[Binoy]

Hello Everyone,

I am working on modeling offshore submarine soil block movement in CFX. I need some suggestions regarding the RMS Residual target value and time step size. I have a block of 1 m length and 0.1 m height. Total simulation time is 4 s. When block starts to move, it forms different pattern (shape) when I use different time steps. Say, when I use 0.01 s as timestep, the shape of the soil block differs from the simulation using 0.05 s as timestep. Also when I change the RMS residual value, block forms different shape for different RMS residual value. For that case I need some suggestions:

a. Timestep: How do I need to fix the time step size? I know about one formula for time step size that related with inlet velocity and smallest length of the element. But in my model left and right boundary conditions are no slip wall. And there is no initial velocity, block in slope will move due to gravity. Is there any formula for this case to fix the time step size.

b. RMS residual target: In CFX solver modeling guide there is a discussion about the residual target value. The default value is 1e-4, and it is for loose convergence. For good convergence value is 1e-5 and 1e-6 is for very tight convergence. But when I used RMS residual value greater than these (1e-2 or 1e-3) the result seems quite well than from result using 1e-4. My question is that, Can I use a larger value of RMS residual target value than the specified values in the CFX guide?

c. Minimum and maximum Coeff. loops.

Please give some suggestions.

I am sorry for my grammatical and spelling mistakes.


Thanks in Advance.

Best regards,
Binoy

[ghorrocks]

To set time step:
Either - Do a simulation at a certain time step, then repeat it with half the time step. If the results are the same within a tolerance you are happy with the time step is OK. If not, then you need to halve the time step again and continue halving until you obtain convergence.

Or - Use adaptive time stepping to find the time step size for you. Set the time step size the adaptive, homing in on 3-5 coeff loops per iteration. Make sure the max and min time step sizes are wide enough it never hits them, and make sure the initial time step is reasonable (if unsure, guess a small time step).

My recommended approach is adaptive time stepping as it easily accounts for changes in the simulation.

You comment that 0.01s and 0.05s time step gives significantly different results says these time steps are too big. You need to go smaller.

Residual - likewise, do a sensitivity study. Do a simulation at a given residual tolerance, then repeat with 10x the tolerance (either tighter or looser). If the results are the same within a tolerance you are happy with then you are OK. If not then tighten the tolerance by 10x and repeat. Keep going until you have obtained the same results on two consecutive refinements.

Maxs/Min coeff loops - for most simulations min=1, max=10 is fine.

[Binoy]

Thank you Glenn for your suggestions.