Boundary conditions in the method of the finite elements
Néstor Durango
Engineering & Development northern University. 1: 24-28, 1995
Introduction
The foundation for the solution of
physical problems by numerical
methods is the correct approach and
the adequate solution of a system of n
equations, that allow to determine the
values that the variable of work takes
in n discrete points or nodes.
These equations are posed for each of
the nodes, and they are of the form.
Ci0i + Ca0a + Cb0b + Cd0d + Ce0e +
Cf0f = O
where:
0j it is the variable or function that is
going to be determined;
Cj are coefficients, whose values
depend on the relative position of the
neighboring swimmers a, b, d, e, f with
respect to node i.
Cjs also depend on the condition of
node i: inner node, or border node.
One of the difficulties in the solution of
physical problems through numerical
methods is the determination of the
coefficients related to the border
nodes. This is particularly true when
the finite difference and finite element
methods are used, in which
techniques that require great care for
their correct application are adopted;
but even then logic and programming
errors are made when these methods
are systematized.
The purpose of this article is to present
the equations that allow to determine
the Cj coefficients for the interior nodes
and for the border nodes, in a way that
show the sequence that facilitates the
user of the finite element method to
propose the algorithms for its
systematization.
The application of the equations
presented here is limited to physical
problems involving the determination
of a potential function 0 in a region or
field R. This situation occurs, for
example, in fluid flow with potential
heat transfer by conduction shear flow
in a continuous medium, current in an
electric field, where it is fulfilled that:
In simple cases, the value of 0 in the R
region can be determined analytically;
or its value can be found graphically, if
it is not required of great accuracy; or
through the use of numerical methods,
in more complex cases. Finite element
method When this method is used, the
R region is divided into subregions R '.
The greater the number of sub-
regions, the greater the accuracy, but
the greater complexity in the approach
and in the solution of the problem.
The subregions are separated by line A= [01(Y2- Y3)+02(Y3- Y1)+03(Y1-
segments, and these concur in the Y2)] ID,
nodes; so each node is associated B= [01(X2-X3) +02(X3-X1) + 03(X1-
with several sub-regions, and each X2)] ID.
sub-region is associated with several
nodes. Developing the above expression for
the node i, which is associated with the
With the finite element method it is subregions delimited by the
assumed that 0 varies in each neighboring nodes a, b, d, e, f, results:
subregion, according to the
relationship
Ci0i + Ca0a + Cb0b + Cd0d + Ce0e +
Cfm = o.

Ca = (Yi- Yf)(Yf- Ya)+(Ya- Yb)(Yb- Yi)+

(Xi - Xf)(Xf - Xa) + (Xa - Xb)(Xb -Xi).
Cb = (Yi - Ya)(Ya - Yb) + (Yb - Yd)(Yd
-Yi) + (Xi - Xa)(Xa - Xb) + (Xb -Xd)(Xd
- Xi).
Cf = (Yi- Ye)(Ye- Yf)+(Yf- Ya)(Ya- Yi)+
(Xi - Xe)(Xe - Xf) + (Xf - Xa)(Xa -Xi).

which is the equation of a plane. A, B,

e, are constants in each subregion and
their values depend on the values of 0
in the nodes associated with R '. If R 'is
associated with nodes 1,2,3, then

01 = AX1 + BY1 + C,
02 = AX2 + BY2 + C,
03 = AX3 + BY3+ C.

This system of equations can be

solved for A and B; if the determinant
of the system is: In each coefficient Cj the X and Y
coordinates of the nodes neighboring
D = X1(Y2- Y3) + X2(Y3- Y1) + X3(Y1-
node j are related.
Y2),
This is how in Ci the X and Y
coordinates of all the neighboring
nodes of i: a, b, d, e, and f appear
related. and in Ca, for example, the
coordinates of the neighboring nodes
are related to node a: i and f.
Border conditions The system of n
equations with n unknowns obtained
from applying the condition '120 = 0 in
the nodes of the R region, has a non-
trivial solution if there are consistent
boundary conditions.
First class frontier conditions The
nodes with known potential meet
them; they are frontiers of first class in
heat transfer, surfaces with known
temperature; In flow of fluids with
potential and in flows in porous media,
it is the impermeable surfaces. This
boundary condition is taken into
account by writing to the node j the
equation
0j =check
Second class border conditions They
are fulfilled by the nodes in which the
potential gradient is known, such as,
for example, surfaces with known heat
flux in heat transfer; cross sections
where the average velocity is known,
in flow of fluids with potential, and in
flows in porous media.
A special case is when the potential
gradient is zero: adiabatic surfaces in
heat transfer.
This boundary condition is taken into
account by applying:
(a0/an)k = V = constant, to the node k
of the figure, in direction n.
In the coefficient of the inner node Cb
the nodes d and a are related; and for
the coefficient Cd the nodes b and e
are related; but neither in one case nor
in the other does node k relate.
Third class border conditions This
condition is fulfilled by the nodes in
which the relation of the potential to its
gradient is known. This is the case of
convective surfaces in heat transfer;
cross sections where both the average
velocity and the pressure are known,
in flows with potential and flows in
porous media.
At the borders of the third class, apply
(O / n) k = - (h / k) Ok to the node k,
shown in the previous figure, obtaining
as a result:
(Ck + Rk [h/kJ )0k + Ca0a + Cb0b
+Cd0d + Ce0e = O
Special border condition
The conditions of the nodes
neighboring a sink or an infinite source
are considered special, where both the
potential and its gradient are known.
In this class of borders V \, 20 = O is
applied to the node m, associated with
the nodes a, b, d, e, obtaining as a
result:
Cm0m + Ca0a + Cb0b + Cd0d + Ce0e
= O.
The coefficients in the previous CONCLUSIONS
equation are:
When the finite element method is
used to solve physical problems with
Cm = (Ya - Yb)2+ (Yb- Yd)2+ (Yd - potential it is important to identify with
Ye)2+(Xa-Xb)2+ (Xb-Xd)2+ (Xd-Xe)2. certainty the different kinds of
boundary involved in the problem.

Ca = (Ya - Yb)(Yb - Ym) + (Xa- Xb)(Xb- Once the boundary classes have been
xm). defined, the equations of
corresponding nodes can be
proposed, carefully following the
sequential relationships of each
Cb= (Ym-Ya)(Ya-Yb)+(Yb-Yd)(Yd-
coefficient
Ym) + (Xm - Xe)(Xa - Xb) + (Xb -
Xd)(Xd - Xm). The rule is that in each coefficient Cj
the X and Y coordinates of the
neighboring rivers are related to j; but
Cd= (Ym-Yd)(Yb-Yd)+(Yd-Ye)(Ye- in the second and third border class
Ym) + (Xm - Xd)(Xb - Xd) + (Xd - conditions, the neighborhood criterion
Xe)(Xe - Xm). is special.

Ce= (Ym-Yd)(Yd-Ye)+(Xm-Xd)(Xd- It is possible to systematize partially, or

Xe). in its entirety, the approach of the
problem and its solution using simple
computational tools.

Bibliography

G. SOTELO. General Hydraulics.

Mexico, Limusa, 1976.
In the proper coefficient, Cm, all the KARLEKAR Heat transfer. Mexico,
neighboring nails are related to the Interamericana, 1986.
node m: a, b / d, e. In the other
coefficients, the coordinates of the
neighboring rivers are related.
RESUMEN.

Podemos observar que en los fundamentos para la solución de problemas físicos

mediante métodos numéricos es el correcto planteamiento y la adecuada solución
de un sistema de n ecuaciones, que permitan determinar los valores que toma la
variable de trabajo en n puntos discretos o nados.

Método de los elementos finitos cuando se emplea este método, se divide la región
R en subregiones R'. A mayor número de sub regiones, mayor exactitud, pero mayor
complejidad en el planteamiento y en la solución del problema.

Las subregiones están separadas por segmentos de recta, y éstos concurren en los
nodos; asi que cada nodo está asociado con varias subregiones, y cada subregión
está asociada con varios nodos.

Con el método de los elementos finitos se supone que 0 varía en cada

subregión, de acuerdo con la relación

0_AX .• DY

El propósito de este artículo es presentar las ecuaciones que permitan determinar

los coeficientes Cj para los nadas interiores y para los de frontera, de una forma
que muestren la secuencia que facilite al usuario del método de elementos finitos
plantear los algoritmos para su sistematización .La aplicación de las ecuaciones
que se presentan aquí se limita a los problemas físicos que involucren la
determinación de una función potencial 0 en una región o campo R.

Esta situación se presenta, por ejemplo, en flujo de fluidos con potencial y

transferencia de calor por conducción, flujo de cortante en un medio continuo,
corriente en un campo eléctrico,donde se cumple que:

Condiciones de frontera

El sistema de n ecuaciones con n incógnitas obtenido de aplicar la condición '120=

Oen los n nodos de la región R, tiene solución no trivial si se dan condiciones
consistentes de frontera.

Condiciones de frontera de primera clase

Las cumplen los nodos con potencial conocido; son fronteras de primera clase en
transferencia de calor, las superficies con temperatura conocida; en flujo de fluidos
con potencial y en flujos en medios porosos, lo son las superficies impermeables.

Esta condición de frontera se tiene en cuenta escribiendo para el nodo j la ecuación

0j = constante

Condiciones de frontera de segunda clase

Las cumplen los nodos en los cuales se conoce el gradiente del potencial, como
por ejemplo, superficies con flujo de calor conocido en transferencia de calor;
secciones transversales donde se conoce la velocidad media, en flujo de fluidos con
potencial, y en flujos en medios porosos.

Un caso especial se tiene cuando el gradiente del potencial es cero: superficies

adiabáticas en transferencia de calor. Esta condición de frontera se tiene en
cuenta aplicando:

(a0/an) k = V = constante,
al nodo k de la figura, en dirección n.

Condiciones de frontera de tercera clase

Esta condición la cumplen los nadas en los cuales se conoce la relación del
potencial con su gradiente. Es el caso de superficies conectivas en transferencia de
calor; secciones transversales donde se conocen tanto la velocidad media como la
presión, en flujos con potencial y flujos
en medios porosos.

CONCLUSIONES

Cuando se usa el método de los elementos finitos para resolver problemas físicos
con potencial es importante identificar con certeza las diferentes clases de
frontera involucradas en el problema.

La regla es que en cada coeficiente Cj se relacionen las coordenadas Xy Y de los

nados vecinos a j; pero en las condiciones de segunda y tercera clase de frontera,
el criterio de vecindad es especial.

Es posible sistematizar parcialmente, o en su totalidad, el planteamiento del

problema y su solución utilizando herramientas computacionales sencillas.