Thermal stresses in an innitely long solid cylinder using

Greens function and the hyperbolic heat equation

J.A. L

OPEZ MOLINA and M. TRUJILLO

Departamento de Matematica Aplicada. IMPA
Universidad Politecnica de Valencia
Camino de Vera 46022 Valencia
SPAIN
jalopez@mat.upv.es matrugui@mat.upv.es
Abstract: - We compute the thermal stresses arising in an homogeneous isotropic innitely long
solid circular cylinder when a constant heat ux is acting on its boundary. We use the hyperbolic
heat equation and Greens function to compute the temperature. A comparison with analogous
results obtained with the classical Fourier heat equation is made.
Key-words: - Greens function, integral transforms, hyperbolic heat equation, thermal stresses.
1 Introduction
The purpose of this communication is to
compute the thermal stresses arising in an in-
nitely long solid circular cylinder when a con-
stant heat ux is applied to its boundary us-
ing the hyperbolic model of heat conduction,
and to compare this results with the computed
ones in the case of using the classical Fourier
heat conduction model.
We consider an homogeneous isotropic solid
innitely long circular cylinder of radius R > 0
at initial temperature T
0
with constant and
independent on the temperature physical pa-
rameters thermal conductivity k, density d,
specic heat c, relaxation parameter , diu-
sivity :=
k
c d
, coecient of linear dilatation
, Youngs modulus E and Poissons coe-
cient . Starting in the time t = 0 a constant
heat ux Q
0
is normally applied in all the
points of its surface. The rst step in order
1
AMS Math. Sub. Class.: 80A20, 74B05.
2
This work was partially supported by FEDER and
MEC, Project MTM2004-02262 and the research net
MTM2006-26627-E.
to compute the thermal stresses developed in
the cylinder is to nd the induced temperature
eld. Choosing the z-cartesian coordinate axis
along the axis of the cylinder, the temperature
will be axisymmetric with respect to the z-axis
(independent on the polar angle ) and inde-
pendent on the z-coordinate, thus the govern-
ing hyperbolic heat equation in cylindrical co-
ordinates will be (see [5])

2
T

2

1

+
1

_
T
t
+

2
T
t
2
_
= 0 (1)
in the domain (, t) ]0, R[]0, [, with the
initial and boundary conditions
T(, 0) = T
0
T
t
(, 0) = 0 ]0, R[, (2)
T

(0, t) = 0 t > 0 (3)

and
T

(R, t) =
Q
0
k
_
H(t) + (t)
_
t > 0. (4)
Following [1] we introduce the dimensionless
Proceedings of the 2nd WSEAS Int. Conference on Applied and Theoretical Mechanics, Venice, Italy, November 20-22, 2006 63
variables
V (, t) = k
T(, t) T
0
R Q
0
, (5)
=

R
, =
t
4 R
2
, =

4 R
2
(6)
which yield to the problem

2
V

2

1

+
1
4
_
V

2
V

2
_
= 0 (7)
in the new domain (, ) ]0, 1[]0, [,
V (, 0) =
V

(, 0) = 0 ]0, 1[, (8)

V

(0, ) = 0 > 0 (9)

and
V

(R, ) = H() + () > 0. (10)

2 Greens function
Problem (7)-(10) is solved in [1] using the
residue and convolution theorems to nd the
Laplace inverse transform of V. We follow a
less involved method, the use of Greens func-
tion, furthermore providing the key to solve
many thermal problems concerning the quoted
geometry in an automatic way (see [3]).
In cylindrical dimensionless coordinates (5)
and (6) Greens function corresponding to
our problem is the distribution G(, ) :=
G(, |
0
,
0
) verifying for every (
0
,
0
)
]0, 1[R the boundary value problem

2
G

2

1

+
1
4
_
G

2
G

2
_
=
=
1

(
0
) (
0
) (, ) ]0, 1[R,
(11)
G

(0, ) =
G

(1, ) = 0 R (12)
and
lim

G(, ) = lim

(, ) = 0 ]0, 1[.
(13)
Denoting by L(, p) the Schwartz-Laplace
transform with respect to of G(, ) we ob-
tain (see [6])

2
L

2
(, p)
1

(, p)+
+
1
4
(p + p
2
) L(, p) =
1

(
0
) e
p
0
,
(14)
L

(0, p) =
L

(1, p) = 0 (15)
and (16)
lim
p0
L(, p) = lim
p0
p L(, p) = 0. (17)
In order to verify (15), condition (17) allow
us to take nite Hankel transforms of second
kind with respect to
H
m
(p) :=
_
1
0
J
0
(
m
) L(, p) d
(m N {0}) where {
m
}

m=0
is the strictly
increasing sequence of non negative zeros of
the equation J
1
() = 0, the Bessel function
of rst kind and order 1. Thus, using equation
(14) and the inversion formula of nite Hankel
transforms (see [7]) we obtain
L(, p) =
= 8

m=0
J
0
(
m

0
) J
0
(
m
)
_
4
2
m
+ p + p
2
_
J
0
(
m
)
2
e
p
0
.
In order to simplify the exposition we assume
that the material of the cylinder is such that
16
2
m
1 = 0 for every m N{0}. Thus,
by inversion of Schwartz-Laplace transforms
we obtain the adimensional Greens function
G(, |
0
,
0
) = 16 H(
0
)

m=0
J
0
(
m

0
) J
0
(
m
)
J
0
(
m
)
2
F
m
(,
0
) (18)
where, for every m N {0}
F
m
(,
0
) = e

0
2

_
H
_
1 16
2
m
_
_
1 16
2
m
sinh
m
(
0
)+
Proceedings of the 2nd WSEAS Int. Conference on Applied and Theoretical Mechanics, Venice, Italy, November 20-22, 2006 64
+
H
_
16
2
m
1
_
_
16
2
m
1
sin
m
(
0
)
_
where we have dened, for every m N

m
:=
_
1 16
2
m
2
,

m
:=
_
16
2
m
1
2
. (19)
3 Temperature in the case of
boundary constant ux
According to the theory of Greens func-
tions, the solution of problem (7)-(10) is given
directly by
(, ) ]0, 1[]0, [ V (, ) =
=
_

G(, |1,
0
)
_
H(
0
) + (
0
)
_
d
0
and after elementary computations and the
application of the equality (see [1])

m=1
J
0
(
m
)

2
m
J
0
(
m
)
=

2
4

1
8
we obtain
= 16

m=0
J
0
(
m
)
J
0
(
m
)
_

0
F
m
(,
0
) d
0
+
+ 16

m=0
J
0
(
m
)
J
0
(
m
)
F
m
(, 0) =
= 8 +
2
2
1
4
e

m=1
J
0
(
m
)

2
m
J
0
(
m
)
R
m
() (20)
where
m N R
m
() = H
_
1 16
2
m
_

_
cosh
m
+
2 16
2
m
_
1 16
2
m
sinh
m

_
+
+ H
_
16
2
m
1
_

_
cos
m
+
2 16
2
m
_
16
2
m
1
sin
m

_
.
It can be checked, after elementary compu-
tations, that this function coincides with the
obtained one in [1].
4 Thermal stresses
We start computing the actual value of
some expressions which will appear frequently
in the sequel. From (5), (6) and (20), by the
well known equality (see [8], page 45)
> 0
_

0
r J
0
( r) dr =

J
1
( )
we obtain
_

0
_
T(, t) T
0
_
rdr =
=
R Q
0
k
_

2
R
2
t +

4
8R
2


2
8
R e

t
2

m=1
J
1
_

m
R

_

3
m
J
0
(
m
)
R
m
_
t
4R
2
_
_
_
(21)
and thus
_
R
0
_
T(, t) T
0
_
d =
RQ
0
k
t. (22)
We are now in a position to compute the
thermal stresses in the cylinder. Since we are
concerned with an innitely long cylinder we
have to deal with a state of plane strain. It is
well known that in this case the components
of the stress tensor in cylindrical coordinates
are given by (see [4], page 206)

(, t) = 0

(, t) =
E
1
_

2
_

0
r
_
T(r, t) T
0
_
dr+
+
1
R
2
_
R
0
r
_
T(r, t) T
0
_
dr
_
,
(23)
Proceedings of the 2nd WSEAS Int. Conference on Applied and Theoretical Mechanics, Venice, Italy, November 20-22, 2006 65

(, t) =
E
1
_
1

2
_

0
r
_
T(r, t) T
0
_
dr+
1
R
2
_
R
0
r
_
T(r, t) T
0
_
dr
_
T(, t) T
0
_
_
(24)
and

zz
(, t) =
E
1
_
2
R
2
_
R
0
r
_
T(r, t) T
0
_
dr

_
T(, t) T
0
__
(25)
having in mind the existence of an axial strain
of magnitude

0
(t) =
2Q
0
kR
t (26)
(see [4], page 205).
After substitution of previous results and
elementary computations we obtain

(, t) =
ERQ
0
k(1 )
_
1
8


2
8R
2
+
+
R

t
2

m=1
J
1
_

m
R

_

3
m
J
0
(
m
)
R
m
_
t
4R
2
_
_
_
,

(, t) =
ERQ
0
k(1 )
_
1
8

3
2
8R
2
R e

t
2

m=1
_
_
J
1
_

R
_

3
m

J
0
_

R
_
R
2
m
_
_
R
m
_
t
4R
2
_
J
0
(
m
)
_
_
,

(, t) = 0
and

zz
(, t) =
ERQ
0
k(1 )
_
1
4


2
2 R
2
+
+ e

t
2

m=1
J
0
_

m
R

_

2
m
J
0
(
m
)
R
m
_
t
4R
2
_
_
_
.
5 Comparison with the classi-
cal case
If we use the classical parabolic Fourier heat
equation to nd the temperature (denoted by
T
F
(, t) to distinguish) we arrive (see [1]) to
T
F
(, t) = T
0
+
RQ
0
k
V
F
_

R
,
t
4R
2
_
where
V
F
(, ) =
= 8 +
2
2
1
4
2

m=1
J
0
(
m
)

2
m
J
0
(
m
)
e
4
2
m

.
It can be seen that this temperature eld
produce the same axial strain
0
given by (26)
and that formulas (22) and

(, t) = 0 still
holds. However in this case we have

(, t) =
ERQ
0
k(1 )
_
1
8


2
8R
2
+
+2
R

m=1
J
1
_

m
R

_

3
m
J
0
(
m
)
e

2
m
R
2
t
_
_
,

(, t) =
ERQ
0
k(1 )
_
1
8

3
2
8R
2
2R

m=1
_
_
J
1
_

R
_

3
m

J
0
_

R
_
R
2
m
_
_
e

2
m
t
R
2
J
0
(
m
)
_
_
and

zz
(, t) =
ERQ
0
k(1 )
_
1
4


2
2 R
2
+
+2

m=1
J
0
_

m
R

_

2
m
J
0
(
m
)
e

2
m
t
R
2
_
_
.
(27)
The quantitative dierences arising in the
stresses values given by the hyperbolic and the
parabolic temperature models are numerically
considerable and thus very important concern-
ing practical industrial applications. Remark
Proceedings of the 2nd WSEAS Int. Conference on Applied and Theoretical Mechanics, Venice, Italy, November 20-22, 2006 66
that these dierences are proportional to the
applied heat ux Q
0
(see the position of the
term Q
0
in all previous formulas of stresses)
and that high uxes applied in small tem-
poral intervals are more and more frequently
used in modern laser technology. As a sim-
ple illustration, working with the dimension-
less stresses deduced from (6) and ignoring the
factor
ERQ
0
k(1)
we have plotted in gures 1, 2
and 3 the radial variation of tension

in the
dimensionless time = 0.15, = 0.158 and
= 0.17 respectively, taking = 0.1 and us-
ing the dashed line for the tension computed
with Fourier heat equation and the solid line
in the case of using the hyperbolic heat equa-
tion.
0.2 0.4 0.6 0.8 1
0.1
0.2
0.3
0.4
Figure 1: Stress

corresponding to = 0.14
and = 0.1
0.2 0.4 0.6 0.8 1
-0.6
-0.4
-0.2
Figure 2: Stress

corresponding to = 0.15
and = 0.1
0.2 0.4 0.6 0.8 1
-0.3
-0.2
-0.1
0.1
Figure 3: Stress

corresponding to = 0.159
and = 0.1
References
[1] Barletta, A., Pulvirenti, B., Hyperbolic
thermal waves in a solid cylinder with a
non-stationary boundary heat ux, Int.
J. Heat Mass Transfer, 41,1, 1998, pp.
107-116.
[2] Ikawa, M., Hyperbolic partial dierential
equations and wave phenomena, Amer-
ican Mathematical Society, Providence,
USA, 2000.
[3] Lopez Molina, J. A., Trujillo, M., Exis-
tence and uniqueness of solutions of the
Neumann problem for the anisotropic
hyperbolic heat equation with non reg-
ular data, Preprint, 2006.
[4] Noda, N., Hetnarski, R. B., Tanigawa,
Y., Thermal stresses, Lastran Corpora-
tion, Rochester, NY, USA, 2000.
[5]

Ozisik, M. N., Tzou, D. Y., On the
wave theory in heat conduction, ASME
J. Heat Transfer, 116, 526-535, (1994).
[6] Pathak, R. S., Integral transforms of
generalized functions and their applica-
tions, Gordon and Breach Science Pub-
lishers, The Netherlands, 1997.
[7] Sneddon, I. N., Fourier transforms,
Dover Publications, Inc., New York,
1995.
[8] Watson, G. N., A treatise on the theory
of Bessel functions, Cambridge Univer-
sity Press, Cambridge, 1966.
Proceedings of the 2nd WSEAS Int. Conference on Applied and Theoretical Mechanics, Venice, Italy, November 20-22, 2006 67