🚀 go-pugleaf

eng10524@nusunix1.nus.sg (RAMASWAMI VISHWANATH) writes:
>Dear nettors , 
>	writing a program for  a fortran77 project , i have run into a
>bad spot . the prg requires a matrix subroutine to calculate values for
>a set of equations . but the equations are not in a square matrix form .
>most books i have referred to give solutions to n x n matrices only . 
>this prg of mine ,( to calcualte deflections of beams under different
>loading conditions) uses finite differentiaiton to get a set of
>equations . thus the equations are of the form 
>		k = x + y + z

>and so on giving me a n x 3 matrix . i am yet to find any concrete
>methods to solve such a system of equations . could anybody suggest
>something ?

    Do you mean that you end up with an nx3 matrix that looks like:

    a1   b1   c1
    a2   b2   c2
...
    aN   bN   cN

or a matrix that looks like:

    b1   c1    0    0   ....
    a2   b2   c2    0   ....
     0   a3   b3   c3   ....
...
     0    0  ...        0   aN   bN


The second is typical of finite difference and finite element methods,
and is referred to as a tri-diagonal matrix. Many numerical methods
texts should give a selection of rapid solution techniques for tri-
diagonal systems.



Dave Halliwell
Department of Geography
University of Alberta
Edmonton, Alberta