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

net/publication/301232489

NUMERICAL SOLUTION OF RADIAL HEAT CONDUCTION IN AN INFINITELY

LONG CYLINDER

Conference Paper · December 2011

CITATIONS READS

0 1,003

1 author:

Nitish Anand
Delft University of Technology
5 PUBLICATIONS   0 CITATIONS   

SEE PROFILE

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

Design tool for Supersonic Turbine using Method of Characteristics View project

All content following this page was uploaded by Nitish Anand on 13 April 2016.

The user has requested enhancement of the downloaded file.

 Proceedings of the
International Conference on Mechanical Engineering 2011
(ICME2011) 18-20 December 2011, Dhaka, Bangladesh

ICME11-TH-032

NUMERICAL SOLUTION OF RADIAL HEAT CONDUCTION IN AN

INFINITELY LONG CYLINDER
D.N. Thatoi*, A. Acharya, N. Anand, P. Acharya, and V. Agrawal

Department of Mechanical Engineering, I.T.E.R, S‘O’A University, Bhubaneswar, Odisha, India

*
Corresponding Author (e-mail: [email protected])

ABSTRACT
Use of modern computational techniques supporting solution of imposed boundary condition in heat
transfer problems is increasing day by day. However determining the appropriate technique in the context
of problem statement, complexity of geometry, simplicity of usage and exacting solution are the challenge
researchers are facing. In this paper, an attempt has been made to find solution through different
computational methods for a one dimensional heat flow problem in steady state. Finite Difference Method
(FDM), Finite volume method (FVM) and Finite Element method (FEM) have been used and a comparative
analysis has been considered to arrive at a desired exactness of the solution. The calculated values from
each of the methods are compared and analyzed. A MATLAB implementation of the numerical solution has
also been provided. Numerical results are presented for a sequence of finer meshes, and the dependency of
the truncation error on mesh size is verified.

Keywords: TDMA, FDM, FVM, FEM

1. INTRODUCTION elements and fields with unregulated border. In this paper

Solutions to conduction problems to establish the the temperature distribution is analyzed in the radial
temperature distribution can be derived from the direction of a solid cylinder. The practical application is
appropriate differential equations with imposed in the electrical wires where heat is generated in the wire
boundary conditions. Analysis of such solutions is better and the same heat it needed to be dissipated to the
done through various computational techniques. Antar surrounding so that the melting of the wire is avoided.
[4] offered a simplified numerical solution of unsteady The cylinder in this case is infinitely long with
heat conduction problem in a short cylinder. This paper isotropic physical characteristics and it has been assumed
proposes simpler solution for heat conduction problem in that the heat is being transferred in the radial direction.
one –dimensional form and compares the results Heat conduction in the axial direction is neglected.
obtained in each method (FDM, FVM, FEM) [5], [12] for
different meshes.
The finite difference method (FDM) [7] is based on 2. THEORY
the differential equation of the heat conduction, which is 2.1 Governing Differential Equation
transformed into a difference equation. The temperature The general heat conduction problem in a 3
values are calculated at the nodes of the network. Using dimensional plane with generation and with time
this method, convergence and stability problem can variation is given by the equation given below.
appear. Han et al. [5] used FDM method with FEM for
analysis of one-dimensional fin. Governing Differential Equation:
Dhawan et al. [2] studied the heat conduction problem
in an aluminium plate using finite element method and (1)
got very good results compared to the exact solution. The
finite element method (FEM) [7] is based on the integral The similar governing differential Equation for cylinder
equation of the heat conduction. This is obtained from is:
the differential equation using variational calculus. In
first case the temperature values are calculated on the (2)
finite elements. Then, based on these partial solutions,
the solution for the entire volume is determined. Using Neglecting d the heat conduction in the axial direction
this method the whole surface has been divided into the equation becomes one dimension, assuming
© ICME2011 1 TH-032
axi-symmetry.
. The similar governing differential equation (GDE) for
sphere is given as:

(3)

In this paper, focus has been given on the GDE for

cylinder.

2.2 Boundary Conditions

a. Symmetric Boundary Condition
This treatment of the boundary condition corresponds
to the physical assumption that, on the two sides of the
boundary, the same physical processes exist. The
variable values at the same distance from the boundary at
the two sides are the same. The function of such a
boundary is that of a mirror that can reflect all the Fig 1. Geometry of cylinder showing 6 different nodes
fluctuations generated in the simulation region. for the finite difference method
In this case, under symmetry boundary condition at
the center i.e. at r=0 , As shown in Fig 1. , the symmetrical cylinder solid
structure is divided into six different nodes for the finite
b. Robbin’s Boundary Condition difference method. The general governing differential
This treatment of the boundary is done when the equation is discretised using FDM is as follows:
boundary is exposed or open to the atmosphere. Hence in For n equal divisions, there will be m nodes where
this kind of boundary condition the conductive heat flux m=n+1
and convective heat flux are made equal. This kind of
boundary condition is also called as mixed type boundary For node i=1:
condition. The governing differential equation can be written as:
Considering the above boundary condition for the
cylinder, (7)
At r = R,
To avoid singularity at the origin , L-Hospitals
(4)
rule is used to get the discretized equation at i=1 as:
2.3 Exact Solution
Integrating the governing differential equation within
the boundary conditions we obtain:
For i = 2 to m-1:
(5)

Putting boundary condition 1: at r=0, , we get For i= m:

C1=0.
Again integrating:
Using boundary condition 2, at r=R 3.2 Finite Volume Method (FVM)
Finite Volume method is another method for
numerical calculation of the differential equation. In
this method the entire volume is divided into smaller
We get C2= volumes and inlet flux is equated with the outlet flux
giving an approximate numerical approach to the
So the exact temperature profile is:
temperature at each mesh [8].
The FVM equations for the solid cylinder considered
(6) in this paper are:

3. NUMERICAL SOLUTION
3.1 Finite Difference Method (FDM)

(8)

© ICME2011 2 TH-032
 solution and then compared with that obtained through
FDM, FVM and FEM.
Following parameters have been considered for
solutions in all techniques as cited above:

Table 1: Parameter matrix

Parameter Value
Radius (m) 0.05
Generation 4x106
(Watt/m3)
k (Watt/mK) 40
Fig 2. Geometry of cylinder showing 7 different shell h (Watt/m2K) 400
T∞ (oC) 20
centers for the finite volume method
Table 2: Comparative results of temperature distribution
For n equal divisions, there will be m shell centers obtained through FDM, FVM and exact
where m=n+2 solution

For i=1: Radi FDM FVM Exact

us (0C) (0C) (0C)
For i=2:
(m)
For i= 3 to m-2: 0.000 332.500 332.500 332.500
0.005 331.875 332.500 331.875
0.010 330.000 330.000 330.000
0.015 326.875 327.500 326.875
For i=m-1: 0.020 322.500 322.500 322.500
0.025 316.875 317.500 316.875
0.030 310.000 310.000 310.000
0.035 301.875 302.500 301.875
For i=m: 0.040 292.500 297.500 292.500
0.045 281.875 292.500 281.875
0.050 270.000 270.000 270.000

As can be seen from the above table (Table 2), the

3.3 Solution Technique FDM solution and the exact solution are exactly match
The governing differential equation has been without any error. This is because of the use of second
discretized using both FDM and FVM. Both the order accurate finite difference for first order derivative
in which case the truncation error is found to be zero [7].
equations are in the form: aiTi=biTi+1+ciTi-1+di
The difference in results obtained through exact
So, Tri-Diagonal Matrix Algorithm (TDMA) [14] is solution and that through FVM is due to the use of first
appropriate to solve the discretized equations. order accurate difference equation, for first order
The corresponding MATLAB Code is appended. The derivative.
TDMA solution follows forward elimination with The results represented in Table 2 have also been
back-substitution using the equation: Ti=PiTi+1+Qi, shown in the form of graphical representation.
where

3.4 Finite Element Method (FEM)

Finite Element Method is another method for
numerical solution the differential equation. ANSYS
13.0 well established commercial software is used in
this paper for required analysis and the details are
produced in the results and discussion section.

4. RESULTS AND DISCUSSION Fig 3. Comparison of FDM, FVM and exact solution
The one-dimensional heat conduction problem having six nodes
considered in this paper has been solved using exact
© ICME2011 3 TH-032
 One of the features of FDM and FVM is that the
relative error between the analytic method and these two
methods decreases as the numbers of nodes (shell
centers) increase. This has been established in the
following graph in which the results have been plotted by
solving the heat conduction problem in FDM and FVM
with 100 nodes.

Fig 7. Cross –sectional temperature contour

Above figure (Fig. 7) shows no direct matching the

results obtained through exact solutions due to the
non-uniformity in meshing.
Fig 4. Comparison of FDM, FVM and exact
solution with 100 nodes
5. CONCLUSION
The numerical modelling with FEM, FVM and FDM
represents an efficient way to obtain temperature
Results obtained through FE Analysis
distribution in steady state conductive heat transfer
ANSYS has been used for analysis of temperature
processes. They also render the solution better to be
distribution through FEM.
automated through computer.
The numerical computation of the temperature field,
on the basis of FDM offer better results for regular
geometry, whereas FEM and FVM can offer better
results for irregular geometry and complex boundary
conditions. Also, choice of finer grids which requires
high computing capability can remove approximation
errors to larger extent.

6. REFERENCES
1. Sarbu, I. and Popina, O., 2001, “Numerical analysis
with finite and boundary elements of thermal fields
in steady state regime”, ARPN Journal of
Engineering and Applied Sciences, Vol. 6, No. 2,
Fig 5. Mesh on the entire rod pp.13-23.
2. Dhawan, S. and Kumar S., 2009, “Comparative
study of numerical techniques for 2D transient heat
conduction equation using finite element method”,
International Journal of Research and Reviews in
Applied Sciences Vol. 1, pp.38-46.
3. Hsu, M.H., 2009, “Differential Quadrature Method
for Solving Hyperbolic Heat Conduction
Problems” Tamkang Journal of Science and
Engineering, Vol. 12, No. 3, pp. 331-338.
4. Antar, M.A., 1999, “A simplified numerical
solution of unsteady heat conduction problems with
an application to a short cylinder”, International
Journal of Mechanical Engineering Education Vol.
Fig 6. Mesh on the cross-section
28, No. 3, pp.201-212.

© ICME2011 4 TH-032
5. Han, Y. M., Cho, J. S. and Kang, H. S., 2005, 7. NOMENCLATURE
“Analysis of a one-dimensional fin using the
analytic method and the finite difference method”, Symbol Parameter Units
J. KSIAM, Vol.9, No.1, pp. 91-98. r Radius Meter
o
6. Dabral, V., Kapoor, S. and Dhawan, S., 2011, T Temperature C
“Numerical Simulation of one dimensional Heat Density kg/m3
Equation: B-Spline Finite Element Method”, t Time sec
C Specific Heat kJ/Kg-K
Indian Journal of Computer Science and
Capacity
Engineering (IJCSE) Vol. 2 No. 2, pp.222-235. k Thermal Watt/m-K
7. Ozisik, M. N., 2000, Finite Difference Methods in Conductivity
Heat Transfer, CRC Press. h Convective heat Watt/m2-K
8. Versteeg, H. K. and Malalasekera, W., 1996, An Transfer Co-efficient
introduction to Computational Fluid Dynamics, g Generation term Watt/m3
o
Longman. Ts Surface temperature C
o
T∞ Surrounding C
9. Chen, T. M. and Chen, C. C., 2010, “ Numerical
temperature
solution for the hyperbolic heat conduction i Node number No unit
problems in the radial-spherical coordinate system
using a hybrid Green's function method”,
International Journal of Thermal Sciences, Vol.49, APPENDIX
pp. 1193-1196.
MATLAB CODE:
10. Chen, G. and Zhou, J.,1992, Boundary element clc
methods. Academic Press, New York. clear all
11. Gafiţanu , M., Poteraşu, V. and Mihalache, N., n=input('Number of nodes:');
1987, Finite and boundary elements with r=input('Radius of the Cylinder(m):');
applications to computation of machine k=input('Thermal Conductivity(Watt/mK):');
components. Technical Press, Bucharest. h=input('Convective heat transfer
12. Wang, B.L. and Mai, Y.W., 2005, “Transient one Co-efficient(Watt/m2):');
dimensional heat conduction problems solved by g=input('Generation term(Watt/m3):');
finite element”, International Journal of Ti=input('Ambient air temperature(0C):');
Mechanical Sciences. Vol. 47, pp. 303-317. m=n+1;
13. Abrate, S. and Newnham, P., 1995, "Finite Element dx=r/n;
Analysis of Triangular Fins Attached to a Thick clc
Wall", Computer & Structures, Vol. 57, No. 6, pp. disp(' FINITE DIFFERENCE SOLUTION');
945-957. a(:,:)=0;
14. Patankar, S.V., 1980, Numerical Heat Transfer and for i=2:m-1
Fluid Flow, Hemisphere Publishing Corporation,
Taylor & Francis Group, New York.
a(i,1)=2;
a(i,2)=(1+(1/(2*(i-1))));
a(i,3)=(1-(1/(2*(i-1))));
a(i,4)=g*dx*dx/k;
end
a(1,1)=4;
a(1,2)=4;
a(1,4)=g*dx*dx/k;
a(m,1)=((2)+(1+(1/(2*(m-1))))*(2*dx*h/k));
a(m,3)=2;
a(m,4)=((1+(1/(2*(m-1))))*(2*dx*h/k)*Ti)+(g*dx*dx/k
);
for i=1:m
if(a(i,3)==0)
p(i)=a(i,2)/a(i,1);
end
if(a(i,3)~=0)
p(i)=a(i,2)/(a(i,1)-(a(i,3)*p(i-1)));
end
end
%disp('Value of P');
© ICME2011 5 TH-032
 %disp(p); %calculation of q
%calculation of q for i=1:m
for i=1:m if(a(i,3)==0)
if(a(i,3)==0) q(i)=a(i,4)/a(i,1);
q(i)=a(i,4)/a(i,1); end
end if(a(i,3)~=0)
if(a(i,3)~=0) q(i)=(a(i,4)+(a(i,3)*q(i-1)))/(a(i,1)-(a(i,3)*p(i-1)));
q(i)=(a(i,4)+(a(i,3)*q(i-1)))/(a(i,1)-(a(i,3)*p(i-1))); end
end end
end % disp('Value of Q'),q
%disp('Value of Q'),q %calculating u
%calculating u T(1:7)=0;
T(1:5)=0; for i=m:-1:1
for i=m:-1:1 if(i==m)
if(i==m) T(i)=q(i);
T(i)=q(i); end
end if(i~=m)
if(i~=m) T(i)=(p(i)*T(i+1))+ q(i);
T(i)=(p(i)*T(i+1))+ q(i); end
end end
end T
T x(1)=0;
for i=1:m x(2)=dx/2;
x(i)=(i-1)*dx; x(m)=r;
end for i=3:m-1
plot(x,T(1:m)); x(i)=x(i-1)+dx;
hold on end
disp(' FINITE VOLUME SOLUTION'); plot(x,T(1:m));
m=n+2; hold on
a(:,:)=0; (g*r/(2*h))+(g*r*r/(4*k))+Ti
for i=3:m-2 a=[-g/(4*k) 0 (g*r/(2*h))+(g*r*r/(4*k))+Ti];
a(i,1)=(2*i)-3; i=1;
a(i,2)=i-1; k1(1)=0;
a(i,3)=i-2; for x=0:0.001:r-0.001
a(i,4)=g*((2*i)-3)*dx*dx/(2*k); i=i+1;
end k1(i)=k1(i-1)+0.001;
a(1,1)=1; end
a(1,2)=1; plot(k1,polyval(a,k1));

a(2,1)=1;
a(2,2)=1;
a(2,4)=g*dx*dx/(2*k);
a(m-1,1)=3*(m-1)-4;
a(m-1,2)=2*(m-2);
a(m-1,3)=m-3;
a(m-1,4)=g*((2*(m-1))-3)*dx*dx/(2*k);
a(m,1)=2+(h*dx/k);
a(m,3)=2;
a(m,4)=h*dx*Ti/k;
for i=1:m
if(a(i,3)==0)
p(i)=a(i,2)/a(i,1);
end
if(a(i,3)~=0)
p(i)=a(i,2)/(a(i,1)-(a(i,3)*p(i-1)));
end
end
% disp('Value of P');
% disp(p);

© ICME2011 6 TH-032

View publication stats