[JefferyMartin]

I'm excited to join the cfd-online.com community! As a new member, I'm eager to learn and contribute. I have a question regarding turbulence modeling: How do you determine the most appropriate turbulence model for an aerospace application involving high-speed flow over a wing?

suika game

[Tobermory]

Wow - that's a BIG question (the hardest in CFD modelling, in my view). There's no silver bullet answer for you, I am afraid. Instead, here are some suggestions:

1. Do your research on the available turbulence models, esp. their shortcomings (they all have these) ... eg. sensitivity to streamline curvature, sensitivity to streamline freestream turbulence, inability to model flow separation etc. etc. Think about your flow scenario and see whether that rules in/out some of the model options.

2. Look in the literature for simulations of flow scenarios like yours and see what they used & how they got on - esp. helpful if they are comparing against expt data.

3. Suck it and see - find a suitable test case and run different models. Test out different meshes. Find what works best, and then apply it to your end case.

[unilord]

Originally Posted by Tobermory

Wow - that's a BIG question (the hardest in CFD modelling, in my view). There's no silver bullet answer for you, I am afraid. Instead, here are some suggestions:

1. Do your research on the available turbulence models, esp. their shortcomings (they all have these) ... eg. sensitivity to streamline curvature, sensitivity to streamline freestream turbulence, inability to model flow separation etc. etc. Think about your flow scenario and see whether that rules in/out some of the model options.

2. Look in the literature for simulations of flow scenarios like yours and see what they used & how they got on - esp. helpful if they are comparing against expt data.

3. Suck it and see - find a suitable test case and run different models. Test out different meshes. Find what works best, and then apply it to your end case.

Hey Tobermory,

Nice reply. I am doing my master thesis regarding CFD in centrifugal pumps, and I came to the conclusion that they use SST and k-epsilon the most. kEpsilon is usefull for design parameters, and SST is more used when off-design is important (because at off-design conditions, several reverse flow regions are observed).

However, when I try to use SST the residuals do not go below 10^-3 and they kind of stabilize in a periodic way (steady state analysis!). However, the kEpsilon residuals go below 10^-5 and stabilize in a "better" way

[Tobermory]

Yep - kEpsilon is certainly the most forgiving & stable to use in my experience ... but it also has the most deficiencies! I often use the realizableKE as an improvement over the basic KE, but it can also lead to troublesome residual convergence like you've noticed for ko-SST.