[ThePhoton]

We are using RPI model in Ansys CFX, and there is a problem that if we reduce mesh size after a limit, the solution doesn't converge, it diverges....

I've read somewhere that RPI model is strongly mesh dependent, so here I am confused as I have to do mesh independent study...


What to do if we hit a design limitation of model?

So can anyone help me out here?

[skewness abyss]

If you are limited on the amount of elements you can use and you are worried that your results will be inaccurate, then the best thing to do is use the grid convergence index (GCI) to quantify the amount of discretization error in your solutions. For instance, if you are calculating the velocity at an important point in your domain and your finest mesh result is 20 m/s, the GCI will let you know its 20 m/s +/- 0.2 m/s, i.e. it estimates the error in your solution. The GCI can be used to provide error intervals on your plots and integrated quantities (forces, average heat transfer, ect.) letting the others know that the exact value has a good chance of being contained in your quoted intervals.

Calculating the GCI requires only two meshes, but three meshes are preferable. Hopefully you have enough resources to generate a coarse mesh and a fine mesh.

If you search "Journal of Fluids engineering Editorial Policy" in google, a pdf should appear giving you the details on how to go about calculating the GCI, and the limitations of the GCI. You can post here if you have questions about the method after reading the PDF.