Hi all

I know the issue of tet vs hex meshes has come-up before, particularly for boundary layer effects, but does anyone have any experience of solving a pure diffusion equation on a tetrahedral mesh (for example, heat conduction) ? I've been told that the inherent non-orthogonality of the mesh causes un-physical values (e.g. of temperature), even for numerical schemes that are supposed to be bounded.

Comments appreciated.

Johnson

You are correct,

The unboundedness is caused in the calculations of gradients over control volume faces. For non-orthogonal meshes the face area vector is decomposed into an orthogonal and non-orthogonal component. The non orthogonal contribution to the face gradient is treated explicitly by evaluating the gradients over the cells straddling the face and then interpolating the gradient to the face. Multiply this by the non-orthogonal component of the face area vector. The orthogonal contribution to the face gradient are treated implicitly.

You might want to try to omit the non-orthogonal contribution to the face gradient all together, although this will introduce additional error into the calculation.

Ragards