Time Step estimators

[AlejandroMunozManterola]

The manual and seemingly random selection of Local Time Stepping (LTS) for problem definition can be trivialised by the implementation of a simple LTS algorithm that takes into account the minimum distance between two nodes found through all the elements of the Finite Element Space, the order of the Finite Element Collection, the speed of the signal and the Courant–Friedrichs–Lewy coefficient.

[AlejandroMunozManterola]

A basic algorithm has been implemented.

If the user chooses to, they can input a manual time stepping value which will be respected. If no input is entered, the solver will automatically calculate the LTS based on the following formula:

$\Delta t = {{CFL {}\cdot{} h} \over {N^{3/2}{}\cdot{} c}}$

Where CFL is the Courant–Friedrichs–Lewy coefficient, which can be manually adjusted with the .setCFL(double) method, h is the minimum internode distance, N is the Finite Element Collection order and c is the signal speed.

[lmdiazangulo]

Maybe we could use power method to extract an estimation of $\Delta t$ . Check mfem's example 9

[AlejandroMunozManterola]

2D estimators have been implemented for triangular meshes, following the inscribed circle estimation.

See commit 926b570 for more info.

Code compiles but feature does not work due to issue with Mesh in const Model.