Development of An Uel Subroutine in Abaqus For A Fully Coupled Heat Transfer, Mass Diffusion and Stress Analysis. Implementation To Elastic Polymers

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

2002 ABAQUS Users Conference 1

DEVELOPMENT OF AN UEL SUBROUTINE IN ABAQUS


FOR A FULLY COUPLED HEAT TRANSFER, MASS
DIFFUSION AND STRESS ANALYSIS.
IMPLEMENTATION TO ELASTIC POLYMERS.
G. RAMBERT
a
, J.C. GRANDIDIER
a
, L. CANGEMI
b
, Y. MEIMON
b

a
Laboratoire de Mcanique et Physique des Matriaux, E.N.S.M.A.
1, Avenue Ader - BP 40109 - 86961 FUTUROSCOPE CHASSENEUIL Cedex France
b
Institut Franais du Ptrole, Division Mcanique Applique
1&4, Avenue du Bois Prau - 92852 RUEIL MALMAISON FRANCE
Abstract: In different industrial fields, polymer structures have to face high temperature and high gas
pressure conditions, inducing gas adsorption and diffusion within them. A sudden drop in external pressure
leads to an irreversible damaging of the material, due to the coupling of mechanical, thermal and gas
diffusion effects. Even if this phenomenon was deeply investigated on an experimental point of view, the
models that were carried out so far do not take all the different coupling possibilities into account. A
thermo-diffuso-mechanical model was thus developed in the framework of the generalized standard media
in order to build up more accurate prediction tools. It compares the system to an open homogeneous two-
component continuous medium. This model was particularly applied to an elastic mechanical behavior of
the material. Its numerical implementation in ABAQUS consisted in developing an UEL subroutine that
takes the fully coupled thermo-diffuso-elastic behavior of the polymer into account. This should lead to a
first qualitative estimation of the coupling effects.
1. Introduction
During the past twenty years, the use of polymers expanded widely in technical structures. In the petroleum
industry for instance, numerous achievements were then made possible among which the one we
particularly focus on: flexible pipes used for oil transportation. In operation, these structures have to face a
wide range of stresses (100Mpa, 350-420K) due to ambient environment and transported gas, which
usually saturates the polymer sheath sealing of the conduit. When the pipe feeding stops, sudden drop of
the inner pressure leads to damages in the sheath. This phenomenon is due to coupling between
mechanical, thermal and gas diffusion effects. Such damage is of course unacceptable for sunk/buried
pipes.
When one aims at foreseeing the in-service behavior of such sheaths, it becomes necessary to build
predictive tools that enable global behavior to report. Thus, we developed a first simple and prospective
model which gives back the behavior of a material subjected to mechanical loads, gas absorption and
temperature gradient (either induced or due to the outer medium). The implementation of such a model in
2 2002 ABAQUS Users Conference
ABAQUS was restricted to elastic material. A first qualitative estimation of the thermo-diffuso-mechanical
coupling is thus possible.
This paper presents the basics of the thermo-diffuso-elastic model as well as the main features of its
development in ABAQUS. Numerical results will be presented during the conference.
2. Presentation of the problem
Generally speaking, a high-pressure gaseous environment leads to gas adsorption then diffusion within the
material it is contacted with. Polymers are subjected to such a phenomenon, which induces the material
dilatation and alteration of its mechanical properties. When the outer pressure is suddenly released, namely
an explosive decompression, gas which is trapped within the material tends to expand and diffuses toward
the outer medium correspondingly (at least often) inducing a local decrease in temperature. Irreversible
mechanical damage generally appears as consequence of coupling between mechanical, thermal and
diffusion effects. Based on the nature of gas-polymer system, decompression rate, thermal conditions, such
damage appear either locally as cracks or in a diffuse way as foam.
On an experimental point of view, this phenomenon was subjected to numerous studies. Two examples are
given hereafter for information. A study carried out by Lorge (1999) reports the behavior of
polyvinylidenefluoride subjected to explosive decompression in a carbon dioxide environment. For several
types of polymer, Dewimilles work (1993) highlights the nature of damage caused by decompression as
well as the modification of global mechanical behavior, which is associated to this phenomenon.
On a theoretical point of view, the three aspects were generally discussed separately: mechanical problem,
thermal problem or diffusion problem. The literature shows a certain amount of mathematical models for
diffusion, based on statistical thermodynamic, and which differ from one another in the choice of internal
variables (Doghieri, 1996). It is important to note that these models do not make it possible to translate the
thermo-diffuso-mechanical coupling.
On a numerical point of view, a first modeling of such a phenomenon was proposed. For instance, Jarrins
and Gaillard-Devauxs works (respectively 1994 and 1995) define a first damage criterion based on
simplifying hypotheses. Nevertheless, the alterations of mechanical properties due to impregnation as well
as temperature variation that occur during decompression are not taken into account. Coupling is moreover
only partially discussed. Actually, ABAQUS provides the *COUPLED TEMPERATURE-
DISPLACEMENT option for analyzing thermo-mechanical or diffuso-mechanical problems with the
possibility of considering respectively the diffusion or temperature effects on the mechanical and thermal
or diffusion properties. But, the fully coupled, simultaneous heat transfer, mass diffusion and stress analysis
is not possible.
As a first step, we found important to build a global model that takes into account the whole coupling panel
(mechanical, thermal and diffusive couplings). Therefore, the generalized standard media were considered.
As a second step, the development of an UEL subroutine was necessary. In order to simplify this model
implementation in ABAQUS, a purely elastic problem was considered. Such a simplification will obviously
not make it possible to bring face to face the computation and experimental results on a quantitative point
of view, but will enable qualitative comparison of the coupling effects.
2002 ABAQUS Users Conference 3
3. System representation: gas and polymer mixture
3.1 Thermodynamic framework
The model set-up is carried out in the framework of generalized standard media and therefore follows a
classical approach. A field of state variables represents the material behavior. Introduction of a first
potential (specific free energy) makes it possible to set laws linking thermodynamic forces (stress, entropy,
chemical potential) to dual variables (strain, temperature, normalized concentration). The dissipations that
are only associated to thermal and diffusion phenomena in an elastic behavior case are discussed using a
second potential (dissipation potential). This potential expression depends on the envisaged coupling (fully
or not). Once defined, the potential expression leads to complementary evolution laws (Fouriers, Ficks).
These different steps are presented in the following paragraphs.
In this thermodynamic framework, every elementary representative volume (ERV) is considered at the
macroscopic scale to be a bi-component continuous, homogenous medium: polymer and gas. Therefore, the
volume characterization can be carried out using average physical value based on each species value. One
focuses on the average behavior of the medium.
Moreover, this ERV is open and thus trades heat power with the surrounding medium as well as matter.
The two-species diffusion is a molecular diffusion, which obeys a Ficks type law and then is driven by
concentration gradients. The sorption mode that is associated to each specie obeys a Henrys law at the
domain boundary.
Besides, the material is assumed to be isotropic and is only studied in the framework of small perturbations.
Finally, a strong development assumption leads to associate each ERV characterization magnitude to a
constant value in time and space. This latter approximation is by no means verified but makes it possible to
strongly simplify the equations.
3.2 Local equations
The behavior of polymer subjected to pressure, gas concentration as well as temperature variation is driven
by diffusion, dynamic and thermal equations. The two first equation families can be classically set up but
the third thermal one is not so easy since it must embed couplings corresponding to dissipative phenomena.
3.2.1 Mass balance
In a bi-component system where no chemical reaction occurs within it, each species mass remains
separately the same, which leads to two mass balance equations.
The species put together are described by the following variables:
average mixture mass density (kg/m
3
).
i C normalized concentration of the i
th
constituent (N/m) and defined as the rate between
concentration (ppm, i.e. mg/kg) and solubility coefficient S
i
(ppm.m
2
/N).
i m J

relative mass flux of the i diffusing phase (kg/m.s).


4 2002 ABAQUS Users Conference
Mass balance for i
th
component leads to the following diffusion equation:
i
i
i m
6
J div -
dt
dC
S 10

= .
This classical balance equation is applied to polymer and gas. Nevertheless, for flux and normalized
concentrations, the following equations come true by definition:
p g m m J - J

= and
dt
dC
S -
dt
dC
S
p
p
g
g = .
Both the diffusion equations are linked and thus only the gas one will be used further in this document:
g
g
g m
6
J div -
dt
dC
S 10

= . (1)
3.2.2 Mechanical balance
The balance equations are derived from fundamental principle of dynamics. When one neglects the inertia
effects (
0


), mechanical balance leads to the following motion equation:
0 f v i d

= + , (2)
where is the total Cauchy stress tensor (N/m
2
), f

is the body force at any point within the ERV per unit
of mass (N/kg) and

the considered particle acceleration (m/s).


3.2.3 Thermodynamic balance First and Second laws
a. Thermodynamic potential, equations of state
It is classically seen in continuous media thermodynamic to introduce the specific free energy of the
mixture (J/kg). For an elastic behavior this potential depends on three state variables: the temperature T
(K), the infinitesimal strain tensor
e
and the gas and polymer normalized concentrations g C and Cp ,
respectively.
) C , C , T , ( p g
e
= .
When a normal configuration is assumed and when one considers the first two thermodynamics laws,
system behavior laws also known as equations of state can be derived from . When specific entropy is
denominated s (J/kg.K), mass chemical potential for i specie (dual variable of the mass fraction) is
denominated i (J/kg), one can write down these laws as follows:
2002 ABAQUS Users Conference 5
p g C , C , T
e




= ,
p g
e
C , C ,
T

s


= and
i k
e
C , , T
i
i
6
i
C


S
10


= .
Remark:
*
.
.

means partial derivation with * as constant value.


When one considers the small perturbations, that is to say infinitely small strains and small differences of
temperature and normalized concentration, the development of potential to the second order make it
possible to build the following linear behavior laws:
I )] C C ( S
10
) ( ) To T ( [ ) 2 3 ( - 2 I ) tr (
o 6
e e
g
g g p g T o + + + + = (3)
) C C ( S
10
D ) tr ( ) 2 3 (
1
) T T ( C
T
1
s s
o 6
e
o
C ,
o
g
g g 1 T
i
e
o + +

+ + =

(4)
) T T ( D - tr ) - ( ) 2 3 (
1
) C C ( S
10
D ) ( - o
e
p
o 6 o o
1 g
g
g g 2
p g
p g +

+ = (5)
where:
o T ,
o
i
c , o , o s and
o
i
respectively represent initial temperature, normalized concentration of
the i
th
constituent, stress, specific entropy and chemical potential at zero strain, thermal and mass
differences.

p g
e
i
e
C , C ,
C ,
T
s
T C

= is the specific heat at constant strain and normalized concentration


(J/kg.K).
and are the Lam elastic coefficients (N/m).
T (K), g and p are the isotropic expansion coefficients linked to respectively heat, gas and
polymer transfer.
The effect of temperature variation (concentration variation) on the chemical potential (entropy) is
linked to 1 D coefficient (J/kg.K).
2 D is defined as function of coefficients DM D g B and A , , , , , p , the two latter ones being
defined further in the document (m/s).
I is identity matrix.

6 2002 ABAQUS Users Conference
b. Heat transfer equation
Let q J

(J/m.s) be the heat flux per unit area, following into the ERV, and r (J/m
3
.s) the volume density of
heat generated externally for instance by an radiating source into, the heat balance equation taking into
account the different couplings has the following shape:
[ ]
dt
dC
S
10
) - ( D T )
dt
d
( tr T ) 2 3 ( r J div -
dt
dT
C
g
g 1 T q
i
e
g
6
p
e
C ,
+

+ + =

(6)
The last part of it corresponds to temperature variation linked to gas diffusion within the polymer material.
Remark: tr represents the trace operator.
c. Dissipations
Thermodynamics Second law makes it possible to define the volume dissipation (positive or nil) linked
to entropy generation. An elastic behavior is characterized by the fact that it has no internal variables,
which leads to a nil intrinsic dissipation. is split into two contributors, 1 and 2 (J/m
3
.s), which are
linked to respectively the temperature and chemical potential gradients, so that:
0 2 1 + = , T d ra g . J - s 1

= and ) - ( d ra g . J - g p g 2 m =

,
where:
=
=
2
1 i
m i q ) J - J (
T
1
J i s


is the entropy flux (

J/s.K.m
2
).
Compliance with Second laws leads to the definition of a dissipation potential d and its dual d* (through
Legendre-Fenchel transformation) (J/m
3
.s). Depending on the considered coupling assumptions, these
scalar value, continuous, convex and positives functions have a different expression. Evolution equations
are given depending on how d* is considered in the following paragraph.
d. Coupling processing
In case of a total coupling between dissipative phenomena, the dual potential turns to be a unique function,
which is given by equation 7. When the normal dissipativit assumption is done, one sets the evolution
complementary equations corresponding to equation 8.
) ] - [ d a gr ,- dT a gr - (
d d
p g
* *

=

(7)
) - ( d a gr -
*
s
p g
) dT a gr (-
d
J

,
T d a gr -
*
m
) ] - [ d a gr (-
d
J
p g
g

= (8)
2002 ABAQUS Users Conference 7
Moreover, when the dissipation phenomena are assumed to be driven by linear and fully coupled laws, the
potential d* has a quadratic form and lead to the following law let us assume the isotropic hypothesis and
the formerly stated expression of the chemical potential (equation 5):
T d a gr A ] [tr d a gr B c d a gr S 10 A - J TD DM g D g
e
6
m

+ + =
] d[tr a gr )] - ( B T C A [ dc a gr S 10 )] - ( A T C D [ - dT a gr )] - ( A - T B [ - J
e
p
6
p p q g DM T DM g g D T 2 g TD TD + + + =


where:
D A is defined to be the product between the diffusion coefficient and the average mixture mass
density (kg/m.s)
DM B corresponds to the volume deformation gradient effect on mass flux (kg/m.s)
DM A is based on and , , , p g coefficients (

m
2
/s
2
)
2 D which was formerly introduced is expressed as a function of the following coefficients:
DM D g B and A , , , , , p (m
2
/s
2
).
TD A corresponds to the temperature gradient effect on the mass flux (kg/m.s.K)
TD B is a coefficient defined out of the former coefficients (W/m.K)
When the total dissipative processes decoupling is assumed, two dual potentials
d
*
1
,
d
*
2
compliant with the
normality assumption are used:
) ] - [ d a gr (-
d
T) d a gr (-
d d
p g
*
2
*
1
*
+ =

.
As formerly stated, an important simplification of these equations consists in considering a quadratic form,
defined as positive, as the dual dissipation potential. One can then recognize well-known evolution
equations, namely Fouriers and Ficks when a simple conduction case is considered for the first one and a
diffusion case is considered for the second one:
T d a gr k - J T s

= and ) - ( d a gr k - J g g g m =


where T k is defined as the rate between thermal conductivity and temperature (J/m.s. K) and k is
easily expressed as a function of DM B and DM A (kg.s/m
3
).
e. Summary: assessment of the problem unknown values
The variables defining the thermo-diffuso-mechanical state or more precisely the thermo-chemico-
mechanical state of the material (gas-polymer mixture) are the components of strain vector (u, v, w for a
3D model and u, v for a 2D model), the temperature T and normalized gas concentration g C . Among all
8 2002 ABAQUS Users Conference
the different coefficients that were introduced, DM B and DM A are the only ones still to be determined
through a parametric study.
4. Development of an UEL subroutine
4.1 General processing
The model implementation in ABAQUS led to program an UEL subroutine, in which elements
contributions to the residual in one hand and to the Jacobian in the other hand were defined respectively by:
{ } F and [ ]
{ }
{ }
{ }
{ }

=
q
F

t
1
-
q
F
- K
n n
,
where { } qn represents the vector of nodal unknown variables associated to the n-node element of the
following dimension:
1 n 5 when three dimensions are considered
{ } { } C ... C , C , T ... T , T , w , v , u ... w , v , u , w , v , u q n n n n n
T
2 1 2 1 2 2 2 1 1 1 n = .
1 n 4 when two dimensions are considered
{ } { } C ... C , C , T ... T , T , v , u ... v , u , v , u q n n n n
T
2 1 2 1 2 2 1 1 n = .
A weak formulation based on the former equations was thus proposed for each problem (mechanical,
thermal and diffusive) considered separately. The spatial and temporal discretization of these three
formulations led to the three following contributions to the residual:
{ } { } { } ( ) q

F

F
n
m m
= , { } { } { } { } ( ) q , q

F

F
n n
T T
= and { } { } { } { } ( ) q , q

F

F n n
g g
= .
The global contribution is thus gotten through composition of the former vectors:
{ } { } { } { } ( )
{ }
{ }
{ }

= =

F

F

F
q , q F F
g
T
m
n n
and its derivation led to build the elements contribution to the Jacbian:
2002 ABAQUS Users Conference 9
[ ]
{ }
{ }
{ }
{ }
[ ] [ ] [ ]
[ ] [ ] [ ]
[ ] [ ] [ ]
(
(
(
(
(
(

=
|
|
|
.
|

\
|

|
|
|
.
|

\
|

=
K K K
K K K
K K K

q
F

t
1
-
q
F
- K
gg gT gm
Tg
TT Tm
mg
mT mm
n n
.
In this matrix, the non-diagonal terms correspond to the different couplings. The subscripts m, T and g
indicate the mechanical, thermal and diffusion processes, respectively. Note that this array is not symmetric
and thus requires the introduction of UNSYMM parameter.
4.2 Remarks about spatial discretization
An iso-parametric interpolation was chosen, in other words the same interpolation for geometry and
unknown fields (displacements, temperatures and normalized concentrations). The proposed elements are
of three kinds:
6-node triangle
4- and 8-node tetrahedron
8- and 20-hexahedron
with for two-dimensional elements a specific process for plane stress, plane strain and axisymmetric
problems.
4.3 Remarks concerning external loading
Mechanical loads as volume, surface or pinpoint force and imposed displacement may act on the ERV all
along the step.
On a thermal point of view, the ERV may be subjected to an outer radiating source. Thermal flux as well as
temperature may be set at one given point or at part of the boundary. Finally it may be subjected to surface
flux (thermal convection, conduction or radiation) defined by the following relations:
) T - T ( h - n . J a c q =

and ) T - T ( h - n . J
4 4
a
r q =

,
where n

is the unit outward normal to the surface, a T being the ambient temperature, c h and r h being
the thermal conduction/convection and radiation coefficients.
In the same way, normalized concentration or matter flux may be applied to the whole REV boundary or
punctually applied. Particularly, diffusion surface flux being compliant with:
) C - C ( h - n . J a g m dif =


was also taken into account, a C being the normalized concentration in the ambient medium and dif h the
associated coefficient.
10 2002 ABAQUS Users Conference
5. Prospects
The first numerical modeling we intend to carry out aims at getting a qualitative definition of the thermo-
diffuso-mechanical coupling effects and at evaluating the magnitude of both the unknown coefficients
through parametric study. The structures we intend to test will be subjected to pressure (thus concentration)
and/or temperature variations. One will estimate how the coupling parameters have influence on the
structures response by letting them vary. These results will be presented during ABAQUS Users
conference.
A thermo-diffuso-visco-elastic model based on Zeners rheology is currently being developed. We expect it
to bring out a good representation of mechanical behavior of polymer subjected to small strain. Then this
model will also be translated in an UEL subroutine. Comparison between numerical and experimental
results will make it possible to get a quantitative estimation of the coupling.
6. References
1. Dewimille, B., J. Martin, and J. Jarrin, Behaviour of thermoplastic polymers during explosive
decompressions in a petroleum environment, Journal de Physique IV, 3, n7, pp.1559-1564, 1993.
2. Doghieri, F., and C. Sarti, Non-equilibrium lattice fluids: a predictive model for the solubility in
glassy polymers, Macromolecules, 29, pp.7885-7896, 1996.
3. Gaillard-Devaux, E, Rupture du polythylne en temprature par dcompression du mthane, Ph. D.
Thesis, ENSM Paris, 1995.
4. Jarrin, J., B. Dewimille, E. Devaux, and J. Martin, Blistering of thermoplastic materials used in the
petroleum industry, Society of Petroleum Engineers, SPE 28482, pp.203-214, 1994.
5. Lorge, O., B.J. Briscoe, and P. Dang, Gas induced damage in poly(vinylidene fluoride) exposed to
decompression, Polymer, 40, pp.2981-2991 1999.
6. Sarti, G., and F. Doghieri, Predictions of the solubility of gases in glassy polymers based on the
NELF model, Chemical Engineering Science, 53, n 19, pp.3435-3447,1998.

You might also like