I used Level set to simulate an uprising bubble in liquid. But the bubble is getting smaller and smaller.

Does anyone has this problem? I do not know the origin of the mass loss. It is from the Level set Eq. or from the accuracy of N-S

Thanks a lot

This is well-known problem in LSM (due to numerical diffusion), (probably your literature review is very poor) thare are several treamtment as cure:

1. increase spatial resolution

2. increase accuracy of your JH solver (e.g. high order WENO + TVD RK)

3. Use Hybrid particle level set method for reducing diffusion (is really effective way)

4. Use spatially addaptive refinement near interface.

5. Use VOF instead of LSM !

Thanks for your suggestion

Then it means it has no connection with NS solver

I used WENO+RT for LS eq. but it does not improve

I am thinking of using higher scheme for convective term in NS solver.

This happen because of the volume dissipation. an alternate method is Allen-cahn model with lagrange multiplier or use of chan-hilliard model to control the volume dissipation of the bubble.

>Then it means it has no connection with NS solver

it is clear that we expect mass-inbalance in the same order as we solve NS, in clear word based on residual of continuty satisfaction,

but LSM has usually severe mass lose, so sometimes bubble can be completely dissipated during advection. This is completely related to LSM not NS, e.g. if u switch from LSM to VOF u can preserve mass in accceptable manner (usually below 1% inbalance can be acheived).

>I used WENO+RT for LS eq. but it does not improve

I am not agree !

do you experience that with increase order of WENO mass loss is decreased, if no, certainly your imlementation is problematic because this fact is well-documented in literatures.

>I am thinking of using higher scheme for convective term >in NS solver.

Why?

if your velosity field is divergence free (bleow prescribed treshold), NS solver is fine ! (from mass-inbalance view point)

For simplicity u can forget NS and stick only to advection step, e.g. use rigid body advection test such as Zalesak's circle in the prescribed divergence-free velosity field, then study effect of high-order HJ solver and sitution of mass loss.

Actually, I am using Cahn-Hilliard equation. The mass loss is severe.

I am trying to find the problem. I used steady flow field, there is no mass lost

Are you redistancing the level set after each time step? I'm asking it because there are some mass correction algorithms that can be used when redistancing/rebuilding the level sets -- I have some references talking about it... the problem is to find ;o) .

Cheers

Renato.

Mr. rt

Thank you so much

If possible, could you give me your mail address, I^d like to send you some results of my simulation.

Thanks

He

rohtav@gmail.com