[Velocity-coupling] - why not least squares ?

Lionel S.

Hi all,

It seems that velocity-coupling is a hudge problem. When using coupled solvers, the linearization can lead to a singular matrix.

SIMPLE is a good idea, but personally I had good results only at low reynolds numbers (less than 200). At high reynolds numbers, it seems that instabilities appears, especially with pressure boundary conditions.

Now I read several articles who told that anoter solution to avoid the problem of singular matrix is getting more equations than unknowns and then solving it with least squares method.

So why not ? Why is it not used more frequently by commercial solvers ? Did anyone try a least square method in CFD ?

Kindest regards

Lionel