Scribd
Upload
33 views

FEM Lecture2

This document discusses variational formulations for computational fluid dynamics problems. It introduces variational principles where the stationarity of a scalar functional leads to the solution. Variational finite element formulations result in integral equations and systems of equations that can be solved for unknown variables. The document outlines how to develop a variational principle for a linear differential equation, and how the finite element approximation is inserted into the variational formulation. It also discusses least squares formulations that can be used when a variational principle cannot be derived. Variational and least squares formulations both result in a symmetric system matrix.

Uploaded by

sanjay
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
33 views

FEM Lecture2

This document discusses variational formulations for computational fluid dynamics problems. It introduces variational principles where the stationarity of a scalar functional leads to the solution. Variational finite element formulations result in integral equations and systems of equations that can be solved for unknown variables. The document outlines how to develop a variational principle for a linear differential equation, and how the finite element approximation is inserted into the variational formulation. It also discusses least squares formulations that can be used when a variational principle cannot be derived. Variational and least squares formulations both result in a symmetric system matrix.

Uploaded by

sanjay
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

COMPUTATIONAL FLUID DYNAMICS: FEM: Variational Formulations

Lecture 21
VARIATIONAL FORMULATIONS

21.1 VARIATIONAL PRINCIPLES


For certain class of problems in physics, it is possible to define a scalar quantity (for example,
potential energy for a problem in mechanics) whose stationarity leads to the solution of the
problem. Formally, a variational principle specifies a scalar functional Π defined as an
integral given by
¶u ¶u ¶u ¶u
P = ò F (x, u, , ,...)dW+ ò G (x, u, , ,...)dG (21.1)
W
¶x ¶y G
¶x ¶y
where u is the dependent variable, and F and G are differential operators. The solution of the
continuum problem is the function u which makes this functional Π stationary with respect to
arbitrary changes δu, i.e.   0 .
Thus, if a variational principle can be found, then it is straightforward to establish an integral
form suitable for finite element analysis. However, in practice, it is very difficult to obtain an
expressions for functions F and G from the differential equation of the problem, except for
linear differential equations. For instance, if L were a linear self-adjoint operator in the
differential equation
L(u ) + f = 0 (21.2)
then F  u ( Lu  2 f ) , i.e. functional Π can defined as

P = ò u ( L(u ) - 2 f )dW (21.3)


W

21.2 VARIATIONAL FINITE ELEMENT FORMULATION


Let us assume that an approximate solution u of Eq. (21.2) expressed in the form

u (x) » u(x) = å Ni (x)u (xi ) (21.4)


i

where Ni are prescribed functions, called interpolation (shape or trial) functions and ui is still
unknown value of variable u at a discrete spatial point xi. We can insert this approximation in
Eq. (21.1) and express the stationarity condition as
¶P
dP = å dui = 0 (21.5)
i ¶ui

The preceding equation holds for any variation δu, which is possible only if

¶P
= 0 for all i. (21.6)
¶ui
Nodal values ui can be determined by solving the system of equations (21.6).

Note that if the functional Π is quadratic (in function u and its derivatives), then Eq. (21.6)
results in a linear system
Ku = b (21.7)

Dr K M Singh, Indian Institute of Technology Roorkee NPTEL 21.1


COMPUTATIONAL FLUID DYNAMICS: FEM: Variational Formulations

where K is a symmetric matrix. In fact, if a variational principle exists, variational finite


element formulation would result in a symmetric system matrix. In such problems, even a
Galerkin finite element formulation would result in a symmetric system.

Further generalizations of the variational formulation include constrained variational forms


based on (i) Lagrange multipliers and (ii) penalty functions. These forms are used to enforce
additional constraints including boundary constraints. For further details, see Zienkiewicz,
Taylor and Zhu (2005).

20.3 LEAST SQUARES FORMULATION


Least squares approach can be used to define a scalar functional for any PDE to circumvent
the difficulty associated with derivation of a variational principle. Thus, for the differential
equation L(u ) + f = 0, the least squares functional can be defined as

P = ò ( L(u ) + f )( L(u ) + f )dW (21.8)


W
Use of usual finite element approximation in conjunction with minimization of the above
functional again leads to a system of equation (21.7) with a symmetric matrix K.

REFERENCES/FURTHER READING
Muralidhar, K. and Sundararajan, T. (2003). Computational Fluid Dynamics and Heat
Transfer, Narosa Publishing House.
Reddy, J. N. (2005). An Introduction to the Finite Element Method. 3rd Ed., McGraw Hill,
New York.
Zienkiewicz, O. C., Taylor, R. L., Zhu, J. Z. (2005). The Finite Element Method: Its Basis
and Fundamentals, 6th Ed., Butterworth-Heinemann (Elsevier).

Dr K M Singh, Indian Institute of Technology Roorkee NPTEL 21.2

You might also like