What happens by doing Bandwidth Reduction?
 

Hey friends,

In fluent, what happens when u do bandwidth reduction by grid>reorder>domain

Why and when do u do it?

Thanks

After you read your mesh in Fluent you can do the bandwidth reduction. Basically Fluent will renumber the nodes of your mesh in order to have a "more diagonal" array. This will allow you to use less memory and also the solution will converge faster.

Do

Hey DoHander,

So basically, it means that one should make it a practice to always do bandwidth reduction after reading any mesh into Fluent...

If you use any of the implicit solvers of Fluent yes, you should always do the bandwidth reduction. For the explicit solver this is useless.

Do

Thanks DoHander :)

I also have a question on this, till what value you need to reorder to achieve the most optimal bandwidth? Seem to me that the lower the bandwidth value the better?

Yup, lower the bandwidth value the better the solution will get in terms of solution time!! But you cannot control the value to which the bandwidth (of matrix) can be reduced, as it depends on the number of non-zero entries in the matrix(the matrix generated as the fluent solves the N-S equations)

details can be found here...

http://en.wikipedia.org/wiki/Banded_matrix

This also clearly shows why you cannot have advantage of reordering for explicit solver...

details can be found here...

http://en.wikipedia.org/wiki/Banded_matrix

This also clearly shows why you cannot have advantage of reordering for explicit solver...

To me it is not clear why there is no advantage to reorder explicit solvers. Would you care to elaborate?

Thanks :)

/Rasmus

Implicit solvers solve liner systems: e.g. Ax=b for the unknown x. When A has a lower bandwidth, certain algorithms (i.e. Gaussian elimination) are faster.

Explicit solvers don't solve linear systems, they simply compute Ax=y for a new y. Regardless of the bandwidth of A, it's simple algebra.

Ah, i see. Thanks for the reply :-)