Divergence problems with interface in 2D transonic external flow
 

I am investigating shockwave unsteadiness on simple axisymmetric rocket shapes (pointed nose cone with cylindrical bodies, see attachment). To do this I am using Fluent 6 and the 2D (non-DP) SST k-omega axisymmetric solver.

The mesh has a non-matching boundary extending radially outward from the point where the nose-cone and cylinder meet as it is only downstream of the junction which is of interest to me. Steady simulations work well and give the expected flow field and pressure distributions were verified by experimental data.

I have unsuccessfully tried to run unsteady simulations using two starting methods:
i) Using the steady solution as a starting point
ii) Re-initialising the whole grid with freestream values
Both methods cause divergence in all monitored parameters (velocities, energy etc) within a few iterations. Looking at the flow field once the solver fails shows that extremely non-physical values for velocities, densities etc are appearing just upstream of the interface.

When switching to the unsteady solver I didn't change any boundary conditions and I am using the first order unsteady method. Expected frequencies are of the order of 10^2 Hz and oscillation distances are of the order of 10^{-2} m.

Any advice?

Thanks,
Daniel

P.S. Attached is the axial (left to right) velocity contour. Note the region of disturbances upstream of the interface.

I deleted and recreated the interface and this seems to have solved the problem. Deary me.