Star-CD transient simulation, problem of loss of mass
 

I have tried several times to simulate a transient turbulent flow through a duck with conjugated heat transfer.
I tried two kinds of BCs, 1) pressure inlet, fixed mass flux as outlet; 2) fixed velocity inlet, flow split (1.0) as outlet.
For both case, the time step was set in the range of micoseconds, thus the maximum Courant number was less than 10. In each simulation, I check the mass conservation, and found out that the mass flux at both inlet and outlet decrease after each time step, therefore, after a long simulation (thousands or hundred of thousand), the mass flux goes to ZERO!!

I am so disappointed with Star-CD. How can it be?

Did anyone here encountered the similar problem with Star-CD transient simulations?

The user manual states "it is particularly important to avoid" a combination of prescribed pressure and flow-split outlet conditions.

Try a fixed mass flux inlet & flow split (1.0) outlet.

Thanks Pauli. Actually I have tried every possible BC, including the fixed inlet mass flux (provide velocity components, density & inlet temperature) and split outlet. It failed!

For a lower velocity or incompressible case, the calculations seem also acceptable. My problem is compressible with a peak velocity of about 400 m/s.

I run the same problem with CFX, it works quite well and much much faster.

For the steady-state, Star-CD runs well. Since in this case the tiny imbalance of mass flux between inlet & outlet is not important.

I have tried to find similar studies or comments on star-cd transient simulations, and found nothing important to me.

Two things to check.

1) For the inlet boundary condition, do you have the flow switch set to mass or velocity. If you want a specific mass flux, it should be mass. Velocity will be adjusted based upon the pressure/density required by the domain.

2) Where is the reference pressure cell & what is it's value. This is very important for compressible flow problems. I don't remember what is the official "correct" method. I tend to place my reference pressure cell close to the outlet. That way I can peg the outlet boundary's pressure to a known value (usually ambient for my work).