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

net/publication/306017722

ANALYTICAL SOLUTION OF 1-D MULTILAYER HEAT CONDUCTION PROBLEM

WITH TIME VARYING AMBIENTS

Conference Paper · December 2015

CITATIONS READS

3 585

2 authors:

Pranay Biswas Suneet Singh

Technion - Israel Institute of Technology Indian Institute of Technology Bombay
7 PUBLICATIONS   25 CITATIONS    158 PUBLICATIONS   1,226 CITATIONS   

SEE PROFILE SEE PROFILE

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

Non-linear dynamics of Pulsating Heat Pipes View project

Bifurcation Analysis of Out-of-Phase Oscillations in Boiling Water Reactors using Multipoint Neutron Kinetics View project

All content following this page was uploaded by Pranay Biswas on 10 August 2016.

The user has requested enhancement of the downloaded file.

 Proceedings of the 23rd National Heat and Mass Transfer Conference and
1st International ISHMT-ASTFE Heat and Mass Transfer Conference
IHMTC2015
17-20 December, 2015, Thiruvananthapuram, India

IHMTC2015-195

ANALYTICAL SOLUTION OF 1D MULTILAYER HEAT CONDUCTION PROBLEM WITH TIME

VARYING AMBIENTS

Pranay Biswas Suneet Singh

Department of Energy Science and Engineering Department of Energy Science and Engineering
Indian Institute of Technology Bombay Indian Institute of Technology Bombay
Mumbai-400076 Mumbai-400076
Email: pranaybiswas@iitb.ac.in Email: suneet.singh@iitb.ac.in
Phone :+91-22-2576-7843
Fax :+91-22-2576-4890

ABSTRACT Keywords: Multilayer composite, Analytical solution,

The present paper discusses the solution of 1-D Ambient with time varying temperature, Auxiliary function
multilayer heat conduction with time dependent ambient
temperatures using orthogonal eigen-function expansion NOMENCLATURE
method (OEEM). It is pointed out that this method is C𝑖,𝑛 , D𝑖,𝑛 Coefficients of eigen-function
applicable only when boundary conditions are homogenous. f𝑛∗ Coefficient used in Eqn. (34)
In the present work, an auxiliary function (satisfying a set of 𝑔𝑖 (𝑟, 𝑡) Volumetric heat generation rate [Wm-3]
conditions) is subtracted from the temperature in such a way 𝑔𝑛∗ (𝑡), I𝑛∗ (𝑡), V𝑛∗ (𝑡): Time Functions used in Eqns. (35-37)
that the modified temperature has homogenous boundary 𝑦𝑛 (𝑡) Time Functions used in Eqn. (42)
conditions. However, it is shown that such a function cannot y𝑛∗ (𝑡) Time Functions used in Eqn. (39)
be uniquely defined. A set of functions can be obtained as h1, hM Heat transfer confidents [Wm-2K-1]
long as they satisfy certain conditions. In fact, there are 𝑘𝑖 Thermal conductivity [Wm-1K-1]
infinite choices for this function. A generalized form of this M Number of layers in the composite
function is being considered in this paper which is valid for 𝑁𝑛 Normalisation factor
Cartesian, cylindrical and spherical coordinate systems.
𝑞𝑖 (𝑟, 𝑡) Auxiliary function [ºC]
The advantage of the generalized form of auxiliary functions
r Radial coordinate or distance [m]
is that these functions need to be defined only in the external
𝑅𝑖,𝑛 (𝑟) Eigen-function
layers which require no further evaluation. This leads to
significant simplification in the solution procedure. These 𝑆𝑖,𝑛 (𝑟), 𝑊𝑖,𝑛 (𝑟): Components eigen-function used in the
simplifications also lead to reduction in the computational solution of eigen-value problem
cost. The abovementioned methodology is implemented with 𝑇𝑖 (𝑟, 𝑡) Temperature [ºC]
eigen-functions and eigen-values obtained by standard 𝑇S1 (𝑡), 𝑇SM (𝑡): Time varying ambient temperatures [ºC]
techniques available in the literature. A few problems are 𝑇0,𝑖 (𝑟) Initial temperature [ºC]
solved with different choices of the auxiliary function. The W Number of terms in eigen-function set (max n)
error estimation shows that the error reduces with
increasing number of terms in the summation series. The GREEK SYMBOLS
error is less than 5%, at midpoints of different layers; 𝛼𝑖 Thermal diffusivity [m2s-1]
number of terms in series is greater than 5. 𝜀𝑖 (𝑟) Error in temperature profile at t = 0 [%]
𝜆𝑖,𝑛 Eigen-value

1
Γ𝑛 (𝑡) Time component of modified temperature dependent and time independent inhomogeneous BCs,
𝜃𝑖 (𝑟, 𝑡) Modified temperature [ºC] respectively[1], [6]–[8]. For the above formulation an
𝜃0,𝑖 (𝑟) Modified initial temperature [ºC] auxiliary function needs to be defined for each layer. The
Ω(𝑟) Special function used in Eqns. (18-20) complexity in evaluation of auxiliary functions increases
rapidly when number of layer in the composite is large. To
SUBSCRIPTS AND SUPERSCRIPTS overcome the abovementioned difficulty, a special choice of
i Layer in the composite auxiliary function is given by Antonopoulos and
n Term in eigen-value set or series summation Tzivanidis[9] for homogenisation of BCs in multilayer
p Index for co-ordinate system [0, 1 & 2] problems. The advantage of their method is that auxiliary
z Power index in Eqns. (18-20) [0, 1 & ≥ 2] function needs to be defined only for innermost and
outermost layer which requires significantly less effort
compared to earlier methods.
INTRODUCTION
In this present work, a generalized form of auxiliary
The use of multilayer composites, such as slabs,
function, introduced by Antonopoulos and Tzivanidis, is
cylinders and spheres, is constantly growing in thermal and
proposed. It can be seen that formulation of auxiliary
thermodynamic systems for engineering application due to
function, by Antonopoulos and Tzivanidis, is a special case
added advantage of combining physical, mechanical and
of the proposed formulation.
thermal properties of different materials. A wide range of
application needs accurate and efficient determination of
heat flux and temperature distributions within each layer of
PROBLEM DESCRIPTION
the composite. In order to find temperature distribution, a
A composite medium consists of M 1-D parallel layers
parabolic partial differential equation (PDE) can be obtained
of slabs, cylinders, or spheres as illustrated in Fig.1. The
by combining Fourier law of heat conduction with 1st law of
assumption here is that the interfaces are in perfect thermal
thermodynamics[1]. These PDEs, both for single layer as
contact. Initially each layer is at a specific temperature
well as multiple layers, have been solved extensively, during
last several decades, using several available methods. 𝑇𝑖 (𝑟, 0) = 𝑇0,𝑖 (𝑟), in 𝑟𝑖−1 < 𝑟 < 𝑟𝑖 , i = 1,..., M. The energy
is dissipated with convection from the two outer boundary
Three major approaches are present to obtain solution surfaces r = r0 and r = rM, into ambient having time varying
of the PDE, i) Numerical method, ii) Approximate method temperatures 𝑇S1 (𝑡) and 𝑇SM (𝑡), with heat transfer
and iii) Analytical method. Enormous effort has been put on coefficients h1 and hM, respectively.
to obtain analytical solution of multilayer heat conduction
problem. The homogenous heat conduction problem can be
solved relatively easily by separation of variables. Similarly,
the problem with a source and homogenous boundary
conditions (BCs) can be solved by expanding the source
(either time dependent or independent) term into
corresponding eigen-functions. Such a method for
multilayer multidimensional system was given by Salt[2],
[3]. However, in case of inhomogeneous BCs these two
relatively straightforward approaches cannot be used. For
such problems, the Finite Integral Transform (FIT) method
can be used which does not require homogenization. A very
generalized FIT approach for multilayer was given by Yener
and Özisik[4] which was further extended by Singh et al. for FIGURE 1. 1-D M-LAYER COMPOSITE
polar coordinates[5]. However, FIT leads to some
difficulties due to mismatch of BCs and corresponding
eigen-functions. Therefore, it is preferable to homogenise The transient heat behavior is modeled by a PDE of
the BCs which leads to a modified source term. The OEEM temperature for each layer of the composite. The
method then can be used to solve the modified problem. To mathematical model for each of the M layers is as follows,
homogenise the BCs, an auxiliary function is subtracted
from the temperature in such a way that the modified 1 𝜕𝑇𝑖 (𝑟,𝑡) 1 𝜕 𝜕𝑇𝑖 (𝑟,𝑡) 𝑔𝑖 (𝑟,𝑡)
= (𝑟 𝑝 )+ (1)
𝛼𝑖 𝜕𝑡 𝑟 𝑝 𝜕𝑟 𝜕𝑟 𝑘𝑖
temperature has homogeneous BCs. This auxiliary function
satisfies a certain set of conditions. It is also noticed that
0 for Cartesian Coordinate
such a function cannot be uniquely defined. In fact there are
infinite choices for this function. Özisik and Singh et. al. where 𝑝 = {1 for Cylindrical Coordinate,
have given some specific auxiliary functions for time 2 for Spherical Coordinate

2
subject to BCs 𝜕𝜃M (𝑟,𝑡)
𝑘M | + ℎM 𝜃M (𝑟, 𝑡)|𝑟M = 0 (12)
At inner surface of the 1st layer, i = 1 𝜕𝑟 𝑟M

𝜕𝑇1 (𝑟,𝑡) and IC

𝑘1 | = ℎ1 [𝑇1 (𝑟, 𝑡) − 𝑇S1 (𝑡)]|𝑟0 (2)
𝜕𝑟 𝑟0
𝜃𝑖 (𝑟, 0) = 𝜃0,𝑖 (𝑟) = 𝑇0,𝑖 (𝑟) − 𝑞𝑖 (𝑟, 0). (13)
At the interface of ith and (i+1)th layer
The modified source terms is given by the following
𝑇𝑖 (𝑟, 𝑡)|𝑟𝑖 = 𝑇𝑖+1 (𝑟, 𝑡)|𝑟𝑖 𝑖 = 1, … . , M − 1 (3) expression

𝜕𝑇𝑖 (𝑟,𝑡) 𝜕𝑇𝑖+1 (𝑟,𝑡) 1 𝜕 𝜕𝑞𝑖 (𝑟,𝑡) 1 𝜕𝑞𝑖 (𝑟,𝑡) 𝑔𝑖 (𝑟,𝑡)

𝑘𝑖 | = 𝑘𝑖+1 | 𝑖 = 1, … . , M − 1 (4) (𝑟 𝑝 )− + .
𝜕𝑟 𝑟𝑖 𝜕𝑟 𝑟𝑖 𝑟 𝑝 𝜕𝑟 𝜕𝑟 𝛼𝑖 𝜕𝑡 𝑘𝑖

At outer surface of Mth layer, i = M Note that, the Eqns. (8-13) are homogeneous only if
𝑞𝑖 (𝑟, 𝑡) satisfies certain conditions given as follows,
𝜕𝑇M (𝑟,𝑡) At inner surface of the 1st layer, i = 1
− 𝑘M | = ℎM [𝑇M (𝑟, 𝑡) − 𝑇SM (𝑡)]|𝑟M (5)
𝜕𝑟 𝑟M
𝜕𝑞1 (𝑟,𝑡)
−𝑘1 | + ℎ1 [𝑞1 (𝑟, 𝑡) − 𝑇S1 (𝑡)]|𝑟0 = 0 (14)
𝜕𝑟 𝑟0
and IC

𝑇𝑖 (𝑟, 0) = 𝑇0,𝑖 (𝑟). (6) At the interface of ith and (i+1)th layer

The BCs need to be homogenized in order to apply 𝑞𝑖 (𝑟, 𝑡)|𝑟𝑖 = 𝑞𝑖+1 (𝑟, 𝑡)|𝑟𝑖 𝑖 = 1, … . , M − 1 (15)
OEEM.
𝜕𝑞𝑖 (𝑟,𝑡) 𝜕𝑞𝑖+1 (𝑟,𝑡)
𝑘𝑖 | = 𝑘𝑖+1 | 𝑖 = 1, … . , M − 1 (16)
𝜕𝑟 𝑟𝑖 𝜕𝑟 𝑟𝑖

HOMOGENIZATION OF BCs
The problem can be homogenized by defining a At outer surface of Mth layer, i = M
modified temperature 𝜃𝑖 (𝑟, 𝑡) for ith layer as follows, 𝜕𝑞M (𝑟,𝑡)
𝑘M | + ℎM [𝑞M (𝑟, 𝑡) − 𝑇SM (𝑡)]|𝑟M = 0. (17)
𝜕𝑟 𝑟M
𝜃𝑖 (𝑟, 𝑡) = 𝑇𝑖 (𝑟, 𝑡) − 𝑞𝑖 (𝑟, 𝑡). (7)

here 𝑞𝑖 (𝑟, 𝑡) is an auxiliary function and is defined CHOICE OF AUXILIARY FUNCTION 𝒒𝒊 (𝒓, 𝒕)
subsequently. Now, substituting 𝑇𝑖 (𝑟, 𝑡) from Eqn. (7) in A generalized formulation is given here, which satisfies
Eqns. (1-6), a modified heat conduction problem is Eqns. (14-17),
formulated with homogeneous BCs as follows,
{Ω(𝑟1 )−Ω(𝑟)}𝑧 ℎ1 𝑇S1 (𝑡)
1 𝜕𝜃𝑖 (𝑟,𝑡) 𝜕𝜃 (𝑟,𝑡)
1 𝜕 1 𝜕 𝜕𝑞 (𝑟,𝑡) 𝑞1 (𝑟, 𝑡) = (18)
= 𝑝 (𝑟 𝑝 𝑖 ) + (𝑟 𝑝 𝑖 ) − {Ω(𝑟1 )−Ω(𝑟0 )}𝑧−1 [ℎ1 {Ω(𝑟1 )−Ω(𝑟0 )}+𝑧𝑘1
𝑑Ω(𝑟)
| ]
𝛼𝑖 𝜕𝑡 𝑟 𝜕𝑟 𝜕𝑟 𝑟 𝑝 𝜕𝑟 𝜕𝑟 𝑑𝑟 𝑟0
1 𝜕𝑞𝑖 (𝑟,𝑡) 𝑔𝑖 (𝑟,𝑡)
+ , (8)
𝛼𝑖 𝜕𝑡 𝑘𝑖 𝑞𝑖 (𝑟, 𝑡) = 0 𝑖 = 2, … , M − 1 (19)
subject to BCs 𝑞M (𝑟, 𝑡) =
At inner surface of the 1st layer, i= 1 {Ω(𝑟)−Ω(𝑟M−1 )}𝑧 ℎM 𝑇SM (𝑡)
𝑑Ω(𝑟)
(20)
{Ω(𝑟M )−Ω(𝑟M−1 )}𝑧−1 [ℎM {Ω(𝑟M )−Ω(𝑟M−1 )}+𝑧𝑘M | ]
𝜕𝜃1 (𝑟,𝑡) 𝑑𝑟 𝑟M
−𝑘1 | + ℎ1 𝜃1 (𝑟, 𝑡)|𝑟0 = 0 (9)
𝜕𝑟 𝑟0
1
where, z = [0, 1 & ≥ 2] and Ω(𝑟) = 𝑟 or ln 𝑟 or − .
th th 𝑟
At the interface of i and (i+1) layer
Antonopoulos used p = 0, z = 2 and Ω(𝑟) = 𝑟[9]. Moreover,
the above formulation of auxiliary function does not require
𝜃𝑖 (𝑟, 𝑡)|𝑟𝑖 = 𝜃𝑖+1 (𝑟, 𝑡)|𝑟𝑖 𝑖 = 1, … . , M − 1 (10) any further evaluation. Özisik[1] introduced a separate form
𝜕𝜃𝑖 (𝑟,𝑡) 𝜕𝜃𝑖+1 (𝑟,𝑡)
of auxiliary function with is defined in each layer of the
𝑘𝑖 | = 𝑘𝑖+1 | 𝑖 = 1, … . , M − 1 (11) composite and need an extra condition, i.e. Laplacian of
𝜕𝑟 𝑟𝑖 𝜕𝑟 𝑟𝑖
auxiliary function is zero, along with conditions in Eqns.
(14-17). It is required to solve two independent multilayer
At outer surface of Mth layer, i = M

3
steady state problem. The solution procedure involves
inverting a 2M×2M matrix. Since, as noted earlier the 𝑅𝑖,𝑛 (𝑟) = C𝑖,𝑛 S𝑖,𝑛 (𝑟) + D𝑖,𝑛 W𝑖,𝑛 (𝑟) (29)
method presented here does not require these evaluations,
the computational procedure is simplified. Moreover, for the where, C𝑖,𝑛 and D𝑖,n are coefficients of the ith layer for nth
same reasons computational cost is also reduced. eigen-value. The special functions S𝑖,𝑛 (𝑟) and W𝑖,𝑛 (𝑟) are
functions of the ith layer for nth eigen-value, and defined in
Eqns. (30-31) as below,
SOLUTION METHOD
Let, the modified temperature for each layer is defined sin(𝜆𝑖,𝑛 𝑟) for 𝑝=0
by the following way (Principle of Superposition)
S𝑖,𝑛 (𝑟) = J0 (𝜆𝑖,𝑛 𝑟) for 𝑝=1 (30)
𝜃𝑖 (𝑟, 𝑡) = ∑∞𝑛=1 Γ𝑛 (𝑡)𝑅𝑖,𝑛 (𝑟), (21) sin(𝜆𝑖,𝑛𝑟)
for 𝑝=2
{ 𝑟

where, 𝑅𝑖,𝑛 (𝑟) is eigen-function and Γ𝑛 (𝑡) is time

and
component of ith layer for nth eigen-value. The following
conditions, in Eqns. (22-23), must hold as temperatures are
cos(𝜆𝑖,𝑛 𝑟) for 𝑝=0
equal and fluxes are continuous at the interfaces (Eqns. (10-
11)) W𝑖,𝑛 (𝑟) = Y0 (𝜆𝑖,𝑛 𝑟) for 𝑝 = 1. (31)
cos(𝜆𝑖,𝑛 𝑟)
{ for 𝑝=2
Γ𝑖,𝑛 (𝑡) = Γ𝑖+1,𝑛 (𝑡) = Γ𝑛 (𝑡) (22) 𝑟

and The main task here is to find 2M number of coefficients

for each eigen-function using homogeneous BCs. The
𝛼
substitution of 𝑅𝑖,𝑛 (𝑟) in Eqns. (29-31) into homogeneous
𝛼𝑖 𝜆2𝑖,𝑛 = 𝛼1 𝜆21,𝑛 ⇒ 𝜆𝑖,𝑛 = 𝜆1,𝑛 √ 1 . (23) BCs in Eqns. (25-28) lead to a system equations (2M
𝛼𝑖
number of homogeneous BCs) for C𝑖,𝑛 and D𝑖,𝑛 , (i = 1, …,
So, time component of modified temperature independent of M). It should be noted that this system of equations is
layer[6], [7]. homogeneous. Therefore, for a non-trivial solution,
determinant of coefficient matrix has to vanish which leads
Evaluation of Eigen-functions 𝑹𝒊,𝒏 (𝒓) to characteristic equation for the eigen-value. The
characteristic equation has M number of variables and
The eigen-function 𝑅𝑖,𝑛 (𝑟) is found after solving the
converted to single variable equation using relation provided
following eigen-value problem,
in Eqn. (23). This equation is transcendental equation and
1 𝑑 𝑑𝑅𝑖,𝑛 (𝑟) numerically solved for multiple eigen-values (𝜆1,𝑛 ). The
(𝑟 𝑝 ) + 𝜆2𝑖,𝑛 𝑅𝑖,𝑛 (𝑟) = 0 (24) methodology used here to find eigen-values and eigen-
𝑟 𝑝 𝑑𝑟 𝑑𝑟
functions is available in literature[6]–[8].
subject to BCs
At inner surface of the 1st layer, 𝑖 = 1 Evaluation of Time-functions 𝚪𝒏 (𝒕)
After finding eigen-function 𝑅𝑖,𝑛 (𝑟) and corresponding
𝑑𝑅1,𝑛 (𝑟) eigen-value as discussed in previous section, the task is here
−𝑘1 | + ℎ1 𝑅1,𝑛 (𝑟)|𝑟 = 0 (25)
𝑑𝑟 𝑟0 0 to find Γ𝑛 (𝑡). The modified temperature 𝜃𝑖 (𝑟, 𝑡) from Eqn.
(21) is substituted in Eqn. (8) and IC in Eqn. (13), as shown
At the interface of 𝑖 th and (𝑖 + 1)th layer in Eqns. (32-33) below.

𝑅𝑖,𝑛 (𝑟)|𝑟 = 𝑅𝑖+1,𝑛 (𝑟)|𝑟 𝑖 = 1, … . , M − 1 (26) 𝑑Γ𝑛 (𝑡) 𝛼𝑖 𝜕 𝜕𝑞𝑖 (𝑟,𝑡)

𝑖 𝑖 ∑∞
𝑛=1 ( + 𝛼𝑖 𝜆2𝑖,𝑛 Γ𝑛 (𝑡)) 𝑅𝑖,𝑛 (𝑟) = (𝑟 𝑝 )−
𝑑𝑡 𝑟 𝑝 𝜕𝑟 𝜕𝑟
𝜕𝑞𝑖 (𝑟,𝑡) 𝛼𝑖
𝑘𝑖
𝑑𝑅𝑖,𝑛 (𝑟)
| = 𝑘𝑖+1
𝑑𝑅𝑖+1,𝑛(𝑟)
| 𝑖 = 1, … . , M − 1 + 𝑔𝑖 (𝑟, 𝑡) (32)
𝑑𝑟 𝑑𝑟
(27) 𝜕𝑡 𝑘𝑖
𝑟𝑖 𝑟𝑖

and
At outer surface of Mth layer, i = M

𝑑𝑅M,𝑛(𝑟)
∑∞
𝑛=1 Γ𝑛 (0)𝑅𝑖,𝑛 (𝑟) = 𝜃0,𝑖 (𝑟) = 𝑇0,𝑖 (𝑟) − 𝑞𝑖 (𝑟, 0). (33)
𝑘M | + ℎM 𝑅M,𝑛 (𝑟)|𝑟 = 0 (28)
𝑑𝑟 𝑟M M
Now, OEEM method is applied to expand different
The above eigen-value problem has a solution in the quantities in terms of eigen-functions. The expansions of
following form, for ith layer quantities are as follows,

4
Modified Initial Temperature 𝑑Γ𝑛 (𝑡)
+ 𝛼1 𝜆21,𝑛 Γ𝑛 (𝑡) = y𝑛∗ (𝑡) (40)
𝑑𝑡

𝜃0,𝑖 (𝑟) = 𝑇0,𝑖 (𝑟) − 𝑞𝑖 (𝑟, 0) = ∑∞ ∗

𝑛=1 f𝑛 𝑅𝑖,𝑛 (𝑟) (34)
The above equation is an ODE (1st order) with
𝑘𝑖 𝑟𝑖 inhomogeneous part. The ODE is solved with following IC,
∑M
𝑖=1𝛼 ∫𝑟 𝑟 𝑝 [𝑇0,𝑖 (𝑟)−𝑞𝑖 (𝑟,0)]𝑅𝑖,𝑛 (𝑟)𝑑𝑟
from Eqns. (33-34),
where, f𝑛∗ = 𝑖 𝑖−1
𝑁𝑛
Γ𝑛 (0) = f𝑛∗ . (41)
Laplacian of Auxiliary Function
The problem in Eqns. (40-41) has solution of following
𝛼𝑖 𝜕 𝜕𝑞𝑖 (𝑟,𝑡)
(𝑟 𝑝 ) = ∑∞ ∗
𝑛=1 V𝑛 (𝑡)𝑅𝑖,𝑛 (𝑟) (35) form,
𝑟 𝑝 𝜕𝑟 𝜕𝑟
2
∑M
𝑟
𝑖 𝜕 𝜕𝑞 (𝑟,𝑡)
(𝑟 𝑝 𝑖 )𝑅𝑖,𝑛 (𝑟)𝑑𝑟
Γ𝑛 (𝑡) = 𝑦𝑛 (𝑡)𝑒 −𝜶1𝜆1,𝑛 𝑡 (42)
𝑖=1 𝑘𝑖 ∫𝑟 𝑖−1 𝜕𝑟 𝜕𝑟
where, V𝑛∗ (𝑡) = ,
𝑁𝑛 𝑡 2
where 𝑦𝑛 (𝑡) = f𝑛∗ + ∫0 {V𝑛∗ (𝜏) − I𝑛∗ (𝜏) + 𝑔𝑛∗ (𝜏)} 𝑒 𝜶1𝜆1,𝑛 𝜏 𝑑𝜏.
Time Derivative of Auxiliary Function The final temperature obtained combining Eqn. (7), Eqns.
(18-21) and Eqn. (42).
𝜕𝑞𝑖 (𝑟,𝑡)
= ∑∞ ∗
𝑛=1 I𝑛 (𝑡)𝑅𝑖,𝑛 (𝑟) (36)
𝜕𝑡

𝑘𝑖 𝑟𝑖 𝜕𝑞𝑖 (𝑟,𝑡)
ESTIMATION OF ERROR 𝜺𝒊 (𝒓, 𝐖)
∑M
𝑖=1𝛼 ∫𝑟 𝑟𝑝 𝑅𝑖,𝑛(𝑟)𝑑𝑟 The terms required in Eqn. (42) are maximum when t =
𝜕𝑡
where, I𝑛∗ (𝑡) = 𝑖 𝑖−1
𝑁𝑛 0. The error between initial temperature and series
and summation of initial temperature in terms of eigen-functions
decreases with increasing number of terms (n). Though
Heat Source theoretically n goes to infinity, for calculation a finite
number (W) of terms are considered in the eigen-function
𝛼𝑖
𝑔𝑖 (𝑟, 𝑡) = ∑∞ ∗
𝑛=1 𝑔𝑛 (𝑡)𝑅𝑖,𝑛 (𝑟) (37) set. Therefore, the error can be expressed as follows,
𝑘𝑖

𝑟 |𝑇0,𝑖 (𝑟)−𝑞𝑖 (𝑟,0)−∑W

𝑛=1 Γ𝑛 (0)𝑅𝑖,𝑛 (𝑟)|
∑M
𝑖=1 ∫𝑟
𝑖 𝑟 𝑝 𝑔𝑖 (𝑟,𝑡)𝑅𝑖,𝑛 (𝑟)𝑑𝑟 𝜀𝑖 (𝑟, W) = (43)
where, 𝑔𝑛∗ (𝑡) = 𝑖−1
. The coefficient 𝑇0,𝑖 (𝑟)
𝑁𝑛
𝑁𝑛 , is normalisation factor and defined as follows, where, summation over n is taken from 1 to W.
𝑘𝑖 𝑟𝑖
𝑁𝑛 = ∑M
𝑖=1
2
∫ 𝑟 𝑝 𝑅𝑖,𝑛 (𝑟)𝑑𝑟 . (38)
𝛼𝑖 𝑟𝑖−1
RESULTS& DISCUSSION WITH EXAMPLE
The quantity V𝑛∗ (𝑡) in Eqn. (36) or quantity I𝑛∗ (𝑡) in In the result, an example problem has been solved for
Eqn. (37), has only two integral terms in the summation 3-layer (M = 3) medium in Cartesian, cylindrical and
series (first term and last term). The intermediate integral spherical co-ordinate system. The parameters and other
terms are equal to zero. But, in case of Özisik form, each specifications are given in Tab.1. The first task is here to find
integral term in the summetion series for the quantity V𝑛∗ (𝑡) eigen-values and associated eigen-functions.
in Eqn. (36) vanishes[1] and quantity I𝑛∗ (𝑡) in Eqn. (37) has
M integral terms (nonzero) in the summation seris.
Therefore, the formulation used here require less TABLE 1: PARAMETERS AND OTHER
computational cost to evaluate modified heat source. SPECIFICATIONS
i ri ki αi hi T0,i (r) 𝒈𝒊 (𝒓, 𝒕)
[m] [Wm-1K-1] [m2s-1] [Wm-2K-1] [°C] [Wm-3]
The Eqns. (32) can be rewritten by taking expansions
from Eqns. (35-37) 0 0.05 - - - - -
-7
1 0.07 1.20 7.73x10 16 25 0
𝑑Γ𝑛 (𝑡)
∑∞
𝑛=1 ( + 𝛼𝑖 𝜆2𝑖,𝑛 Γ𝑛 (𝑡) − y𝑛∗ (𝑡)) 𝑅𝑖,𝑛 (𝑟) = 0 (39)
𝑑𝑡 2 0.25 0.72 4.46x10-7 - 25 0
-7
3 0.27 1.20 7.73x10 8 25 0
where, y𝑛∗ (𝑡) = V𝑛∗ (𝜏) − I𝑛∗ (𝜏) + 𝑔𝑛∗ (𝜏). Now, terms in Eqn.
(39) are mutually independent and true for every n. So,
every individual term is equal to zero and following
equation can be obtained,

5
 The ambient with time dependent temperatures are
given here in Eqns. (44-45) as follows,

2𝜋(𝑡−54000)
𝑇S1 (𝑡) = 25 + 10 cos [ ] (44)
84600

2𝜋(𝑡−54000)
𝑇SM (𝑡) = 20 + 3 cos [ ]. (45)
84600

The initial temperature is kept same and constant in space

for every layer and that is 𝑇0,𝑖 (𝑟) = 25℃.

To obtain the eigen-values of the system, the

characteristic or transcendental equation is solved using the
Mathematica 9 software. After finding the eigen-values and FIGURE 3. EIGEN-FUNCTION Ri,n = 10 VS. r FOR
setting C1,n = 1, the remaining coefficients in the eigen- SLABS
functions (C𝑖,𝑛 for i = 2, …, M and D𝑖,𝑛 for i = 1, …, M) can
Temperature Profile 𝑻𝒊 (𝒓, 𝒕)
be obtained by any 2M-1 successive equations from the
The estimation of final temperature distribution, within
equation set given by Eqns. (25-28).
the composite medium, depends on the choices of z as
defined in Eqns. (18-20). For the case where z = 0,
It is obvious that (2M-1) (Eqns. (25-28)) boundary /
𝑞1 (𝑟, 𝑡) = 𝑇S1 (𝑡) and 𝑞M (𝑟, 𝑡) = 𝑇SM (𝑡), which makes the
interface conditions are satisfied while obtaining the
coefficients as discussed above. The remaining boundary / modified initial temperature 𝜃0,𝑖 (𝑟) and 𝑞𝑖 (𝑟, 𝑡)
interface condition is satisfied only if eigen-values are discontinuous. As a result, choice z = 0 is unable to estimate
correctly obtained. In the present work, (2M-1) equations temperature distribution properly. For the case where z = 1,
1 𝜕 𝜕𝑞 (𝑟,𝑡)
used are Eqns. (25-27). The residual of Eqn. (28) is plotted the Laplacian of auxiliary function, i.e 𝑝 (𝑟 𝑝 𝑖 )
𝑟 𝜕𝑟 𝜕𝑟
in Fig.2 for different values of 𝜆1,𝑛 to show the accuracy of does not exist at the interfaces. For this choice, the
the obtained eigenvalues. From the very small values of estimation of temperature distribution is inaccurate. Now,
residual the accuracy of the eigenvalues is established. for the cases where z ≥ 2, estimates of temperature
Moreover, for the nth eigen-value, the corresponding eigen- distribution are found to be quite accurate. However, higher
functions cross the horizontal axis (n-1) times. Therefore, by values of z (> 3), makes 𝑞𝑖 (𝑟, 𝑡) complicated function of r.
plotting the eigen-functions in the domain, for each eigen- Therefore, computational cost for the integrals needed in the
value, any missing eigen-value can be detected. For the sake solution procedure is quite high for z > 3, without any
of brevity, the plot for only n = 10 is shown in Fig.3 and it significant increase in accuracy. In view of the above
is seen that it crosses the horizontal axis 9 times. discussion, the optimum value for z is chosen as 2. The
temperature profile within the material is shown in Fig.4.
Here, Fig.4.1 shows temperature at r = 0.05m, 0.06m,
0.07m, 0.16m, 0.25m, 0.26m and 0.27m for 3-layer slabs
with z = 1, 2 and 3. The Fig.4.2 and Fig.4.3 are the
corresponding figures for hollow concentric cylinders and
for hollow concentric spheres, respectively, at the same
1
locations. The choices of function Ω(r) are r, ln r and − for
𝑟
slabs, cylinders and spheres, respectively.

Estimation of Error 𝜺𝒊 (𝒓, W)

The temperature profile at t = 0 requires maximum
number of terms in the series summation. The error between
initial temperature and series summation in Eqn. (21)
decreases with increasing number of terms (n). The error
specified in Eqn. (43) for each layer and for nth terms is
FIGURE 2. RESIDUAL OF Eqn. (28) VS. λ1,n FOR estimated and shown in Fig.5, for p = 0, 1 and 2,
SLABS respectively. The choice is z = 2 for all these three cases. The
errors are shown in Fig.5 at r = 0.06m, 0.16m and 0.26m
(midpoint of each layer). The error is less than 5% for all
cases when n ≥ 5.

6
FIGURE 4. TEMPERATURE PROFILE OF (1. a-c) SLABS FOR Ω(r) = r, (2. a-c) HOLLOW CYLINDERS FOR Ω(r) = ln
𝟏
r AND (3. a-c) HOLLOW SPHERES FOR Ω(r) = −
𝒓

FIGURE 5. εi (r, W) VS. n FOR (a) SLABS, (b) HOLLOW CYLINDERS AND (c) HOLLOW SPHERES

7
 CONCLUIONS [4] Y. Yener, M. N. Özişik, and M. N. Ozisik, “On the
The formulation of auxiliary function for external Solution of Unsteady Heat Conduction in Multi-
layers by Antonopoulos and Tzivanidis[9] is proposed in region Media with Time-dependent Heat Transfer
generalized form. The OEEM is applied successfully to Coefficient,” in Proceedings of the 5th International
estimate temperature distribution within an M-layer Heat Transfer Conference, 1974, pp. 188–192.
composite medium. The advantage of this formulation is
that the auxiliary function need to be defined only in the two [5] S. Singh, P. K. Jain, and Rizwan-uddin, “Finite
external layers. There are two important factors which are Integral Transform Method to Solve Asymmetric
playing important role in the computational cost required to Heat Conduction in a Multilayer Annulus With
obtain solutions. First, the auxiliary function need not be Time-Dependent Boundary Conditions,” Nucl. Eng.
evaluated for this formulation. Second, the term in modified Des., vol. 241, no. 1, pp. 144–154, Jan. 2011.
source, which is derived from auxiliary function, has only
two integrals in the summation series compared to Özisik
[6] S. Singh, P. K. Jain, and Rizwan-uddin, “Analytical
form where number of integrals is obviously equal to Solution to Transient Heat Conduction in Polar
number of layers. These two important factors lead to Coordinates with Multiple Layers in Radial
significant reduction in the complexity of the code and also
Direction,” Int. J. Therm. Sci., vol. 47, no. 3, pp.
reduces the overall computational cost. The error between
261–273, 2008.
initial temperature and estimated temperature at t = 0
decreases with increasing number of terms in the summation
series. The error for all cases, for example problem solved [7] P. K. Jain, S. Singh, and Rizwan-uddin, “Analytical
in this work, is less than 5% for n ≥ 5. Solution to Transient Asymmetric Heat Conduction
in a Multilayer Annulus,” J. Heat Transfer, vol. 131,
no. 1, pp. 0113040–0113047, 2009.
REFERENCES
[8] P. K. Jain, S. Singh, and Rizwan-uddin, “An Exact
[1] M. N. Özisik, Heat Conduction, 2nd ed. John Wiley Analytical Solution for Two-Dimensional,
& Sons, Inc., 1993. Unsteady, Multilayer Heat Conduction in Spherical
Coordinates,” Int. J. Heat Mass Transf., vol. 53, no.
9–10, pp. 2133–2142, Apr. 2010.
[2] H. Salt, “Transient Conduction in a Two-
Dimensional Composite Slab—I. Theoretical
Development of Temperature Modes,” Int. J. Heat [9] K. A. Antonopoulos and C. Tzivanidis, “Analytical
Mass Transf., vol. 26, no. 11, pp. 1611–1616, Nov. Solution of Boundary Value Problems of Heat
1983. Conduction in Composite Regions with Arbitrary
Convection Coundary Conditions,” Acta Mech., vol.
118, pp. 65–78, 1996.
[3] H. Salt, “Transient Conduction in a Two-
Dimensional Composite Slab—II. Physical
Interpretation of Temperature Modes,” Int. J. Heat
Mass Transf., vol. 26, no. 11, pp. 1617–1623, Nov.
1983.

View publication stats