[harshavardhanps3839]

I am using ANSYS FLUENT for my simulation. I am designing low speed open type sub sonic wind tunnel(Mach NO~0.1). I completed my geometry and meshing. Currently I am using the following boundary conditions.
At Inlet: Patm(static pressure inlet= atmospheric gauge pressure i.e Patm=0 pa).
At outlet: mass flow rate boundary condition.

Along with pressure inlet I am giving 1% turbulent intensity and 1 cm turbulent length scale(honey comb cell characteristic length).I have some questions can you please answer them. Thanks in advance for the help.

1)Which Pressure-velocity coupling scheme should i use for better convergence. I have set the convergence residual criteria for continuity and momentum as 10^-6? Should i have to go for SIMPLIC, PISO or COUPLED SCHEME.
2)My interest is in predicting the location of the flow separation point in the diffuser accurately. Kindly advice suitable scheme. What uncertainty can i except in the location predicted by fluent simulation
and what range of uncertainty should i except?

[DnyanMiri]

I would say go with SIMPLEC if you have a normal computing capability and if you have good computational capacity go with coupled.

About residuals, residuals are not the only parameters you have to look up to for convergence. You should create report definitions for expected important parameters and check their behavior, if they becomes steady as the time progresses then you can say that solution has reached to steady state. For example create mass imbalance at inlet and outlet.

I would suggest to provide residual criteria of less than 10e-9 for first timestep and give max no of iteration per time step 1000. See how your residuals are behaving. When residuals stops decreasing (nearly become horizontal), assume that these are the limits you have to set for residuals. These values depend upon various things like mesh, initialization etc.

Have a good day ahead.

[LoGaL]

Use coupled whenever possible, but be aware that it requires more RAM than simple. Coupled is usually more stable and requires less iterations to reach convergence.

You may as well use pseudo transient method.

Use second order scheme for all quantities, leave pressure scheme to standard. Use k-omega SST, it is the best RANS model to predict separation, even though you should remember that the most accurate option is alwais to run LES

As somebody else alrdy said, to judge convergence, monitor quantities of interest at certain points in the domain, residuals can lie

[harshavardhanps3839]

Thank you for your reply, sir. If we use the coupled scheme how to decide pseudo transient options like the time-step method, Time scale factor, and verbosity.