Math 615 (Applied) Continuum Numerical Analysis

Ed Bueler, Spring 2010 UAF


Instructor:     Ed Bueler       Chapman 301C
Phone: 474-7693   eMail: elbueler@alaska.edu
Office Hours at
  www.dms.uaf.edu/~bueler/OffHrs.htm
Class Time: MWF 1:00--2:00
Classroom: Chapman 104.
Web Site: http://www.dms.uaf.edu/~bueler/

Course Description, Goals, etc:  3.0 credits.  Methods for approximating partial differential equations (PDEs) and related problems on computers.  PDEs are the underlying equations for most problems of flow, fields, thermodynamics, deformation, quantum stuff, curvature, and etc.   Mathematical analysis of these methods.  Conceptual frameworks for understanding numerical analysis of such problems.  Both practical and abstract approaches to problems, including discussion of application contexts.

Most classtime will be spent with my lecture, with Matlab/Octave/Pylab (MOP) demonstrations when I can fit them in.  I will help you started with MOP, but you must show initiative in learning to do actual numerical computation.  Homework assignments and a student-chosen project will include practical parts and will require actual implementation (in MOP).  Abstract thought is, however, essential in order to understand the choices (e.g. among numerical methods) one faces in solving major problems.  Thus all homework assignments will have mathematical exercises, and in these you will frequently be asked to "show" and "prove".  Formal proof style is not important, but clear presentation of sufficiently-general logical arguments is important.

The emphasis is on finite difference methods.  I will also gloss spectral methods and finite elements.  We will seek to think in terms of vectors and matrices, and not just manipulate lots of numbers in a computer program.  Instead of a list of finite difference schemes, for instance, we will seek the underlying linear algebra structure.  The course will include some real nonlinear examples, for which one inevitably uses a sequence of approximating linear problems.

Thus, at the end this course you will not be a professional numerical analyst.  But you will be able to evaluate and use many numerical tools for solving scientific and engineering problems, and you will be able to code some of the basic methods (i.e. for the purpose of prototyping more serious solutions).  Furthermore you will have some of the mathematics necessary to take the next steps to learn mathematical finite element method, spectral methods, parallel-izable matrix methods, and so on, and some appreciation of the mathematics of Fourier analysis and stability analysis.

Calendar:  A day-to-day tentative schedule for the semester is a
http://www.dms.uaf.edu/~bueler/Math615S10.htm

Prerequisites:  Informally: undergraduate ordinary differential equations, undergraduate linear algebra, exposure to the basic ideas of numerical analysis, and exposure to Fourier series and separation of variables (for solving the classical linear PDE boundary value problems).  Also some exposure to computer programming.

Formally
: The prerequisites are MATH 302, MATH 310, MATH 314, and MATH 421 or permission of the instructor.  The list in the catalogue is in error.  CS 201 and MATH 422 are not specifically needed, though they are nice things to know.

Textbook:  The required text is K. Morton and D. Mayers, Numerical Solution of Partial Differential Equations, 2nd ed. Cambridge Univ. Press 2005. There are, of course, many other textbooks on numerical analysis of PDEs, but I actually like this one.  We will cover chapters 1, 2, 3, 4, first half of 5, some of 6, and a little of 7.  Four other texts are recommended, of which two are freely available online a page at a time.  See http://www.dms.uaf.edu/~bueler/Math615S10.htm

Your Grade = Homework + Project:  Sixty percent of the course, and the grade, will be based on nearly weekly homework assignments.  Here is where you will learn the mathematics and gain breadth and perspective.  You will be asked to think abstractly on some problems and to use MOP on many others.   (Expect to produce at most a couple of half-page-long MOP programs per assignment.)

The last assignment will be worth double the value of a previous assignment, and it must be worked on solo.  It may be regarded as a take-home final exam.

It is assumed that students in this class have in mind (or can acquire) specific continuum  modelling problems in applied fields.   These will mostly, but not exclusively, be PDE problems, and they are supposed to be nontrivial problems.  Frequently they are a significant simplification of a thesis/dissertation project, for instance.  I am eager to help and advise on choosing and refining such problems.  Forty percent of the grade in the course will be on a project based on such a problem.  Two project assignments will be given, one due midsemester, and one due at the finals time, with the first part a preparatory stage for the more complete second part.  Both mathematical analysis and actual numerical computation will be required on your project.  Furthermore, at least one presentation of the project will be required during the semester.  The presentation can be either oral or on a poster.  These presentations are important to the class, as the class will act as consultants to the presenter.
The course grade will be determined by points on the homework and project, according to the schedule at right  --->
I will use plus/minus grades as indicated.
Percent
91 --100 %
 
88 --  90 %
84 --  87 %
  76 --  83 %

73 --  75 %

  69 --  72 %
 
61 --  68 %
 
57 --  60 %
  41 --  56 %
 
0 --  40 %
Grade
A
A-
B+
B
B-
C+
C
C-
D
F

Policies and makeup exams:
   The department has specific policies on incompletes, late withdrawals, and early final examinations, etc; see http://www.dms.uaf.edu/dms/Policies.html.  You are covered by the UAF Honor Code.  I will work with the Office of Disabilities Services (208 Whitman, 474-5655) to provide reasonable accommodation to student with disabilities.

Programming in the course:  You will use Matlab, Octave, or Pylab (=Python+scipy+matplotlib), called "MOP" from now on.  Matlab is commercial while Octave and Pylab are free and open source.  Octave is so close to Matlab that I can confidently recommend it as a very usable clone.  Programs in MOP will appear on my website, and these can be used in homework problems and in projects. Copious resources are available for learning Matlab/Octave , with many for Pylab also, but in more geeky form.  See my modest Matlab/Octave tutorial and links page (www.math.uaf.edu/~bueler/MatlabEx.htm).

Students with no programming experience, or fear of fiddling around at a command line, will have a high hurdle to cross.  The programming experienced in Math 310 should be sufficient as preparation, however, as should be any standard computer science programming course.  Students who are very well-established and secure in programming environments are encouraged to learn both Matlab/Octave and the Pylab tools.  (Use of other languages than these, like C and FORTRAN, is fully the students responsibility, and, in fact, may cause substantial disadvantage.)

An ad for the MOP languages might look like this:  MOP is a language designed to do numerical analysis coursework.  Programs can be written and run in MOP in a highly traditional programming style.  Mathematical and graphical inputs and outputs can be handled more directly in MOP than in most compiled programming languages.  Matrices appearing in problems can be easily analyzed.  Many of the operations appearing in numerical problems are natural and quick in MOP, and require much more work in compiled languages like C or FORTRAN.  Even established compiled-language programmers will find it a desirable prototyping tool.