Hi folks, just wondering if anybody has experience with simulating laminar hypersonic compression corners at high Mach number. There is this phenomenon that the separation bubble becomes larger as the mesh is refined. the streamlines reveal vortices, some of which suggest the presence of dipole vortices (vortices within a larger vortex with counter rotation). The bubble is almost absent when a turbulent simulation is carried out. a transitional simulation shows a bubble smaller than that seen in the schlieren image. heat flux measurements suggest transition at the corner. my questions: 1. how is the size of the mesh related with the bubble size. i hear that some researchers report cases where the bubble increase until a ceertain grid refinement is reached, the opposite is then observed i.e. the bubble decreases with further refinement. this makes grid convergence based on the resolution of the bubble impossible. I know that at some grid level numerical dissipation are almost comparable with the dffusive terms, but then the accuracy of the discretisation could play a more important rule. Can someone please, share experience and thoughts on this with me. 2. any ideas on vortex dynamics in shock boundary layer interactions will be appreciated. regards, ben

hi i want any information in cfd bubble column specialy in bubble formation by water electricity

hi there rana, i am trying to be impolite but i don't see how your question relates to the topic in this thread. anyway i have no experience with electrolysis and bubble formation. this is something entirely different.

Hi,

Your comments about grid refinement seem strange. The numerical dissipation of progressively finer grids should reduce, meaning that you will eventually converge on a true grid refined solution. This applies even for highly dissipative schemes like upwinding, you just need a grid fine enough.

My first choice in your case would be to use the high-resolution scheme due to the high mach number, but if low dissipation is critical you can move to higher order schemes like quick or even central differencing. Central differencing has no numerical dissipation but is very numerically unstable and will be difficult to get a converged solution.

Glenn Horrocks