solving mixed convection flows

tom

Hi,

I was recently faced with a problem solving the flow in a ventilated room, with the heat source represented by a high temperature floor. I got a converged solution with the std k-e model and 2nd order convection. However, when I tried to improve the accuracy by using the realisable k-e model, the solution became highly unsteady. I tried time stepping with the segregated solver using reasonably small time-steps but could never get a steady solution. Finally, I read in the Ansys manual online(I'm using starccm)that if a large time step of similar size to the residence time in the CFD domain was chosen then it would be possible to push the unsteadiness caused by buoyancy out of the domain, and once this has been done smaller time steps could be used.

This technique worked for me. However, it meant that I needed to use the coupled solver, as time steps as large as the residence time could not be solved with the segregated solver. My question is, what would happen if the CFD code did not contain a coupled solver, or if I had not enough memory to run the coupled solver, how then would I get a steady-state solution?

Thanks for any advice