Local vs Global CFL

[jose.simongreminger]

Dear all, 

I am running a transient simulation of a partially-filled rotating cylinder with a explicit VOF formulation. Therefore I a monitoring both the global CFL and the maximum local CFL. My global CFL is around 0.15 but my maximum local CFL is around 0.5. I wanted to ask you if I can increase my maximum local CFL, as long as I keep my global CFL below 0.5 (to keep a stable simulation). Moreover, I also trying to imlpement surface tension effect. However, when I introduce surface tension between the water and the air, I have a large numerical diffusion at the interface when I start rotating. This numerical diffusion at the interface only starts when I start rotating the cylinder, and seems to be independent of the timestep that I take. I am using a constant value for the surface of  0.072N/m, CSF model, Wall adhesion.

[Rob]

How are you rotating the cylinder? Time step will be limited by cell rotation, in that you can't pass cells in a step. How well resolved is the free surface region, and how big (roughly) is the domain? 

[jose.simongreminger]

Hello Rob, 

Thank you for your answer. 

I am rotating the cylinder by applying a circular acceleration on the XY plane, and tilted gravity force on the Z axis. My domain is a cylinder with a 10mm diameter and a 15mm height. The resolution of the free surface is 0.1mm. If I understand you correctly I should then always keep the maximal local CFL below 0.5, even if the global CFL is at 0.15 and could be increased, right?

I needed to go this resolution to avoid numerical diffusion (without surface tension). But then when I introduce again surface tension, I have numerical diffusion at the interface again. 

[Rob]

The resolution may need to be increased around the free surface but surface tension forces are going to dominate. Are you using geo-reconstruct for interface tracking? Compressive is good, but you may be on the limits for it's accuracy. 

CFL is a useful measure, but also check the time the rotation takes to cross 0.1mm (cell size) at the (mesh) interface as that's not checked by VOF.