See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/290170320

Transient heat transfer across a multi-layer wall

Article · January 2003

CITATION READS

1 1,349

3 authors:

N. Simões Antonio Tadeu

University of Coimbra University of Coimbra
105 PUBLICATIONS   476 CITATIONS    309 PUBLICATIONS   2,179 CITATIONS   

SEE PROFILE SEE PROFILE

L. Godinho
University of Coimbra
199 PUBLICATIONS   1,061 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

NVTRail: Noise and Vibrations induced by railway traffic in tunnels: an integrated approach View project

PhD work View project

All content following this page was uploaded by N. Simões on 11 October 2017.

The user has requested enhancement of the downloaded file.

 TRANSIENT HEAT TRANSFER ACROSS A MULTI-LAYER WALL

N. Simões, A. Tadeu and L. Godinho

Civil Engineering Department
Faculty of Sciences and Technology, University of Coimbra
Polo II - Pinhal de Marrocos, 3030 Coimbra, Portugal

ABSTRACT

This paper presents analytical expressions for computing the thermal insulation provided by
solid layered walls, when subjected to a plane heat source. The solution is first computed in the
frequency domain for a wide range of frequencies. Time responses are then obtained by means
of (fast) inverse Fourier transforms into space-time. These functions are established using the
required boundary conditions at the various interfaces: continuity of heat fluxes and
temperatures.
A set of applications is used to illustrate the importance of the proposed analytical solutions.
Three different types of wall construction are analysed. Two of the models assume the existence
of insulating material, in different positions, while the third model has no insulating material.

Introduction

In building physics the thermal behaviour of the construction elements is an important issue with
respect to the comfort inside dwellings and the manifestation of pathologies related with vapor
condensation. So transient heat propagation through solid layers is a fundamental problem in
several fields.

The book by Carslaw and Jaeger [1] is a standard reference on heat transfer, since it includes
analytical solutions and Green’s functions for the diffusion equation. The numerical methods are
usually grouped according to the approach used to deal with the time-dependent terms: the “time
marching” approach starts from a specified initial state of the system and evaluates the solution
step by step, at successive time intervals; the Laplace transform modifies the time domain
diffusion equation into an elliptical equation. The former method requires a numerical transform
inversion to calculate the physical variables in the time domain, once the solution has been
obtained for a sequence of values of the transform parameter.

A negative aspect of the “time marching” schemes is that they can produce unstable solutions,
which can be substituted by converting the time in a transform variable. The first boundary
integral representation for studying transient heat conduction, based on the Laplace transform,
 was presented by Rizzo and Shippy [2]. Their numerical approach used a Laplace transform
tocreate a time-independent boundary integration in a transform domain. Different authors have
proposed various solutions for analyzing the diffusion type problem using Laplace transforms
such as those presented by Cheng et al. [3] and Zhu et al. [4]. Most of the techniques reported in
the literature use the Laplace transform to move the solution from the time domain to a transform
domain. However, the inversion of Laplace transforms is associated with loss of accuracy, which
may lead to the amplification of small truncation errors. Several researchers have tried to
overcome this major negative aspect as the algorithm for the Laplace inversion proposed by
Stehfest shows [5].

The present work provides analytical functions for computing the heat emitted by a spatially
sinusoidal, harmonic heat line source placed in solid layered formations. These fundamental
solutions make it possible to compute the heat field variables (fluxes or temperatures) at some
location in the solid domain caused by a heat source placed elsewhere in the media. These
expressions do not need full discretization of the domain, which is required by numerical
techniques such as finite elements and finite differences, or even by discretizing the free surface
using boundary elements techniques.

The fundamental solution for a heat point source is obtained by first applying a time Fourier
transform and then a spatial Fourier transform in the direction in which the geometry does not
vary ( z ). The 3D heat field in the time domain can be then synthesized by using the inverse
Fourier transforms in the frequency and wavenumber ( k z ) domain. If we assume the existence of
virtual heat sources equally spaced, Lz, along z , this inverse Fourier transformation becomes a
discrete summation, which allows the solution to be obtained by solving a limited number of
two-dimensional problems.

The derivation of the Green’s functions for the layered solid formations requires expressing the
Green’s function for the unbounded media as a superposition of heat plane waves. This follows
an approach similar to the one used first by Lamb [6] for the propagation of elastodynamic
waves in two-dimensional media, and Tadeu et al. [7] to compute three-dimensional
elastodynamic fields using a discrete wave number representation.

The final heat field for a layered formation is computed as the sum of the heat source terms equal
to those in the full-space and a set of surface terms, generated within each solid layer and at each
interface. The amplitudes of these surface terms are defined after ascribing specific boundary
conditions: continuity of temperatures and normal fluxes between solid layers.

3D problem formulation

The diffusion equation for transient heat conduction in solids is expressed as:
1 T
 2T  (1)
K t
 2 2 2  k
where  2   2  2  2  , t is time, T (t , x, y, z ) is temperature, K  is the thermal
 x y z  c
diffusivity, k is
the thermal conductivity,  is the density and c is the specific heat. Applying a
Fourier transform in the time domain, one obtains
 2
 2   i 
  ˆ
    K   T ( , x, y, z )  0 (2)
   

where i   1 and  is the frequency. Equation 2 is a Helmholtz equation comparable to the one
applied when solving acoustic problems, where  velocity of pressure waves  corresponds to
 i
in the diffusion equation.
K

The fundamental solution of equation 2 for a heat point source in an unbounded medium,
located at x0 , y 0 , 0 , T  , x, y, z , t    x  x 0   y  y 0  z e i t  where  x  x0  ,   y  y 0  and  z 
are Dirac-delta functions, can be expressed as
i
1   x  x0 2  y  y0 2  z 2
Tˆ f ( , x, y, z )  e K
(3)
2k x  x0 2   y  y0 2  z 2
Owing to the computational cost involved in solving many 3D problems, when the geometry
of the problem remains constant along one direction (z) it is easier to express the full 3D problem
as a summation of simpler 2D solutions. A Fourier transformation along that direction is applied
to the 3D solution and this is expressed as a summation of 2D solutions, with different spatial
wavenumbers k z (Tadeu & Kausel [8]). The application of a spatial Fourier transformation along
the z direction to the equation 2 leads to the following equation

 2
 ~ 2   i  ~
    K  k z 
2 
  T ( , x, y, k z )  0 (4)
   
 

~  2 2 
with  2   2  2  .
 x y 
The solution for the fundamental equation for a heat line source is obtained from equation (3),
applying a spatial Fourier transform to it in the z direction,
~ i   i 
T f ( , x, y, k z )  H0  k z 2 r0  (5)
4k  
 K 

where H n   are Hankel functions of the second kind and order n , and r0  x  x0 2   y  y 0 2 .
This response can be seen as the response to a spatially varying heat line source of the form
T  , x, y, k z , t    x  x0   y  y 0 e i t k z z  .

The full three-dimensional solution is then achieved by applying an inverse Fourier transform
along the k z domain. This inverse Fourier transformation can be expressed as a discrete
summation if we assume the existence of virtual sources, at equal distances apart, L z , along z,
which allows the solution to be obtained by solving a limited number of two-dimensional
problems:
2
M
~
Tˆ ( , x, y, z ) 
L  T ( , x, y, k
m M
zm ) e
ik zm z
(6)
 2
with k zm being the axial wavenumber given by k zm  m. The distance Lz must be large enough
Lz
to avoid spatial contamination from the virtual sources (Bouchon & Aki [9]). The authors have
applied a similar method to study the wave propagation inside seismic prospecting boreholes
(Tadeu et al [10]).

These same equations can be written as a continuous superposition of heat plane waves.
Equation 5, which results from a spatially sinusoidal harmonic heat line source along the z
direction, applied at the point x0 , y0  , is then given by the expressions,

~ i  e  i y  y 0   ik  x  x 
T f ( , x, y, k z ) 
4 k 



 e x 0 dk


x (7)

 i
where    k z  k x with Im   0 , and the integration is performed with respect to the
2 2

K
horizontal wave number ( k x ) along the x direction.

In the above equation, the integral can be transformed into a summation if an infinite number of
such sources are distributed along the x direction, at equal intervals Lx .
n  
~  E
T f ( , x, y, k z )  E 0   
n  
 E d
n 
(8)

i  x  x0   i 2
where E 0  , E  e  i n y  y0
, E d  e  ik xn
, n  2
 k z  k xn
2
with Im n   0 , k xn  n,
2kLx K Lx
which can in turn be approximated by a finite sum of equations ( N ). Note that k z  0
corresponds to the two-dimensional case.

The heat in the spatial-temporal domain is calculated by applying a numerical inverse fast
Fourier transform in k z and in the frequency domain. The aliasing phenomenon is prevented by
using complex frequencies with a small imaginary part of the form  c    i (with   0.7 ,
being  the frequency step). The effect of the complex frequency is later taken into account in
the time domain by rescaling the time response by an exponential factor of the form e t .

Green’s functions in a layered formation delimited by two unbounded media

The system to be analyzed consists of a set of m solid flat layers of infinite extent bounded by
flat, semi-infinite, solid media (top semi-infinite medium (medium 0 ) and bottom semi-infinite
medium (medium m  1 )). The layers can have different thermal material properties and different
thicknesses. This system is excited by a spatially sinusoidal heat source placed in the first layer
(medium 1) (see figure 1). The response is achieved by adding together the direct contribution of
the heat source and the surface heat terms originated at all interfaces.
 Medium 0
Interface 1 X
h1 Medium 1
 Source Interface 2

Interface m
Medium m
Interface m+1
Medium m+1

Fig.1: Geometry of the problem for a multi-layer bounded by two semi-infinite solid media.

For the solid layer j , the heat surface terms on the upper and lower interfaces can be expressed
as
n     n    E 
~  j1 Anjt  E d , T~j 2 ( , x, y, k z )  E 0 j
E
T j1 ( , x, y, k z )  E 0 j    nj
n   


 j 2 Anjb  E d
  nj
n   


 (9)

j 1 j
 i n y  hl  i n y  hl  i
i
 
j j

where E0 j  , E j1  e l 1
, E j2  e l 1
,  nj   k z 2  k xn 2 , with Im  nj  0
2k j L x Kj
kj
and hl is the thickness of the layer l . Meanwhile, K j  is the thermal diffusivity in the
j cj
solid medium, j , ( k j ,  j and c j are the thermal conductivity, the density and the specific heat of
the material in the solid medium j , respectively). The heat surface terms produced at interfaces
1 and m  1 , governing the heat that propagates through the top and bottom semi-infinite media,
are respectively expressed by
n   n    
~  E 01 b  ~ E ( m 1) 2
T02 ( , x, y, k z )  E 00  

n    n 0
An 0  E d ,

T( m 1) 2 ( , x, y, k z )  E 0( m 1)   
n   n ( m 1)
Ant ( m1)  E d


(10)

The final system matrix established takes into account the coupling between the different layers,
so that the heat created simultaneously by the source and surface terms leads to the continuity of
fluxes and temperatures along the m  1 solid interfaces. Thus a system of 2(m  1) equations in
the 2(m  1) unknown coefficients is defined ( F a  b ) for each value of n ,

 1 1 e i n1h1 ... 0 0 0 

 1 1 ei    e  i  n1 y 0 
   ... 0 0 0   e  i  n1 y 0 
 k0 n 0 k1 n1 k1 n1   
 0   b   k1 n1 
e i n1h1 1 ... 0 0 0 
A n0

  At  i n1 h1  y0 
 0 e i n1h1 1   n 1    e 
... 0 0 0
  Anb1   e n1 1 0
 i h  y
 k1 n1 k1 n1 
 ... ... ... ... ... ... ...     (11)
   ...    k1 n1 
 0 0 0 ... 1 ei nm hm 0   Anm t   ... 
 
 1 ei nm hm 
 
 0 
 0 0 0 ...   0  Anm
b
 
 k1 nm km nm   At   0 
 0 0 0 ... e i nm hm1 1 1   n ( m 1)   
   0 
e i nm hm 1 1  
 0 0 0 ...   0 
 km nm km nm km 1 n ( m 1) 

Solving the system gives the amplitude of the surface terms in each solid interface. The Green’s
functions for a solid layer formation are expressed as the sum of the source terms and these
surface terms, leading to the following expressions,
 top semi-infinite medium (medium 0 )
n  
~  E 01 
T ( , x, y, k z )  E 00   
n   n0
Anb0  E d

if y0

solid layer 1 (source position)

n  
~ i  E11 t 
H 0 K t1 r0   E 01 
E
T ( , x, y, k z )   An1  12 Anb1  E d if 0  y  h1
4k1 
n    n1
 n1 

solid layer j ( j  1)
n     j 1 j
~ E E
T ( , x, y, k z )  E 0 j   j1 Anjt  j 2 Anjb  E d if
  nj
n   
 nj 


l 1
hl  y  h
l 1
l

bottom semi-infinite medium (medium m  1 )

n    
~ E ( m 1) 2
T( m 1) 2 ( , x, y, k z )  E 0( m 1)   
n   n ( m 1)
Ant ( m 1)  E d


(12)

The solution for a heat source located in a different solid layer can be calculated maintaining the
same matrix system ( F ) and changing only the independent terms corresponding to the direct
incident field ( b ). The derivation of the total system of equations is thus quite straightforward,
and for this reason it is not presented here.

The results provided by the analytical expressions presented here were compared with those
obtained by applying the Boundary Element Method (BEM) model, which requires the
discretization of all solid interfaces, using the Green’s functions for a full space. The solutions
were found to be in very close agreement with those obtained using the BEM model (not
included here).

Applications

To illustrate the applicability of the proposed analytical solutions, the thermal insulation
provided by solid layered walls when subjected to a heat plane source is computed. Assuming
that the multi-layer wall is bounded by two semi-infinite media (water) allows us to assume that
the convection and radiation phenomena are not involved. Three different models of a 0.32 m
thick wall are considered. The geometry of the models is illustrated in Figure 2. Models 1 and 2
assume the existence of insulating material plus a concrete layer. To allow a better understanding
of the problem’s physics we have included a third model, that does not have insulating material.
Both faces of the wall, for all models, are covered with a 2 cm layer of cement mortar . The
material’s thermal properties are listed in Table I. At time t  1.1 h , a heat source at a point O
creates a plane heat pulse that evolves as plotted in Figure 3, propagating away from O with a
power that increases linearly from 0 W to 1000.0 W .

The computations are performed in the frequency range ( 0 , 128  105 Hz ), with a frequency
increment of 1  10 5 Hz , which determines the total time duration of 27.78h for the analyses in the
time domain.
 0,32 0,32 0,32

Concrete
Polyestyrene
Cement mortar

O O O
x x x
R3 R4 R1 R2 R3 R4 R1 R2 R3 R4
R1 R2
Water Water
medium medium

0,08 0,2 0,02 0,28 0,02

0,2 0,08 0,02

a) b) c)
Fig.2: Geometry of the problem: a) Model 1; b) Model 2; c) Model 3.

1200

1000
Source power (W)

800

600

400

200

0
0 5 10 15 20 25
Time (h)

Fig.3: Temporal evolution of the heat source.

Table I: Material’s thermal properties.

Thermal Conductivity ( k ) Density (  ) Specific heat ( c p )
Concrete 1.4 W .m 1 .º C 1 2300 Kg .m 3 880 J .Kg 1 .º C 1
Cement mortar 0.72 W .m 1 .º C 1 1860 Kg.m 3 780 J .Kg 1 .º C 1
Extruded 0.027 W .m 1 .º C 1 55 Kg .m 3 1210 J .Kg 1 .º C 1
Polystyrene
Water 0.606 W .m 1 .º C 1 998 Kg .m 3 780 J .Kg 1 .º C 1

Figure 4 illustrates the results obtained at receivers R1, R2, R3 and R4, placed in the middle of
each layer of Models 1 and 2. For all plots, the time response begins at null temperature,
corresponding to the initial conditions defined for the present problem. As the source starts
emitting energy ( t  1.1 h ), the temperature increases progressively at the receivers. Receiver R1,
located in the middle of the cement mortar layer closer to the heat source, is the first to record a
distinct change in temperature for all models. The temperature registered at this point increases
smoothly as the energy generated at the source point increases from 0 W to 1000.0 W . After the
source reaches maximum power, the receiver R1 of Model 1 (see Figure 4a) registers a maximum
temperature of 26.5º C , while the same receiver of Model 2 (see Figure 4b) reaches 49.0º C .
Comparing these temperature values with the result obtained for Model 3 (see Figure 4c), it is
clear having the insulating layer closer to the heat source leads to a higher value of temperature
in the cement mortar layer (receiver R1).

Considering the above observations, and comparing the response of receiver R2 of Model 1 with
that computed for Model 3, we can conclude that the presence of an insulating layer results in
higher temperature values on the source side of the insulating material, due to an energy
accumulation phenomenon. Whatever its position, the insulating layer establishes a similar
response for Models 1 and 2 at receiver R4; this receiver clearly registers lower temperatures
than the values obtained for Model 3.
 60 60 60
Rec. 1 Rec. 1 Rec. 1
Rec. 2 Rec. 2 Rec. 2
50 Rec. 3 50 Rec. 3 50 Rec. 3
Rec. 4 Rec. 4 Rec. 4
Temperature (ºc)

Temperature (ºc)

Temperature (ºc)
40 40 40

30 30 30

20 20 20

10 10 10

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

a) b) c)
Fig. 4: Heat curves registered at R1, R2, R3 and R4: a) Model 1; b) Model 2; c) Model 3.

The sequence of curves in Figure 5 displays the temperature distribution at a line of receivers
parallel to the heat flow and perpendicular to the layers, at different times ( t  5 h , t  15 h and
t  25 h ).

Analysing the behaviour of the three models, it is evident that the distinct configuration of the
temperature distribution is due to the presence and position of the insulating material.

60 60 60
t=5 h t=5 h t=5 h
t=15 h t=15 h t=15 h
t=25 h t=25 h t=25 h
50 50 50
Temperature (ºC)
Temperature (ºC)

Temperature (ºC)

40 40 40

30 30 30

20 20 20

10 10 10

0 0 0
0 0.1 0.2 0.3 0 0.1 0.2 0.3 0 0.1 0.2 0.3
y (m) y (m) y (m)

a) b) c)
Fig. 5: Distribution of temperature registered at a line of receivers across the wall for different times
( t  5 h , t  15 h and t  25 h ): a) Model 1; b) Model 2; c) Model 3;

At t  5 h the temperature curves of Models 1 and 3 (see Figure 5a and 5c) are very similar, while
in Model 2 the influence of the insulating layer is clearly visible. As the heat propagates away
from the source the energy continues to spread through the domain, generating a progressive
temperature increase. However, the differences between the models become more evident and
the thermal gradient obtained in the insulating material is clearly marked. The receivers placed
behind the insulating material register lower temperatures than the other regions. This layer
retains the energy for a longer time in regions placed near the source.

As the source power drops to 0 W , the energy continues to propagate through the media and a
fall in temperature values is registered at receivers placed closer to the source. At receivers
located further away, a slight rise of temperature is perceptible, as shown in Figure 5 for t  25 h .

Conclusion

The analytical solutions developed for calculating the heat propagation in layered media, in the
presence of a spatially sinusoidal harmonic heat plane source, seem to be very appropriate for
analysing transient heat conduction. A discrete integration over frequencies has been used to
 compute the heat field generated by a heat point source heating a multi-layer wall in an
unbounded medium. Time signatures were calculated from the frequency responses by applying
inverse (Fast) Fourier transformations, using complex frequencies in order to remove the effects
of the aliasing phenomenon.

The time responses computed for different models of walls confirm that the method is useful and
can be used in many engineering applications, since it allows calculation of the temperature
distribution at different times, and evaluation of the thermal insulation provided by construction
elements.

References

1. H. S. Carslaw, J. C. Jaeger, Conduction of Heat in Solids, second edition, Oxford University

Press (1959)

2. F. J. Rizzo, D. J. Shippy, A Method of Solution for Certain Problems of Transient Heat

Conduction, AIAA Journal, Vol. 8, pp. 2004-2009 (1970)

3. A. H. D. Cheng, Y. Abousleiman, T. Badmus, A Laplace Transform BEM for Axysymmetric

Diffusion Utilizing Pre-tabulated Green’s Function, Engineering Analysis with Boundary
Elements, Vol. 9, pp.39-46 (1992)

4. S. P. Zhu, P. Satravaha, X. Lu, Solving the Linear Diffusion Equations with the Dual
Reciprocity Methods in Laplace Space, Engineering Analysis with Boundary Elements, Vol.
13, pp. 1-10 (1994)

5. H. Stehfest, Algorithm 368: Numerical Inversion of Laplace Transform, Communications of

the Association for Computing Machinery, Vol. 13, no. 1, pp. 47-49 (1970)

6. H. Lamb, On the propagation of tremors at the surface of an elastic solid, Phil. Trans. Roy.
Soc. London, A203:1-42 (1904)

7. A. Tadeu, J. António, 2.5D Green’s Functions for Elastodynamic Problems in Layered

Acoustic and Elastic Formations, Journal of Computer Modeling in Engineering and
Sciences, CMES, Vol. 2, n.º 4, pag. 477-495 (2001)

8. A. Tadeu, E. Kausel, Green´s functions for two-and-a-half dimensional elastodynamic

problems, Journal of Engineering Mechanics – ASCE 126(10), 1093-1097 (2000)

9. M. Bouchon, Aki Keiti, Time-domain transient Elastodynamic Analysis of 3D Solids by

BEM, Int. J. Numer. Methods in Eng., vol. 26, pp. 1709-1728 (1977)

10. A. Tadeu, L. Godinho, P. Santos, Wave motion between two fluid filled boreholes in an
elastic medium, EABE – Engineering Analysis with Boundary Elements 26(2), 101-117
(2002)

View publication stats