Applied Mathematical Modelling 31 (2007) 1444–1459

www.elsevier.com/locate/apm

Conduction and convection phenomena through a slab

with thermal heterogeneities
A. Tadeu *, N. Simões
Department of Civil Engineering, Faculty of Sciences and Technology, University of Coimbra, Polo II,
Pinhal de Marrocos, 3030-290 Coimbra, Portugal

Received 1 June 2005; received in revised form 1 November 2005; accepted 12 April 2006
Available online 30 June 2006

Abstract

This paper addresses the computation of the three-dimensional transient heat transfer through a layered solid and/or
fluid formation containing irregular inclusions. The use of appropriate Green’s functions for a flat layer formation in a
boundary element method formulation avoids the discretization of the layer interface boundaries.
Both conduction and convection are taken into account in the heat diffusion generated by a source, placed somewhere
in the layered system, assuming a known convection velocity field. This work is an extension of earlier work by the authors,
in which only the conduction phenomenon was considered. As before, the calculation process is defined first in the
frequency domain, while the final time series are later computed by applying inverse (fast) Fourier transforms. Once
again, complex frequencies are used in order to avoid aliasing phenomena.
The applicability of the model is illustrated by solving the case of a solid layer submerged in a fluid medium and
containing multiple circular cylindrical thermal heterogeneities. The importance of convection phenomena is studied for
different inclusions’ thermal properties.
 2006 Elsevier Inc. All rights reserved.

Keywords: Transient heat transfer; Conduction; Convection; Boundary element method; 2.5D Green’s functions; Frequency domain;
Fourier transform

1. Introduction

Analytical solutions can only be used to solve simple problems, while the boundary element method (BEM)
can be applied to deal with general problems. The BEM provides an alternative to the well established finite
element method (FEM) for solving physical problems, such as that of heat diffusion, for which BEM codes
have been developed by various authors, including Brebbia et al. [1], Pina and Fernandez [2]. Most of the
known techniques that use BEM to solve transient heat transfer problems utilize ‘‘time marching schemes’’
or Laplace transforms.

*
Corresponding author. Tel.: +351 239797204; fax: +351 239797190.
E-mail address: [email protected] (A. Tadeu).

0307-904X/$ - see front matter  2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.apm.2006.04.010
 A. Tadeu, N. Simões / Applied Mathematical Modelling 31 (2007) 1444–1459 1445

As examples of ‘‘time marching’’ approaches, several works can be cited: Chang et al. [3] and Shaw [4] used
a time-dependent fundamental solution for studying transient heat processes; later, Wrobel and Brebbia [5]
described a formulation for axisymmetric diffusion problems; Carini et al. [6] implemented a symmetric
boundary element method for studying the transient heat conduction over a two-dimensional homogeneous
domain using semi-analytical integrations; also using a time marching boundary element method, Lesnic
et al. [7] solved the unsteady diffusion equation in both one and two dimensions, taking into account the treat-
ment of singularities.
As mentioned above, an alternative numerical scheme, the Laplace transform technique, can be imple-
mented, too, to solve heat transfer problems. The purpose is to remove the time-dependent derivative, using
instead a transform variable. However, this process then requires an inverse transform to find the solution
in the time domain. A Laplace transform boundary element method approach was used, for example, by
Sutradhar et al. [8,9] to solve the three-dimensional transient heat conduction in functionally graded
materials.
Several numerical schemes have been proposed to improve the efficiency of the BEM and its applicability to
more general problems, such as those involving nonlinearities. The dual reciprocity boundary method
(DRBEM) is one of these techniques, and this was originally proposed by Nardini and Brebbia [10]. A number
of works were subsequently published, including those by Satravaha and Zhu [11,12], applying the Laplace
transform dual reciprocity method to the case of transient heat conduction in the presence of nonlinear mate-
rial properties (thermal conductivity, density and specific heat coefficients were all assumed to be functions of
temperature), boundary conditions and sources. Guven and Madenci [13] developed a coupled finite element–
boundary element analysis method for the solution of transient two-dimensional heat conduction of domains
with dissimilar materials and geometric discontinuities. Tanaka and Tanaka [14] studied the heat conduction
in anisotropic non-homogeneous media using a BEM formulation based on the application of fundamental
solutions for a fictitious homogeneous medium. This approach was first used by Butterfield for potential flow
problems [15]. In order to study the transient heat conduction with nonlinear source terms in a domain where
the thermal material properties change in spatial co-ordinates Sladek et al. [16] applied a local boundary inte-
gral equation method.
The BEM requires the discretization of the solid and fluid thermal interfaces and the knowledge of funda-
mental solutions. The definition of suitable Green’s functions can avoid the discretization of some of those
thermal interfaces, leading to a more efficient formulation. This paper computes the three-dimensional tran-
sient heat transfer through a flat-layered formation that contains heterogeneities, using Green’s functions that
avoid the discretization of the flat interfaces.
The technique proposed in this paper is an adaptation and extension of a formulation used by the authors
to solve a similar problem, where only the conduction phenomenon was addressed [17]. In the work described
here, both the conduction and the convection phenomena are taken into account with a pre-prescribed con-
vection velocity.
The technique proposed makes use of time Fourier transforms to allow the calculations to be made in the
frequency domain. This procedure overcomes some of the difficulties posed by the ‘‘time marching’’ and
Laplace transforms approach, which may lead to loss of accuracy and the amplification of small truncation
errors. If a spatial Fourier transform is then applied in the z and x directions the three-dimensional solution
can be obtained as a sum of 2D problems with different spatial wavenumbers kz and kx.
The Green’s functions for a layered medium, without heterogeneities, can be obtained as a superposition of
heat plane sources, as was described originally by Lamb [18] for the propagation of elastodynamic waves in
two-dimensional media. This approach was later adopted by other authors, such as Bouchon [19] and Tadeu
and António [20] to compute three-dimensional elastodynamic fields using a discrete wave number represen-
tation. The solution can be expressed as the sum of the heat source terms equal to those in the unbounded
space and the surface terms. These last terms need to satisfy the boundary conditions at the flat layer inter-
faces: continuity of normal fluxes and temperatures.
This paper first defines the three-dimensional problem and describes how the solution for a point source,
applied in an infinite domain, can be written as a continuous superposition of heat plane terms, in the fre-
quency domain. The time domain solutions, obtained after applying inverse frequency and spatial Fourier
transforms, are compared with analytical solutions [21,22]. Next, the analytical Green’s functions for a layer
1446 A. Tadeu, N. Simões / Applied Mathematical Modelling 31 (2007) 1444–1459

formation bounded by two semi-infinite media are described, implemented and compared with those com-
puted by the BEM model, which incorporates Green’s functions for an unbounded medium and have the dis-
advantage of requiring the discretization of the boundary interfaces. Notice that the extension of the
interfaces’ discretization is limited by introducing damping; otherwise the system of equations involved would
be too large to be solved.
These Green’s functions are then incorporated into a BEM code to compute the heat transfer in the pres-
ence of cylindrical thermal heterogeneities placed in a solid layer bounded by two semi-infinite layers. Different
simulations are performed to illustrate the applicability of the proposed model, and to evaluate the importance
of the convection phenomena and the presence of thermal heterogeneities in heat diffusion across a solid layer
bounded by fluid media.

2. 3D problem formulation

Transient heat transfer by conduction and convection in a homogeneous, isotropic body can be modelled
by
   
o2 o2 o2 1 o o o 1 oT
2
þ 2
þ 2
T  V x þ V y þ V z T ¼ ; ð1Þ
ox oy oz K ox oy oz K ot
in which Vx, Vy and Vz are the pre-prescribed velocity components in directions x, y and z respectively, t is
time, T(t, x, y, z) is temperature, K = k/(qc) is the thermal diffusivity, k is the thermal conductivity, q is the den-
sity and c is the specific heat. A Fourier transformation in the time domain applied to Eq. (1) gives the equa-
tion below, expressed in the frequency domain
0 rffiffiffiffiffiffiffiffiffi!2 1
 2 2 2  
@ o þ o þ o 1 o o o ix A b
 Vx þVy þVz þ T ðx; x; y; zÞ ¼ 0; ð2Þ
ox2 oy 2 oz2 K ox oy oz K
pffiffiffiffiffiffiffi
where i ¼ 1 and x is the frequency. Eq. (2) differs from the Helmholtz equation by the presence of a con-
vective term. For a heat point source, applied at (0, 0, 0) in an unbounded medium, of the form
p(x, x, y, z, t) = d(x)d(y)d(z)ei(xt), where d(y) and d(z) are Dirac-delta functions, the fundamental solution of
Eq. (2) (see [22]) can be expressed as
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
V x xþV y yþV z z
V 2x þV 2y þV 2z pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
e 2K i  ix x2 þy 2 þz2
Tb f ðx; x; y; zÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi e 4K 2 K
: ð3Þ
2
2k x þ y þ z 2 2

To avoid the computational requirements of 3D problem formulation when the geometry of the problem re-
mains constant along the z direction, the full 3D problem can be expressed as a summation of simpler 2D solu-
tions. This requires the application of a Fourier transformation along that direction, writing this as a
summation of 2D solutions with different spatial wavenumbers kz (see [23]). The application of a spatial Fou-
rier transformation to
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
V2 2 2
x þV y þV z ix
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
i  K x2 þy 2 þz2
e 4K 2
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; ð4Þ
x2 þ y 2 þ z2

along the z direction, leads to this fundamental solution

V x xþV y yþV z z
0sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
ie 2K V 2x þ V 2y þ V 2z ix
Te f ðx; x; y; k z Þ ¼ H0 @    ðk z Þ r0 A;
2
ð5Þ
4k 4K 2 K
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where H0( ) are Hankel functions of the second kind and order 0, and r0 ¼ x2 þ y 2 .
This response is related to a spatially varying heat line source of the type pðx; x; y; k z ; tÞ ¼ dðxÞdðyÞeiðxtkz zÞ
(see Fig. 1).
 A. Tadeu, N. Simões / Applied Mathematical Modelling 31 (2007) 1444–1459 1447

(0, 0 , 0 )

x
z
y

Fig. 1. Spatially harmonic varying line load.

The full three-dimensional solution can be synthesized by applying an inverse Fourier transform along the
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 
V 2 þV 2 þV 2 2
kz domain to the expression i2 H0  x 4Ky2 z  ix K
 ðk z Þ r0 ,
0sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
V x xþV y yþV z z Z
e 2K 1
i @ V 2x þ V 2y þ V 2z ix
Te ðx; x; y; zÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi H0    ðk z Þ2 r0 Aeikz z dk z : ð6Þ
2k x2 þ y 2 þ z2 1 2 4K 2 K

Assuming the existence of virtual sources, equally spaced at Lz, along z, then Eq. (6) changes into
0sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
V x xþV y yþV z z Z 1 X
e 2K i @ V 2x þ V 2y þ V 2z ix 1
Te ðx; x; y; zÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi   ðk z Þ r0 Aeikz z
2
H0  2
eikzm Lz dk z : ð7Þ
2
2k x þ y þ z 12 2 2 4K K m¼1

Using the results from distribution theory (e.g. [24]) this equation can be expressed as
0sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
V x xþV y yþV z z
2p e 2K X
1
V 2x þ V 2y þ V 2z ix
Te ðx; x; y; zÞ ¼ H0 @    ðk zm Þ r0 Aeikzm z ;
2
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð8Þ
Lz 2k x2 þ y 2 þ z2 m¼1 4K 2 K

where kzm is the axial wavenumber given by k zm ¼ 2p Lz

m. Eq. (8) can be approximated in turn by a finite discrete
summation, which enables the solution to be obtained by solving a limited number of two-dimensional
problems,
0sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
V x xþV y yþV z z
2p e 2K X
M
V 2x þ V 2y þ V 2z ix
Te ðx; x; y; zÞ ¼ H0 @    ðk zm Þ r0 Aeikzm z :
2
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð9Þ
Lz 2k x2 þ y 2 þ z2 m¼M 4K 2 K

The distance Lz must be large enough to prevent spatial contamination from the virtual sources [25]. An anal-
ogous approach has been used by Tadeu et al. [26] and Godinho et al. [27] to solve problems of wave
propagation.
The fundamental solution of the differential equation obtained from Eq. (2) after the application of a spa-
tial Fourier transformation along the z direction (see Eq. (10)) is Eq. (5), with Vz = 0.
0 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!2 1
 2 2   
@ o o 1 o o ix 2 A Te ðx; x; y; k z Þ ¼ 0:
2
þ 2  Vx þVy þ  ðk z Þ ð10Þ
ox oy K ox oy K

Eq. (5), for a spatially sinusoidal harmonic heat line source applied at the point (0, 0) along the z direction,
subject to convection velocities Vx, Vy and Vz, can be further manipulated and written as a continuous super-
position of heat plane phenomena, as in Garvin [28],
V x xþV y yþV z z Z þ1  imjyj 
ie 2K e
Te f ðx; x; y; k z Þ ¼ eikx ðxÞ dk x ; ð11Þ
4pk 1 m
1448 A. Tadeu, N. Simões / Applied Mathematical Modelling 31 (2007) 1444–1459
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
V 2 þV 2 þV 2 2 2
where m ¼  x 2Ky z  ix K
 ðk z Þ  k x with (Im(m) 6 0), and the integration is related to the horizontal wave
number (kx) along the x direction. The use of the expansion of the Hankel function is described by Morse and
Feshbach [29].
Assuming the existence of an infinite number of virtual sources, these continuous integrals can be trans-
formed into a summation if an infinite number of such sources is distributed along the x direction, spaced
at equal intervals Lx. The above equation can then be written as
 X 
V x xþV y yþV z z
n¼þ1
ie 2K E
Te f ðx; x; y; k z Þ ¼ E0Ed ; ð12Þ
4k n¼1
mn
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
V 2 þV 2 þV 2 2
i
where E0 ¼ 2kL x
, E ¼ e imn jyj
, E d ¼ e ik xn ðxÞ
, m n ¼  x 4Ky2 z  ix K
 ðk z Þ  k 2xn with (Im(mn) 6 0), k xn ¼ 2p
Lx
n,
which can in turn be approximated by a finite sum of equations (N). Note that kz = 0 corresponds to the
two-dimensional case.

2.1. Responses in the time domain

The heat responses in the spatial–temporal domain are obtained by means of an inverse fast Fourier trans-
form in kz and kx and in the frequency domain. In order to prevent the aliasing phenomena, complex frequen-
cies with a small imaginary part of the form xc = x  ig (with g = 0.7Dx, and Dx being the frequency step)
are used in the computation procedure. The constant g cannot be made arbitrarily large, since this leads either
to severe loss of numerical precision, or to underflows and overflows in the evaluation of the exponential win-
dows (see [30]). The time evolution of the heat source amplitude can be diversified. The time Fourier transfor-
mation of the incident heat field defines the frequency domain where the BEM solution needs to be computed.
The response may need to be computed from 0.0 Hz up to very high frequencies. An intrinsic characteristic of
this problem is that the heat responses decay very fast as the frequency increases, which allows us to limit the
upper frequency for the solution. The static response can be computed when the frequency is zero, since the
use of complex frequencies leads to arguments for the Hankel function that are different from zero (xc = ig
for 0.0 Hz).

2.2. Verification of the solution

The formulation described above was verified by computing heat propagation in an unbounded medium
when conduction and convection are considered. The results obtained were compared with the analytical
response in the time domain.
The exact solution of the three, two or one-dimensional convective diffusion, expressed by Eq. (1), in an
unbounded medium subjected to a unit heat source can be found in the literature, see Banerjee [21]. The time
solution at (x, y, z) for a unit heat source placed at (0, 0, 0) at time t = t0 in an unbounded medium is given by
the expression
ðsV x þxÞ2 ðsV y þyÞ2 ðsV z þzÞ2
e 4Ks
T ðt; x; y; zÞ ¼ d=2
; if t > t0 ; ð13Þ
qcð4pKsÞ

where s = t  t0; the parameter d can be 3, 2 or 1 depending on whether we are in the presence of a three, two
or one-dimensional problem, respectively.
The verification example assumes that a homogeneous unbounded medium was excited by a cylindrical
heat source (d = 2) placed at (0, 0, 0). The material’s density (q), specific heat (c), and the thermal conductivity
(k) attributed to the medium were 2300 kg m3, 880.0 J kg1 C1 and 1.4 W m1 C1, respectively. The con-
vection velocity ascribed in the x, y directions was 1 · 106 m/s.
The heat responses were calculated along a line of 40 receivers placed from (x = 1.5, y = 0.35, z = 0) to
(x = 1.5, y = 0.35, z = 0). The calculations, using the formulation described above, were done in the frequency
range [0, 1024 · 107 Hz] with an increment of Dx = 107 Hz, which defines a time window of T = 2777.8 h.
 A. Tadeu, N. Simões / Applied Mathematical Modelling 31 (2007) 1444–1459 1449

0.5
350 h
450 h
550 h
0.4 650 h

Temperature (°C)
0.3

0.2

0.1

0
-1.5 -1.0 -0.5 0 0.5 1.0 1.5
X (m)

Fig. 2. Temperature distribution along a line of 40 receivers, at different times (350 h, 450 h, 550 h and 650 h) for a cylindrical (d = 2) unit
heat source.

The correspondent time domain responses were obtained by applying a numerical inverse fast Fourier
transform in kz and in the frequency domain. These responses are illustrated in Fig. 2 by marks, while the solid
lines indicate the solution computed using Eq. (13). The two solutions show good agreement. It is interesting
to note that the response is not symmetric because of the convection phenomenon in the x direction.

2.3. 2.5D heat diffusion Green’s functions in a flat layer bounded by two unbounded media

The Green’s functions for a system composed of a layer placed between two semi-infinite media (see Fig. 3),
with known vertical convection velocity, can be established by imposing the required boundary conditions at
the interfaces, which are in this case the continuity of temperatures and heat fluxes. So, the heat will flow from
the source position layer through the interfaces to the other media. The conduction and convection phenom-
ena in this propagation process can be modelled. This process assumes that there is an inflowing mass along
the boundary that verifies the mass conservation.
The definition of the solution involves two different kinds of terms: the source terms and the surface terms.
The source terms, which can also be designated as the incident field, are equal to those in the full-space, while
the surface terms need to satisfy the continuity of temperatures and normal fluxes at the interfaces. These two
terms can be expressed in a similar form. For a heat source located at (x0, y0, z0), the terms generated in either
side of the interfaces (see Fig. 3) are expressed as follows:

~ Medium 0
T 02
X
Interface 1

~
T 11
Medium 1
h1
~
T 12
Interface 2
~
T 21
Medium 2

Fig. 3. Geometry of the problem for a flat solid or fluid layer bounded by two semi-infinite solid or fluid media.
1450 A. Tadeu, N. Simões / Applied Mathematical Modelling 31 (2007) 1444–1459

Upper semi-infinite space (interface 1)

V y0 ðyy 0 Þ X 
n¼þ1
E01 b

Te 02 ðx; x; y; k z Þ ¼ E00 e 2K 0
An0 Ed ; ð14Þ
n¼1
mn0
Solid layer (interface 1)
V y1 ðyy 0 Þ X 
n¼þ1
E11 t

Te 11 ðx; x; y; k z Þ ¼ E01 e 2K 1
An1 Ed ; ð15Þ
n¼1
mn1
Solid layer (interface 2)
V y1 ðyy 0 Þ X 
n¼þ1
E12 b

Te 12 ðx; x; y; k z Þ ¼ E01 e 2K 1
An1 Ed ; ð16Þ
n¼1
mn1
Lower semi-infinite space (interface 2)
V y2 ðyy 0 Þ X 
n¼þ1
E21 t

Te 21 ðx; x; y; k z Þ ¼ E02 e 2K 2
An2 Ed ; ð17Þ
n¼1
mn2
where E0j =  i/2kjLx, E00 ¼ eimn0 y , E11 ¼ eimn1 y , E12 ¼ eimn1 jyh1 j , E21 ¼ eimn2 jyh1 j and mnj ¼
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
ðV yj =2K j Þ  ix=K j  k 2z  k 2xn with Im(mnj) 6 0 (j = 1 corresponds to the intermediate layer (medium 1),
while j = 0 and j = 2 indicate the upper and lower semi-infinite media, respectively (medium 0, 2)). Meanwhile,
Kj = kj/qjcj is the thermal diffusivity in the medium j (kj, qj and cj are the thermal conductivity, the density and
the specific heat of the material in the medium, j, respectively) and Vyj is the convection in the y direction in the
fluid medium j.
The unknown coefficients Abn0 , Atn1 , Abn1 and Atn2 are computed by establishing the appropriate boundary con-
ditions, so that the field produced simultaneously by the source and surface terms leads to the continuity of
heat fluxes and temperatures at y = h1 and y = 0. Assuming that the heat source is in the intermediate layer,
the following system results when the boundary conditions are imposed for each value of n:
2h i h i h i 3
V y0 V V
þ imn0 cc1021  2Ky11  imn1  2Ky11 þ imn1 eimn1 h1 0 2 3 2 3
6 2K 0
7 Abn0 b1
6 k c
1 30
1 eimn1 h1 0 76 t 7 6 7
6 k 0 c41 h i h i h 7
i 76 n1 A 7 ¼ 6 b2 7 ;
6 ð18Þ
6 V y1 imn1 h1 V y1 V y2 c22 74 Ab 5 4 b3 5
4 0 2K 1
 im n1 e 2K 1
þ im n1  2K 2
 im n2 c11 5 n1
Atn2 b4
0 eimn1 h1 1  kk12 cc22
11
Pj Pj1
V yj ð h y Þ V yj ð h y Þ
l¼1 l 0 l¼1 l 0
where c1j ¼ m1nj e 2K j
, c2j ¼ m1nj e 2K j
, while b1 ¼ ½V y1 =2K 1 þ imn1 eimn1 y 0 , b2 ¼ eimn1 y 0 , b3 ¼
½V y1 =2K 1  imn1 eimn1 jh1 y 0 j and b4 ¼ eimn1 jh1 y 0 j , if the source is in the intermediate layer (0 < y0 < h1).
The temperature for the three-layer media is computed by adding the contribution of the source terms to
that associated with the surface terms originated at the various interfaces. This leads to the following expres-
sions for the temperatures in the three media, when the source is in the intermediate layer,
X E01
V y0 ðyy 0 Þ n¼þ1

e
T ðx; x; y; k z Þ ¼ E00 e 2K 0 b
A Ed ; if y < 0;
n¼1
mn0 n0
i V y12Kðyy 0 Þ V y1 ðyy 0 Þ

Te ðx; x; y; k z Þ ¼ e 1 H0 ðK t1 r1 Þ þ E01 e 2K 1
4k 1
X E11
n¼þ1
E12 b
 ð19Þ
t
 An1 þ An1 Ed ; if 0 < y < h1 ;
n¼1
mn1 mn1
X E21
V y2 ðyy 0 Þ n¼þ1

e
T ðx; x; y; k z Þ ¼ E02 e 2K 2 t
A Ed ; if y > h1 ;
n¼1
mn2 n2
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
V2
where K t1 ¼  4Ky12  Kix1  ðk z Þ2 and r1 ¼ ðx  x0 Þ2 þ ðy  y 0 Þ2 .
1
 A. Tadeu, N. Simões / Applied Mathematical Modelling 31 (2007) 1444–1459 1451

This derivation assumes that the spatially sinusoidal harmonic heat source is located in the intermediate
layer. However, the equations can be easily manipulated to cater for another position of the source.

3. Boundary element formulation

The fundamental BEM equations are not described in detail here, since they can be found in Wrobel [5].
The boundary integral equations for a homogeneous isotropic medium layer that is embedded by an infinite
medium and contains a cylindrical body (bounded by a surface S), when this system is subjected to an incident
heat field given by Te inc , are expressed as follows:
along the exterior domain
Z
p Te ðextÞ ðx0 ; y 0 ; k z ; xÞ ¼ qðextÞ ðx; y; gn ; k z ; xÞGðextÞ ðx; y; x0 ; y 0 ; k z ; xÞ ds
S
Z
 H ðextÞ ðx; y; gn ; x0 ; y 0 ; k z ; xÞ Te ðextÞ ðx; y; k z ; xÞ ds
S
Z
 GðextÞ ðx; y; x0 ; y 0 ; k z ; xÞ Te ðextÞ ðx; y; k z ; xÞV ðextÞ
n ds þ Te inc ðx0 ; y 0 ; k z ; xÞ; ð20Þ
S

along the interior domain

Z
p Te ðintÞ ðx0 ; y 0 ; k z ; xÞ ¼ qðintÞ ðx; y; gn ; k z ; xÞGðintÞ ðx; y; x0 ; y 0 ; k z ; xÞ ds
S
Z
 H ðintÞ ðx; y; gn ; x0 ; y 0 ; k z ; xÞ Te ðintÞ ðx; y; k z ; xÞ ds
S
Z
 GðintÞ ðx; y; x0 ; y 0 ; k z ; xÞ Te ðintÞ ðx; y; k z ; xÞV ðintÞ
n ds: ð21Þ
S

The boundary integral equations incorporate a convective term, where Vn = Vxnx + Vyny. In these equations,
the superscripts ‘‘int’’ and ‘‘ext’’ refer to the interior and exterior domains respectively, gn is the unit outward
normal along the boundary, G and H are respectively the fundamental solutions (Green’s functions) for the
temperature ( Te Þ and heat flux (q), at (x, y) due to a virtual point heat load at (x0, y0). The factor p is a constant
defined by the shape of the boundary, taking the value 1/2 if the shape is smooth and (x0, y0) 2 S. Note that
this formulation assumes initial conditions of null temperatures and null heat fluxes throughout the domain.
Other initial conditions would require the evaluation of surface or volume integrals.
If the boundary is discretized into N straight boundary elements, with one nodal point in the middle of each
element, Eqs. (20) and (21) take the form:
along the exterior domain
X
N X
N X
N
qðextÞkl GðextÞkl  Te ðextÞkl H ðextÞkl  GðextÞkl Te ðextÞkl V nðextÞ þ Te kincl ¼ pkl Te ðextÞkl ; ð22Þ
l¼1 l¼1 l¼1

along the interior domain

X
N X
N X
N
qðintÞkl GðintÞkl  Te ðintÞkl H ðintÞkl  GðintÞkl Te ðintÞkl VnðintÞ ¼ pkl Te ðintÞkl ; ð23Þ
l¼1 l¼1 l¼1

where

q(ext)kl and Te ðextÞkl are the nodal heat fluxes and temperatures in the exterior domain,
q(int)kl andR Te ðintÞkl are the nodal heat fluxes and temperatures R in the interior domain,
H ðextÞkl ¼ Cl H ðextÞ ðx; xl ; y l ; gl ; xk ; y k ; k z Þ dC l ; H ðintÞkl ¼ Cl H ðintÞ ðx; xl ; y l ; gl ; xk ; y k ; k z ÞdC l ;
R R
GðextÞkl ¼ Cl GðextÞ ðx; xl ; y l ; xk ; y k ; k z Þ dC l ; GðintÞkl ¼ Cl GðintÞ ðx; xl ; y l ; xk ; y k ; k z Þ dC l ;
1452 A. Tadeu, N. Simões / Applied Mathematical Modelling 31 (2007) 1444–1459

where gl is the unit outward normal for the lth boundary segment Cl. In Eqs. (22) and (23), H(ext) (x, xl, yl, gl,
xk, yk, kz) and G(ext) (x, xl, yl, xk, yk, kz) are respectively the Green’s functions for the heat fluxes and tempera-
ture components in the exterior medium of the inclusion, that is, for a flat layer bounded by two unbounded
media. H(int)(x, xl, yl, gl, xk, yk, kz) and G(int) (x, xl, yl, xk, yk, kz) are respectively the Green’s functions for the
heat fluxes and temperature components in the interior medium of the inclusion, that is, for an unbounded
medium, at point (xl, yl), caused by a concentrated heat load acting at the source point (xk, yk). If the loaded
element coincides with the element being integrated, the factor pkl takes the value 1/2.
The two-and-a-half dimensional Green’s functions for temperature and heat fluxes in Cartesian co-ordi-
nates are those for an unbounded medium,

i Vr0
Gðx; y; x0 ; y 0 ; k z ; xÞ ¼ e 2K H0 ðk tr rÞ;
4k      ð24Þ
i Vr0 V or0 or
H ðx; xl ; y l ; gl ; xk ; y k ; k z Þ ¼ e 2K H0 ðk tr rÞ  k tr H 1 ðk tr rÞ ;
4 2K ogl ogl
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
ðxl  xk Þ2 þ ðy l  y k Þ2 ,
2
where V is the assumed radial convection velocity, k tr ¼ V 4K 2 þ ix
K
 ðk z Þ , r ¼
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r0 ¼ x2 þ y 2 is the distance to the convection source position, and Hn( ) are Hankel functions of the second
kind and order n.
If the element to be integrated is not the loaded element, the integrations in Eqs. (22) and (23) are evaluated
using a Gaussian quadrature scheme, while for the loaded element, the existing singular integrands in the
source terms of the Green’s functions are calculated in closed form (see Tadeu et al. [31,32]).
The final system of equations is assembled assuring the continuity of temperatures and heat fluxes along the
boundary of the inclusion. The unknown nodal temperatures and heat fluxes are obtained by solving this sys-
tem of equations, allowing the heat field along the domain to be defined.
The final integral equations are manipulated and combined so as to impose the continuity of temperatures
and heat fluxes along the boundary of the inclusion, and a system of equations is assembled. The solution of
this system of equations gives the nodal temperatures and heat fluxes, which allow the reflected heat field to be
defined.

4. Verification of the Green’s functions

In order to verify the accuracy of the Green’s functions, the results were compared with those arrived by
applying the BEM model using the Green’s solutions for an unbounded medium. BEM applications imply the
discretization of all material interfaces. Notice that the BEM code implemented and used here was first tested
using a simple problem geometry (circular cylindrical geometries), for which analytical solutions are known
(not included here).

Medium 1
Rec. 1
• Source X
Interface 1

Rec. 2 Medium 2
h1 =0.50 m

Interface 2

Rec. 3
Medium 1

Fig. 4. Geometry of the problem used in the verification of the Green’s functions. A 0.5 m thick layer (medium 2), bounded by two semi-
infinite media (media 1).
 A. Tadeu, N. Simões / Applied Mathematical Modelling 31 (2007) 1444–1459 1453

Table 1
Material’s thermal properties used for the verification of the Green’s functions
Medium 1 Medium 2
Thermal conductivity, k [W m1 C1] 0.72 1.4
Density, q [kg m3] 1860.0 2300.0
Specific heat, cp [J kg1 C1] 780.0 880.0

The performance of the analytical Green’s functions is illustrated using a system built by a layer (medium
2), 0.5 m thick, placed between two unbounded media (media 1), as shown in Fig. 4. A harmonic heat line
source is applied in medium 2 at point (x = 0.1 m, y = 0.1 m). The thermal material properties of the
layered formation are listed in Table 1. The convection velocities chosen for medium 1 and 2 were
1 · 107 m/s and 5 · 107 m/s, respectively.
Since complex frequencies are used, the full discretization of the layer interfaces is avoided, as described by
Bouchon and Aki [25] and Phinney [33]. The distance that has been imposed for the discretization is given by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the expression, Ldist ¼ 2 k j =ðqj cj Df Þ. In order to get the largest spatial distance, the thermal material prop-
erties used were those of medium 2, which allow the highest thermal diffusivity.
The computations were performed in the frequency range (0, 32 · 0.5 · 105 Hz), with a frequency incre-
ment of Dx = 0.5 · 105 Hz. Fig. 5 shows the real and imaginary parts of the heat responses at receivers
Rec. 1 (x = 0.1 m, y = 0.15 m), Rec. 2 (x = 0.1 m, y = 0.25 m) and Rec. 3 (x = 0.1 m, y = 0.65 m) when
kz = 0.4 rad/m. The analytical responses are displayed by the solid lines, while the marks correspond to the
BEM solution. The square and round marks correspond to the real and imaginary parts of the responses,
respectively. The BEM model uses 200 constant boundary elements to discretize the layer interfaces.

0.004 0.015

0.002 0.010
Amplitude (°C)

Amplitude (°C)

0 0.005

-0.002 0

-0.004 -0.005

-0.006 -0.010
0 0.00005 0.00010 0.00015 0 0.00005 0.00010 0.00015
(a) Frequency (Hz) (b) Frequency (Hz)

0.0015

0.0010
Amplitude (°C)

0.0005

-0.0005

-0.0010
0 0.00005 0.00010 0.00015
(c) Frequency (Hz)

Fig. 5. Validation for a layer without inclusions bounded by two semi-infinite media. Frequency responses (real and imaginary parts): (a)
Receiver Rec. 1; (b) Receiver Rec. 2; (c) Receiver Rec. 3.
1454 A. Tadeu, N. Simões / Applied Mathematical Modelling 31 (2007) 1444–1459

These two solutions are in very close agreement. Equally good results were obtained from tests in which
receivers and sources were placed at different points.

5. Verification of the BEM model using the proposed Green’s functions

The BEM formulation using the Green’s functions for a layered formation is implemented to compute the
three-dimensional heat field generated by a spatially sinusoidal harmonic heat line source placed in a two-
dimensional horizontal layer, 0.5 m thick, submerged in an unbounded fluid medium. This solid layer contains
a circular cylindrical inclusion with radius 0.1 m and placed parallel to the solid layer interfaces (along the z

Medium 1
Rec. 1
• Source X
Interface 1

Rec. 2
Medium 2
h1 =0.50 m Medium 3
Rec. 3
Interface 2

Rec. 4
Medium 1

Fig. 6. Geometry of the problem for a layer (0.5 m thick) with a cylindrical circular inclusion (0.1 m radius), bounded by two semi-infinite
media. Position of the heat source and of the receivers Rec. 1–4.

0.02 0.0010

0.0005
Amplitude (°C)

0.01
Amplitude (°C)

0
-0.0005

-0.01 -0.0010
0 0.00005 0.00010 0.00015 0 0.00005 0.00010 0.00015
(a) Frequency (Hz) (b) Frequency (Hz)

6x10 -7

1x10 -5
4x10 -7
Amplitude (°C)
Amplitude (°C)

2x10 -7
0

-1x10-5 -2x10-7

0 0.00005 0.00010 0.00015 -4x10-7

0 0.00005 0.00010 0.00015
(c) Frequency (Hz)
(d) Frequency (Hz)

Fig. 7. Verification of the proposed BEM model applied to the case of a solid layer containing an inclusion bounded by two semi-infinite
fluid media. Frequency responses at: (a) Receiver Rec. 1; (b) Receiver Rec. 2; (c) Receiver Rec. 3; (d) Receiver Rec. 4.
 A. Tadeu, N. Simões / Applied Mathematical Modelling 31 (2007) 1444–1459 1455

direction) (see Fig. 6). Incorporating the Green’s functions, derived above, into the BEM formulation, the dis-
cretization is limited to the inclusion surface. The solutions obtained are compared with those given by the
BEM formulation which uses the Green’s functions only for the unbounded media and requires the additional
discretization of the flat interfaces.
The properties and the assumed convection velocities of media 1 and 2 are the same as those used in ver-
ification section above, while the thermal properties and the convection velocity in the inclusion are assumed
to be k3 = 0.12 W m1 C1, c3 = 1380.0 J kg1 C1 and q3 = 510 kg m3 and 2 · 107 m/s, respectively.
This system is heated by a harmonic line source located at (x = 0.1 m, y = 0.1 m). All the calculations
are also performed in the frequency range [0, 32 · 0.5 · 105 Hz] with a frequency increment of Dx = 105 Hz
including a complex component given by g (g = 0.7Dx). Fig. 7 displays the real and imaginary parts of the
responses at receivers 1–4 when kz = 0.4 rad/m. The receivers Rec. 1–4 are placed at (x = 0.1 m,
y = 0.15 m), (x = 0.1 m, y = 0.1 m), (x = 0.1 m, y = 0.35 m) and (x = 0.1 m, y = 0.6 m), respectively.
The solid lines represent responses of the proposed formulation, while the marks correspond to the BEM
solutions using the full boundary discretization. The square and round marks indicate the real and imaginary
parts of the BEM responses, respectively. The standard BEM uses 250 constant boundary elements to model
the layer interfaces and the inclusion boundary, while the proposed model uses only 50 constant boundary
elements to discretize the inclusion. All the plots reveal an excellent agreement between the two solutions
presented.

6. Applications

In order to illustrate the applicability of the formulation presented above, several simulations have been
performed. They use a layered system built with a solid concrete layer (medium 2) containing three inclusions
(medium 3), and buried in two semi-infinite water media (medium 1). Inside the solid layer, 0.3 m thick, three
cylindrical circular inclusions with 0.055 m of radius are modelled, as shown in Fig. 8a. The semi-infinite water
media were assumed to allow both conduction and convection phenomena, while only the conduction phe-
nomenon exists in the solid layer. This process assumes that there is an inflowing mass along the top of the
upper and the bottom of the lower boundaries that verifies the mass conservation.
Two different systems are modelled: in the first (Case 1) the inclusions are made of polystyrene (with ther-
mal properties k = 0.027 W m1 C1, c = 1210 J kg1 C1 and q = 55 kg m3); in the second (Case 2) steel
properties (with thermal properties k = 63.9 W m1 C1, c = 434.0 J kg1 C1 and q = 7832.0 kg m3) were
prescribed for the inclusions. The vertical convection velocities allowed at the top and bottom water medium
were 2 · 106 m/s and 1 · 106 m/s, respectively.
The thermal conductivity, specific heat and density of the concrete layer are k = 1.4 W m1 C1,
c = 880 J kg1 C1 and q = 2300 kg m3, respectively. The thermal properties of the top and bottom

Y 1200
(a) (b)
Medium 1 1000
Vyt = 2× 10−6 m /s
X
Interface 1
Source power (W)

800
Medium 2
Medium 3 Rec. 3
Rec. 2 600
h1 =0.30 m
Rec. 1
400
Interface 2
0.1 m
Medium 1 200
Heat plane source
Vyb = 1× 10 −6m/s
0
0 5 10 15 20 25
Time (h)

Fig. 8. Geometry and source power information of the problem: (a) Solid layer with three circular cylindrical inclusions, bounded by two
unbounded fluid media; (b) Heating curve of the source.
1456 A. Tadeu, N. Simões / Applied Mathematical Modelling 31 (2007) 1444–1459

Case 1 Case 2
0.4 0.1 0.4 0.1
0.25 0.25
0.3 0.3 0.5
0.5
1 1
0.2 1.5 0.2 1.5

y(m)
y(m)

2 2
3 3
0.1 5 4 0.1 5
4
7.5 7.5
0 10 0 10
20
15 20 15
-0.1 -0.1
-0.5 0 0.5 -0.5 0 0.5
x(m) x(m)
(a)
0.4 0.4 1.5
1.5 2
2
3 3
0.3 0.3 4
4
5
5
0.2 0.2

y(m)
y(m)

7.5 7.5
0.1 10 0.1
10
0 0

-0.1 -0.1
-0.5 0 0.5 -0.5 0 0.5
x(m) x(m)
(b)

Fig. 9. Distribution of temperature registered at a grid of receivers for a homogeneous concrete layer with circular inclusions of
polystyrene (Case1) and steel (Case2), when convection velocity is ascribed to the top and bottom media: (a) t = 10 h; (b) t = 20 h.

unbounded media were kept constant and similar to those of the water: k = 0.606 W m1 C1,
c = 4181 J kg1 C1 and q = 998.0 kg m3.
The calculations are first performed in the frequency domain [0, 128 · 0.5 · 105 Hz], with a frequency
increment of 0.5 · 105 Hz. This defines a time window with a total duration of 55.56 h. Null initial temper-
atures and heat fluxes are prescribed along the full domain.
These systems were subjected to a plane heat source placed in the lower medium (y0 = 0.1 m). The energy
emitted by the plane heat source was assumed to have a trapezoidal evolution, as shown in Fig. 8b. The heat
source starts emitting energy at t  0.76 h and its power is increased linearly from 0.0 W to 1000.0 W, reaching
maximum power at t  3.46 h; the source keeps introducing heat into the system, continuously, for a period of
t  2.72 h; the power then falls linearly; it reaches 0.0 W at t  8.89 h. Note that different temporal power
source evolutions could be easily implemented.
Fig. 9 presents contour plots obtained at t = 10 h and t = 20 h, using the temperature amplitudes obtained
over a fine grid of receivers. The convection phenomenon is modelled assuming that its origin coincides with
the position of the heat source. The marked differences observed between the computations performed for
Case 1 and Case 2 scenarios are attributed to the different thermal properties of the inclusions, since the rest
of system remains the same. The temperature curves in the presence of the polystyrene inclusions (Case 1)
reveal higher temperatures at the bottom of the inclusions and lower temperatures behind the inclusion when
compared with the Case 2 (steel inclusions). The energy being accumulated at the source side of the inclusions,
observed in the Case 1, is due to the low thermal diffusivity coefficient of the polystyrene.
For comparison, Fig. 10 displays similar results computed when no convection is allowed in the fluid media.
Comparing the two cases in Figs. 9 and 10 at t = 10 h, it is interesting to note that the heat spreads faster when
positive vertical convection velocities are ascribed to the top and bottom media. Higher temperatures are
therefore registered along the solid media. At t = 20 h, the heating sources are already switched off. However,
as can be seen for both cases analysed, the energy is still propagating across the domain, since the heat equi-
librium is not completed.
To better illustrate the time evolution of the heat diffusion, the results computed at three receivers are
shown in Fig. 11. The receivers Rec. 1, Rec. 2 and Rec. 3 are placed at (x = 0.0 m, y = 0.073 m),
(x = 0.0 m, y = 0.145 m) and (x = 0.0 m, y = 0.216 m), respectively, as displayed in Fig. 8a.
 A. Tadeu, N. Simões / Applied Mathematical Modelling 31 (2007) 1444–1459 1457

Case 1 Case 2
0.4 0.05 0.4
0.05
0.1 0.1
0.3 0.25 0.3 0.25
0.5 0.5
0.2 1 0.2 1
1.5

y(m)
y(m)

1.5 2
2
0.1 3 0.1 3
4 4
5 5
7.5 7.5
0 10 0 10
15 15
20 20
-0.1 -0.1
-0.5 0 0.5 -0.5 0 0.5
x(m) x(m)
(a)
0.4 0.4 0.5
1 1
2 1.5 1.5
0.3 0.3 2
3 3
4 4
0.2 5 0.2 5

y(m)
y(m)

0.1 0.1 7.5

7.5
10
0 10 0

-0.1 -0.1 15
-0.5 0 0.5 -0.5 0 0.5
x(m) (b) x(m)

Fig. 10. Distribution of temperature registered at a grid of receivers for a homogeneous concrete layer with circular inclusions of
polystyrene (Case1) and steel (Case2), considering only the conduction phenomenon: (a) t = 10 h; (b) t = 20 h.

Conduction phenomenon Conduction and convection phenomena

15 15
Rec. 1 Rec. 1
Rec. 2 Rec. 2
Rec. 3 Rec. 3
Temperature (°c)

Temperature (°c)

10 10

5 5

0 0
0 5 10 15 20 25 0 5 10 15 20 25
Time (h) (a) Time (h)

15 15
Rec. 1 Rec. 1
Rec. 2 Rec. 2
Rec. 3 Rec. 3
Temperature (°c)

Temperature (°c)

10 10

5 5

0 0
0 5 10 15 20 25 0 5 10 15 20 25
Time (h) (b) Time (h)

Fig. 11. Temperature evolution at the receivers Rec. 1, 2 and 3: (a) Case 1 (polystyrene inclusions); (b) Case 2 (steel inclusions).
1458 A. Tadeu, N. Simões / Applied Mathematical Modelling 31 (2007) 1444–1459

The receiver Rec.1, which is closer to the position of the heat plane source, is the first to record a progres-
sive change of temperature. The higher temperatures registered at the Rec. 1 when the inclusion is made of
polystyrene (Case 1) confirm its insulating properties. The receiver placed inside the polystyrene inclusion also
records higher temperatures than those registered in the steel inclusion, because the energy does not propagate
easily through this material. Notice also that, the receiver located immediately behind the inclusion (receiver 3)
registers lower temperatures when the inclusion is made of polystyrene (Case 1).
The results illustrated in Fig. 11 show that, in the time window being studied, only the receiver nearest the
heat source records a fall in temperature after the source power has dropped to 0 W. In fact, the temperature is
still increasing at the receivers placed further away from the heat source. The energy equilibrium along the
domain takes a long time to be established, as can be predicted from the results obtained for a time window
of 25.55 h.
The modelling of the positive convection velocity along the y direction introduces marked differences in the
thermal response along the domain. For example the receiver Rec. 1 registers a maximum temperature of
11.4 C when both conduction and convection phenomena are modelled, while the same receiver in the pres-
ence of the conduction alone records a temperature of 9.75 C. The presence of the convection enables the heat
to flow faster, and so the temperature falls sooner.
Once the source has reached the maximum power, the differences between the temperatures registered at the
receivers is seen to be higher in the presence of the extruded polystyrene inclusions (Case 1). The opposite
behaviour (small thermal gradient) is found in Case 2 because the steel allows higher thermal diffusivity.
The temperature variation inside the steel inclusion is very small; as a consequence, the temperature curve
for Rec. 2 and 3 are very close and similar (see Fig. 11b).
Finally, it is important to remember that the heat transfer across a layer formation can be affected to a con-
siderable extent by the presence of thermal heterogeneities and the convection phenomenon. The proposed
formulation can help to model this kind of problem with greater efficiency.

7. Conclusions

The transient heat transfer by conduction and convection across a solid layer with heterogeneities, bounded
by two semi-infinite media has been described.
This paper has illustrated how heat diffusion can be determined in the frequency domain, reformulating the
mathematical and numerical formulations found in other fields such as wave problems. The BEM technique
proposed can handle any type of heat source and allows the modelling of layered media without having to
discretize the solid interfaces. A series of verifications were performed to assure the formulation’s accuracy:
the verification of the frequency domain approach using the Green’s functions for the unbounded medium;
the verification of the Green’s functions for a layered formation; and the verification of the formulation based
on a BEM model that includes the Green’s functions for the layered formation.
The proposed formulation was used to compute the temperature distribution evolution for a solid layer
containing several thermal heterogeneities. Different thermal properties were prescribed for the inclusions
and both the conduction and convection phenomena were modelled. The results obtained demonstrate the
particular importance of the convection phenomenon, and the marked influence of the thermal heterogeneities
on the heat transfer across a layered formation.

References

[1] C.A. Brebbia, J.C.F. Telles, L.C. Wrobel, Boundary Element Techniques: Theory and Applications in Engineering, Springer, Berlin,
1984.
[2] M.L.G. Pina, J.L.M. Fernandez, Applications in heat conduction by BEM, in: C.A. Brebbia (Ed.), Topics in Boundary Element
Research, Springer, Berlin, 1984.
[3] Y.P. Chang, C.S. Kang, D.J. Chen, The use of fundamental Green’s functions for solution of problems of heat conduction in
anisotropic media, Int. J. Heat Mass Transfer 16 (1973) 1905–1918.
[4] R.P. Shaw, An integral equation approach to diffusion, Int. J. Heat Mass Transfer 17 (1974) 693–699.
[5] L.C. Wrobel, C.A. Brebbia, A formulation of the boundary element method for axisymmetric transient heat conduction, Int. J. Heat
Mass Transfer 24 (1981) 843–850.
 A. Tadeu, N. Simões / Applied Mathematical Modelling 31 (2007) 1444–1459 1459

[6] A. Carini, M. Diligenti, A. Salvadori, Implementation of a symmetric boundary element method in transient heat conduction with
semi-analytical integrations, Int. J. Numer. Methods Eng. 46 (1999) 1819–1843.
[7] D. Lesnic, L. Elliot, D.B. Ingham, Treatment of singularities in time-dependent problems using the boundary element method, Eng.
Anal. Bound. Elem. EABE 16 (1995) 65–70.
[8] A. Sutradhar, G.H. Paulino, L.J. Gray, Transient heat conduction in homogeneous and non-homogeneous materials by the Laplace
transform Galerkin boundary element method, Eng. Anal. Bound. Elem. EABE 26 (2) (2002) 119–132.
[9] A. Sutradhar, G.H. Paulino, The simple boundary element method for transient heat conduction in functionally graded materials,
Comput. Methods Appl. Mech. Eng. 193 (2004) 4511–4539.
[10] D. Nardini, C.A. Brebbia, A new approach to free vibration analysis using boundary elements, Boundary Element Methods in
Engineering, Computational Mechanics Publications/Springer-Verlag, Southampton/Berlin, 1982.
[11] S.P. Zhu, P. Satravaha, An efficient computational method for nonlinear transient heat conduction problems, Appl. Math. Model. 20
(1996) 513–522.
[12] P. Satravaha, S. Zhu, An application of the LTDRM to transient diffusion problems with nonlinear material properties and nonlinear
boundary conditions, Appl. Math. Comput. 87 (1997) 127–160.
[13] I. Guven, E. Madenci, Transient heat conduction analysis in a piecewise homogeneous domain by a coupled boundary and finite
element method, Int. J. Numer. Methods. Eng. 56 (2003) 351–380.
[14] M. Tanaka, K. Tanaka, Transient heat conduction problems in inhomogeneous media discretized by means of boundary volume
element, Nucl. Eng. Des. 60 (1980) 381–387.
[15] R. Butterfield, An application of the boundary element method to potential flow problems in generally inhomogeneous bodies, in:
C.A. Brebbia (Ed.), Recent Advances in Boundary Element Method, Pentech Press, London, 1978.
[16] J. Sladek, V. Sladek, Ch. Zhang, A local BIEM for analysis of transient heat conduction with nonlinear source terms in FGMs, Eng.
Anal. Bound. Elem. EABE 28 (2004) 1–11.
[17] A. Tadeu, J. António, L. Godinho, N. Simões, Boundary element method analyses of transient heat conduction in an unbounded
solid layer containing inclusions, Comput. Mech. 34 (2004) 99–110.
[18] H. Lamb, On the propagation of tremors at the surface of an elastic solid, Philos. Trans. Roy. Soc. London A203 (1904) 1–42.
[19] M. Bouchon, Discrete wave number representation of elastic wave fields in three-space dimensions, J. Geophys. Res. 84 (1979) 3609–
3614.
[20] A. Tadeu, J. António, 2.5D Green’s functions for elastodynamic problems in layered acoustic and elastic formations, J. Comput.
Model. Eng. Sci. CMES 2 (4) (2001) 477–495.
[21] P.K. Banerjee, The Boundary Element Methods in Engineering, McGraw-Hill, England, 1981.
[22] H.S. Carslaw, J.C. Jaeger, Conduction of Heat in Solids, second ed., Oxford University Press, 1959.
[23] Gradshteyn, Ryzhik, Tables of integrals, series, and products, 1980.
[24] L. Schwartz, Theorie des distributions, Hermann, Paris, 1966.
[25] M. Bouchon, K. Aki, Time-domain transient elastodynamic analysis of 3D solids by BEM, Int. J. Numer. Methods Eng. 26 (1977)
1709–1728.
[26] A. Tadeu, L. Godinho, P. Santos, Wave motion between two fluid filled boreholes in an elastic medium, Eng. Anal. Bound. Elem.
EABE 26 (2) (2002) 101–117.
[27] L. Godinho, J. António, A. Tadeu, 3D sound scattering by rigid barriers in the vicinity of tall buildings, J. Appl. Acoust. 62 (11)
(2001) 1229–1248.
[28] W. Garvin, Exact transient solution of the buried line source problem, Proc. Roy. Soc. London, Ser. A – Math. Phys. Sci. 234 (1199)
(1955) 528–541.
[29] P.M. Morse, Feshbach Methods of Theoretical Physics, McGraw-Hill, New York, 1953, p. 823.
[30] E. Kausel, J.M. Roësset, Frequency domain analysis of undamped systems, J. Eng. Mech. ASCE 118 (4) (1992) 721–734.
[31] A. Tadeu, P. Santos, E. Kausel, Closed-form integration of singular terms for constant, linear and quadratic boundary elements –
Part I: SH wave propagation, Eng. Anal. Bound. Elem. EABE 23 (8) (1999) 671–681.
[32] A. Tadeu, P. Santos, E. Kausel, Closed-form integration of singular terms for constant, linear and quadratic boundary elements –
Part II: SV-P wave propagation, Eng. Anal. Bound. Elem. EABE 23 (9) (1999) 757–768.
[33] R.A. Phinney, Theoretical calculation of the spectrum of first arrivals in the layered elastic medium, J. Geophys. Res. 70 (1965) 5107–
5123.