[sreevaibhav]

Greetings,
I'm trying to simulate the effectiveness of various turbulence models over RAE2822 airfoil at Transonic Speeds. I've used SKE,RKE,SKW,KWSST,TRKKLW and RSM. All of them have successfully simulated the flow while SA's nut(Turbulent Kinematic Viscosity) is diverging. Please give me your suggestions.

Thank you.

[juliom]

The good thing is that is only one that is not converging. It means that your problem is well posed. Thus, I guess that your mesh and boundary conditions are also right. This is what I would do if I were you. Use the solution of any of the converged solution as initial condition and solve the problem in a transient approach. Sometimes, the only way to solve a problem is using transient approach. I'm sure it will work out.

[sbaffini]

Well, while the suggestion from Julio is generally valid, my experience implementing 1 and 2 equations turbulence models is that SA usually is the only one you get right straight out of the box, but only if you closely follow the reference.

As all your other turbulence models are working, I assume you know how to properly handle positive and negative contributions of turbulence source terms to the matrix coefficients (i.e., how to only consider contributions which enhance stability), as well as all the other common features for such models.

Then, if we exclude a specific implementation bug for SA (which only you can be aware of), this only leaves out one thing you might not be aware of about SA. That is how to properly implement its diffusion term; more specifically its non conservative part. How did you implement it?

[sreevaibhav]

Originally Posted by sbaffini

Well, while the suggestion from Julio is generally valid, my experience implementing 1 and 2 equations turbulence models is that SA usually is the only one you get right straight out of the box, but only if you closely follow the reference.

As all your other turbulence models are working, I assume you know how to properly handle positive and negative contributions of turbulence source terms to the matrix coefficients (i.e., how to only consider contributions which enhance stability), as well as all the other common features for such models.

Then, if we exclude a specific implementation bug for SA (which only you can be aware of), this only leaves out one thing you might not be aware of about SA. That is how to properly implement its diffusion term; more specifically its non conservative part. How did you implement it?

I'm using fluent for this study, So i'm not quite sure how i can contemplate the diffusion term in the solver for better stability of the solution.

[sreevaibhav]

Originally Posted by juliom

The good thing is that is only one that is not converging. It means that your problem is well posed. Thus, I guess that your mesh and boundary conditions are also right. This is what I would do if I were you. Use the solution of any of the converged solution as initial condition and solve the problem in a transient approach. Sometimes, the only way to solve a problem is using transient approach. I'm sure it will work out.

Thank you for your input, I'll initialize the unsteady case of SA with a steady state solution from a well converged model

[sbaffini]

Originally Posted by sreevaibhav

I'm using fluent for this study, So i'm not quite sure how i can contemplate the diffusion term in the solver for better stability of the solution.

Well, then forget what I wrote. Still, it remains true that SA is quite robust with respect to other models. I would investigate your y+ (try avoiding intermediate values in the range 10-30) and inlet boundary conditions (check with values suggested here https://turbmodels.larc.nasa.gov/spalart.html).

[sreevaibhav]

Thank you for your input. Much appreciated

[arjun]

Originally Posted by sreevaibhav

I'm using fluent for this study, So i'm not quite sure how i can contemplate the diffusion term in the solver for better stability of the solution.

Which flow model are you using with Fluent?

Fluent with SA and coupled flow model on rae2822 was present as tutorial case so it does converge and converge pretty well if i remember correctly.

With segregated version it has issues because when i made tutorial out of rae2822 for wildkatze (using SA model ) i noted that fluent's segregated solver had hard times (if at all converged).

Wildkatze converges (with SA model) on rae2822 segregated solver but could have issues if settings are fiddled.