[Savio Poovathingal]

Hi All,

1) I was hoping to get help with the meaning of the output Time(s) when the solver is run in steady state mode.

2) I am trying to simulate flow in a nozzle. There is high pressure at one end and very low pressure at the other and I am trying to determine the density in the nozzle after 1 micro second. I am running the code with no multi-grid, euler explicit time integration, and no CFL adaptation. Based on the output, min dT = 1.6145e-08. If I run the solver for 100 time steps and exit, my physical time should be 1.6145e-06 s. Am I correct? I am confused because the solver outputs min dT and max dT and I am struggling with the meaning of Time(s).

3) Is there an elegant way of customizing what can be output in the convergence history file and the screen? I managed to hack the source code but if there is a smarter way, without rebuilding every time, it would be helpful.

Thanks,
Savio

[talbring]

Hi Savio,

the min DT and and max DT are related to the local time stepping technique for the pseudo-time. Every cell uses its own time step to advance. minDT and maxDT just represent the minimal and maximal time-step over the whole mesh.

So it is not related to the physical time. Same for the number in "Time(s)". This is just the wall-clock time of one iteration.

Your last question: Unfortunately that is not possible at the moment.

Tim

[mhd_mrt]

Hello guys!

Quick question related to time stepping:

I am running Navier-Stokes Laminar simulation of moving normal shock in a moderately well refined mesh (~10million).
I am using the simple “time stepping” option for my unsteady simulation. And I define a time step. As Savio mentioned, unlike the inviscid case, it throws two values of min DT and max DT and a user specified DT. Are the min and max DT automatically calculated by the code using the specified CFL number and the min and max cell size of the .su2 mesh?

For the inviscid (EULER) case, my mesh was uniform and structured (squares everywhere) and the simulation took only a few hours for the normal shock to move from left to right (in the case of a shock tube). However, for the laminar viscous case (NAVIER-STOKES) I refined the mesh close the wall (higher concentration of points to capture boundary layer= 4times smaller dx near the wall compared to the uniform dx of inviscid case) and my simulation is taking wayyyyyy wayyyyyyy longer (in fact I think more than few weeks to simulate the same case of notmal shock passing from left to right).

It clearly seems to me that SU2 is ignoring my specified time step and using the min DT which is why my simulation is taking forever??! So my question is does SU2 automatically figure out that my solution will diverge if it uses my specified DT and hence decides to use the min DT instead or is there any control over how I can strictly run the simulation with my specified time step? (Even though it will diverge)

I should mention that for the sake of comparison, I am using explicit RK for both inviscid and viscous case (hence trying to avoid implicit time stepping) which could be one of the contributing reasons??!

I understand that the simulation will take longer compared to inviscid because of the extra terms (viscosity ,...) calculation.
However, I didn’t expect that it would be represented in the total iterations (in this case every 30 viscous iterations = 1 inviscid iteration in terms of same distance travelled by shocks in both cases) and would only take longer (physical time) for the CPU Cores to write the equal number of iterations as compared to the inviscid case.

I hope I am clear, if not I can post my .cfg file if that will help explaining it better.


Any advice / suggestions / similar experience is highly appreciated as I am a fairly new user of SU2 and Im trying to simulate moving shocks.

Cheers,
Mehdi

[Savio Poovathingal]

Tim: Thanks for your help. Things make sense now.

Mehdi: I suggest using the dual_timestepping scheme. Combining it with multi grid worked well for me. I am still running my case but the solution so far looks meaningful.

Savio