x ðn; gÞ þ U yy T y ðn; gÞ dl  Rxy ux ðn; gÞ þ Ryy uy ðn; gÞ dl;

L L

for any point ðx; yÞ in the considered domain. Here vector T ¼ fT x ðn; gÞ; T y ðn; gÞg designates the outer load applied and dis-
tributed over the boundary contour: ðn; gÞ 2 L.
The fundamental Green’s displacement tensor in (4.6) is
50 M.A. Sumbatyan, M. Brigante / Engineering Fracture Mechanics 145 (2015) 43–53

" #
1 ðn  xÞ2
U xx ðn; g; x; yÞ ¼ ð1  c2 Þ  ð1 þ c 2
Þ ln r ;
4pl r2
1  c2 ðn  xÞðg  yÞ
U yx ðn; g; x; yÞ ¼ U xy ðn; g; x; yÞ ¼ ; ð4:7Þ
4pl r2
" #
1 ðg  yÞ2
U yy ðn; g; x; yÞ ¼ ð1  c2 Þ  ð1 þ c2 Þ ln r :
4pl r2

The components of the fundamental Green’s stress tensor are

   
Rxx @U xx @U yx @U xx @U yx
¼ þ ð1  2c2 Þ nx ðn; gÞ þ c2 þ ny ðn; gÞ;
k þ 2l @n @g @g @n
   
Ryx @U yx @U xx @U yx @U xx
¼ c2 þ nx ðn; gÞ þ þ ð1  2c2 Þ ny ðn; gÞ;
k þ 2l @n @g @g @n
    ð4:8Þ
Rxy @U xy @U yy @U xy @U yy
¼ þ ð1  2c2 Þ nx ðn; gÞ þ c2 þ ny ðn; gÞ;
k þ 2l @n @g @g @n
   
Ryy @U yy @U xy @U yy @U xy
¼ c2 þ nx ðn; gÞ þ þ ð1  2c2 Þ ny ðn; gÞ:
k þ 2l @n @g @g @n
In frames of a standard treatment, with the use of Eqs. (4.6)–(4.8), the boundary conditions (2.9) and expressions (2.8)
result in a system of BIEs over the set L of all circular boundary contours. If this system is solved, stress vector
fT x ðx; yÞ; T y ðx; yÞg at any point inside the medium can simply be calculated by using Eq. (4.6).

5. Results of numerical simulation and comparison with ultrasonic evaluation

In the case of cracked medium a (virtual) specimen contains a large number N of cracks of equal length 2 mm randomly
distributed (by their location and direction) in the rectangular 5 cm  5 cm domain, see Fig. 2. The number of cracks N is
varied from N ¼ 1 up to N ¼ 1000. For each crack M ¼ 20 boundary elements is chosen in the way described above. Then
the total number of elements can be as large as I ¼ NM ¼ 1000  20 ¼ 20; 000, with dimension of respective LAS
20; 000  20; 000. The computational implementation of such calculations requires an appropriate powerful computer
platform.
The algorithm proposed works as follows. At the initial step a single horizontal crack is generated in the central position
of the chosen 5 cm  5 cm domain. The stress intensity energetic coefficient k1 for this single crack is simply calculated as the
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
quantity k1 ¼ K 2I þ K 2II , see the paragraph just after Eq. (3.15), if we speak about the in-plane problem. In the anti-plane
problem it is defined as k1 ¼ jK z j, where K z is the tangential stress intensity factor. After that, step by step, there is sequen-
tially generated one additional crack, so that at N-th step the algorithm generates the crack number N, by using three random
numbers, the first and second ones – for the Cartesian coordinates of a crack’s tip somewhere inside the chosen domain, and
the third one – for the polar angle defining the line segment connecting the tips of the crack. Then the so-constructed crack is
verified for potential intersections with all other cracks already constructed over the previous steps of the algorithm. If there
is any intersection, then such a crack is unsuitable, and new three random numbers are generated again, until the new crack
becomes suitable. Further, for each N-th step of the algorithm there is calculated the maximum intensity coefficient
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kN ¼ max K 2m I þ K 2m II ; ðm ¼ 1; . . . ; NÞ determined as a maximum value among all 2N tips of N cracks already generated.
We thus assume that with the increasing number of cracks located in the domain under consideration the maximum stress
intensity factor kN must increase too (and this is confirmed by our calculations), so that the ratio k1 =kN can be accepted as an
adequate parameter to characterize the strength of the cracked medium. An example of the so-calculated strength versus
number of cracks for the case of in-plane problem is shown in Fig. 5 by black disks, and in a smoothed form – by a certain
curve.
For the problem with a set of round voids several criteria can be accepted to assess the strength of the damaged medium.
We apply the one which seems to be quite reasonable. Let us assume that the critical value of stress, which leads to a crash of
the material, is known as r . Then in the undamaged medium (no voids), where the maximum interior stress is ryy ¼ r0
everywhere inside the medium, ratio ðr =r0 Þ defines the factor in front of applied load resulting in the crash of the material
under a monotonically increasing loading. For the medium with voids loaded by the same outer load r0 , one may assume
that the most dangerous component of the stress tensor is again ryy (see Fig. 4), the quantity which must obviously increase
with the increasing number of voids N. If rN ¼ max ryy inside the medium (out of any void) containing the set of N voids,
then ratio ðrN =r Þ indicates again the factor in front of the outer load which results in the crash, this time – in the damaged
medium. Under such conditions, ratio ðrN =r Þ=ðr =r0 Þ ¼ ðr0 =rN Þ may be accepted as a strength factor, it thus must decrease
with the increasing parameter N.
In the proposed approach the algorithm, predicting the strength with increasing N, works as follows. At the initial step
there is generated an extremely dense set of equal round voids of radius 1 mm in a 5 cm  5 cm domain, in the way
 M.A. Sumbatyan, M. Brigante / Engineering Fracture Mechanics 145 (2015) 43–53 51

Fig. 5. Strength and ultrasonic velocity versus number of cracks, in-pane problem.

analogous to that applied for the cracked medium. The maximum value attained in our numerical experiments is around
N ¼ 440. After that there are randomly chosen 3000 points in the material (i.e. between neighbor voids), so that the distance
between all these points and any void’s center is greater than 1:3 R. Then, by removing randomly a part of voids for less dense
geometries, this can guarantee that each trial point is located relatively far from all voids’ boundaries also for arbitrary value
of N. The so-constructed mesh of 3000 points is used, to calculate the maximum value rN of ryy over the points of this mesh.
The results of such calculations versus number of cracks are shown by black disks in Fig. 6, in the anti-plane case. The respec-
tive curve represents this relation as a smoothed pattern.
Let us note that in the algorithm for the medium with round voids there are used M ¼ 40 boundary elements for each
circular boundary. Then the total number of elements can reach the value I ¼ NM ¼ 440  40 ¼ 17; 600, with the maximal
dimension of respective LAS being 17; 600  17; 600. Applying such an algorithm, by removing a part of voids for less dense
geometries, one can evaluate the relation strength versus ‘‘porosity’’.
As declared in the title, the principal goal of the present work is to establish a certain correlation between the strength
and the variation of the ultrasonic velocity with the increasing number of cracks. For this reason we have performed a
numerical simulation of the through transmission ultrasonic technique, by the method proposed in the authors’ recent
works [13,17], on the basis of the Ray Tracing algorithm. The algorithm implies the modeling of US rays propagating in
the damaged medium. Two ultrasonic transducers, Source (S) and Receiver (R) are placed on the opposite lines y ¼ h above
and below the chosen domain containing defects, compare with Figs. 2 and 4. All US rays are registered in their propagation
from S to R. In the undamaged medium all rays of the US beam radiated from the source S, which can in principle arrive at the
receiver R, have no obstacle over their trajectories. With the increasing number of voids, less number of rays can reach the
receiver directly, without re-refections. For every particular geometry with voids, we choose a uniform grid of nodes over the
base of the probe S, which contains K = 300 nodes, and J = 360 rays are irradiated from each of these nodes in all directions
inside the medium. Then the algorithm follows each ray over its flight, with all its reflections. If any ray passes the outer
boundary of the domain not crossing the base of the receiver R, then this ray is rejected as inappropriate. Only those rays
are accepted as appropriate, which, after all refections, arrive at the base of the receiver R. Note that the total number of rays
is K  J ¼ 300  360 ¼ 108; 000. It is assumed that for each reflection the reflection angle is equal to the incident one. Hence,
the algorithm can operate with specular refections only.
In the anti-plane problem it is well known that the reflection coefficient on the free boundary of an acoustic medium is a
unit value for arbitrary angle of incidence [28]. Therefore, in this case the amplitude of any sound ray arrived at the base of

Fig. 6. Strength and ultrasonic velocity versus number of round voids, anti-plane problem.
52 M.A. Sumbatyan, M. Brigante / Engineering Fracture Mechanics 145 (2015) 43–53

the receiving probe R is pk ¼ 1, k is the sequence number of the registered ray. With increasing number of defects, the ampli-
tude decay is caused by the decreasing number of the arriving rays, due to multiple obstacles over the track of the radiated
rays.
For the in-plane simulation, physical properties of the material should be taken into account, since the reflection coeffi-
cient on the free boundary of the elastic medium depends on physical parameters [28]. In our simulation we accept the
hypothesis that the reflection coefficient for the normal reflection from the free boundary surface is V ¼ 0:95, and this does
not depend upon the angle of incidence. If there are q reflections of a certain ray, over its whole trajectory, then the full
amplitude of this ray registered at the receiver is pk ¼ V q .
In the previous works [13,17] there is introduced a concept of the so-called ‘‘mean time delay’’:
P
tk p
T s ¼ Pk k ; ð5:1Þ
k pk

where t k is the delay of the k-th ultrasonic ray and pk is the respective registered amplitude. The introduced time delay leads
to a certain decrease of the ‘‘seeming’’ wave velocity. If in the undamaged medium the wave velocity is v 0 then in the dam-
aged medium one obtains the following value of the ultrasonic through-transmitted wave speed:

L v0
vN ¼ v0 ¼ ; ð5:2Þ
L þ v0 Ts 1 þ v 0 T s =L

where L is the distance between the transmitter and the receiver (which is 5 cm in the examples presented in Figs. 5 and 6),
and N indicates the number of defects in the configuration under simulation.
The values of ratio v 0 =v N versus number of cracks calculated according to the algorithm described above are represented
in Figs. 5 and 6 for the solid sample containing N cracks and round voids, respectively, by the set of transparent disks, and the
thin lines represent these patterns as certain smoothed curves.
In the discretization of the basic BIEs the precision of the calculations should be controlled by convergence tests with
increasing number of nodes. The test case with N ¼ 75 defects is chosen for this aim, both for crack-like and round flaws
– the cases presented in Figs. 5 and 6, now the number of boundary nodes per every defect varies from M ¼ 10 up to
M ¼ 70. The results of the calculations are presented in Fig. 7. It is clearly seen that in the case of cracked medium the con-
vergence to the true value is monotonic, that is why we accept M ¼ 20 as an appropriate quantity for the line k1 =kN in Fig. 5,
as indicated above. With this assumption the relative error between M ¼ 20 and respective asymptotic value in Fig. 7 is
around 7%. In the case of round voids the convergence is not monotonic, and we choose M ¼ 40, as a more reliable number
of nodes over each circular boundary. For this choice, the precision turns out even less than 5%. In both the cases the pre-
cision of the calculations is comparable with what can be attained in the US experiments. It is interesting to notice from
Fig. 7 that too rough boundary mesh, say M ¼ 10 nodes per crack, may even result in the physically unfeasible value
k1 =kN > 1.

Fig. 7. Strength for N ¼ 75 defects with variation of the number of mesh nodes M: cracked medium – upper line, round voids – lower line.
 M.A. Sumbatyan, M. Brigante / Engineering Fracture Mechanics 145 (2015) 43–53 53

6. Physical conclusions

The principal goal of the theoretical simulation with the proposed numerical algorithms is to give an efficient instrument
for strength analysis of the damaged elastic materials. Obviously, in the engineering applications the number of flaws is not
known a priori. Therefore, a certain correlation is established between strength, outer load and the through-transmission wave
velocity. If such patterns of correlation are constructed correctly, then the qualitative data taken from the ultrasonic inspection
can reliably predict strength properties of the material. The following conclusions can be extracted from the obtained results:

1. Physically, it is quite evident that on the initial stage of the loading the wave velocity weakly depends upon the number of
defects, being approximately constant. In fact, this is because for small number of flaws there exist a lot of ultrasonic rays
which, being emitted from the transmitter, hit receiver’s base like in the case of undamaged medium, hence the time
delay is very small in this case. This implies no change in the through-transmission velocity.
2. This evident observation leads to weak dependence of v N versus N, for small values of parameter N. Perhaps, just this fact
is the real reason of the above discussed criticism in literature [7], and this certainly justifies the point of view that the
value of the wave speed cannot be a measure of strength for the damaged materials. In fact, this is so indeed at the initial
stage of the loading.
3. Thereagainst, the strength of materials is much more sensitive than the wave speed, with respect to the change of number
of defects, even for relatively small N. Figs. 5 and 6 validate numerous experimental observations presented in detail in
literature, namely, the obvious correlation between decreasing strength and decreasing wave velocity, with parameter N
increasing.
4. It is very interesting to see from Figs. 5 and 6 that with increasing N, the diagram for the decreasing wave speed is con-
siderably smoother than the one for the decreasing strength. This is quite natural from the physical point of view.
5. It should be noted that in any damaged solid, under increasing loading, there always arise the defects of both the types –
cracks and voids simultaneously, hence the experiments typically show a decreasing strength and wave velocity some-
where between two opposite cases studied in the present work.

Acknowledgment

The first author is grateful to the Russian Scientific Foundation (RSCF), for its support by Project 15-19-10008.

References

[1] Budynas RG. Advanced strength and applied stress analysis. New York: McGraw-Hill; 1998.
[2] Anderson T. Fracture mechanics: fundamentals and applications. Boca Raton, FL: CRC Press; 1994.
[3] Karihaloo BL. Fracture mechanics and structural concrete. Reading, MA: Addison Wesley; 1995.
[4] Shah SP, Swartz SE, Ouyang C. Fracture mechanics of concrete: applications of fracture mechanics to concrete, rock and other quasi-brittle
materials. New York: Wiley & Sons; 1995.
[5] Krautkramer J. Utrasonic testing of materials. Berlin: Springer-Verlag; 1990.
[6] Malhotra VM, Carino NJ. Handbook on nondestructive testing of concrete. New York: CRC Press; 1991.
[7] Popovics S, Popovics JS. A critique of the ultrasonic pulse velocity method for testing concrete. NDT E Int 1997;30:260.
[8] Ishimaru A. Wave propagation and scattering in random media, vols. 1 and 2. New York: Academic; 1978.
[9] Liu W, Kriz RD. Multiple wave scattering in fiber-reinforced composites: micromechanical point of view. Wave Motion 1998;27:223–44.
[10] Davis PM, Knopoff L. The elastic modulus of media containing strongly interacting antiplane cracks. J Geophys Res 1995;100:18253–8.
[11] Tourin A, Fink M, Derode A. Multiple scattering of sound. Waves Random Media 2000;10:R31–60.
[12] van Baren GB, Mulder WA, Herman GC. Finite-difference modeling of scalar wave propagation in cracked media. Geophysics 2001;66:267–76.
[13] Brigante M, Modano M. Theoretical models and experimental techniques in nondestructive evaluation of concrete. Key Engng Mater 2005;293–
294:207–16.
[14] Chen J-T, Shen W-C, Wu A-C. Null-field integral equations for stress field around circular holes under antiplane shear. Engng Anal Bound Elem
2006;30:205–17.
[15] Chen YZ. Integral equation methods for multiple crack problems and related topics. Appl Mech Rev 2007;60:172–94.
[16] Davis PM, Knopoff L. The elastic modulus, percolation, and disaggregation of strongly interacting, intersecting antiplane cracks. Proc Natl Acad Sci USA
2009;106:12634–9.
[17] Brigante M. On multiple scattering in acoustic media: a deterministic Ray Tracing method for random structures. Ultrasonics 2013;53:652–7.
[18] Scarpetta E, Sumbatyan MA. On the oblique wave penetration in elastic solids with a doubly periodic array of cracks. Quart Appl Math
2000;58:239–50.
[19] Sumbatyan MA, Brigante M. An efficient representation for kernels in the 2d dynamic displacement discontinuity method for cracked elastic materials.
ZAMM 2011;91:516–22.
[20] Brigante M, Sumbatyan MA. Reconstruction of crack clusters in the rectangular domain by ultrasonic waves. Research NDE 2010;21:193–212.
[21] Brigante M, Sumbatyan MA. Application of the inverse diffraction technique to the problem of reconstruction of complexly shaped flaws by ultrasonic
methods. Russ J NDT 2010;46:98–112.
[22] Brigante M, Sumbatyan MA. An efficient numerical algorithm for crack reconstruction in elastic media by the circular US scanning. Inverse Prob Sci Eng
2010;18:361–79.
[23] Slaughter WS. The linearized theory of elasticity. Basel: Birkhäuser; 2002.
[24] Crouch SL, Starfield AM. Boundary element methods in solid mechanics. London: George Allen & Unwin; 1983.
[25] Banerjee PK, Butterfield R. Boundary element methods in engineering sciences. Oxford: McGraw-Hill; 1981.
[26] Ciarletta M, Iovane G, Sumbatyan MA. On stress analysis for cracks in elastic materials with voids. Int J Eng Sci 2003;41:2447–61.
[27] Cherepanov GP. Mechanics of brittle fracture. Oxford: McGraw-Hill; 1979.
[28] Brekhovskikh LM. Waves in layered media. 2nd ed. New York: Academic Press; 1980.