I only know the application of graphic visual effects. There are several major schools. Stanford stream likes to use level set/particle level set. The accuracy of Level Set is obviously not enough, and the problem of volume reduction will occur during simulation, that is, the more simulated water, the less it will be, and finally it will be gone. The advantage is that it is relatively simple to implement, and you need to get an extrapolation when simulating water anyway. The accuracy of particle level set is very good, but the disadvantage is that it is really difficult to realize and the parameters are chaotic, which directly affects the final effect. My boss is a graduate student, and he strongly advises me not to try, because he has never seen anyone succeed in repro except Stanford. . However, I later found out that UCLA actually has a library: the particle level collection library, which has never been used. It's hard to evaluate. You can have a look if you are interested. ? Chris Wojtan invented direct display grid to simulate free-form surfaces, because the characteristics of free-form surfaces have to deal with various complex topological changes such as fusion and splitting, and I think it will be really painful to realize. UBC generally uses FLIP/PIC based on particle mesh mixing, and then uses (current state of the art) methods to reconstruct surfaces through particles. Personally, I like this routine best, although the calculation of anisotropic matrix during reconstruction is quite broken. The following is the effect I achieved with this method, which is quite satisfactory. Finally, this method of surface reconstruction is also used in real-time rendering. Combined with the point Splatting method, the ellipse stretched by anisotropic matrix is rendered on the depth buffer, and then blurred, and then delayed coloring is done normally, but this is just rendering, which is a bit off topic. There is also a comment about the method based on location and screen space. What they are after is not the accuracy of simulation at all. Their contribution is practical in interactive real-time rendering. I believe that no special effects field will use these methods. I can fully understand that real numerical simulation experts are disdainful of the method of simulating honey juice and such an inaccurate rendering method, but it is still valuable. Obviously, not everyone cares whether the water is like honey.