I'm very new to Level Set Method. Thanks to very insightful feedback/discussion in this forum, I got some idea about pros and cons of Level Set method over VOF. I have some additional questions, and would greatly appreciate your kind response.

1. In general, how faster is level set method over VOF? (I finally understood that level set is an interface tracking method, plugged into NS solver for flow field calculation, but would like to have any kind of idea about speed...)

2. Related with question #1, for capillary driven flow, how faster would be level set method than VOF?

3. Regarding mass loss in level set, how significant is it? Any rough idea in any kind of examplary case?

Many thanks~!

regarding to speed: no matter (share of interface capturing method in a NS solver is indeed neglegible), i think that CPUof PLIC vof cold be comparable with level set, note that level set needs reinitialization step which is time consuming,

mass loss is level set is very serious, if conservation is essential for you, you sould use more advance level set like particle level set, but non of level set methods can compete with VOF in terms of mass conservation

the most interesting feature of level set is its higher accuracy to compute normal and curvature and also ease of implementation.

Thank you very much for kind response. It really helps me a lot. I appreciate it...

Is the mass loss no. 1 reason that level set is not so popular in industry/engineering, though its pros?

regarding to incompressible flow, u r right, mass loss of LSM force us to use VOF

but not that LSM has its own prefarnce and is the most accepted method in free/moving boundary comunity

you may wonder if i say that, volume of publications in applied science on LSM is fairly more than couple of thousands times of VOF (google on it)

I hate to bother you, but one more question...

I saw very nice animations in Prof. Fedkiw's website, such as "Two-way solid fluid coupling with thin rigid and deformable solids (with Eran Guendelman, Andrew Selle and Frank Losasso)".

For such a simulation, what is typical CPU time on what kind of computing power (for paralell processing, how many nodes...) again, hope this question doesn't bother you. Becuase it seems that if it was modeled in VOF, it would become a huge model to capture that much of details...

Many thanks again...

Can I ask your opinion on the part "regarding to incompressible flow, u r right, mass loss of LSM force us to use VOF" you wrote. If the fluid flow is compressible, I wonder if mass loss due to LSM does not cause troubles.

Dear Fredi. Can I ask your opinion on the part "regarding to incompressible flow, u r right, mass loss of LSM force us to use VOF" you wrote. If the fluid flow is compressible, I wonder if mass loss due to LSM does not cause troubles.

>I hate to bother you, but one more question...

no, problem, u'r welcome!

> I saw very nice animations in Prof. Fedkiw's

ok, i saw that (i'm quit familiar with Ron works and track his progress), first note that Ron is more active in the field of computer graphics in which actual physics is not so important.

regarding to that simulation, i do not know details about cpu etc, but VOF also can repeat such simulation in competetive cpu time,

as a final note i should emphasis that speed of simulation in a incompressible solver (in particular for a large problem) is almost independent from interface tracking method (it takes about 5-10% cpu) (i'm sure that a particle level set method takes 3-4 times more cpu and memory than plic-vof, you can try by this code: http://www.magix.ucla.edu/software/levelSetLibrary/)

also alternative method is CIP (see Yabe papers in JCP) and if you look for fast and approximate method, you maybe interested to see this pasper: http://graphics.snu.ac.kr/publications/journals/2005%20Song%20ACM%20TOG.pdf

if u describe your application in details as well as your computing resource i can guide you more

jinwon: indeed a good question, i do not know its answer but interested to know

to my knowledge there is not some helpful comparison between methods to track comprressible flow interface

but i think that, in compressible flow rule of interface position is somehow like other quantity (not so serious which is matter in incompressible one) and if we treat level set method like momentum convective terms etc, method would be consistent

i think a good comparison can be an interesting original work

for comparison between VOF, levels set (LS) and particle LS (PLS) see this recent paper:

http://css.engineering.uiowa.edu/~yn...A-2008-530.pdf

if you read conc. section, it implies that cpu of particle level set is very high in contrsat to vof, also results of vof is more match with physics and mass conservation of PLs is comparable to vof, not better (all in agreement with what i expect)!

i'm sure that in general vof wins in terms of mass conservation

Fredi, Thanks million! It really gives me a clear idea again. If you don't mind, can I have your email address, so that I can share the short probelm description?

I guess you still have not looked at Anne Bourlioux's papers from the early 1990s. She 2D used shock fitting for detonations. This is essentially "level sets" for detonations which incorporates the Rankine-Hugoniot relations to strictly enforce conservation.