the inversion of large impedance matrix in MOM

Hi all

I try to analyse a structure using method of moment which produces large dense squared impedance matrix. when I try to solve the matrix equation A*X=B, the inversion procedure takes more than 4GB RAMs.Is there any idea to solve this matrix equation in reasonable time with less RAMs?

[Walter Roberson]

How are you solving A*X=B ?

[rahman]

X = (A^-1)*B !!!

[Walter Roberson]

Use X = A \ B; instead.

[rahman]

I try it. But unfortunately nothing changed. It used same amount of RAMs! Is there any programming or mathematical techniques?

[Steven Lord]

How large is your "large dense" matrix?

Are you using a 32-bit or 64-bit version of MATLAB?

Can you easily compute A*x for a vector x without explicitly creating the matrix A? If so, use one of the iterative solvers like GMRES.

[rahman]

Tnx Steven Lord, GMRES is a good idea. It reduced about 2GB RAMs.

Answer 1

[Walter Roberson]

svd with 'econ' mode and see http://adrianboeing.blogspot.ca/2010/05/inverting-matrix-svd-singular-value.html

[rahman]

I try it but it takes more time with same RAMs!

[Walter Roberson]

Then the answer to your original question is NO. Install more RAM if the current size is a problem.