I am modelling the thermal field on the water jacket in the cylinder head by the fluid-solid coupling method. But I find the simulation results vary greatly when I try changing the number of the prism layer on the waterjacket. So who can give me some advice on how many prism layers should we generate when we are doing the simulation to get an accurate result?
This is an application of the more general topic of grid discretization. The brute force method--which doesn't always work--is to keep refining the grid until the answer quits changing. Many CFD formulations will diverge or even become unstable as the grid is made smaller and smaller. In that case, you must refine it and look for a minimum or maximum of something, such as the overall heat transfer or average temperature of a surface or fluid. There are also non-dimensional parameters that characterize the scale of the elements relative to properties and velocity. In general you will want to stay at or below some critical value based on the geometry. There are entire books devoted to such non-dimensional parameters.
This is an application of the more general topic of grid discretization. The brute force method--which doesn't always work--is to keep refining the grid until the answer quits changing. Many CFD formulations will diverge or even become unstable as the grid is made smaller and smaller. In that case, you must refine it and look for a minimum or maximum of something, such as the overall heat transfer or average temperature of a surface or fluid. There are also non-dimensional parameters that characterize the scale of the elements relative to properties and velocity. In general you will want to stay at or below some critical value based on the geometry. There are entire books devoted to such non-dimensional parameters.
If nothing else, there is always the trial-and-error system. Start from coarse to finer grid and follow up the output parameters. The results combined with common sense and physics will give you a guess. As professors said before me, if you expect gradients (changes) you need finer grid. But also, a too strong refinement could produce an error.
Also, if you could calculate the y+ field by CFD (which you could calculate in general too), and it could also give you a clue how much you need to refine (or not) the grid.
It depends on the model you used for the fluid field. You can calculate the Yplus in your different cases. Generally, we recommend y+~1 and ten or more prism layers in the viscous sublayer is appropriate for the enhance wall treatment . While for the standard wall function, a much larger y+~30 is sufficient. I suggest you attach some detailed information of your simulation below for getting a targeted answer.
hi, I am modelling the fluid-solid coupling on the water-jacket of cylinder head. But I am not sure how many prism layers should I generate to get an accurate result. Most importantly, I put the toughness of the wall into consideration. So i have to use the high Y+ mesh. However, when I have generated 5 prism layers(about 3.5mm in total), the Y+ value is still smaller than 5. So I dont how to increase the Y+ now. The simulation software I use here is Star-ccm+.
- Badges
- Science topic