Scribd
Upload
3 views

ProjectA

The document outlines a project for MATH-H-301 students focused on implementing numerical methods in C or FORTRAN. Students must submit their work by December 5, 2021, and present it on December 13 or 15, with evaluations based on code correctness, efficiency, and clarity. The project involves discretization of a boundary value problem, implementing a two-grid method, a multigrid method, and adapting the multigrid method for preconditioning, with specific requirements for documenting and presenting the work.

Uploaded by

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

ProjectA

The document outlines a project for MATH-H-301 students focused on implementing numerical methods in C or FORTRAN. Students must submit their work by December 5, 2021, and present it on December 13 or 15, with evaluations based on code correctness, efficiency, and clarity. The project involves discretization of a boundary value problem, implementing a two-grid method, a multigrid method, and adapting the multigrid method for preconditioning, with specific requirements for documenting and presenting the work.

Uploaded by

hinoxgame
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Numerical Methods

(2022 – 2023)
Project

This project statement is for students who were previously


registered for the MATH-H-301 course.

Rules

Projects are strictly individual.


The work consists in a properly documented software code written in C or FORTRAN
which implements the numerical method described in the project statement, as well as in
a short report summarizing the experiments made with the code, including the answer to
the specific questions raised in the project statement.
The work has to be submitted by December 5, 2021, before noon (12 hours), by email to
Artem Napov ([email protected]).
Each student will have to present shortly (10 min.) his/her project on December 13 or 15,
according to a schedule that will be announced by December 9. In the subsequent question
& answer session, the jury of the project will in particular make sure that the student is
the author of the code.
Beside the presentation itself, the jury of the project will also evaluate the correctness of
the code, its efficiency, generality and clarity, as well as the content of the report.
Remark: if the submitted project is obviously below minimal requirements, an insufficient
mark will be attributed based on the submitted material only (without oral defense).
It is allowed to use or take inspiration from general purpose routines available on the
Internet. In such a case, however, proper credits should be given in the report, which
should also document which modifications where made and how the interface with the
borrowed routines was implemented.

1
Project statement

1 Discretization (0 points)

Consider the discretization of the boundary value problem



−∆u = 0 on Ω ⊂ R2 ,
boundary conditions on ∂Ω ,

with the domain Ω and the problem parameters as specified in your MATH-H-301 project.
The boundary conditions are also the same as in your MATH-H-301 project, except if
they are all homogeneous (because the problem in this project was initially an eigenvalue
problem). In (and only in) this latter case, exchange homogeneous boundary conditions
on ∂Ω p
u = exp( x2 + y 2 ) on ∂Ω .

2
2 Two-grid method (4 points)

Implement the solution of the linear system with a proper geometric two-grid method,
using Gauss–Seidel smoothing (either standard or in symmetric/symmetrized form) and
solving the coarse grid problem with the direct solver UMFPACK.
For UMFPACK, in http://metronu.ulb.ac.be/MATH-H-301/1819/projet.v0.tar.gz,
you will find the package together with an example of use.
Hint: start with the implementation of stationary iterations with (symmetric) Gauss-
Seidel method (either as a preconditioner or as a solver).
Check the convergence of the method by plotting the residual norm against the number of
two-grid iterations; use semi-logarithmic scale.

3 Multigrid method (12 points)

On the basis of the previously obtained two-grid method, implement the solution of the
linear system with a multigrid method.
Compare the convergence with that of the two-grid method by plotting the corresponding
residual norms against the number of iterations; use semi-logarithmic scale.
Note : your multigrid method implementation is expected to run on large enough instances
of your discretized PDE problem. However, limiting the code to a series of selected problem
sizes, chosen to enable a generation of a proper multigrid hierarchy, is advised.

4 Multigrid preconditioning (4 points)

Adapt your implementation of the multigrid method to use it as a preconditioner.


Implement the preconditioned conjugate gradient method, and use it with the multigrid
method as a preconditioner.
Compare the convergence with that of the standalone multigrid method by plotting the
corresponding residual norms against the number of iterations; use semi-logarithmic scale.

You might also like