Suggested procedure to produce training data from hydro simulations

[tgrassi]

Hydro codes with operator splitting have (1) (thermo)chemistry and (2) hydro operators.
The aim is to emulate the (thermo)chemistry but not the hydro.
Hydro codes usually produce initial and final values at the beginning and at the end of the hydro timestep.
Since CODES needs temporal sequence values (trajectories), the default output produced by hydro codes is not useful.
A possible overcome could be:

• Run the hydro simulation normally and store the variables at the beginning of each hydro timestep for every cell or particle
• Optional: reduce the stored data (there will be a lot of similar if not identical data)
• Find the largest hydro timestep of the hydro simulation (dt_max)
• Run (thermo)chemistry as an IVP for each point for dt_max with N number of intermediate steps (e.g., 10 or 100)
• You now have a set of trajectories of length dt_max, with N timesteps each, one trajectory for every initial condition stored.