[ganesh]

Dear Friends,

I am working on developing an unstructured staggered grid based incompressible flow solver. In testing the solver for accuracy, I chose the Taylor-Green vortex problem at Re=100. For unstructured meshes at t=1 on very fine meshes I see that the errors in velocity fall at a rate close to 1. My convective discretisation uses a linear reconstruction and viscous discretisation is based on Blazek's procedure (ie. gradient at face is average of gradients in cells sharing the face + a correction term to avoid odd-even decoupling). Has anyone tested this problem (not in the inviscid limit, but that is also fine) on 2D unstructured meshes for accuracy ? Is it possible to get more details if this is the case.

P.S.: I get a second order accuracy in global velocity error on structured meshes, so that is fine. Also for a steady state problem (with an exact solution at Re=100, different problem not the Taylor vortex), unstructured meshes give me close to 1.8 in the error fall rate

Thanks and Regards,

Ganesh