International Conference on Computer Engineering, Information Science & Application Technology (ICCIA 2016)

Numerical Method for Three-Dimensional Heat Conduction in Cylindrical

and Spherical Coordinates

Xinxin Jia, Zongrui Hao, Hu Wang, Lei Wang, Xin Wang

Instrumentation of Shandong Academy of Sciences, Qingdao 266061, China
[email protected]

Abstract. According to the differential equations of heat conduction on cylindrical and spherical
coordinate system, numerical solution of the discrete formula on cylindrical and spherical coordinate
system with high accuracy were derived. Compared with the analytical solution, this discrete formula
was verified with a high degree of accuracy. To make the complex dispersion coefficient of diffusion
term θ more concrete in spherical coordinates, this paper derived the discretion coefficient of
diffusion term θ by the first mean value theorem of integral. The accurate schemes provide a good
reference for researchers whose work in solving the equation of heat conduction of three-dimensional
cylindrical coordinates and spherical coordinates, and it will provide accurate numerical schemes and
the theoretical basis for solving practical engineering problems.
Keywords: Differential equation of heat conduction; Spherical coordinate; Numerical heat transfer;
the first mean value theorem for integral.

1. Introduction
Numerical Heat Transfer has been widely used to solve practical complex heat transfer
calculations [1-3]. The important thing to solve the problem was how to discrete the thermal
conductivity differential equations. Numerical Heat Transfer in two-dimensional cylindrical
coordinates and polar coordinates equation of heat conduction were applied widely. When referring
to the discretion schemes of three-dimensional cylindrical and spherical coordinates, now there is not
a relatively discrete format for it [4, 5].
At the present, when it comes to the questions of cylinder and sphere in numerical heat transfer, it
is just simplified into the radial or two-dimensional polar coordinates, which caused a lot of
inconvenience in computation, promotion and application. Therefore, it is very important for us to
solve to the three-dimensional cylinder coordinates thermal conductivity partial differential
equations and thermal conductivity of the sphere of partial differential equations. This paper derived
high precision cylindrical coordinates and spherical coordinate’s equation of heat conduction discrete
schemes and compared with one-dimensional analytical solution to verify the accuracy of the
numerical solution.

2. Problem description
The partial differential equations for heat conduction was established in the Cartesian coordinate
system and the energy principle and Fourier's [6] law was used. Then we will have thermal
conductivity of partial differential Eq. (1) as follows:
  2T  2T  2T
( ρcT )  λ 2  λ 2  λ 2  S (1)
τ x y z
Where λ is thermal conductivity, c is the heat capacity of the thermal conductivity, S is inner heat
source.
The cylindrical coordinate system Eq.(2) and the spherical coordinate system Eq.(3) of the
differential equations[6] are described as follows:
 1  T 1  T  T
( cT )  (r ) 2 ( ) ( )  S , 0    2 (2)
 r r r r   z z

© 2016. The authors - Published by Atlantis Press 156

  1  T 1  T 1  T
( cT )  2 (r 2 ) 2 ( ) 2 ( sin  )S (3)
 r r r r sin 2    r sin   
0     ,0    2
In general, there are two methods to discrete the equation of heat conduction, that is, Taylor series
expansion method and finite volume method. In order to ensure the physical significance [7, 8], this
paper will derive the thermal conductivity of partial differential equations by finite volume method.

3. Literature References
First, multiply r on both sides of the Eq.(2), then the integral equation in control volume and
unsteady time items are shown in Figure1. The Eq.(2) changed into Eq.(4) as follows:
t  t t  t
  T 1  T  T
  r τ ( ρcT )drd dz    [ r ( λr r )  r 
t V t V
(λ

)  r ( λ )]drd dz
z z
t  t (4)
   rSdrd dz
t V

Fig.1 Cylindrical coordinates control volume

3.1 The integral of the unsteady terms
After integrating we can gain the unsteady term in the following forms:
t  t n e d
T (rn  rs )r
t s w b r τ drd dz  2  z(TP  TP )
0
(5)

3.2 The integral of the diffusion terms r, φ, z

After integrating the diffusion terms into the following forms:
t  t n e d
 T
n
 T  TN  TP T  TS 
    r (r r )drddzdt  zt  r (r r )dr  zt  r
t s wb s
n n
(r ) n
 s rs P 
(r ) s 
(6)
t t n e d
1  T r
e
 T r  T T T T 
  r 
t s wb
(λ

)drd dzdt  ln( n )zt  ( λ )d  ln( n )zt  λe E P  λw P W 
rs w
  rs  ( δ ) e (δ ) w 
(7)
t  t n e b
 T (r  r )r  T
b

 r ( λ )drd dzdt  n s

z z 2
 t  ( λ )dz
z z
t s wd d (8)
(r  r )r  T T T T 
 n s  t  λb B P  λd P D 
2  ( δz ) b (δz ) d 
3.3 The integral of the source term
t  t n e d
(rn  rs )r
t s w b rSdrddzdt  2 zt (S C  S PTP ) (9)

Here, the discrete equations will be solved by the techniques for linear equation, it is sufficient to
express the average value S as:
S  S C  S PTP (10)
157
 where, SC stands for the constant part of S, SP is the coefficient of TP.
3.4 The discrete Equation for cylinder
With the linearized source expression, the discrete equation would changed into the Eq.(4) as
having the form:
aPTP  aETE  aW TW  aN TN  aS TS  aBTB  aDTD  b (11)
Where:
rn r
)z
ln( ln( n )z
r z
aE 
rs , aW  rs , a N  rn z , a S  s ,
( ) e e ( ) w  w (r ) n n (r ) s  s
(rn  rs )r (r  r )r
aB  , aD  n s , b  S C V  a P0 TP0
2(δz ) b λb 2(δz ) d λd
aP  aE  aW  aN  aS  aB  aD  aP0  SP V , V  (rn  rs )r  z ,
2
( ρc) P V
a P0  .
t
At this point, it is interesting to examine the physical significance of the various coefficients in the
discrete equation. The neighbor coefficients aE, aW, aN, aS, aE, aD represent the conductance between
the point P and the corresponding neighbors. The term aP0 TP0 is the internal energy contained in the
control volume at time t. The constant term b consists of this internal energy and the rate of heat
generation in the control volume resulting from SC. The ΔV is the volume of the control volume. The
center-point coefficient aP is the sum of all neighbor coefficients and contains a contribution from the
linearized source term.

4. The discrete control equation of spherical coordinates

At this stage, multiply r2sin2θ on both sides of the Eq.(2), then the integral equation that was in
control volume and unsteady time items points are shown in Figure 2.The new equation is as follows:

Fig.2 Spherical coordinate control volume

t  t t  t
   T  T
  r sin θ ( ρcT )dtdrdθd    sin θ ( λr 2 )  (λ )
2 2 2

t V
τ t V
r r   (12)
t  t
 T 
 sin θ ( λ sin θ ) dtdrdθd    S  r sin 2 θdtdrdθd
2

θ θ  t V

4.1 The integral of the unsteady terms

After integrating the unsteady term into the following forms:
t t n e d
 (r 3  r 3 )
  r sin 2  ( cT )dtdrdd  n s (  sin  w cos  w  sin e cos e ) (TP  TP0 ) (13)
2

t s wb
 6

158
4.2 The integral of the diffusion terms r, φ
After integrating the diffusion term r, φ becomes the following forms:
t  t n e d
 T 1  T T T T  (14)
    sin  (r 2 )dtdrdd  (  sin  w cos  w  sin  e cos  e )t rn2 N P  rs2 P S 
2

t s wb
r r 2  (r ) n (r ) s 
t  t n e d
 T T  T T  T 
  
t s wb
(λ

)dtdrdθd   B P  P D θrt
 (δ ) b (δ ) d 
(15)

4.3 Application integral mean value theorem derive the diffusion term θ
In the derivation of the integral term, this paper use the integral mean value theorem which will
derive the diffusion term θ scientifically and reasonably.
Obviously, the heat flux that leaves one control volume through a particular face must be identical
to the flux that enters the next control volume through the same face. Otherwise, the overall balance
would not be satisfied. It is the fact that the energy across the interface θe to θw continuously, so the
heat conduction equation can be thought as a continuity equation. Because of the diffusion
term θ  [0, π ] , it happens for θ [0, π ] , and we have sin θ  0 . Based on the above two conditions of
continuity and sin   0 , which satisfied the integral mean value theorem[9], the diffusion term θ and
Eq.(3) are changed as the following   [w, e] shows, so that the θ conduction term denotes:
t  t n e d t  t n e d
 T  T
 t
s w b sin   ( sin   )dtdrdd  sin       ( sin   )dtdrdd
t s wb
(16)

It can be seen that when  is much smaller, the value    P .

t  t n e d t  t n e d
 T  T
sin θξ   θ
( λ sin θ )dtdrdθd  sin θP
θ   θ ( λ sin θ θ )dtdrdθd
t s wb t s wb (17)
 T  TP T T 
 sin θP sin θe E  sin θw P W   rt
 (δθ )e (δθ ) w 
4.4 The integral of the source term
After integrating the source term S becomes the following forms:
t  t n e d
(rn3  rs3 )
    r sin   Sdtdrdd  (  sin  w cos  w  sin  e cos  e )t ( S C  S PTP ) (18)
2 2

t s wb
6
4.5 The discrete Equation for sphere
The discrete equation can easily be seen as follows:
aPTP  aETE  aW TW  aNTN  aSTS  aBTB  aDTD  b (19)
Where
r sin  P sin  e r sin  P sin  w
aE  , aW  ,
( ) e  ( ) w 
(θ  sin θ w cos θ w  sin θe cos θe )  rn2
aN  ,
2(δr ) n λ
(  sin  w cos  w  sin  e cos  e )  rs2
aS  ,
2(r ) s 
θr r
aB  , aD  , b  S C V  a P0 TP0 ,
(δ ) b λb ( ) d d
aP  aE  aW  aN  aS  aB  aD  aP0  SP V ,
(rn3  rs3 )
V  (  sin  w cos  w  sin  e cos  e ) ,
6
( ρc) P V
a P0  , 0  θ  π,0   2π .
t
With the different express, the terms of a E , aW , a N , a S , a B , a D , P , a P0 TP0 etc. are the same as the
part 3 shown.
159
5. Numerical value comparison
5.1 Application of discrete equations and grid independent analysis
To illustrate the workings of the discrete equations this paper gives a detailed example of
cylindrical heat conduction. The inner and outer diameter of the cylinder were R1 and R2 The
temperature of the inside and outside of the cylinder was T1 and T2 and the radial heat conduction
analytical solution is as follows:
ln( R / R1 )
T  T1  (T2  T1 ) (20)
ln( R2 / R1 )
We consider it as a steady and three-dimensional cylinder, and the heat constant cylindrical
boundary conditions (the problem data) are as follows:
1  T 1  T  T
 r r ( λr r )  r 2  ( λ  )  z ( λ z )  0

 R1  1, T1  1
(21)
 R  2, T  1000
 2 2

0   2π
The numerical computed in this paper as well as the analytical solutions are compared in Figure3.
Given the coarseness of the triangle that numerical solution gives reasonable agreement with the
analytical solution, the error between the numerical and analytical solution was 8.86×10-3%.
5
1000

4
dimensionless temperature

800
2-norm error (%)

3
600

2
analytical solution Grid systgem used in computation
400
numerical solution
1
200
0
0
0 2 4 6 8 10 0 5000 10000 15000 20000 25000
dimensionless radius Grid number
Fig.3 Analytical solution and numerical solution Fig.4 Grids number and error relations

To quantify the accuracy of the reconstruction formulas between numerical solution and analytical
solution we define the relative error as:
T  Tnum 2
error   100% (22)
T 2
Where the error means the error between the solutions of the governing equations, and T refers to
the solutions by means of the analytical solution; Tnum refers to the solutions by means of the
numerical solution.
n 1
 ( xi )
2 2
The error means L2-norm [10]: T 2
i 1
Based on the error obtained from grid-independent solution [11], the relation between the equation
error and grid number is shown in Figure 4. When the grid number reaching 5000, the calculation
error control is less than 1%; with the encryption of the grid, the calculation error is reduced gradually.
The analytical solution error is 8.86×10-3%, the choice of computational grid number is
37×20×16=11840.

160
5.2 The validation of the thermal equation of the sphere
To illustrate the workings of the sphere equations, this paper gives a detailed example of sphere
heat conduction. The inner and outer diameter were R1 and R2. The temperature of the inside and
outside of the sphere was T1 and T2.
We consider it as a steady and three-dimensional sphere, and the heat constant boundary
conditions (the problem data) are as follows:
1  2 T 1  T 1  T
 r 2 r ( λr r )  r 2 sin 2 θ  ( λ  )  r 2 sin θ θ ( λ sin θ θ )  0

 R1  1, T1  1 (23)
 R  2, T  100
 2 2

0   2π ,0  θ  π
The numerical computed in this paper as well as analytical solutions are compared in Figure5.
Given the coarseness of the triangle that numerical solution gives reasonable agreement with the
analytical solution, the error between the numerical and analytical solution was 0.53%.
3.0

100
2.5
dimensionless temperature

80
2.0
Grid systgem used in computation
2-norm error (%)

60
1.5

numerical solution
40 1.0
analytical solution

20 0.5

0 0.0
0 2 4 6 8 10 0 4000 8000 12000 16000 20000
dimensionless radius Grid number

Fig.5 Analytical solution and numerical solution Fig.6 Grids number and error relations

Based on the error obtained from grid-independent solution, the relation among the equation error
and grid number is shown in Figure 6. When the grid number reaching 12000, the calculation error
control is less than 1%; with the encryption of the grid, the calculation error is reduced gradually. The
analytical solution error is 0.5%, the choice of computational grid number is 47×20×12=11280.

6. Conclusion
According to the three-dimensional equation of heat conduction of the cylinder and sphere, this
paper derived the high-accuracy computational discrete scheme with the finite volume method. In the
process of the discrete sphere, this paper uses the integral mean value theorem reasonably, which has
derived the complex diffusion term more strictly. The new three-dimensional cylindrical coordinates
and spherical coordinates provide numerical calculation discrete scheme, which will provide a good
reference for the researchers. The application of numerical calculation discrete scheme is written in
FORTRAN computer language, which is of stable operation and simple application. This paper have
contrasted and analyzed the error, which was controlled within 0.5%. All the work has provided a
high-precision numerical calculation discrete scheme for teaching and scientific research.

Acknowledgements
The work was supported by the youth national natural science fund project of China No.
(51504146; 51206101) and the Dr Funds of Shandong Academy of Sciences No. (2015QN023).
I wish to acknowledge the help of Xiaoling SUN and Zunwei LI in Correction translation errors
and Mr. Xu provides a lot of help.
161
References
[1]. W. Q. Tao. Numerical Heat Transfer, 2nd ed., Xi’an Jiaotong University Press, China, 2001.
[2]. P. Ding, W. Q. Tao, “Reduced order modeling of fluid flow and heat transfer in tube-fin heat
exchanger”, Journal of China University of Petroleum (Edition of Natural Science), vol. 35, no. 2,
pp. 137-140, 2011.
[3]. S.B. Huang, Z.M. Li, “Numerical analysis of laminar pulsating flow for power law fluid in
concentric annulus”, Journal of China University of Petroleum (Edition of Natural Science), vol.
36, no 2, pp. 127-136, 2011.
[4]. Y. H. Jin, Q. Guang, Q. Y. Cao, et a.l, “Numerical simulation of gas-phase flow field in vortex
finder of cyclone separators”, Journal of China University of Petroleum (Edition of Natural
Science),vol. 32, no. 6, pp. 109-112, 2008.
[5]. P, Ding, X. Wu, Y. He, et a.l, “A fast and efficient method for predicting fluid flow and heat
transfer problems”, ASM E Journal of H eat Transfer, vol. 130, no. 7, pp. 1-17, 2008.
[6]. X. C. Deng, “Explore the establishment of thermal conductivity in the spherical coordinate
system of differential equations”, An Hui Inst Technology, vol. 6, no. 4, pp.58-70, 1987.
[7]. PATANKA SV, SPALDING DB, “A calculation procedure for heat, mass and momentum
transfer in three-dimensional parabolic flow”, International Journal Heat Mass Transfer, vol. 15,
pp. 1787-1806, 1972.
[8]. Patankar, S. V., “Numerical Heat Transfer and Fluid Flow”, McGraw-Hill, New York, 1980.
[9]. D. K. Tong, S. T. Zhou, A. D. Wu, L. H. Zhang, “Computational Method,” Computational
Method, University of Petroleum Press, 2004.
[10]. M. H. Xu, Y. B. Zhang, W. Q. TAO, “A modified Delaunay mesh generation method for 2-D
domain”, Journal of the University of Petroleum, China( Edition of Natural Science), vol. 25, no.
2, pp.100-105, 2001.
[11]. LEE C K, HOBBS R E. “Automatic adaptive finite element mesh generation over arbitrary
two-dimensional domain using advancing front technique”, Computers& Structure, vol. 71,
pp.9-34, 1999.

162