[aerosayan]

Hello everyone,

I'm looking for a simple code example for only 2D linear convection on unstructured triangular grids. Not convection-diffusion. Only linear convection, and nothing else. I understand how to solve convection diffusion on 2D structured grids with FTBS scheme, but I don't understand how to solve them on unstructured triangular grids. Since every example for unstructured grid shows combined convection-diffusion, it kind of becomes hard to understand the basic concept.

[sbaffini]

What is your difficulty? Because, in general, it really sums up to simply not using the diffusion part from the convection-diffusion example.

For explicit schemes, of course, there are caveats, and not all the convective schemes are compatible with all the time advancement schemes. But, for a classical, implicit, cell-centered, finite volume discretization on unstructured grids, you really just need to neglect diffusion from a convection-diffusion example.

[arjun]

I think you can derive by simply setting diffusion coefficient D=0.


On unstructured grids the face values will be from left if the fluxes are positive and from the right if fluxes are negative.

In the end you will get the sources computed as

Src_left -= phi_upwind * flux

Src_right += phi_upwind * flux


In the end you loop through all the cells and add

Src += phi * mass_imbalance


This is how the source terms are computed in Wildkatze.