[djerezg]

Hi to all,

anyone can help me to understand better how "local-time stepping (LTS)" works?

I know its a kind of pseudo transient method but actually i dont know how this method choose his delta t for each iteration (for example in ansys fluent you know how the method set this delta t.)

If some one have some document or information, i'd be very grateful.

Greetings!

[clapointe]

This information was provided when it was first incorporated : https://openfoam.org/release/2-0-0/steady-state-vof/. Basically, the solution is accelerated to steady state because a time step "field" is computed, so each cell has its own timestep instead of one global timestep. The timestep field is smoothed (and changes between iterations are damped) to maintain stability.

Caelan

[djerezg]

Originally Posted by clapointe

This information was provided when it was first incorporated : https://openfoam.org/release/2-0-0/steady-state-vof/. Basically, the solution is accelerated to steady state because a time step "field" is computed, so each cell has its own timestep instead of one global timestep. The timestep field is smoothed (and changes between iterations are damped) to maintain stability.

Caelan

TY for you answer!
i already knew that info, but how this time step field for each element is calculated? or just take the maximun time to keep the courant number in each element?

Greetings!

[clapointe]

At the bottom of the link there are paths to source files -- so you know that there is one to compute the time step "field" on the solver end and other files in src to facilitate. So check out interFoam for a modern version and we see a setRDeltaT.H file (actually, there is a general collection of VoF-related files in the multiphase solvers folder). This is where the (inverse) time step field is computed. It is commented to designate smoothing, etc.

Caelan

[djerezg]

Originally Posted by clapointe

At the bottom of the link there are paths to source files -- so you know that there is one to compute the time step "field" on the solver end and other files in src to facilitate. So check out interFoam for a modern version and we see a setRDeltaT.H file (actually, there is a general collection of VoF-related files in the multiphase solvers folder). This is where the (inverse) time step field is computed. It is commented to designate smoothing, etc.

Caelan

I tried to read that code, but was difficult to me understand very well that info.
Can you (or anyone who knows and want to) help me to understand better that code with words, please?

Greetings!

[clapointe]

The comments give us the gist of what is happening in setRDeltaT.H. Working through essentially gives us -- compute "timesteps" for each cell based off a max courant number, smooth this timestep field (in space) to improve stability, and then damp this timestep field (in time, as in limit changes) to further improve stability. As far as I'm aware, these steps are included in each instance of setRDeltaT; other damping may be applied, e.g. for the multiphase solvers based on interface location, or for reacting solvers based on heat release. Hope this was a bit clearer.

Caelan

[djerezg]

Great! that explain a lot.

Now im wondering, what is the criteria in Ansys Fluent, because when you use pseudo transient method in a steady simulation, you cant choose a Courant number. In that case, the under-relaxation is controlled through the pseudo time step size, but there is not relation with courant number, or in that case just use a time for each iteration until the program converge?

thank you for your help, is very usefull for me.

Greetings!

Edit: I found this https://www.afs.enea.it/project/nept...underrelax-eqn
I hope this i'd be usefull for someone else.

Cheers!

[granzer]

Originally Posted by djerezg

Great! that explain a lot.

Now im wondering, what is the criteria in Ansys Fluent, because when you use pseudo transient method in a steady simulation, you cant choose a Courant number. In that case, the under-relaxation is controlled through the pseudo time step size, but there is not relation with courant number, or in that case just use a time for each iteration until the program converge?

thank you for your help, is very usefull for me.

Greetings!

Edit: I found this https://www.afs.enea.it/project/nept...underrelax-eqn
I hope this i'd be usefull for someone else.

Cheers!

Here is another discussion that I found helpful: What are "Pseudo Time URFs"?