Discussion Forum

Ok, after lots of playing with parameters I solved my own problem? Under the Study>Solver Sequence>Stationary>Segregated>Segregated Step 2 (mod1_ep, mod1_k for my turbulent flow problem), I changed the number of iterations to 5 instead of 3.

Also, what led me to this solution was the discovery that the entire error message might be important. The error was something like

"Singular matrix.
Segregated solver"

...so I started to play with the segregated solver.

Thanks!

Joe

Link to COMSOL's explanation of the singular matrix error

www.comsol.com/support/knowledgebase/953/

In case anyone is having problems like this, I'm just going to keep posting solutions that I find in my COMSOL experiences.

Singular Matrix? Try running the simulation a second time. Sometimes it will solve the second time without error

Thanks,

Joe

Hi

try a "get initial values (3.5) or compute to here step by step, and check your initial conditions, this is the simplest way I have found to avoid singular matrices dues to missing data or BC inconsistencies.

Probably youmena run a "restart" the second time (from first initial conditiones updated), because just running twice from the same situation should not change the issue

have fun Comsoling
ivar

I'm not sure how to "restart" my solution in COMSOL 4.0..

As for BCs, in case someone notices an inconsistency or possible issue, I have
fluid inlet condition of average normal velocity
fluid outlet condition of no viscous stress and 1[atm] of pressure
HT BC 1000W in at a 1/8" diameter spot, and a
HT BC temperature BC of 287K at the flow inlet.

So, so far my cooling is incredibly poor, causing my block with 1000W to be ~1.5e5 K. Is it possible that this insane temperature is causing problems?

Thank You,

Joe

These are the things I've learned about Singular Matrix errors so far. Maybe this will be helpful to someone else, or just reading them in a group like this I will find the common denominator

-Sometimes just changing initial conditions a little (i.e. initial temp 5000K to 10000K) will allow the problem to solve

-Is a mesh is too coarse, or a mesh steps from one domain to another, this will cause a singular matrix problem

-According to Matlab:
-- A determinant w/ a row or column of zeros has a value of 0 -> singular matrix
-- A determinant w/ two rows or columns equal has a value of 0 -> singular matrix
--(I wish i knew how these applied to CFD/HT in comsol)

-Material properties can affect if you get a singular matrix result (i.e. use of molybdenum instead of copper)

-Removing the heat transfer section of my conjugate HT and flow problem while maintaining the same mesh results in a converging solution, where it would normally solve as a singular matrix problem.

Thanks,

Joe