[quarkz]

Hi,

I tried to run a simple 6dof simulation, allowing only 1dof in pitch for an airfoil. The problem setup comes from the wingMotion2D.

I ran for 0.5s and the simulation completes w/o error.

However, I realised that if I stop at 0.1s and then resumes, at the first few steps, the forces and moments become very large. At the 1st step, it is of the order 1e5 or 1e6. It decreases as the simulation runs and got back to normal after abt 10+ to 20 timesteps. In comparison, at the same time step, if I run from zero, it is of the order 1e2.

delta_t for the resumed case is 1e-7 and it slowly becomes larger. on the other hand, if I run from zero, delta_t ~ 1e-5.

Is this expected or normal? I'm thinking some values in the calculation are different in the resumed case, at least in the beginning.

Can anyone shed light on this?

And how can I reduce such spikes?

[JulioPieri]

I'm not sure about the 6dof solver, but with combustion/radiation I also noticed that some variables did oscillate when resuming a case (Q_dot, G, etc). My conclusion was that some intermediate variables were not written, so those results were reset every begin of run.

Check how the force is calculated and see if the variables are written in files. If they aren't, you can try writing them to see if it solves the problem - the IO must be read_if_present on them.

[quarkz]

Hi JulioPieri, thanks for the info.

So what I did is to mix the old and new values with decreasing influence on the old variables as simulation proceeds. This seems to help to prevent divergence.