[roddyalan]

Hi all,

When I compute, the velocity residuals go to 10^-6 values, but the Continuity Residual stays constant at 10^-2 (SA and kEpsil Models). This is not enough for my professor. What can I do to lower it.

Thanks

[Chris D]

Can you provide some more information about the problem? Also, residuals don't really tell the entire story with respect to convergence. Are you using any other monitors, such as mass flow rate, drag, etc., to monitor your solution?

[roddyalan]

I am using Flow Rate - Velocity. It is a Backward Facing Step. (X, Y Velocity Residuals become const at 10^-6. Continuity gets constant at 10^-2). If that satisfactory for convergence.

[bramv101]

I rarely use the residuals as a convergence criteria, so I would ask yourself what are good criteria for your model and focus on these...
However I work from an industry approach so time is money, therefore we make sure that our solutions never run to long, if its converged on the criteria that is satisfying then we are happy.
If your professor has no feeling whatsoever with industry (and they usually do) then you should focus on residuals.
In that case recheck your meshsize and boundary conditions, there the first items that will influence your residuals convergence

[roddyalan]

Hi,
To be honest, i am flying blind. I am not sure what Residuals I should be looking out for. If someone can point me in the right direction, I would appreciate it.
The problem is a Backward Facing Step, using Spalart Allmaras and Realisable k-Ep Models with Boundary Conditions as inlet = velocity_inlet and outlet = pressure_outlet. The inlet velocity is 0.04421m/s and medium is water.
I am assuming he wants Continuity Residual on 10^-6, but it becomes CONSTANT at 10^-1, 10^-2. I shall check with the prof about this.

Appreciate the time and help. Thanks guys

[-mAx-]

Residuals should at least decrease below 1e-3
The double precision solver may help in your case (2ddp / 3ddp).
Convergence criteria are not only Residuals.
You may monitor the goal-value you want to compute (pressure drop, mass flow, etc...)
Your solution converged only if the monitor curve goes "flat"
Bramv101 is right, in the industry we don't have time.... but I insist you may monitor any variable.
I already experienced a well converged calculation (below 1e-5, with still-oscillating monitor (steady solution expected))

[jsm]

Hi,

Just few more informations. I think that you are solving the simulation in segregated solver. Change it to coupled solver and also play with changing URF & discretization scheme to higher order.

I hope it will help U