You are on page 1of 11

Nuclear Engineering and Design 241 (2011) 144–154

Contents lists available at ScienceDirect

Nuclear Engineering and Design


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

Finite integral transform method to solve asymmetric heat conduction in a


multilayer annulus with time-dependent boundary conditions
Suneet Singh a , Prashant K. Jain b,∗ , Rizwan-uddin c
a
Energy Science and Engineering, Indian Institute of Technology (IIT) Bombay, Mumbai, India
b
Reactor and Nuclear Systems Division, Oak Ridge National Laboratory, 1 Bethel Valley Road, Oak Ridge, TN, USA
c
Nuclear, Plasma and Radiological Engineering, University of Illinois at Urbana–Champaign, USA

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

Article history: The separation of variables (SOV) method has recently been applied to solve time-dependent heat conduc-
Received 21 October 2008 tion problem in a multilayer annulus. It is observed that the transverse (radial) eigenvalues for the solution
Received in revised form 16 August 2010 in polar (r-) coordinate system are always real numbers (unlike in the case of similar multidimensional
Accepted 18 October 2010
Cartesian problems where they may be imaginary) allowing one to obtain the solution analytically. How-
ever, the SOV method cannot be applied when the boundary conditions and/or the source terms are
time-dependent, for example, in a nuclear fuel rod subjected to time-dependent boundaries or heat
sources. In this paper, we present an alternative approach using the finite integral transform method to
solve the asymmetric heat conduction problem in a multilayer annulus with time-dependent boundary
conditions and/or heat sources. An eigenfunction expansion approach satisfying periodic boundary con-
dition in the angular direction is used. After integral transformation and subsequent weighted summation
over the radial layers, partial derivative with respect to r-variable is eliminated and, first order ordinary
differential equations (ODEs) are formed for the transformed temperatures. Solutions of ODEs are then
inverted to obtain the temperature distribution in each layer. Since the proposed solution requires the
same eigenfunctions as those in the similar problem with time-independent sources and/or boundary
conditions—a problem solved using the SOV method—it is also “free” from imaginary eigenvalues.
© 2010 Elsevier B.V. All rights reserved.

1. Introduction

In modern engineering applications, multilayer components are extensively used due to the added advantage of combining physical,
mechanical and thermal properties of different materials. Many of these applications (for example, in various automotive, space, chemical,
civil and nuclear industries) require a detailed knowledge of transient temperature and heat-flux distribution within the component layers.
Time-dependent temperature distribution in such components, with the presence of sources (with all three types of boundary conditions)
may either be obtained using analytical or numerical methods. Nonetheless, numerical solutions are preferred and prevalent in practice,
due to either unavailability or higher computational complexity of the corresponding analytical (or exact) solutions. Exact solutions, when
available, are however very useful in analyzing governing physics of the problem and thus, are more insightful compared to the numerical
solutions. Nowadays, such solutions also find their applications in validating and comparing various numerical algorithms, which may
help improve computational efficiency of computer codes that currently rely on numerical techniques.
Analytical solutions for 1D time-dependent multilayer heat conduction problems were developed several decades ago. Mathematical
theory for such problems in more than one dimension was also developed during the same time (Ozisik, 1993; Salt, 1983a,b; Mikhailov
and Ozisik, 1986; Yener and Ozisik, 1974). Several of these approaches were based on the separation of variables (SOV) and the finite integral
transform (FIT) methods. However, application of these approaches in producing exact solutions was hindered due to the difficulty of solving
the corresponding eigenvalue problems (which are an essential part of such solution methodologies). Furthermore, in 2D and 3D Cartesian
coordinates, these eigenvalues can be imaginary, rendering the solutions even more difficult. It has recently been shown that similar
problems in 2D cylindrical (r,), 2D spherical (r,) and 3D spherical (r, , ) coordinate systems do not possess imaginary eigenvalues (Singh
et al., 2008; Jain et al., 2009, 2010) and therefore, exact solutions can be obtained using the softwares capable of symbolic manipulations
(for example, MapleTM (Maple, 2010), Mathematica (Mathematica, 2010) etc.).

∗ Corresponding author. Tel.: +1 865 574 6272; fax: +1 865 574 2032.
E-mail address: jainpk@ornl.gov (P.K. Jain).

0029-5493/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.nucengdes.2010.10.010
S. Singh et al. / Nuclear Engineering and Design 241 (2011) 144–154 145

Nomenclature

aimp , bimp coefficients of Bessel functions in transverse (radial) eigenfunction (see Eq. (29))
Ain , Bin , Cin coefficients in inner surface boundary condition (see Eqs. (2), (18) and (25))
Aout , Bout , Cout coefficients in outer surface boundary condition (see Eqs. (3), (19) and (26))
fi (r, ) initial temperature distribution in the ith layer at t = 0
gi (r, , t) time-dependent volumetric heat source distribution in the ith layer
Jm Bessel function of the first kind of order m
ki thermal conductivity of the ith layer
M number of angular eigenfunction used in the solution
Nmp norm for r-direction (see Eq. (30))
P number of radial eigenfunction used in the solution corresponding to each angular eigenvalue
q heat-flux
q0 magnitude of heat flux in the illustrative example (see Eq. (48))
r radial coordinate
ri outer radius for the ith layer
Rimp (imp r) transverse eigenfunctions for the ith layer
t time
Ti (r, , t) temperature distribution for the ith layer
Ym Bessel function of the second kind of order m

Greek symbols
˛i thermal diffusivity of the ith layer
 angular coordinate
imp transverse (radial) eigenvalues
 angle subtended by the multilayers

Abbreviations
1D, 2D, 3D one-dimensional, two-dimensional, three-dimensional
FIT finite integral transform method
ODEs ordinary differential equations
RHS right hand side
SOV separation of variables method

Subscripts and superscripts


i layer or interface number
n outermost layer number

Recent advances in computational resources for symbolic manipulations have created renewed interest among researchers (Singh
et al., 2008; Jain et al., 2009, 2010; Lu and Viljanen, 2006; de Monte, 2006, 2000, 2003, 2002; Lu et al., 2006a,b, 2005; Haji-Sheikh
and Beck, 2002) in developing exact analytical solutions of problems for which numerical solutions are currently more prevalent.
Although multilayer heat conduction problems have been studied in great detail and various solution methods have been devel-
oped in the past (for review, see Ozisik, 1993), there is a continued need to explore and develop novel methods to solve problems
for which exact solutions are still in infancy. One such problem is to determine the unsteady temperature distribution in cylindri-
cal polar coordinates (r,) with multiple layers in the radial direction exposed to time-dependent boundary conditions and/or heat
sources.
Salt (1983a,b) solved the time-dependent heat conduction problem by applying an orthogonal expansion technique in a 2D composite
slab with no internal heat source, and homogenous boundary conditions. Mikhailov and Ozisik (1986) solved an analogous 3D problem in
a Cartesian nonhomogenous finite medium. Recently, Lu et al. (Lu and Viljanen, 2006; Lu et al., 2006a, 2005) combined a Laplace Transform
approach and the separation of variables method for a rectangular and cylindrical multilayer slab with time-dependent periodic boundary
conditions. Treatment in the cylindrical coordinates is however restricted to the r − z coordinates only. Haji-Sheikh and Beck (2002) applied
a Green’s function approach to solve a 3D Cartesian two-layer orthotropic heat conduction problem. They also accounted for the effects of
contact resistance in their solution. Eigenfunction expansion method is applied by de Monte (2003) to solve the unsteady heat conduction
problem in a 2D two-layer isotropic slab with homogenous boundary conditions.
An exact solution based on the separation of variables (SOV) method is recently developed by Singh et al. (2008) for multilayer heat
conduction in polar coordinates. However, that exact solution is applicable only to the domains with pie slice geometry ( < 2, where  is
the azimuthal angle subtended by the layers) and time-independent boundary conditions. Recently, the same approach, based on the SOV
method, is extended to determine the temperature distribution in complete disc-type (i.e.,  = 2) layers by Jain et al. (2009). However, due
to the limitation of the SOV method, it cannot be conveniently extended to include the effects of time-dependent boundary conditions and/or
sources. One typical example of such a problem is a nuclear fuel rod, which consists of concentric layers of different materials and often
subjected to asymmetric time-dependent boundary conditions. Moreover, numerous other applications including multilayer insulation
materials, several cryogenic systems, tall buildings and other cylindrical structures may be subjected to time-dependent sources and/or
boundary conditions, and may benefit from having an exact solution. This paper presents an approach based on the finite integral transform
(FIT) method to solve for such temperature distributions. Solution is valid for any combinations of time-dependent, inhomogeneous
146 S. Singh et al. / Nuclear Engineering and Design 241 (2011) 144–154

Fig. 1. Schematic representation of an n-layer annulus.

boundary conditions at inner and outer radii of the domain. Results for an illustrative problem involving a three-layer annulus subjected
to asymmetric, time-dependent heat-flux are also presented.
It should be noted that the generalized integral-transform method for solving heat-conduction in arbitrary finite composite regions is
developed by Yener and Ozisik (Ozisik, 1993; Yener and Ozisik, 1974). However, reducing the scheme to closed-form solutions for specific
geometry and coordinate system is not trivial which is particularly true in the context of cylindrical polar coordinates. However, recent
progress in computer based symbolic manipulations (Maple, 2010; Mathematica, 2010) and increase in computational power have now
made it possible to override such computational difficulties, and yield closed-form analytical solutions (Singh et al., 2008; Jain et al., 2009,
2010).

2. Mathematical formulation

Consider an n-layer annulus (r0 ≤ r ≤ rn ) as shown schematically in Fig. 1. All the layers are assumed to be isotropic in thermal properties
and are in perfect thermal contact. Let ki and ˛i be the temperature independent thermal conductivity and thermal diffusivity of the ith
layer. At t = 0, each ith layer is at a specified temperature fi (r, ) and time-dependent heat sources gi (r, , t) are switched on for t > 0. Both,
the inner (i = 1, r = r0 ) as well as the outer (i = n, r = rn ) surfaces of the annulus may be subjected to any combination of time-dependent
boundary conditions of the first, second or the third kind.
Under these assumptions, the governing heat conduction equation along with the boundary and initial conditions, are as follows:
Governing equation:
 
1 ∂Ti 1 ∂ ∂Ti 1 ∂2 Ti g (r, , t)
(r, , t) = r (r, , t) + (r, , t) + i , ri−1 ≤ r ≤ ri , 1 ≤ i ≤ n (1)
˛i ∂t r ∂r ∂r r 2 ∂ 2 ki

Boundary conditions:

• Inner surface of the 1st layer (i = 1)

∂T1
Ain (r0 , , t) + Bin T1 (r0 , , t) = Cin (, t) (2)
∂r
• Outer surface of the nth layer (i = n)

∂Tn
Aout (rn , , t) + Bout Tn (rn , , t) = Cout (, t) (3)
∂r
• Periodic boundary conditions (i = 1, 2, ..., n)

Ti (r,  = 0, t) = Ti (r,  = 2, t) (4)

∂Ti ∂Ti
(r,  = 0, t) = (r,  = 2, t) (5)
∂ ∂
• Interface of the ith layer (i = 2, ..., n)

Ti (ri−1 , , t) = Ti−1 (ri−1 , , t) (6)

∂Ti ∂T
ki (ri−1 , , t) = ki−1 i−1 (ri−1 , , t) (7)
∂r ∂r
S. Singh et al. / Nuclear Engineering and Design 241 (2011) 144–154 147

Initial condition:

Ti (r, , t = 0) = fi (r, ). (8)

Boundary conditions either of the first, second or third kind may be imposed at r = r0 and r = rn by choosing the appropriate coefficients in
Eqs. (2) and (3). However, the case in which Bin and Bout are simultaneously zero is not considered. In addition, asymmetric, time-dependent
boundary conditions can be applied by choosing - and t-dependent Cin and Cout . Furthermore, temperature distribution in multiple layers
with zero inner radius (r0 = 0) can be solved by assigning zero values to constants Bin and Cin in Eq. (2).

3. Eigenfunction expansion in the Â-direction

Due to periodic boundary conditions in the -direction, Ti (r, , t) can be expanded into angular eigenfunctions (cos(m), sin(m) and a
constant) as follows:




  
Ti (r, , t) = Ti0 (r, t) + Timc (r, t) cos(m) + Tims (r, t) sin(m). (9)
m=1 m=1

Similarly, specified expressions for the sources, boundary conditions, and initial condition, gi (r, , t), Cin (, t), Cout (, t) and fi (r, ), are also
expanded as:




  
gi (r, , t) = gi0 (r, t) + gimc (r, t) cos(m) + gims (r, t) sin(m) (10)
m=1 m=1


∞ 

  
Cin (, t) = Cin,0 (t) + Cin,mc (t) cos(m) + Cin,ms (t) sin(m) (11)
m=1 m=1


∞ 

  
Cout (, t) = Cout,0 (t) + Cout,mc (t) cos(m) + Cout,ms (t) sin(m) (12)
m=1 m=1





  
fi (r, ) = fi0 (r) + fimc (r) cos(m) + fims (r) sin(m). (13)
m=1 m=1

Coefficients in Eqs. (10)–(13) can be evaluated by applying the orthogonality conditions in the -direction. For example, in Eq. (10):
 2
 1
gi0 (r, t) = gi (r, , t)d (14)
2 0
 2
 1
gimc (r, t) = gi (r, , t) cos(m)d (15)
 0
 2
 1
gims (r, t) = gi (r, , t) sin(m)d. (16)
 0

Substituting Eqs. (9)–(13) in Eqs. (1)–(8) yields the following set of equations for coefficients in Eq. (9) for each positive integer m [Note that
in the formulation that follows, subscripts 0, c or s are omitted for clarity because coefficients with these different subscripts are essentially
governed by the same set of equations as shown below (see Eqs. (17)–(22)). Moreover, in the equations for coefficient with subscript 0,
m = 0 is taken.]:

 
 
1 ∂Tim 1 ∂ ∂Tim m2  gim (r, t)
(r, t) = r (r, t) − T (r, t) + , ri−1 ≤ r ≤ ri , 1 ≤ i ≤ n (17)
˛i ∂t r ∂r ∂r r 2 im ki

Boundary conditions:

• Inner surface of the 1st layer (i = 1)



∂T1m  
Ain (r0 , t) + Bin T1m (r0 , t) = Cin,m (t) (18)
∂r
• Outer surface of the nth layer (i = n)

∂Tnm  
Aout (rn , t) + Bout Tnm (rn , t) = Cout,m (t) (19)
∂r
• Interface between ith and (i − 1)st layer (i = 2, ..., n)
 
Tim (ri−1 , t) = Ti−1,m (ri−1 , t) (20)

 
∂Tim ∂Ti−1,m
ki (ri−1 , t) = ki−1 (ri−1 , t) (21)
∂r ∂r
148 S. Singh et al. / Nuclear Engineering and Design 241 (2011) 144–154

• Initial condition:
 
Tim (r, t = 0) = fim (r). (22)

4. Finite integral transform in the r-direction


 ri
Operating Eq. (17) by ri−1
rRim (r)dr and using integration of parts twice on the first term on the right hand side (RHS), one gets (see
Appendix A for details):
 ri  
  ri   dR  
1 ∂Tim 1 d im m2 
(r, t) rRim (r)dr = r (r) − 2 Rim (r) rTim (r, t)dr
ri−1
˛i ∂t ri−1
r dr dr r


ri  ri 
∂Tim  ∂Rim gim (r, t)
+ rRim (r) (r, t) − rTim (r) (r) + rRim (r) dr. (23)
∂r ∂r ri−1
ki
ri−1

4.1. Eigenvalue problem in the r-direction

Rim (r) in Eq. (23) is chosen so as to satisfy,

1 d
 dR   m2 
im
r (r) + − + 2im Rim (r) = 0 (24)
r dr dr r2

and boundary conditions:

• Inner surface of the 1st layer (i = 1)

dR1m
Ain (r = r0 ) + Bin R1m (r = r0 ) = 0 (25)
dr
• Outer surface of the nth layer (i = n)

dRnm
Aout (r = rn ) + Bout Rnm (r = rn ) = 0 (26)
dr
• Interface between ith and (i − 1)st layer (i = 2, ..., n)

Rim (ri−1 ) = Ri−1,m (ri−1 ) (27)

∂Rim ∂Ri−1,m
ki (ri−1 ) = ki−1 (ri−1 ). (28)
∂r ∂r

Solutions of the above equations are eigenfunctions Rimp (r) corresponding to the eigenvalues imp , and given by:

Rimp (imp r) = aimp Jm (imp r) + bimp Ym (imp r). (29)

The eigenfunctions imp satisfy the following orthogonality condition:


n  ri
k i 0 if p =
/ q
rRimp (imp r)Rimq (imq r)dr = . (30)
˛i Nmp if p = q
ri−1
i=1

Note that the above condition is satisfied if and only if (Singh et al., 2008):

˛i 2imp = ˛1 21mp . (31)

Application of the boundary conditions (Eqs. (25) and (26)) and interface conditions (Eqs. (27) and (28)) to the transverse eigenfunction
Rimp (imp r) yields a (2n × 2n) matrix for each integer value of m. Eigencondition for the transverse direction is obtained by setting the
determinant of this matrix equal to zero. Roots of which, in turn, yield the infinite number of eigenvalues 1mp corresponding to the first
layer for each integer value of m (similar to Singh et al., 2008).
It has been shown earlier by Singh et al. (2008) and Jain et al. (2009) that in polar coordinates, dependence of the eigenvalues in the
transverse direction on those in the other direction is not explicit. Absence of explicit dependence leads to a complete solution which does
not have imaginary transverse eigenvalues, and thus imp are real.

4.2. Formulation of the time-dependent ODEs and their solution

In view of Eq. (24), Eq. (23) can be written as,


 ri  
 

ri  ri 
1 ∂Tim ∂Tim ∂Rimp gim (r, t)
(r, t) + 2imp Tim

(r, t) rRimp (r)dr = rRimp (r) 
(r, t) − rTim (r) (r) + rRimp (r) dr. (32)
ri−1
˛i ∂t ∂r ∂r ri−1
ki
ri−1
S. Singh et al. / Nuclear Engineering and Design 241 (2011) 144–154 149

Now using Eq. (31), one gets:


 ri  
 

ri  ri 
1 ∂Tim ∂Tim ∂Rimp gim (r, t)
(r, t) + ˛1 21mp Tim

(r, t) rRimp (r)dr = rRimp (r) 
(r, t) − rTim (r) (r) + rRimp (r) dr. (33)
˛i ri−1
∂t ∂r ∂r ri−1
ki
ri−1

Multiplying the above equation by ki and summing over all the n layers, one gets,

n  ri  
 
n 

ri
k i
∂Tim ∂Tim ∂Rimp
(r, t) + ˛1 21mp Tim

(r, t) rRimp (r)dr = ki rRimp (r) 
(r, t) − rTim (r) (r)
˛i ri−1
∂t ∂r ∂r
i=1 i=1 ri−1

n 
 ri

+ rRimp (r)gim (r, t)dr. (34)
ri−1
i=1

Now defining,

n  ri
r k i 
Tmp (t) ≡ rRimp (r)Tim (r, t)dr (35)
˛i ri−1
i=1

and,
n 
 ri
r 
gmp (t) ≡ rRimp (r)gim (r, t)dr (36)
ri−1
i=1

Eq. (34) reduces to:


r 
n 

ri
dTmp ∂Tim ∂Rimp
(t) + ˛1 21mp Tmp
r
(t) = ki rRimp (r) 
(r, t) − rTim (r, t) (r) r
+ gmp (t). (37)
dt ∂r ∂r
i=1 ri−1

Now application of the interface conditions, Eqs. (20) and (21) and (27) and (28) yields (see Appendix B for details):
r
Cout,mp (t)


r
dTmp ∂T  dRnmp
(t) + ˛1 21mp Tmp
r
(t) = kn rn Rnmp (rn ) nm (rn , t) − Tnm

(rn ) (rn )
dt ∂r dr


∂T1m  dR1mp r
− k1 r0 R1mp (r0 ) (r0 , t) − T1m (r0 ) (r0 ) + gmp (t) (38)
∂r dr

C r (t)
in,mp

r r
Defining Cout,mp (t) and Cin,mp (t) as the first and second square bracket on the RHS of the above equation, one gets:

r
dTmp
(t) + ˛1 21mp Tmp
r r
(t) = kn rn Cout,mp r
(t) − k1 r0 Cin,mp r
(t) + gmp (t), (39)
dt
r
where Cin,mp (t) can be evaluated from Eqs. (18) and (25) as,
⎡ 
Cin,m (t)R1mp (r0 )
, if Ain =
/ 0
⎢ Ain
r
Cin,mp (t) = ⎣  (40)
Cin,m (t)((dR1mp )/(dr))(r0 )
− , if Ain = 0
Bin
r
and similarly, Cout,mp (t) can be evaluated from Eqs. (19) and (26) as,
⎡ 
Cout,m (t)Rnmp (rn )
⎢ , if Aout =
/ 0
r
Cout,mp (t) = ⎣ Aout . (41)

Cout,m (t)((dRnmp )/(dr))(rn )
− , if Aout = 0
Bout
Solution of Eq. (39) is:
 t
r r −˛1 2 t −˛1 2 t r r r ˛1 2 
Tmp (t) = Tmp (0)e 1mp +e 1mp (kn rn Cout,mp () − k1 r0 Cin,mp () + gmp ())e 1mp d, (42)
0
r (0) is evaluated using the initial condition (Eq. (22)) and the definition in Eq. (35) as:
where Tmp


n  ri
r k i 
Tmp (t = 0) = rRimp (r)fim (r)dr. (43)
˛i ri−1
i=1
150 S. Singh et al. / Nuclear Engineering and Design 241 (2011) 144–154

Fig. 2. Asymmetric heat conduction in a three-layer annulus. Each layer has a different thermal conductivity (ki ) and thermal diffusivity (˛i ). The lower-half of the annulus
( ≤  ≤ 2) is kept insulated, while the upper-half (0 ≤  ≤ ) is subjected to a (, t)-dependent incoming heat-flux.

5. Inversion formula


Tim (r, t) can be expanded in the generalized Fourier series as follows:




Tim (r, t) = cmp (t)Rimp (r). (44)
p=1

The time-dependent coefficients are found by applying the orthogonality condition (Eq. (30)),


n
ki
 ri 
˛i ri−1
rRimp (r)Tim (r, t)dr
i=1
cmp (t) = . (45)
Nmp

Now using the definition in Eq. (35),


r (t)
Tmp
cmp (t) = . (46)
Nmp
  
As stated before, the above formulation and solution is valid for Ti0 (r, t), Timc (r, t) and Tims (r, t) in Eq. (9). Substituting Eq. (44) with
appropriate subscripts (0, c or s) in Eq. (9), one gets:


∞ r
T0p (t)   Tmpc
∞ ∞ r (t)   Tmps
∞ ∞ r (t)
Ti (r, , t) = Ri0p (r) + Rimp (r) cos(m) + Rimp (r) sin(m). (47)
N0p Nmp Nmp
p=1 m=1 p=1 m=1 p=1

6. Illustrative example and results

A three-layer annulus (r0 ≤ r ≤ r3 , 0 ≤  ≤ 2; see Fig. 2) is initially at a uniform zero temperature. For time t > 0, following - and
t-dependent heat-flux,
2
 q0  2 ( − ) f (t), 0≤≤
q (r = r3 , , t) = (48)
0,  ≤  ≤ 2

is applied at the outer surface (r = r3 ) while the inner surface (r = r0 ) is maintained isothermal at zero temperature. The details of f(t) are
given later in this section. This leads to the coefficients Ain = 0, Bin = 1, Aout = k3 , Bout = 0, Cin (, t) = 0 and Cout (, t) = q (r3 , , t) in the respective
boundary condition equations. There is no volumetric heat generation in any of the layers, i.e., gi (r, , t) = 0.
Parameter values used in this problem are, k2 /k1 = 2, k3 /k1 = 4; ˛2 /˛1 = 4, ˛3 /˛1 = 9; r1 /r0 = 2, r2 /r0 = 4, r3 /r0 = 6.
It should be noted that, in the results that follow, r, t, and Ti (r, , t) are in the units of r0 , r02 /˛1 and, q0 r0 /k1 , respectively.
S. Singh et al. / Nuclear Engineering and Design 241 (2011) 144–154 151

Fig. 3. A comparison of temporal and steady state radial temperature distributions at different angular positions: (a)  = 0, (b)  = /4, (c)  = /2, (d)  = 3/2, (e)  = 7/4,
with the results reported in Jain et al. (2009). Lines represent data from Jain et al. (2009) and circles are from data obtained in current work.
152 S. Singh et al. / Nuclear Engineering and Design 241 (2011) 144–154

Fig. 4. Temporal variation of temperature at  = /2 and mid-point of different layers. Non-oscillating plots are for f(t) = 1 and oscillating ones are for f(t) = sin(t).

For this particular problem, the infinite series solution for the temperature Ti (r, , t) is truncated at p = P and m = M, leading to,


P r
T0p (t) 
M 
P r (t)
Tmpc 
M 
P r (t)
Tmps
Ti (r, , t) = Ri0p (r) + Rimp (r) cos(m) + Rimp (r) sin(m) + εi (r, , t; M, P) (49)
N0p Nmp Nmp
p=1 m=1 p=1 m=1 p=1

where εi (r, , t ; M, P) is the truncation error. Both, M and P are chosen to be 15 for the results shown in this paper. [The change in the
values of temperature for (M = 10 and P = 10) and those obtained with (M = 15 and P = 15) is of the order of 10−4 . Since, series is sufficiently
converged; we have used 15 terms for both M and P.]
The analysis is carried out for the following two cases:

Case 1: f(t) = 1
Case 2: f(t) = sin(t).

In the Case 1, boundary condition is time-independent. However, solution of the problem is obviously time-dependent as temperature
increases with time from its initial value to finally attain a steady state temperature distribution. This problem can be solved by the separation
of variables (SOV) method as well as the finite integral transform (FIT) method discussed in this paper. Therefore, solution obtained from
these two methods are compared to show the accuracy and correctness of the method developed in the current work. Radial temperature
distributions at different angular positions and for different values of t obtained from applying the SOV as well as the FIT method are shown
in Fig. 3. Results obtained using the FIT method based time-dependent formulation are in excellent agreement with those obtained earlier
using the SOV method for the time-independent boundary conditions. It should be noted that correctness and accuracy of the SOV method
have already been established by Jain et al. (2009) for the same problem.
Solution for the Case 2 is used to show the application of the methodology for time dependent boundary condition using the FIT method.
Note that it is not possible to solve this problem using the SOV method. Temporal variations of temperature for the Case 2, at  = /2 and
mid-point of different layers (r = 1.5, 3.0 and 5.0) are shown in Fig. 4. Solution for the case 1 (at the same points) is also plotted in the same
graph. As expected, the solution for f(t) = sin(t) oscillates about the solution for f(t) = 1. Frequency of oscillations at these three different
points are the same, however, there is a phase difference due to diffusion lag.

7. Conclusions

In this paper, an analytical solution is presented for the asymmetric transient heat conduction in a multilayer annulus subjected to time-
dependent boundary conditions. Each layer can have spatially varying as well as time-dependent volumetric heat sources. Inhomogeneous
boundary condition of the first, second or third kind can be applied in the radial direction. Proposed solution is also applicable to the
layered-structures with zero inner radius (r0 = 0). In polar coordinates, dependence of the eigenvalues in the transverse direction on those
in the other direction is not explicit. Absence of explicit dependence leads to a complete solution which does not have imaginary transverse
eigenvalues. Numerical evaluation of the series solution shows that a reasonable number of terms are sufficient to obtain results with
acceptable errors for engineering applications.
S. Singh et al. / Nuclear Engineering and Design 241 (2011) 144–154 153

Appendix A.
 ri
Operating Eq. (17) by ri−1
rRim (r)dr leads to:

 ri  
  ri  
  ri    ri 
1 ∂Tim d dTim m2 
gim (r, t)
(r, t) rRim (r)dr = r (r) Rim (r, t)dr − Rim (r) rTim (r, t)dr + rRim (r) dr. (A1)
ri−1
˛i ∂t ri−1
dr dr ri−1
r2 ri−1
ki

Term A

Using integration by parts on Term A in the above equation gives:




ri  ri 
dTim dTim dRim (r)
Term A = r (r, t)Rim (r) − r (r, t) dr (A2)
dr ri−1
dr dr
ri−1

Term B

Now, using integration by parts on Term B, we get:




ri  ri  ri  dR 
dTim  dRim d im 
Term A = r (r, t)Rim (r) − rTim (r) (r) + r (r) Tim (r, t)dr . (A3)
dr dr ri−1 ri−1
dr dr
ri−1

Substitution of Eq. (A3) into Eq. (A1) yields Eq. (23).

Appendix B.

Consider the first term on the RHS of Eq. (37), which is as follows:


n 

ri
∂Tim 
∂Rimp
ki rRimp (r) (r, t) − rTim (r, t) (r) . (B1)
∂r ∂r
i=1 ri−1

Evaluating the above expression at ri and ri−1 , one gets:


n 


n 

∂Tim 
∂Rimp ∂Tim 
∂Rimp
ki ri Rimp (ri ) (ri , t) − ri Tim (ri ) (ri ) − ki ri−1 Rimp (ri−1 ) (ri−1 , t) − ri−1 Tim (ri−1 ) (ri−1 ) . (B2)
∂r ∂r ∂r ∂r
i=1 i=1

Separating out the last term in the first summation and first term in the second summation yields:


n−1 

∂Tim ∂Rimp ∂T  dRnmp


ki ri Rimp (ri ) 
(ri , t) − ri Tim (ri ) (ri ) + kn rn Rnmp (rn ) nm (rn , t) − Tnm

(rn ) (rn )
∂r ∂r ∂r dr
i=1



n 

∂T1m  dR1mp ∂Tim 


∂Rimp
− k1 r0 R1mp (r0 ) (r0 , t) − T1m (r0 ) (r0 ) − ki ri−1 Rimp (ri−1 ) (ri−1 , t) − ri−1 Tim (ri−1 ) (ri−1 ) . (B3)
∂r dr ∂r ∂r
i=2

Appropriately changing the indexing and range of first summation in the above expression, leads to:
 


n 
∂Ti−1,m ∂Ri−1,mp ∂T  dRnmp
ki−1 ri−1 Ri−1,mp (ri−1 ) 
(ri−1 , t) − ri−1 Ti−1,m (ri−1, ) (ri−1 ) + kn rn Rnmp (rn ) nm (rn , t) − Tnm

(rn ) (rn )
∂r ∂r ∂r dr
i=2



n 

∂T1m  dR1mp ∂Tim 


∂Rimp
− k1 r0 R1mp (r0 ) (r0 , t) − T1m (r0 ) (r0 ) − ki ri−1 Rimp (ri−1 ) (ri−1 , t) − ri−1 Tim (ri−1 ) (ri−1 ) . (B4)
∂r dr ∂r ∂r
i=2

Notice the same range of both the summations in the above expression. Combining the two summations together, one gets:
    

n 
∂Ti−1,m ∂Ri−1,mp 
∂Tim ∂Rimp
 
ki−1 ri−1 Ri−1,mp (ri−1 ) (ri−1 , t) − ri−1 Ti−1,m (ri−1, ) (ri−1 ) −ki ri−1 Rimp (ri−1 ) (ri−1 , t) − ri−1 Tim (ri−1 ) (ri−1 )
∂r ∂r ∂r ∂r
i=2


∂Tnm dRnmp ∂T  dR1mp
+ kn rn Rnmp (rn ) 
(rn , t) − Tnm (rn ) (rn ) − k1 r0 R1mp (r0 ) 1m (r0 , t) − T1m

(r0 ) (r0 ) . (B5)
∂r dr ∂r dr

From the interface conditions (Eqs. (20), (21), (27) and (28)), it can be immediately seen that following holds for i = 2, ..., n:
   
 
∂Ti−1,m ∂Ri−1,mp ∂Tim ∂Rimp
 
ki−1 ri−1 Ri−1,mp (ri−1 ) (ri−1 , t) − ri−1 Ti−1,m (ri−1, ) (ri−1 ) = ki ri−1 Rimp (ri−1 ) (ri−1 , t) − ri−1 Tim (ri−1 ) (ri−1 ) .
∂r ∂r ∂r ∂r
(B6)
154 S. Singh et al. / Nuclear Engineering and Design 241 (2011) 144–154

Therefore, each term in the summation in (B5) is zero. Hence, expression in (B5) reduces to:

∂T  dRnmp ∂T  dR1mp
kn rn Rnmp (rn ) nm (rn , t) − Tnm

(rn ) (rn ) − k1 r0 R1mp (r0 ) 1m (r0 , t) − T1m

(r0 ) (r0 ) (B7)
∂r dr ∂r dr

The above discussion shows that the expression in (B1) and (B7) are equal. Hence, one can get Eq. (38) from Eq. (37).

References

de Monte, F., 2000. Transient heat conduction in one-dimensional composite slab – a ‘natural’ analytic approach. Int. J. Heat Mass Transfer 43 (19), 3607–3619.
de Monte, F., 2002. An analytic approach to the unsteady heat conduction processes in one-dimensional composite media. Int. J. Heat Mass Transfer 45 (6), 1333–1343.
de Monte, F., 2003. Unsteady heat conduction in two-dimensional two slab-shaped regions – exact closed-form solution and results. Int. J. Heat Mass Transfer 46 (8),
1455–1469.
de Monte, F., 2006. Multi-layer transient heat conduction using transition time scales. Int. J. Therm. Sci. 45, 882–892.
Haji-Sheikh, A., Beck, J.V., 2002. Temperature solution in multi-dimensional multi-layer bodies. Int. J. Heat Mass Transfer 45 (9), 1865–1877.
Jain, P.K., Singh, S., Rizwan-uddin, 2009. Transient analytical solution to asymmetric heat conduction in a multilayer annulus. J. Heat Transfer 131 (1), 011304.
Jain, P.K., Singh, S., Rizwan-uddin, 2010. An exact analytical solution for two-dimensional, unsteady, multilayer heat conduction in spherical coordinates. Int. J. Heat Mass
Transfer 53, 2133–2142.
Lu, X., Viljanen, M., 2006. An analytical method to solve heat conduction in layered spheres with time-dependent boundary conditions. Phys. Lett. A 351, 274–282.
Lu, X., Tervola, P., Viljanen, M., 2005. A new analytical method to solve heat equation for multi-dimensional composite slab. J. Phys. A: Math. Gen. 38, 2873–2890.
Lu, X., Tervola, P., Viljanen, M., 2006a. Transient analytical solution to heat conduction in composite circular cylinder. Int. J. Heat Mass Transfer 49, 341–348.
Lu, X., Tervola, P., Viljanen, M., 2006b. Transient analytical solution to heat conduction in multi-dimensional composite cylinder slab. Int. J. Heat Mass Transfer 49, 1107–1114.
http://www.maplesoft.com/ (last accessed on 09 August 2010).
2010. Mathematica, Version 7.0. Wolfram Research, Inc., Champaign, IL.
Mikhailov, M.D., Ozisik, M.N., 1986. Transient conduction in a three-dimensional composite slab. Int. J. Heat Mass Transfer 29 (2), 340–342.
Ozisik, M.N., 1993. Heat Conduction. Wiley, New York.
Salt, H., 1983a. Transient heat conduction in a two-dimensional composite slab. I. Theoretical development of temperatures modes. Int. J. Heat Mass Transfer 26 (11),
1611–1616.
Salt, H., 1983b. Transient heat conduction in a two-dimensional composite slab. II. Physical interpretation of temperatures modes. Int. J. Heat Mass Transfer 26 (11), 1617–1623.
Singh, S., Jain, P.K., Rizwan-uddin, 2008. Analytical solution to transient heat conduction in polar coordinates with multiple layers in radial direction. Int. J. Therm. Sci. 47 (3),
261–273.
Yener, Y., Ozisik, M.N., 1974. On the solution of unsteady heat conduction in multi-region media with time-dependent heat transfer coefficient. In: Proceedings of the 5th
International Heat Transfer Conference, Tokyo, pp. 188–192, Cu 2.5.

You might also like