I'm trying to compute a channel flow with nonlinear k-eplison model ,which is added as a source term to the momentum equations of the standard k-epsilon model. The source term involves derivatives of Reynolds stress,and which are derived by the nonlinear k-epsilon model.I used Gauss theorum to calculate the derivatives,and assumed zero Reynolds stress at the walls and Neumann boundary condition at the inlet and outlet. I started with standard k-epsilon model,but when I switched to the nonlinear model,divergence occurred to the x-monmentum equation immediately. What's the problem?

Try to reduce the under relazation factor of the momentum equation and of k and eps equations. Try values around 0.4.

If this doesn't work, check boundary conditions.

Hi

ap

I used an outflow profile of a sufficiently long channel flow computed by standard k-epsilon model as the inlet boundary condition,and "outflow" as the outlet boundary condition. The same problem occurred even when the relaxation factor is reduced to 0.01.

Check the source term implementation.

Be sure to explicitly define the term dS in the DEFINE_SOURCE macro, which is used by FLUENT to stabilize the solution.

Try to iterate using first order discretization scheme for turbulent variables, at least for the first iterations.

P.S. An underrelaxation factor of 0.01 is too low, in my opinion. You risk to obtain unphysical results using such a low value. If divergence problem persist even with low values of the underrelaxation factor, usually there is some problem in the implementation of the source term or in boundary conditions.

Hi and sorry for my late answer

ap

I'm sorry to say that my problem was due to my carelessness in programming the source code.I did not not notice an extremly small number in some denominators. Anyway,thank you so much!I have learned a lot from you. Now the problem is that the results oscillate wildly. Another question:what's the accuracy of Gauss theorum for calculating derivatives?I think it's 2nd order accuracy.Is it right?