Chapter 1

1D Theory of Thermoelasticity
1.1 The governing equations of the linear theory of thermoelasticity
The governing equation of the general theory of thermoelasticity reduce to a much simple form in the case some simplifying assumptions are considered to be valid. This is particularly true if we assume the linearity conditions for the equilibrium equations, the straindisplacement (compatibility equations and the constitutive equations). According to what already established for the general theory, we assume as state variables of the problem, and as unknowns of the mathematical model of the thermoelastic behavior: the displacement components ui = ui (xi , t) of the material point having coordinates xi in the reference conguration (lagrangian approach); the Green-Lagrange strain components ij = ij (xi , t); the Cauchy stress components ij = ij (xi , t). The forcing actions on the 3D continuum body are represented by: the external volume forces per unit volume Xi (xi , t) acting on the volume V occupied by the body; the external surface forces fi (xi , t) acting on the surface Sf of V where the external forces are applied; the applied displacements ui (xi , t), acting on Su , the portion of S where the displacement are prescribed with Sf + Su = S ;

1. 1D Theory of Thermoelasticity

the variation of temperature T (xi , t), assumed to be given for any point P of the body, in the hypothesis that the mechanical problem has no inuence on the heat conduction problem. Also initial conditions are prescribed for displacement ui (xi , 0) = uo i (xi ) and velocities u i (xi , 0) = u o i (xi ) (1.2) (1.1)

We now assume that the displacement ui are small as compared to the dimension of the body. In this circumstance, for the establishment of the equilibrium conditions of the body, the actual conguration of the body at time t, C t can be considered coincident with the reference conguration C o , that is assumed to be known. The equilibrium equation can be written with reference to the initial volume V , and also the Cauchy stress, dened in general on the actual conguration, is now referred to a known conguration. The equilibrium equations read ij,j + X i = 0 with X i = X i u i (1.4) i = 1, 2, 3 on V (1.3)

being the density (mass per unit volume) of the body. In general we might assume a non homogeneous body with = (xi ) If also a small displacement gradient assumption with respect to unity is formulated, that is a small rotations and small strain hypothesis, we can disregard the second (non linear) terms of the Green Lagrange strain tensor that reduces to the engineering strain term, which can be expressed now, in terms of displacement as: ij = 1 (ui,j + uj,i ) 2 i = 1, 2, 3 on V (1.5)

for a total of 6 equations, considering the symmetry condition of ij = ji . The boundary conditions for the stress can be expressed as ij nj = fi i = 1, 2, 3 on Sf (1.6)

with nj = cos nxj being n the normal to the boundary surface, and the one on displacement can be written as ui = ui i = 1, 2, 3 on Su (1.7)

1. 1D Theory of Thermoelasticity

The constitutive relations assume the form, for an isotropic body, ij = Eijhk (hk hk T ) i, j = 1, 2, 3 (1.8)

where is the thermal expansion coecient, hk is the Kronecker (hk = 0 for h = k , hk = 1 for h = k ), Eijhk is the elastic tensor. In conclusion, for a total of 15 unknown components, 15 equations represent the physics of the problem. One possible technique of solution is to keep the displacement ui as the unknown and express the equilibrium equation in terms of the 3 displacement components after expressing the stress as formulated in the constitutive relations and the strain as expressed in the strain-displacement equations. With this substitutions the equilibrium equations, expressed in displacement components, read By assuming and , that is an homogeneous thermoelastic body, the equations will now read
[ ]

{Eijhk

1 (uh, k + uk,h ) hk T },j + Xi u i = 0 2

i = 1, 2, 3

(1.9)

By assuming Eijhk = const and = const, that is an homogeneous thermoelastic body, the equations will now read 1 Eijhk (uh, kj + uk,hj ) Eijhk hk T,j + Xi u i = 0 2
[ ]

i = 1, 2, 3

(1.10)

The stress boundary conditions expressed in terms of displacement read Eijhk 1 (uh, k + uk,h ) hk T nj = fi 2 i = 1, 2, 3 on Sf (1.11)

While on Su ui = ui and the initial conditions ui (xi , 0) = uo i (xi ) and u i (xi , 0) = u o i (xi ) (1.12)

The governing equations represents a system of three partial dierential equations of second order in time t and of second order in space variables xi . The temperature variation term appears in the eld equation derived at the rst order while appears underived in the boundary conditions. Since the variation of the temperature is a given function of x and t, one can also see this term as a special component of the body forces Xi and the surface forces fi In fact if we assume novel forcing terms Xi = Xi Eijhk hk T,j and fi = fi Eijhk hk T nj (1.14) (1.13)

1. 1D Theory of Thermoelasticity

the governing equations of the linear thermoelastic problem will reduce to the linear elastic 3D problem, provided that Xi and fi are assumed as forcing terms. Once solution is obtained for the displacement ui , the strain ij can be obtained from the strain-displacement relations and the stress from the constitutive equations. In general it is extremely dicult to obtain a closed form solution for the governing equations in terms of displacement. This is due to the intrinsic diculty of nding solutions of partial dierential equation for general shape of the volume occupied by the body. Alternative approaches to the solution can also be followed, as the one in which the stress ij or the strain ij are assumed as unknown of the problem. In this case the compatibility equations in terms of strain, that represents the integrability conditions of the strain-displacement equations for obtaining continuous ui from assigned ij , need to be imposed too. As well known closed form solution can be obtained for just few cases of linear elastic problem. The same is true for the linear thermoelastic problem. For practical purposes of using their model of the thermoelastic physics for quantitative evaluation of the mechanical response of a body under thermal actions, procedures for approximate solutions need to be set up. One family of numerical methods consider directly the governing equations of the problem and, in order to transform the dierential nature of the equation and reduce the complexity of a volume of general shape substitute the dierentials with the nite increment of the unknown functions in a nite increment of the variables xi . The governing equations are written many times in correspondence to every point identied by the grid of increments xi . A family of mechanical techniques, called nite dierences, follows this kind of approach. Intuitively, the consequences to the closed form solution is reached progressively as the nite increments and on a smaller and smaller grid of xi variables. An alternative approach for nding approximate solution is based on the use of another form of the governing equations, that is possible to demonstrate to be equivalent to them, the virtual work equation.

1.2

Thermoelastic problems in 1D solids

Let us now consider the thermo-elastic problem in a 1D solid, like the one depicted in gure, a bar subjected to a load per unit length p(x, t) and to a temperature variation T (x, t) with respect the ambient temperature. A xed end is assumed at the left and x = 0, with no lack of generality, since only rigid body motions are suppressed due to it. The body has section A(x), density per unit length (x) and is thermally isolated in any directions normal to x and is supposed to have a linear thermoelastic behavior with an elastic modulus E (x) and an the thermoelastic response of their body expansion coecient (x) can be studied by assuming as state variables the displacement function u(x, t), assumed to

1. 1D Theory of Thermoelasticity

p(x,t)

T(x,t)

Figure 1.1: The thermoelastic problem in a 1D solid.

be continuous along x, the axial strain (x, t) and the normal force N (x, t). The governing equations of the problem can be set up as follows: strain/displacement relation u(x, t) x equilibrium equation (x, t) = N (x, t) = p(x, t) x obtained by setting the translation equation along x N (x, t) dx + p(x, t)dx = 0 x
2

(1.15)

(1.16)

(1.17)

u where p(x, t) = p(x, t) including the inertial term constitutive equation (linear thert2

moelasticity) N (x, t) = E (x){(x) (x)T (x, t)} (1.18)

where T (x, t) = (x)T (x, t) can be dened as thermal strain and the dierence (x, t) T (x, t) = M (x, t) as mechanical strain. Considering the boundary conditions we can either apply the imposed displacement, the imposed force or a combination of the two: u(0, t) = uo (t) u(l, t) = ul (t) N (0, t) = N o (t) N (l, t) = N l (t) (1.19)

1. 1D Theory of Thermoelasticity

In the case an elastic spring of stiness K or a pointwise translational mass M is applied to one of the ends, say the left one at x = l, we will have

K
ul -Kul

M
ul
M 2u t 2

Figure 1.2: Dierent boundary conditions at one end of the structure.

N (l, t) = Kul as to the initial condition u(x, 0) = ui (x)

or

N (l, t) = M

2u t2

(1.20)

(1.21)

The resulting governing equation can be obtained by substituting in the equilibrium equation the remaining two, thus obtaining, expressed in the primitive state variable u(x, t), the following expression u(x, t) 2 u(x, t) {E (x)A(x) (x)T (x, t) } = p(x, t) + x x t2 with, as initial condition on x = 0 u(0, t) = u0 (t) or E (0){u(x, t)x |0 (0)T (0, t)}A(0) = N0 (1.23)
[ ]

(1.22)

and as boundary condition on x = l u(l, t) = ul (t) or E (l){u(x, t)x |l (l)T (l, t)}A(l) = Nl (1.24)

or other combination of the primitive and its rst derivative as in the case corresponding to the spring or the point mass. The governing equations described alone (eld equation and limit condition) have a linear nature. In the static case the equation reads as follows u(x) {E (x)A(x) (x)T (x) } = p(x) x x
[ ]

(1.25)

if the body has an homogeneous nature both for the elasticity and for the thermal expansion phenomena we can further simplify the expression: E u(x) A(x) A(x)(x)T (x) = p(x) x x
( )

(1.26)

1. 1D Theory of Thermoelasticity

with the boundary conditions u(0) = u0 u(l) = ul or or EA(0){ux |0 T (0)} = N0 EA(l){ux |l T (l)} = Nl (1.27) (1.28)

Note that T (x) appears under a rst order derivation in the eld equation. The governing equation is in this case an ordinary second order dierential equation that, if A(x) = const, is also characterized by constant coecients. Let us now try to nd the solution in simple cases for which p(x) is assumed to be zero. If also A(x) = const, the eld equation reads EAu (x) EAT (x) = 0 with A(x) = const T (x) = T = N (l ) = 0
x ()

(1.29)

u(0) = 0

(1.30)

In this case the equations read, with () = EAu (x) = 0 u(0) = 0 N (l) = EA u l T = 0
( )

and () =

2 () x2

(1.31) (1.32) (1.33)

The solution has the form u(x) = A + Bx with A and B obtained by imposing the boundary conditions.

1. 1D Theory of Thermoelasticity

1.3

Examples

Example n.1

l T(x)
T

u(0) = 0

O u(x)
T

x
l EA

N (l) = 0 T (x) = T p(x) = 0 u(0) = A = 0

O (x)
T

x
u (l) T = B T B = T

O N(x)

u(x) = T x (x) = T
( )

N (x) = EA T T = 0

O (x)
T

x
T

1. 1D Theory of Thermoelasticity

Example n.2

l T(x)
T

u(0) = 0

O (x)
T

x
T

u(l) = 0 T (x) = T p(x) = 0

O u(x) O (x) O N(x) O

x
u(x) = A + Bx u(0) = A = 0

u (x) = B

u(l) = Bl = 0 u(x) = 0 (x) = 0

N (x) = EA T = EAT

x
EA T

1. 1D Theory of Thermoelasticity

10

Example n.3

l T(x)
al

x
u(0) = 0 N (l) = 0 T (x) = ax

O T(x)
al

p(x) = 0 EAu (x) = EAT (x) = EAa

O u(x) O (x)
al

x
1 al 2 2

u(x) = A + Bx + cX 2 u (x) = B + 2Cx u (x) = 2Cx = a u(0) = A = 0 u (x) = B = 0 1 ax2 u(x) = 2 (x) = ax N (x) = EA (ax ax) = 0 C= a 2

O N(x) O

1. 1D Theory of Thermoelasticity

11

Example n.4

l T(x)
al

x
u(0) = 0 u(l) = 0 T (x) = ax

O T(x)
al

p(x) = 0 EAu (x) = EAT (x) = EAa

O u(x) O (x)

l2 a 8

u(x) = A + Bx + cX 2 u (x) = B + 2Cx u (x) = 2Cx = a C= a 2

x
a 2

O N(x) O

x
EA a 2

u(0) = A = 0 a a 2 l B= l u(l) = Bl + 2 2 a a 2 u(x) = lx + x 2 2 a 1 u (x) = l + ax u ( ) = 0 2( 2 ) a a N (x) = EA l + ax +ax = EA l 2 2

The expression for T (x) can also be written as T (x) =

Tb l x

1. 1D Theory of Thermoelasticity

12

Example n.5 u(0) = 0 u(l) = 0 T (x) = Ta + p(x) = 0 This case can be solved by superimposition of eects by superimposing the solution for Tb Ta x l

l T(x)
al

x
T1 (x) = Ta and the one for

O (x)
T

x
al

T2 (x) =

Tb Ta x l

obtained for the example 2 and 4 respectively, as follows

O u(x)
a l2 8

u(x) = u1 (x) + u2 (x) that is

O (x)

a 2

u1 (x) = 0 1 (x) = 0
( )

O N(x) O

N1 (x) = EA T = EAT

x
u2 (x) = 2

x
EA a 2

Tb Ta Tb Ta lx + x2 l 2l ( ) ( ) Tb Ta Tb Ta 2 (x) = + x 2 l ( ) a Tb Ta N2 (x) = EA l 2 2l In conclusion Tb Ta Tb Ta lx + x2 l 2l ( ) ( ) Tb Ta Tb Ta (x) = + x 2 l a N (x) = EA (Ta + Tb ) 2 u(x) = 2

1. 1D Theory of Thermoelasticity

13

Example n.6 u(0) = 0

K
l T(x)
T

N (l) = kul

T (x) = T p(x) = 0 u(x) = A + Bx

O T(x)
T

u(0) = A = 0 N (l)) = EA u l T u (x) = B

O u(x) O (x) O N(x) O

x
1 EA Tl 1 + kl / EA

u(l) = Bl from which we obtain EA B T B=

x
1 EA T 1 + kl / EA

= kBl

T EA/l EA/l + k

x
EA/l T x EA/l + k EA/l T (x) = EA/l + k ( ) EA/l N (x) = EA T T EA EA/l + k u(x) = that for k 0 will be equal to N (x) = 0 while for k will be equal to N (x) = T EA

x
1 1 EA T 1 + kl / EA

The stiness k may represent the stiness of another bar of dierent characteristics, measured at its end, as in the following picture. In this case the bar a has a variation in temperature T whereas b remains indisturbed from the thermal viewpoint.
T 0 T =0

x la lb

1. 1D Theory of Thermoelasticity

14

Example n.7

e x l
In the presence of a small gap e the boundary conditions change from N (l) = 0 to u(l + e) after the displacement has reached the value e. It is a nonlinear case for which, in case e is small as compared to l one way just charge the boundary condition and add the solution obtained for displacement and strain to the one with xed right end. Example n.8 T (x) = T

l T(x) O T(x) O u(x) O (x) O N(x)

x
u(0) = 0 N (l) = 0 x T (x) = sin l p(x) = 0
( )

The solution is of the kind u(x) = A + Bx x cos /l l ( ) x B + sin l ( ) EA u (x) T (x) = EAB =0 A= A /l /l EAB = 0 B = 0 ( ( )) x 1 cos /l l ( ) x sin l ( ( ) ( )) x x EA sin sin =0 l l
( )

u (x) = N (x) =

u(0) = N (l) = u(x) =

(x) = N (x) =

1. 1D Theory of Thermoelasticity

15

Example n.9 u(0) = 0 u(l) = 0 x T (x) = sin l p(x) = 0 EAu

l T(x) O (x)
T

upart (x) =

x
EAa

u part (x) = 2 x cos l2 l

= a =

O u(x)

upart (x) =

x EAT = EA cos l l ( ) x a cos l ( ) 2 x a 2 cos l l ( ) x EA cos l l /l ( ) x cos /l l

O (x) O N(x)

u(x) = u (x) =

u(0) = u(l) =

u(x) = (x) = N (x) =

x cos A + Bx /l l ( ) x B + sin l A =0 A= /l /l 1 + Bl + = 0 B = 2 /l /l /l l ( ( )) x x 1 2 cos /l l l ( ( )) x 2 + sin /l l l l 2EA

1.4

Thermoelastic eect in a 2D bending beam

Let us consider a beam of unitary width in the x, y plane subjected to a variation of the temperature T (y ) = T (h/2) = T1 . +
T1 +T2 h z

uniform in the x direction, with T (h/2) = T2 ,

1. 1D Theory of Thermoelasticity

16

y,v T2 h z l T1 x,u +

T1 + T2 2

Figure 1.3: Thermoelastic eect in a 2D beam.

The corresponding thermal strain will be

T = T (y ) = with

T 1 + T 2 T + 2 h

(1.34)

T = T2 T1

(1.35)

The case can be studied by separately consider the uniform strain along y (that has been already considered in the analysis of the bar) and the linear symmetric variation along y . Let us now briey recall some elements of the 2D beam theory. DISPLACEMENT FIELD ASSUMPTIONS

y,v

dw0 dx

x,u

Figure 1.4: The rotation eld in a 2D beam.

u(x, y ) = u0 (x) (x)y v (x, y ) = v0 (x) STRAIN DISPLACEMENT RELATIONS

1. 1D Theory of Thermoelasticity

17

x (x, y ) =

du0 (x) d(x) y x (x, y ) = 0 (x) y dx dx y (x, y ) = 0

xy (x, y ) = 0 with d2 w0 Curvature dx2 du0 Axial Strain dx

= 0 =

EQUILIBRIUM EQUATIONS (bending equations only)

y M

p(x) M+dM x

T+dT

Figure 1.5: Forces and moments acting on an dx element of a 2D beam.

M (x) =

dT (x) dx dp(x) T (x) = dx

from 2D continua:

h 2

M (x) =

h 2

x (x) y dy

(1.36)

CONSTITUTIVE EQUATIONS From 2D continua

x = E (x T ) (notice that T expresses a variation of the temperature)

(1.37)

1. 1D Theory of Thermoelasticity

18

M (x) =

h 2

h 2

Ex ydy +

h 2

h 2

ET ydy

(1.38)

and by substituting the strain-displacement equations:

M (x) =

h 2

h 2

E0 (x)y Ey 2 dy +

h 2

h 2

h 2

T 2 y dy = h

(1.39)

= E

h 2

T y 2 dy + E h

h 2

T y 2 dy = EI + E h

In case p(x) = const = 0 and T (0) = T (l) = 0 also T (x) = 0 and M (x) = 0. In this case (thermal action only acting on the beam) we will have
( )

T EI + h

= 0 = alpha

T h

(1.40)

that in fact, corresponded to a curvature like in gure, where the top bers of the beam elongate more due to the higher temperature.

y,v

x,u R

Figure 1.6: The deformed 2D beam.