Hello everyone,

I am simulating the flow of water in a composed channel (with a bank) and I am using the VOF model to determine the water surface (water-air). I start my simulation with the RNG k-e model, an initial of water height, "k" and "e". All the spatial discretization schemes used are second order. When the simulation converges with this turbulence model (normalized errors are around 10e-03) I change to RSM model. Until this point, all the under-relaxation factors were divided by 2.

When I start to simulate with the RSM model (using the solution obtained with RNG k-e) and all  the under-relaxation factors divided by 2, it diverges. I tried to divide the these factors by 4 and 8, but the simulation is still diverging. Do I have to set all these parameters to 0.01? or, do you know some correct combination of these factors in order to obtain a fast and converged solution?

Do you know the impact of starting the solution with the RSM model and first order spatial discretization for the first 1000-2000 iterations and then change to a second order scheme? Do it will implement a false diffusion in the model?

Thank you some much for reading this question.

Alejandro