KikeM / romtime

Unsteady Finite Element Reduced Order Model for Time Moving Domains
1 stars 1 forks source link

ENH : BDF-2 implementation #2

Closed KikeM closed 3 years ago

KikeM commented 3 years ago

See references:

KikeM commented 3 years ago

I am finally implementing this because I have realised how slow it is to produce results due to the small time step I need to take (1e-3) to produce meaningful results.

Reminds me of Steven's words about better algorithms: "Computers have peaked, so we can only reach more detailed solutions with better algorithms".

KikeM commented 3 years ago

Not surprisingly, I can now cut by half the number of timesteps and yet obtain a better mass conservation.

Well done!

I am quite happy with this feature, because I consider it is the kind of functionality one might inject since the beginning but it doesn't actually change anything about the Reduced Order Model by itself. Thus, I only implemented it when the number of time steps became a problem.

Let problems show up before you fix them ;)