[schnitzlein]

As a newbie to OpenFoam I started to set up a simple example, i.e. laminar flow (no turbulence model) in a rectangular channel with inlet at one end and outflow at the other. At the wall the respective boundary condition is prescribed.
Despite the solvers for U seem to converge quite rapidly (I set the relative errors to 0 for U and p iterations) I encountered an ever increasing value for the continuity error. I tried to modify the grid, the superficial velocity, etc. but I failed to obtain a converged solution.
Any help is greatly appreciated.

[mattijs]

Compare your case to e.g. simpleFoam/pitzDaily. Run that one without turbulence and see what happens.

Do you have empty patches? If so is your mesh only one cell thick and perfectly aligned?

[schnitzlein]

I ran pitzDaily without turbulence, i.e.
turbulenceModel laminar
turbulence off
and obtained similar results, i.e. ever increasing
values for the time continuity errors.

[mattijs]

So is your case steady 'in real life'? Is it turbulent? Are you trying to simulate a non-physical problem (e.g. simulate a turbulent flow without a model for the turbulence)

simpleFoam might converge if your flow is steady. But if it isn't you can run turbFoam and see if that reaches a steady state.

[eugene]

Not surpised that pitzDaily blows up, it is a turbulent flow. If you do not have enough damping in the form of modelling and/or numerics you will get a build up of turbulent energy in your velocity field. Basically, the flow will just become more and more unstable untill it blows up. There are many good reasons why turbulence models have to be used, and this is one of them.

Try increasing your viscosity until your duct is in the laminar regime i.e. Re(b) < 1000~3000.

[eugene]

And/Or the instabilities are causing inflow through your outlet boundary, which is probably why you are getting the continuity errors. Using a properly specified inletOutlet boundary should "fix" this.

[schnitzlein]

Thank you for your comments and suggestions.
According to preliminary test runs the problem
may be may be solved at least partially by specifying an inletOutlet boundary.