[juho]

How could I create an adaptive timestepping that would restart the calculation from the previous time step?

I'm having trouble making a satisfactory system just based on the values of the already calculated timestep (like the Courant number) and would like to adjust the time step and resolve if nonconverging/nonphysical values are met.

Time.H seems to have functions to reset the time and to create subcycles. Could these be used? How are the previous timesteps stored?

[benk]

I know this post is a ocuple of years old now but I'm interested in doing this too.

I want to use large time steps in my model but if I find that the model starts to diverge, I want to restart based on the last succesful time step and readjust my dt to a smaller one.

I know I can use runTime.setDeltaT() to set my time step, but how do I tell the solver to restart based on the last succesful time iteration?

[benk]

So I think I figured out how to do this but maybe there's a better way.

The command runTime.setTime(lastTime, lastTime); will reset the simulation to the last good time step as long as you set the argument lastTime equal to the last good time step using something like lastTime = runTime.value(); after each succesful time iteration.

[benk]

Actually, upon further investigation, it seems as if the method I described above will reset the time but it doesn't take the solution from that old timestep, it will continue with the previous bad solution...so now the question is: How can I use the last good solution?