Numerical Heat Transfer, Part B: Fundamentals

ISSN: 1040-7790 (Print) 1521-0626 (Online) Journal homepage: https://www.tandfonline.com/loi/unhb20

A Distributed Transfer Function Method for Heat

Conduction Problems in Multilayer Composites

Bingen Yang

To cite this article: Bingen Yang (2008) A Distributed Transfer Function Method for Heat
Conduction Problems in Multilayer Composites, Numerical Heat Transfer, Part B: Fundamentals,
54:4, 314-337, DOI: 10.1080/10407790802359038

To link to this article: https://doi.org/10.1080/10407790802359038

Published online: 11 Sep 2008.

Submit your article to this journal

Article views: 139

Citing articles: 9 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=unhb20
Numerical Heat Transfer, Part B, 54: 314–337, 2008
Copyright # Taylor & Francis Group, LLC
ISSN: 1040-7790 print=1521-0626 online
DOI: 10.1080/10407790802359038

A DISTRIBUTED TRANSFER FUNCTION METHOD

FOR HEAT CONDUCTION PROBLEMS IN
MULTILAYER COMPOSITES

Bingen Yang
Department of Aerospace and Mechanical Engineering, University of Southern
California, Los Angeles, California, USA

A new analytical solution method is developed for problems of conduction heat transfer
in one-dimensional multilayer composite bodies. In this method, the original governing
equations are first cast in a spatial state form in the Laplace transform domain; the solution
of the state equation is then obtained in terms of distributed transfer functions. With a new
residue formula, inverse Laplace transform of the s-domain solution yields exact transient
temperature and heat flux. Exact steady-state solutions are also obtained with the transfer
function formulation. The proposed method is numerically efficient as it only requires
simple operations of two-by-two matrices.

1. INTRODUCTION
Heat conduction in multilayer composite solids has a variety of applications in
design and development of buildings, industrial furnaces, nuclear reactors, turbines,
rockets, space craft, and the like [1–3]. Recent advancements in new composite
materials, high-tech devices and instruments, and industrial processes have further
enhanced the desire and effort to better understand the behaviors of temperature
distributions and heat flows in composite solids [4–16]. While numerical methods
are widely used to solve this type of heat transfer problems, analytical methods
are still in demand. Compared with numerical methods, analytical methods provide
more information and deeper physical insight, offer more accurate solutions, and are
numerically more efficient. In addition, analytical solutions can be applied to verify
numerical algorithms for complicated engineering problems. The current investi-
gation is concerned with analytical methods for conduction heat transfer in one-
dimensional multilayer slabs, with special attention given to closed-form transient
solutions that are expressed in terms of explicit functions.
Several analytical methods are available for heat conduction problems in
composite solids, among which are the orthogonal expansion technique [17, 18],

Received 30 April 2008; accepted 11 July 2008.

This work was a result of the author’s previous projects partially sponsored by the U.S. Army
Research Office and NASA’s Jet Propulsion Laboratory. The author would like to thank Professor
S. S. Sadhal for discussions on the subject.
Address correspondence to Bingen Yang, Department of Aerospace and Mechanical Engineering,
University of Southern California, 3650 McClintock Avenue, Room 430, Los Angeles, CA 90089-1453,
USA. E-mail: [email protected]

314
 TRANSFER FUNCTION METHOD FOR HEAT CONDUCTION 315

Green’s function method [19, 20], Laplace transform method [1, 21], and the method
of separation of variables [3, 22]. These techniques are well documented in a few clas-
sic textbooks [1–3] and have been used by many researchers. In theory, existing ana-
lytical techniques should give exact solutions for any one-dimensional composite
body. In practice, if the number of layers is just more than a few the utility of the
techniques becomes difficult. In the literature, results on exact transient solutions
have been limited to slabs with two or three layers [13, 23]. Some steady-state solu-
tions are also feasible for regularly arranged two-material laminates [20, 24, 25].
There are several issues that restrict the application of the analytical techniques
to multilayer composite solids. First, the orthogonal expansion technique requires
establishment of orthogonality and normalization conditions for system eigenfunc-
tions, and evaluation of related spatial integrals. This involves tedious derivations even
for slabs with several layers, and often yields expressions too complicated to handle in
computation. The Green’s function method and the method of separation of variables,
which also use system eigenfunctions to construction solutions, bear similar problems.
Second, the Laplace transform method has to deal with s-domain (Laplace transform
domain) solutions that consist of complicated expressions, which makes inverse
Laplace transform impractical, if not impossible. In the literature, exact transient solu-
tions by Laplace transform are only available for two-layer slabs [1]. For slabs of more
layers, approximations are usually made in a solution process, as in computation of
residues [26] and in evaluation of Laplace transform integrals [27]. Third, conventional
analytical techniques assign two unknown parameters for each layer of a composite
solid, which could be challenging in both derivation and computation if the number
of layers is very large. This is especially true when transient solutions are determined.
On top of the above-mentioned issues, solution by most analytical techniques is
problem-dependent, requesting different derivations and solution algorithms for dif-
ferent slab configurations (number of layers, boundary conditions, layer interfaces,
etc.). This inflexibility in treating composite slabs makes analytical techniques less
user-friendly, compared to numerical methods.
In this article, a new analytical solution method is proposed for the problems
of conduction heat transfer in one-dimensional multilayer composite bodies. The
method, called the distributed transfer function method (DTFM), is capable of
obtaining exact solutions for a composite slab with any number of layers and at
the same time avoids the aforementioned issues of existing analytical techniques.
Distributed transfer functions have been developed for vibration analysis and
feedback control of elastic continua [28–31]. While distributed transfer functions
are useful in determination of eigensolutions and steady-state solutions, they have
not been applied in transient response analysis in terms of inverse Laplace transform.
In this work, distributed transfer functions are introduced to the heat conduction
problems of composite slabs for the first time. In the development, the heat equa-
tions of a composite slab are cast in a spatial space equation, from which the s-
domain solution is determined by a distributed transfer function formulation. Based
on this formulation, a new residue formula is devised for inverse Laplace transform
of the s-domain solution, which yields exact temperature and heat flux of the slab.
Eigensolutions and steady-state responses are also given in exact and closed form.
The DTFM-based analysis only involves simple manipulations of two-by-two
matrices, and thus is highly efficient in computation.
316 B. YANG

This article is arranged as follows. The governing equations for a one-

dimensional composite slab are presented in Section 2. A distributed transfer func-
tion formulation is given in Section 3. With the transfer function formulation, exact
eigensolutions and steady-state responses are determined in Sections 4 and 5,
respectively. In Section 6, a formula for precise evaluation of transfer function resi-
dues is derived, which eventually leads to the exact transient solutions of the slab
subject to arbitrary heat generation, boundary excitations, and initial temperature.
The solution method is extended to composite bodies with thermal resistance at layer
interfaces in Section 7. The proposed DTFM is illustrated in two numerical examples
in Section 8.

2. STATEMENT OF PROBLEM
A one-dimensional n-layer composite slab is shown in Figure 1, where lj is the
thickness of the jth layer, x0 and xn mark the left and right boundaries of the body,
and xj is the spatial coordinate of the interface between the jth and ( j þ 1)th layers.
Here x0 ¼ 0 and xj ¼ l1 þ l2 þ


þ lj for j ¼ 1; 2; . . . ; n. The heat conduction in the
slab is governed by

q2 Tj ðx; tÞ gj ðx; tÞ 1 qTj ðx; tÞ

þ ¼ x 2 ½xj1 ; xj  j ¼ 1; 2; . . . ; n ð1Þ
qx2 kj aj qt

where Tj ðx; tÞ, kj and aj are the temperature, thermal conductivity, and thermal
diffusivity of the jth layer, respectively, and gj ðx; tÞ describes heat generation within
the layer. The boundary conditions of the slab are of the general form
qT1
a1 þ a0 T1 ¼ cL ðtÞ at x ¼ x0 ð2aÞ
qx

qTn
b1 þ b0 Tn ¼ cR ðtÞ at x ¼ xn ð2bÞ
qx

Figure 1. A one-dimensional n-layer composite slab.

 TRANSFER FUNCTION METHOD FOR HEAT CONDUCTION 317

where a0 , a1 , b0 , and b1 are constants that are properly assigned to characterize dif-
ferent types of boundaries, and cL ðtÞ and cR ðtÞ are related to prescribed boundary
temperature and=or heat flux. Under the assumption of perfect thermal contact,
the matching conditions at the layer interfaces are
Tj ðxj ; tÞ ¼ Tjþ1 ðxj ; tÞ ð3aÞ

qTj ðxj ; tÞ qTjþ1 ðxj ; tÞ

kj ¼ kjþ1 ð3bÞ
qx qx

for j ¼ 1; 2; . . . ; n  1. (Bodies with thermal resistance at layer interfaces will be dealt

with in Section 7.) In addition, the initial condition is
Tj ðx; 0Þ ¼ hj ðxÞ x 2 ½xj1 ; xj  j ¼ 1; 2; . . . ; n ð4Þ

with hj ðxÞ being a given temperature distribution in the jth layer.

The objective of this work is to solve the boundary-initial value problem
formed by Eqs. (1)–(4). With the proposed distributed transfer function method,
exact closed-form temperature and heat flux will be obtained for a composite slab
with any number of layers and subject to arbitrary energy sources gj ðx; tÞ, boundary
excitations cL ðtÞ and cR ðtÞ, and initial temperature distribution hj ðxÞ.

3. TRANSFER FUNCTION FORMULATION

In this section, the s-domain solution of the composite slab is derived through
establishment of a spatial state representation and a distributed transfer function for-
mulation. For basic concepts of distributed transfer functions, refer to refs. [28–31].

3.1. Spatial State Representation

Laplace transform of Eqs. (1)–(3) with respect to time gives

q2 T j ðx; sÞ s gj ðx; sÞ hj ðxÞ

2
¼ T j ðx; sÞ   x 2 ½xj1 ; xj  ð5Þ
qx aj kj aj

qT 1 ðx0 ; sÞ
a1 þ a0 T 1 ðx0 ; sÞ ¼ cL ðsÞ
qx ð6Þ
qT 1 ðxn ; sÞ
b1 þ b0 T n ðxn ; sÞ ¼ cR ðsÞ
qx

T j ðxj ; sÞ ¼ T jþ1 ðxj ; sÞ

qT j ðxj ; sÞ qT jþ1 ðxj ; sÞ ð7Þ
 kj ¼ kjþ1
qx qx

where the over-bar stands for Laplace transform, s is the complex Laplace transform
parameter, and the initial condition (4) has been used. For the jth layer, defining the
318 B. YANG

spatial state vector

!
T j ðx; sÞ
f^
gj ðx; sÞg ¼
j ðx; sÞ
qT ð8Þ
kj qx

converts Eq. (5) to the first-order state equation

q
f^
g ðx; sÞg ¼ ½Fj ðsÞf^
gj ðx; sÞg þ f^pj ðx; sÞg x 2 ½xj1 ; xj  ð9Þ
qx j

where

   
0 1=kj kj 0
½Fj ðsÞ ¼ pj ðx; sÞg ¼ 
gj ðx; sÞ þ hj ðxÞ
f^ ð10Þ
kj s=aj 0 aj 1

These state vectors should satisfy the matching conditions (7); namely,
gj ðxj ; sÞg ¼ f^
f^ gjþ1 ðxj ; sÞg j ¼ 1; 2; . . . ; n  1 ð11Þ

For convenience of analysis, define the following global quantities.

f^ gj ðx; sÞg
gðx; sÞg ¼ f^ ½F ðx; sÞ ¼ ½Fj ðsÞ x 2 ðxj1 ; xj Þ
ð12Þ
f^
pðx; sÞg ¼ f^
pj ðx; sÞg x 2 ðxj1 ; xj Þ

for j ¼ 1; 2; . . . ; n. Also, to satisfy the matching conditions (11), f^

gðx; sÞg must be
continuous at the interfaces xj ; j ¼ 1; 2; . . . ; n  1. With these global quantities,
Eq. (9) is converted to a global state equation over the entire region of the slab
q
f^
gðx; sÞg ¼ ½F ðx; sÞf^
gðx; sÞg þ f^pðx; sÞg x 2 ½x0 ; xn  ð13aÞ
qx

Likewise, the boundary conditions (6) can be cast in the matrix form
½Mb f^
gðx0 ; sÞg þ ½Nb f^
gðxn ; sÞg ¼ f
cb ðsÞg ð13bÞ

where


  
a a1 =k1 0 0
cL ðsÞ
½Mb  ¼ 0 ½Nb  ¼ f
cb ðsÞg ¼ ð14Þ
0 0 b0 b1 =kn
cR ðsÞ

Thus, the original boundary-initial value problem is reduced to the global state
Eq. (13a) subject to the boundary condition (13b).

3.2. Distributed Transfer Functions

The composite slab is a nonuniformly distributed parameter system because
½F ðx; sÞ is a piecewise step function in region ½x0 ; xn . In the proposed method,
Eq. (13) is solved through use of the state transition matrix ½Uðx; n; sÞ of the slab,
 TRANSFER FUNCTION METHOD FOR HEAT CONDUCTION 319

which is the unique solution of [32].

q
½Uðx; n; sÞ ¼ ½F ðx; sÞ½Uðx; n; sÞ x; n 2 ½x0 ; xn  ð15Þ
qx
subject to the condition ½Uðx; x; sÞ ¼ ½I, with ½I being the two-by-two identity
matrix. The state transition matrix has the properties
½Uðx; n; sÞ1 ¼ ½Uðn; x; sÞ
ð16Þ
½Uðx; z; sÞ ¼ ½Uðx; y; sÞ½Uðy; z; sÞ
The state transition matrix can be expressed by
½Uðx; n; sÞ ¼ ½Uðx; sÞ½Uðn; sÞ1 ð17Þ
where ½Uðx; sÞ is any fundamental matrix that is a nonsingular solution of
q
½Uðx; sÞ ¼ ½F ðx; sÞ½Uðx; sÞ x 2 ½x0 ; xn  ð18Þ
qx
It is easy to show that a fundamental matrix for the heat conduction problem is
(
e½F1 ðsÞðxx0 Þ x 2 ½x0 ; x1 
½Uðx; sÞ ¼ ð19Þ
e½Fj ðsÞðxxj1 Þ e½Fj1 ðsÞlj1 . . . e½F1 ðsÞl1 x 2 ½xj1 ; xj  j ¼ 2; 3; . . . ; n
where the exponential matrices are given by
" pffiffiffiffiffiffiffiffiffi  pffiffiffiffiffiffiffiffiffi  #
cosh s=aj x p1ffiffiffiffiffiffi sinh s=aj x
e ½ Fj ðsÞx
¼ pffiffiffiffiffiffiffiffiffi kj s=aj ð20Þ
pffiffiffiffiffiffiffiffiffi  pffiffiffiffiffiffiffiffiffi 
kj s=aj sinh s=aj x cosh s=aj x
Substitute Eq. (19) into Eq. (17) to obtain

½Uðx; n; sÞ ¼ e½Fj ðsÞðxxj1 Þ ½Uðxj1 ; sÞ½Uðxk1 ; sÞ1 e½Fk ðsÞðnxk1 Þ ð21Þ
where x 2 ½xj1 ; xj  and n 2 ½xk1 ; xk  for 1  j; k  n. Because ½Uðx0 ; sÞ ¼ ½I,
½Uðx; x0 ; sÞ ¼ ½Uðx; sÞ ð22Þ
Once the transition matrix is computed, the unique solution to Eq. (13) can be
taken as
Z xn
f^
gðx; sÞg ¼ b ðx; n; sÞf^
½G b ðx; sÞf
cb ðsÞg x 2 ½x0 ; xn 
pðn; sÞgdn þ ½H ð23Þ
x0

where
(
b ðx; sÞ½Mb ½Uðx0 ; n; sÞ n < x
½H
b ðx; n; sÞ ¼
½G
b ðx; sÞ½Nb ½Uðxn ; n; sÞ n > x
½H ð24aÞ
b ðx; sÞ ¼ ½Uðx; x0 ; sÞ½ZðsÞ
½H 1

with
½ZðsÞ ¼ ½Mb  þ ½Nb ½Uðxn ; x0 ; sÞ ð24bÞ
320 B. YANG

See the Appendix for the proof of Eqs. (23) and (24). Here matrices ½G b  and ½H
b  are
the distributed transfer functions of the composite slab. Because the fundamental
matrix ½Uðx; sÞ is continuous at all layer interfaces, the solution given in Eqs. (23)
and (24) automatically satisfies the matching conditions (11).
The distributed transfer functions in Eq. (23) are the s-domain Green’s
functions of the boundary-initial value problem, Eqs. (1)–(4). The spatial state
representation and distributed transfer function formulation derived in this sec-
tion lay a foundation for development of exact and closed-form heat conduction
solutions in both transient and steady states, as shall be seen in the subsequent
sections.

4. EIGENSOLUTIONS
In the DTFM, system eigenvalues are needed for transient solutions. Vanishing
f^
pg and f
cb g in Eq. (13) leads to the eigenvalue problem defined by
q
fwðxÞg ¼ ½F ðx; sÞfwðxÞg x 2 ½x0 ; xn  ð25aÞ
qx
subject to
½Mb fwðx0 Þg þ ½Nb fwðxn Þg ¼ 0 ð25bÞ
where s is an eigenvalue and fwðxÞg is the eigenfunction associated with s. Assume
that
fwðxÞg ¼ ½Uðx; x0 ; sÞfag ð26Þ
with fag being a constant vector to be determined. The fwðxÞg, which satisfies
Eq. (25a), is plugged in the boundary condition (25b), yielding
ð½Mb  þ ½Nb ½Uðxn ; x0 ; sÞÞfag ¼ 0 ð27Þ
This gives a transcendental characteristic equation
detð½Mb  þ ½Nb ½Uðxn ; x0 ; sÞÞ ¼ 0 ð28Þ
whose roots, sk ; k ¼ 1; 2; . . . , are the eigenvalues of the composite slab. The sk are
also the poles of the transfer functions ½G b  and ½Hb  because det½Zðsk Þ ¼ 0 by
Eq. (24b).
By the nature of heat conduction, the eigenvalues can be written as
sk ¼ b2k k ¼ 1; 2; . . . ð29Þ
where bk are non-negative roots of
 2 2 2 l1
DðbÞ  det ½Mb  þ ½Nb e½Fn ðb Þln . . . e½F2 ðb Þl2 e½F1 ðb Þ ¼0 ð30Þ

with the exponential matrices given by

2  pffiffiffi 3
pffiffiffiffi aj  pffiffiffiffi
2 cos bx= aj kj b sin bx= a j
e ½Fj ðb Þx
¼ 4  pffiffiffiffi  pffiffiffiffi 5 ð31Þ
kb
 pjffiffiffi
aj sin bx= aj cos bx= aj
 TRANSFER FUNCTION METHOD FOR HEAT CONDUCTION 321

It is easy to see that Dð0Þ is finite. Also, it can be verified that


2 2 2 U11 ðbÞ b1 U12 ðbÞ
e½Fn ðb Þln


e½F2 ðb Þl2 e½F1 ðb Þl1 ¼ ð32Þ
bU21 ðbÞ U22 ðbÞ
where


U11 U12
¼ ½Dn 


½D2 ½D1  ð33Þ
U21 U22
with
2  pffiffiffi 3
pffiffiffiffi aj  pffiffiffiffi
cos blj = aj kj sin blj = aj 5
½Dj  ¼ 4 kj  pffiffiffiffi  pffiffiffiffi ð34Þ
 pffiffiffi
aj sin blj = aj cos blj = aj

Note that Uij ðbÞ only contain sine and cosine functions. Substitute Eq. (32) into
Eq. (30) to obtain
a0 b1 a1 b0 a1 b1 1
DðbÞ ¼ U22 ðbÞ  U11 ðbÞ  b U21 ðbÞ þ a0 b0 U12 ðbÞ b0 ð35Þ
kn k1 k1 kn b
where Eq. (14) has been used. Equation (35) indicates that DðbÞ is smooth and well-
behaved in any given finite region. The characteristic function can be scaled, such as
DðbÞ=ð1 þ bÞ, to render it bounded in the entire region ½0; 1Þ. Hence, standard
root-searching techniques are applicable to Eq. (30) for accurate solution of the
eigenvalues.
The eigenfunction associated with b2k , although not needed in the DTFM-
based analysis, can be obtained by
fwk ðxÞg ¼ Uðx; x0 ; b2k Þ fak g ð36Þ
where fak g is a nonzero solution of Eq. (27) with s ¼ b2k . The fwk ðxÞg gives both
temperature and heat flux distributions in the kth mode.
It should be noted that the eigenequation (27) and the characteristic
equation (30) treat different boundary conditions and any number of layers system-
atically with the same format. This symbolic feature of the solution technique totally
eliminates tedious derivations and messy expressions that are encountered in many
existing analytical methods.

5. STEADY-STATE RESPONSE
5.1. Solution of Temperature Problem
The governing equations for the steady-state temperature of the composite slab
are derivable from Eqs. (1)–(4) by dropping the time derivative term in Eq. (1), the
initial condition (4), and the time dependence of all the quantities. The exact and
closed-form solution of such a temperature problem is given by Eq. (23) with s ¼ 0,
Z xn
fgss ðxÞg ¼ b ðx; n; 0ÞfpðnÞgdn þ ½H
½G b ðx; 0Þfcb g x 2 ½x0 ; xn  ð37Þ
x0
322 B. YANG

where
   
Tj ðxÞ 0
fgss ðxÞg ¼ dTj ðxÞ fpðxÞg ¼ gj ðxÞ for x 2 ½xj1 ; xj  ð38Þ
kj dx 1
and fcb g is a constant vector. The ½G b  and ½H
b  can be easily computed as the
exponential matrices that are contained in them become


1 x=kj
e½Fj ð0Þx ¼
0 1

5.2. Solution to Periodic Excitations

Assume that the composite slab is subject to sinusoidal excitations

gj ðx; tÞ ¼ fj ðxÞeixt j ¼ 1; 2; . . . ; n
ixt
ð39Þ
cL ðtÞ ¼ vL e cR ðtÞ ¼ vR eixt
pffiffiffiffiffiffiffi
where i ¼ 1, and x is the excitation frequency. The cyclic steady-state solution
for the heat conduction problem is
Z xn
fgðx; tÞg ¼ b ðx; n; ixÞfpðnÞgdn eixt þ ½H
½G b ðx; ixÞfvb g eixt x 2 ½x0 ; xn  ð40Þ
x0

   
0 vL b ðx; n; ixÞ
where fpðnÞg ¼ fj ðnÞ for n 2 ½xj1 ; xj  and fvb g ¼ . The ½G
1 vR
and ½Hb ðx; ixÞ are known as frequency response functions [33].
If internal energy sources and boundary excitations are periodic functions,
expand them in Fourier series and obtain a solution for each harmonic component
in the series. Then by the principle of superposition, the steady-state temperature and
heat flux of the composite slab are the sums of these harmonic solutions. Again, the
transfer function formulation provides exact periodic solutions.

6. TRANSIENT RESPONSE
6.1. Green’s Function Formula
Inverse Laplace transform of Eq. (23) gives the Green’s function formula
Z tZ xn
fgðx; tÞg ¼ ½Gðx; n; t  sÞffg ðn; sÞgdnds
0 x0
Z t Z xn
þ ½Hðx; t  sÞfcb ðsÞgds þ ½Gðx; n; tÞ ffh ðnÞgdn ð41Þ
0 x0

where fgðx; tÞg is the inverse Laplace transform of f^gðx; sÞg; the Green’s functions
b  and ½H
[G] and [H] are the inverse Laplace transforms of the transfer functions ½G b ,
 TRANSFER FUNCTION METHOD FOR HEAT CONDUCTION 323

respectively,
 
cL ðtÞ
fcb ðtÞg ¼ ð42aÞ
cR ðtÞ
and
   
0 kj 0
ffg ðx; tÞg ¼ gj ðx; tÞ ffh ðx; tÞg ¼  hj ðxÞ ð42bÞ
1 aj 1
for x 2 ½xj1 ; xj . The terms on the right-hand side of Eq. (41) represent the effects of
internal heat generation, boundary excitations, and initial temperature on the tran-
sient response.
According to Eqs. (24) and (28), the singularities of the distributed transfer
functions at b2k come from ½ZðsÞ1 . Assume that the eigenvalues of the composite
slab are distinct. By the theorem of residues [1, 2], the Green’s functions in Eq. (41)
are of the form
X
1
2
½Gðx; n; tÞ ¼ ½Uðx; x0 ; b2k Þ½Rk ½Dðx; n; b2k Þ ebk t ð43aÞ
k¼1

X
1
2
½Hðx; tÞ ¼ ½Uðx; x0 ; b2k Þ½Rk  ebk t ð43bÞ
k¼1

where
(
½Mb ½Uðx0 ; n; b2k Þ n<x
½Dðx; n; b2k Þ ¼ ð44Þ
½Nb ½Uðxn ; n; b2k Þ n>x

½Rk  ¼ Res ð½ZðsÞ1 Þ ð45Þ

s¼b2k

and Eqs. (16) and (24) have been used. The ½Rk  shall be called transfer function
residues.

6.2. Transfer Function Residues

If the transfer function residues are known, the exact transient solution of the
composite slab is readily computed by Eqs. (41)–(43). The ½Rk  is expressed as [2]

adj½Zðb2k Þ
½Rk  ¼ d
ð46Þ
ds jZðsÞjs¼b2 k

where adj[Z] is the adjoint of [Z] and jZj ¼ det[Z]. It can be shown that
 
d d
jZðsÞj ¼ det ½Mb  þ ½Nb  ½Uðxn ; sÞ ð47Þ
ds ds
324 B. YANG

where
d d
½Uðxn ; sÞ ¼ ðe½Fn ðsÞln Þe½Fn1 ðsÞln1


e½F2 ðsÞl2 e½F1 ðsÞl1
ds ds
d
þ e½Fn ðsÞln ðe½Fn1 ðsÞln1 Þ


e½F2 ðsÞl2 e½F1 ðsÞl1
ds
d
þ





þ e½Fn ðsÞln e½Fn1 ðsÞln1


e½F2 ðsÞl2 ðe½F1 ðsÞl1 Þ ð48Þ
ds

The derivatives in the previous equation are obtained as


d ½Fj ðsÞlj aj ðsÞ bj ðsÞ
ðe Þ¼ ð49Þ
ds cj ðsÞ aj ðsÞ

where
lj qffiffiffiffiffiffiffiffiffi
aj ðsÞ ¼ pffiffiffiffiffiffi sinh s=aj lj
2 aj s

lj qffiffiffiffiffiffiffiffiffi 1 qffiffiffiffiffiffiffiffiffi
bj ðsÞ ¼ cosh s=aj lj  pffiffiffiffiffiffiffiffiffi sinh s=aj lj ð50Þ
2kj s 2kj s s=aj

k j lj qffiffiffiffiffiffiffiffiffi kj qffiffiffiffiffiffiffiffiffi
cj ðsÞ ¼ cosh s=aj lj þ pffiffiffiffiffiffi sinh s=aj lj
2aj 2 aj s

It follows that the transfer function residues are given by

adj½Zðb2k Þ
½Rk  ¼   ð51Þ
d
det ½Mb  þ ½Nb  ds ½Uðxn ; b2k Þ

where the derivative d½Uðxn ; sÞ=ds is given in Eqs. (48)–(50).

The residue formula (51) is exact; no approximation has been made in its deri-
d
vation. For an n-layer slab, evaluation of ds ½Uðxn ; b2k Þ only requests summation of
n products of two-by-two matrices; see Eq. (48). As such, estimation of the transfer
function residues for a composite slab is extremely simple, and does not encounter
any singularities of ½ZðsÞ1 .

6.3. Exact Transient Solution

In the literature, exact transient heat-conduction solutions by inverse Laplace
transform have been limited to slabs with two layers [1]. The transfer function
formulation and the new residue formula developed in this work allow easy deter-
mination of exact and closed-form transient solutions for composite slabs with
many layers.
With transfer function residues obtained, the exact closed-form transient
solution of the boundary-initial value problem can now be derived. By Eq. (51)
 TRANSFER FUNCTION METHOD FOR HEAT CONDUCTION 325

and det½Zðb2k Þ ¼ 0, it can be shown that ½Rk ½Zðb2k Þ ¼ 0. As a result,

½Rk ½Dðx; n; b2k Þ ¼ ½Rk ½Mb ½Uðx0 ; n; b2k Þ x; n 2 ½x0 ; xn  ð52Þ

where Eq. (44) has been used. Thus, the Green’s function in Eq. (43a) becomes

X
1
2
½Gðx; n; tÞ ¼ ½Uðx; x0 ; b2k Þ½Rk ½Mb ½Uðx0 ; n; b2k Þ ebk t ð53Þ
k¼1

It follows from Eqs. (19), (22), (41), (43b), and (53) that

X
1
fgðx; tÞg ¼ ½Uðx; b2k Þ½Rk  ð½Mb fIg;k ðtÞg þ fIb;k ðtÞg þ ½Mb fIh;k ðtÞgÞ ð54Þ
k¼1

where

X
n Z Z
t xj  
2 2 0
fIg;k ðtÞg ¼  ½Uðxj1 ; b2k Þ1 ebk ðtsÞ e½Fj ðbk Þðnxj1 Þ gj ðn; sÞdnds
j¼1 0 xj1 1
Z t  
2 cL ðsÞ
fIb;k ðtÞg ¼ ebk ðtsÞ ds
0 cR ðsÞ
X
n Z xj  
b2k t kj ½Fj ðb2k Þðnxj1 Þ 0
fIh;k ðtÞg ¼ e ½Uðxj1 ; b2k Þ1 e hj ðnÞdn
j¼1
aj xj1 1
ð55Þ

The ½U, ½U1 , and exponential matrices in the previous equation are given in
explicit expressions by Eqs. (19) and (20). In many cases, the integrals in Eq. (55)
can be evaluated by exact quadrature.
In summary, the DTFM-based transient analysis takes the following three
steps.

1) Determine the eigenvalues b2k of the composite slab by Eq. (30).

2) Obtain the residues ½Rk  by Eq. (51).
3) For given excitations, compute the transient response by Eqs. (54) and (55).

Solutions from this process are of closed form in the sense that every term in Eqs. (54)
and (55) is given by explicit functions.

7. THERMAL RESISTANCE AT LAYER INTERFACES

For an n-layer composite slab with thermal resistance at its layer interfaces, the
heat equations and boundary conditions are the same as Eqs. (1) and (2), but the
326 B. YANG

matching conditions become

qTj ðxj ; tÞ
 kj ¼ hj ðTj ðxj ; tÞ  Tjþ1 ðxj ; tÞÞ
qx ð56Þ
qTj ðxj ; tÞ qTjþ1 ðxj ; tÞ
 kj ¼ kjþ1
qx qx

for j ¼ 1; 2; . . . ; n  1, where hj is a conductance coefficient, with h1j describing the

thermal resistance between the jth and ( j þ 1)th layers [1, 3]. For perfect thermal
contact, hj ! 1 and temperature continuity at the interface is resumed.
Following Section 3, the matching conditions (56) are cast into the spatial state
form

gjþ1 ðxj ; sÞg ¼ ½Cj f^

f^ gj ðxj ; sÞg j ¼ 1; 2; . . . ; n  1 ð57Þ

where


1 1=hj
½Cj  ¼ ð58Þ
0 1

Matrix ½Cj  reduces to an identity matrix when hj ! 1 (for perfect thermal contact).
The state equation and boundary condition are the same as Eqs. (13a) and (13b). It is
easy to show that a fundamental matrix of the composite body is

e½F1 ðsÞðxx0 Þ x 2 ðx0 ; x1 Þ

½Uðx; sÞ ¼ ð59Þ
e½Fj ðsÞðxxj1 Þ ½Cj1 e½Fj1 ðsÞlj1


½C1 e½F1 ðsÞl1 x 2 ðxj1 ; xj Þ

Because matrices ½Cj  do not contain parameter s, the results in the previous sections
are directly applicable here. This is done by replacing Eq. (19) with Eq. (59). For
instance, instead of Eq. (30), the characteristic equation of the composite slab is
2 2 2
DðbÞ  detð½Mb  þ ½Nb e½Fn ðb Þln
½Cn1 


e½F2 ðb Þl2
½C1 e½F1 ðb Þl1
Þ¼0 ð60Þ

Also, the residue formula (51) is valid if the derivative in it is taken as

d d  ½Fn ðsÞln
½Uðxn ; sÞ ¼ e ½Cn1 e½Fn1 ðsÞln1


e½F2ðsÞ l2 ½C1 e½F1 ðsÞl1
ds ds
d  ½Fn1 ðsÞln1
þ e½Fn ðsÞln ½Cn1  e


e½F2 ðsÞl2 ½C1 e½F1 ðsÞl1
ds
d  ½F1 ðsÞ1
þ





þ e½Fn ðsÞln ½Cn1 e½Fn1 ðsÞln1


e½F2 ðsÞl2 ½C1  e ð61Þ
ds

Thus, for bodies with thermal resistance at interfaces exact and closed-form
solutions can be obtained by the proposed DTFM with no additional difficulty.
 TRANSFER FUNCTION METHOD FOR HEAT CONDUCTION 327

8. NUMERICAL EXAMPLES
The proposed DTFM is illustrated in two examples: a two-layer slab in which
the method is validated with some existing results in the literature; and a seven-layer
slab in which the DTFM is shown to be convenient and efficient in modeling and
analysis of heat in conduction in composite slabs of many layers. In numerical simu-
lation, all physical parameters are assigned nondimensional values. In determination
of the eigenvalues of a slab (for transient analysis) the bisection method is used.

8.1. A Two-Layer Slab

Consider a two-layer slab with boundary conditions

qTðx0 ; tÞ
k1 þ h1 Tðx0 ; tÞ ¼ 0
qx ð62Þ
qTðx2 ; tÞ
k2 þ h2 Tðx2 ; tÞ ¼ 0
qx

where x0 ¼ 0 and x2 ¼ l1 þ l2 . Let the system parameters have the following values:
k1 ¼ 1; a1 ¼ 1; l1 ¼ 1; k2 ¼ 2; a2 ¼ 1; l2 ¼ 2; h1 ¼ 1; h2 ¼ 2. The same
problem is studied in ref. [8]. The eigenvalues of the slab are determined by solving
the characteristic equation (30). Table 1 lists the first 20 roots and the corresponding
eigenvalues. The results by the DTFM match those given in ref. [8].

Table 1. First 20 eigenvalues of the two-layer slab

Root bk from Root bk by Eigenvalue b2k by

k ref. [8] DTFM DTFM

1 0.6150667 0.6150667 0.3783071

2 1.5436569 1.5436570 2.3828768
3 2.2845548 2.2845549 5.2191910
4 3.3173428 3.3173428 11.0047635
5 4.4454150 4.4454150 19.7617147
6 5.2652385 5.2652385 27.7227366
7 6.3765322 6.3765322 40.6601627
8 7.5254517 7.5254517 56.6324237
9 8.3582468 8.3582468 69.8602897
10 9.4877882 9.4877882 90.0181255
11 10.6403975 10.6403975 113.2180589
12 11.4771473 11.4771473 131.7249097
13 12.6138397 12.6138397 159.1089525
14 13.7672471 13.7672471 189.5370939
15 14.6056780 14.6056780 213.3258301
16 15.7460178 15.7460179 247.9370782
17 16.8995043 16.8995043 285.5932457
18 17.7387985 314.6649730
19 18.8813042 356.5036483
20 20.0346615 401.3876599
328 B. YANG

Figure 2. Transient temperature of the two-layer slab subject to a uniform initial temperature h0 ¼ 1.

Now consider the transient response of the slab subject to a uniform initial
temperature Tðx; tÞ ¼ h0 , x 2 ½x0 ; x2 . According to Eqs. (54) and (55),
X
1
2 X
2
kj
fgðx; tÞg ¼ h0 ebk t ½Uðx; b2k Þ½Rk ½Mb  ½Uðxj1 ; b2k Þ1 fqk;j g ð63Þ
k¼1 j¼1
aj

where
0 a   pffiffiffiffi 1
 k bj 2 1  cos bk lj = aj
fqk;j g ¼ @ A
j k
pffiffiffi  ð64Þ
aj pffiffiffiffi
b sin bk lj = aj
k

For h0 ¼ 1, the transient temperature of the slab is simulated by the series (63) with
the first 20 terms. The temperature distributions at times t ¼ 0, 0.1, 0.5, 1, 2, 3, 5, and
100 are plotted in Figure 2, which are in good agreement with the results presented in
ref. [8].

8.2. A Seven-Layer Slab

A seven-layer composite slab has the boundary conditions

qTðx0 ; tÞ
 k1 ¼ pB ðtÞ
qx ð65Þ
qTðx7 ; tÞ
k7 þ hTðx7 ; tÞ ¼ 0
qx

The values of the layer parameters used are given in Table 2, and the heat transfer
coefficient at x7 is chosen as h ¼ 2. The eigenvalues of the slab are first completed.
 TRANSFER FUNCTION METHOD FOR HEAT CONDUCTION 329

Table 2. Physical parameters of the seven-layer slab

Layer #1 #2 #3 #4 #5 #6 #7

kj 2 3 5 5 3 2 1.5
aj 6 8 10 14 11 9 7
lj 1 1.5 2 1.5 2 1 1

The characteristic function DðbÞ of the slab is plotted in Figure 3. Also in the figure is
a scaled characteristic function DðbÞ=ð1 þ bÞ, which is bounded for any b. Either
function leads to the same results in numerical simulation. Table 3 gives the first
20 eigenvalues of the slab. For comparison, the table also lists the results obtained
by the finite element method (FEM). The finite element prediction converges to
the DTFM solution as the number of elements increases. The FEM with 20, 100,
and 500 elements has maximum relative errors of 37.4%, 3.3%, and 0.13%, respect-
ively. As can been seen, a large number of elements is needed for accurate finite
element analysis.
The transient and steady-state responses of the slab are computed in three cases
of excitations: internal heat generation, time-dependent boundary excitations, and
2
nonuniform initial temperature profiles. Due to fast decay of ebk t with a higher
mode number k, a precursor convergence study reveals that, after a very short time
(say t  0:001), there is no significant difference between the results obtained by a
model with the first 50 terms from the series (54) and those by a model with the first
20 terms. For this reason, the first 20 terms in the series are used in all the transient
response simulations. The steady-state solution is obtained in exact and closed form,
Eq. (37).

Figure 3. Characteristic functions of the seven-layer slab versus b.

330 B. YANG

Table 3. First 20 eigenvalues of the seven-layer slab

Finite element method (m ¼ number of elements)

k DTFM (b2k ) m ¼ 20 m ¼ 100 m ¼ 500

1 0.128655 0.128692 0.128656 0.128655

2 2.077742 2.088418 2.078168 2.077759
3 4.838283 4.896827 4.840611 4.838376
4 10.110454 10.355875 10.120166 10.110842
5 16.603052 17.286810 16.629935 16.604126
6 23.876771 25.284265 23.932111 23.878983
7 34.699051 37.581853 34.811666 34.703551
8 45.779431 50.773509 45.972773 45.787153
9 61.574859 70.833202 61.934578 61.589224
10 74.582489 88.278180 75.108645 74.603484
11 93.794677 114.662632 94.605335 93.827021
12 116.392302 150.802082 117.691983 116.444095
13 133.878533 179.736120 135.676870 133.950168
14 153.951864 209.699406 156.226680 154.04235
15 180.286324 249.290177 183.248580 180.40425
16 207.970426 290.427476 212.101273 208.134782
17 236.951027 327.621219 242.327219 237.164559
18 261.172873 357.082145 267.538110 261.425527
19 302.766554 414.980344 311.172051 303.100724
20 334.159209 459.132198 345.209196 334.597603

Case 1. Response to Internal Heat Generation

The slab has an energy source uniformly distributed in the fourth layer

g4 ðx; tÞ ¼ g0 x 2 ½x3 ; x4  t0 ð66Þ

where g0 is a constant. Assuming zero boundary excitations and zero initial

temperature, the transient response of the slab by Eq. (54) is
X
1
1 2
fgðx; tÞg ¼ g0 ½Uðx; b2k Þ½Rk ½Mb ½Uðx3 ; b2k Þ1 fqk;4 g 2
ð1  ebk t Þ ð67Þ
k¼1 bk

where fqk;4 g is given in Eq. (64). The steady-state response by Eq. (37) is
Z x4
fgss ðxÞg ¼ g0 ½Uðx; 0Þ½Zð0Þ1 ½Eðx; nÞ ½Uðn; 0Þ1 dn ð68Þ
x3

with

½Mb  nx
½Eðx; nÞ ¼ ð69Þ
½Mb   ½Zð0Þ n>x

Figure 4 shows the transient temperature of the slab (solid lines) at times t ¼ 0,
0.5, 2, 4, 7, and 30 and the steady-state temperature (dashed line) for g0 ¼ 1.
 TRANSFER FUNCTION METHOD FOR HEAT CONDUCTION 331

Figure 4. Temperature distributions of the seven-layer slab with internal heat generation in the fourth
layer, g0 ¼ 1: solid line – transient response and dashed line – steady-state response.

Also, the heat flux in the body, qðx; tÞ ¼ ½ 0 1 fgðx; tÞg, is plotted in
Figure 5. As time increases, these transient solutions converge to the steady-
state ones.

Figure 5. Distributions of heat flux of the seven-layer slab with internal heat generation in the fourth
layer, g0 ¼ 1: solid line – transient response and dashed line – steady-state response.
332 B. YANG

Case 2. Response to Boundary Excitation

The boundary heat flux in Eq. (61) is time-dependent

pB ðtÞ ¼ q0 ð1  ert Þ t0 ð70Þ

where q0 and r are positive constants. The transient response of the slab with zero
initial temperature and zero inner sources is

X1  (  )
2 1 1 b2k t
fgðx; tÞg ¼ q0 ½Uðx; bk Þ½Rk  1e  ak ðtÞ ð71Þ
k¼1
0 b2k

with
(  2
1
b2k r
ert  ebk t for r 6¼ b2k
ak ðtÞ ¼ 2
ð72Þ
tebk t for r ¼ b2k

In steady-state, the boundary heat flux lim pB ðtÞ ¼ q0 and the solution by Eq. (37) is
t!1
 
1 1
fgss ðxÞg ¼ q0 ½Uðx; 0Þ½Zð0Þ ð73Þ
0

Let q0 ¼ 1 and r ¼ 0:15. The transient and steady-state temperature profiles of the
slab are computed by Eqs. (71) and (73) and plotted in Figure 6.

Figure 6. Temperature distributions of the seven-layer slab with the boundary excitation given in Eq. (70),
q0 ¼ 1, r ¼ 0:15: solid line – transient response and dashed line – steady-state response.
 TRANSFER FUNCTION METHOD FOR HEAT CONDUCTION 333

Case 3. Transient Response to Initial Temperature

Nonuniform initial temperature profiles are selected to show the capability of
the proposed method in dealing with complicated spatial distributions of excitations.
First, consider a parabolic distribution of initial temperature

Tðx; 0Þ ¼ 6  1:6x þ 0:16x2 x 2 ½0; 10 ð74Þ

The transient temperature of the composite slab under such an excitation is plotted
in Figure 7, where the dashed line represents the initial temperature profile.
Next, consider a piecewise continuous profile of initial temperature.

Layer 1: T1 ðn; 0Þ ¼ 4  n n 2 ½0; 1

Layer 2: T2 ðn; 0Þ ¼ 2 n 2 ½0; 1:5
Layer 3: T3 ðn; 0Þ ¼ 3 n 2 ½0; 2
Layer 4: T4 ðn; 0Þ ¼ 4 n 2 ½0; 1:5 ð75Þ
2
Layer 5: T5 ðn; 0Þ ¼ 6 þ 0:5n  n n 2 ½0; 2
Layer 6: T6 ðn; 0Þ ¼ 3 n 2 ½0; 1
Layer 7: T7 ðn; 0Þ ¼ 1 n 2 ½0; 1

where n is a local coordinate which for the jth layer is n ¼ x  xj1 . The initial
temperature is continuous within each layer, but has jumps at the layer interfaces.

Figure 7. Transient temperature of the seven-layer slab subject to the parabolic distribution of initial
temperature as given in Eq. (74): solid line – transient response; dashed line – initial temperature profile.
334 B. YANG

Figure 8. Transient temperature of the seven-layer slab subject to the initial temperature distribution as
given in Eq. (75): solid line – transient response and dashed line – initial temperature profile.

The transient response of the slab is presented in Figure 8, where the dashed line
portraits the initial temperature profile.
In the above transient Ranalysis, no approximation has been made. The key in
computation is to evaluate 0j e½Fj ðbk Þðnxj1 Þ hj ðnÞdn in Eq. (55). For the initial
l 2

temperature profiles given in Eqs. (74) and (75), exact quadrature has been used
for these integrals.

9. CONCLUSIONS
A distributed transfer function method (DTFM) has been developed for con-
duction heat transfer in one-dimensional multilayer slabs. The DTFM is capable
of delivering exact analytical solutions for slabs composed of any number of layers
and subject to arbitrary internal heat generation, boundary excitations, and initial
temperature distribution. The DTFM is applicable to composite bodies with
thermal resistance at layer interfaces.
The thrust of the DTFM is that in transient analysis it gives exact transfer
function residues without having to deal with tedious derivations and complicated
expressions. The utility of the residue formula (51) requires little computational
effort and avoids errors caused by singularities of system transfer functions. To
the author’s knowledge, the DTFM is the first Laplace-transform-based analytical
method that practically delivers exact and closed-form transient solutions for heat
condition in one-dimensional composite slabs with any number of layers.
With a spatial state formulation, the DTFM models composite slabs in a
compact manner. Physical parameters and boundary conditions are systematically
 TRANSFER FUNCTION METHOD FOR HEAT CONDUCTION 335

treated through easy assignment of state matrices ½Fj  and boundary matrices
½Mb  and ½Nb . The solution algorithms remain the same for different slab
configurations and various excitations. The DTFM-based analysis only involves
simple mathematical operations of two-by-two matrices. The symbolic feature
and low-order matrix manipulation make the proposed solution method user-
friendly and highly efficient in numerical simulation, as has been demonstrated
in Section 8.
Root searching by the proposed method makes use of a smooth and bounded
characteristic function DðbÞ (see Figure 3 for instance). Because of this, the eigenva-
lues of a composite slab can be accurately computed without the issues of numerical
instability and missing roots. It should be pointed out that classical boundary value
approach also yields a characteristic equation that contains sine and cosine
functions [3], although the determinant of a 2n-by-2n matrix (for an n-layer slab)
has to be computed during root searching. The DTFM only needs to evaluate a pro-
duct of n two-by-two matrices (see Eq. (30) or (35)), and therefore is numerically
more efficient for composite bodies of many layers.
The DTFM is different from existing analytical methods in the following
aspects. First, unlike the orthogonal expansion technique and the Green’s function
method, the DTFM does not require the knowledge on system eigenfunctions,
and does not need to deal with related spatial integrals of system eigenfunctions.
Second, different from traditional Laplace transform method, the DTFM does not
involve complicated expressions in inverse Laplace transform, and obtains transient
solutions without any approximation. Third, many analytical techniques assign two
unknown parameters for each layer of a composite slab, which could be challenging
if the number of layers is large. The DTFM formulates and solves the heat conduc-
tion problem through simple operations of two-by-two matrices, regardless of the
number of layers. As a result, accurate transient and steady-state response can be
easily computed.
The proposed method is extensible to multilayer composites of other
geometries. A follow-up study is underway.

REFERENCES
1. H. S. Carslaw and J. C. Jaeger, Conduction of Heat in Solids, 2nd ed., Oxford University
Press, Oxford, 1959.
2. P. M. Morse and H. Feshbach, Methods of Theoretical Physics, McGraw-Hill, New York,
1953.
3. M. N. Ozisik, Boundary Value Problems of Heat Conduction, International Textbook,
Scranton, PA, 1968.
4. F. M. Ramos and A. Giovannini, Finite Analytic Numerical Method for Transient Heat
Diffusion in Layered Composite Materials, Numer. Heat Transfer B, vol. 22, pp. 305–319,
1992.
5. J. Pakanen, Conduction of Heat through One- and Multilayer Slabs: A Differential
Difference Approach, Numer. Heat Transfer B, vol. 30, pp. 67–91, 1996.
6. A. K. Das and S. S. Sadhal, A Note on the Evaluation of Thermal Constriction Resistance
for Finite Thickness Gaps, J. Heat Transfer, vol. 119, pp. 177–180, 1997.
7. R. Siegel, Transient Thermal Analysis of Parallel Translucent Layers by Using Green’s
Functions, J. Thermophysics and Heat Transfer, vol. 13, pp. 10–17, 1999.
336 B. YANG

8. F. de Monte, Transient Heat Conduction in One-Dimensional Composite Slab, A ‘Natural’

Analytic Approach, Inter. J. Heat and Mass Transfer, vol. 43, pp. 3607–3619, 2000.
9. R. P. A. Rocha, M. Anuel, and E. Cruz, Computation of the Effective Conductivity of
Unidirectional Fibrous Composites with an Interfacial Thermal Resistance, Numer. Heat
Transfer A, vol. 39, pp. 179–203, 2001.
10. B. Ya. Kantor, N. V. Smetankina, and A. N. Shupikov, Analysis of Non-Stationary
Temperature Fields in Laminated Strips and Plates, Inter. J. Solids and Structures, vol. 38,
pp. 8673–8684, 2001.
11. A. Sutradhar, G. H. Paulino, and L. J. Gray, Transient Heat Conduction in Homo-
geneous and Non-Homogeneous Materials by The Laplace Transform Galerkin Bound-
ary Element Method, Eng. Anal. with Boundary Elements, vol. 26, pp. 119–132, 2002.
12. T. P. Fredman, An Analytical Solution Method for Composite Layer Diffusion Problems
with An Application in Metallurgy, Heat and Mass Transfer, vol. 39, pp. 285–295, 2003.
13. Y. Sun and I. S. Wichman, On Transient Heat Conduction in a One-Dimensional
Composite Slab, International J. Heat and Mass Transfer, vol. 47, pp. 1555–1559, 2004.
14. X. Lu and P. Tervola, Transient Heat Conduction in the Composite Slab-Analytical
Method, J. Physics A: Math. and Gen., vol. 38, pp. 81–96, 2005.
15. J. Li, P. Cheng, G. P. Peterson, and J. Z. Xu, Rapid Transient Heat Conduction in
Multilayer Materials with Pulsed Heating Boundary, Numer. Heat Transfer A, vol. 47,
pp. 633–652, 2005.
16. Z.-H. Wang and K. H. Tan, Green’s Function Solution for Transient Heat Conduction in
Concrete-Filled CHS Subjected to Fire, Eng. Structures, vol. 28, pp. 1574–1585, 2006.
17. V. Vodicka, Warmeleitung in Geschichteten Kugel-und Zylinderkorpern, Schweizer
Archiv, vol. 10, pp. 297–304, 1950.
18. C. W. Tittle, Boundary Value Problems in Composite Media: Quasi-Orthogonal
Functions, J. App. Physics, vol. 36, pp. 1486–1488, 1965.
19. J. V. Beck, Green’s Function Solution for Transient Heat Conduction Problems, Inter. J.
Heat and Mass Transfer, vol. 27, pp. 1235–1244, 1984.
20. S. C. Huang and Y. P. Chang, Heat Conduction in Unsteady, Periodic and Steady States
in Laminated Composites, ASME J. Heat Transfer, vol. 102, pp. 742–748, 1990.
21. A. C. Giere, 1964=1965, Transient Heat Flow in a Composite Slab—Constant Flux, Zero
Flux Boundary Conditions, Appl. Sci. Research, Section A, vol. 14, pp. 191–198, 1965.
22. P. E. Bulavin and V. M. Kashcheev, Solution of Nonhomogeneous Heat-Conduction
Equation for Multilayered Bodies, Int. Chem. Eng., vol. 5, no. 1, pp. 112–115, 1965.
23. G. P. Mulholland and M. H. Cobble, Diffusion Through Composite Media, Inter. J. Heat
and Mass Transfer, vol. 15, pp. 147–160, 1972.
24. M. Jin and S. S. Sadhal, Thermal Boundary Conditions for Heterogeneous Solids,
TMS Annual Meeting, Properties of Composites Session, Anaheim, Calif., February 4–8,
1996.
25. A. Haji-Sheikh, J. V. Beck, and D. Agonafer, Steady-State Heat Conduction in Multi-
layer Bodies, Inter. J. Heat and Mass Transfer, vol. 46, pp. 2363–2379, 2003.
26. M. C. B. Gough, Modeling Heat Flow in Buildings: An Eigenfunction Approach, Ph.D.
Thesis, Cambridge University Press, Cambridge, 1982.
27. A. Al-Mujahid and M. E. Zedan, Transient Heat-Conduction Response of a Composite
Plane Wall, Wärme- und Stoffübertragung, vol. 26, pp. 33–39, 1990.
28. B. Yang, Transfer Functions of Constrained=Combined One-Dimensional Continuous
Dynamic Systems, J. Sound and Vibration, vol. 156, pp. 425–443, 1992.
29. B. Yang and C. A. Tan, Transfer Functions of One-Dimensional Distributed Parameter
Systems, ASME J. Applied Mechanics, vol. 59, pp. 1009–1014, 1992.
30. B. Yang and H. Fang, Transfer Function Formulation of Non-Uniformly Distributed
Parameter Systems, ASME J. Vibration and Acoustics, vol. 116, pp. 426–432, 1994.
 TRANSFER FUNCTION METHOD FOR HEAT CONDUCTION 337

31. B. Yang, Stress, Strain, and Structural Dynamics: An Interactive Handbook of Formulas,
Solutions, and MATLAB Toolboxes, Elsevier Science, Boston, 2005.
32. C. T. Chen, Linear Systems Theory and Design, 3rd ed., Oxford University Press,
New York, 1999.
33. R. C. Dorf and R. H. Bishop, Modern Control Systems, 10th ed., Prentice Hall, Upper
Saddle River, New Jersey, 2005.

APPENDIX
Proof of the s-Domain Solution
By Eqs. (23) and (24)
Z x
gðx; sÞg ¼ ½Uðx; x0 ; sÞ½ZðsÞ1 ½Mb 
f^ ½Uðx0 ; n; sÞf^pðn; sÞgdn
x0
Z xn
½Nb  ½Uðxn ; n; sÞf^pðn; sÞgdn þ f
cb ðsÞg
x
ðA1Þ

Differentiation of the both sides of Eq. (A1) and use of the properties (15) and (16),
yields
q
f^ gðx; sÞg þ ½Uðx; x0 ; sÞ½ZðsÞ1 ð½Mb ½Uðx0 ; x; sÞ
gðx; sÞg ¼ ½F ðx; sÞf^
qx
þ ½Nb ½Uðxn ; x; sÞÞf^
pðx; sÞg
gðx; sÞg þ ½Uðx; x0 ; sÞ½ZðsÞ1 ½ZðsÞ½Uðx0 ; x; sÞf^pðx; sÞg
¼ ½F ðx; sÞf^
¼ ½F ðx; sÞf^
gðx; sÞg þ f^
pðx; sÞg

which means that the expression in Eq. (23) satisfies Eq. (13a). Now, use Eq. (A1) to
compute

½Mb f^ gðxn ; sÞg ¼ ½Mb ½ZðsÞ1 ðfcb ðsÞg  ½Nb ½Uðxn ; x0 ; sÞfagÞ
gðx0 ; sÞg þ ½Nb f^
þ ½Nb ½Uðxn ; x0 ; sÞ½ZðsÞ1 ðfcb ðsÞg þ ½Mb fagÞ
¼ fcb ðsÞg
Rx
where fag ¼ x0n ½Uðx0 ; n; sÞf^
pðn; sÞgdn, and ½Nb ½Uðxn ; x0 ; sÞ ¼ ½ZðsÞ  ½Mb  has been
used. This indicates that the expression in Eq. (23) also satisfies the boundary
condition (13b). Because the fundamental matrix given in Eq. (19) is continuous
at all layer interfaces, the state vector given in Eq. (23) automatically satisfies the
mating conditions (11). Therefore, Eqs. (23) and (24) provide the unique solution
of the problem (13).