[TKSRichy]

which turbulence models do you guys use when running simulations? and for what kind of studies?

I mainly use Phoenics for air ventilation assessment, and indoor air movement simulation, I wonder which model should I use for each type.

[phoenics_cfd]

Originally Posted by TKSRichy

which turbulence models do you guys use when running simulations? and for what kind of studies?

I mainly use Phoenics for air ventilation assessment, and indoor air movement simulation, I wonder which model should I use for each type.

Hi Richard,

This is a list of the turbulence models in PHOENICS: http://www.cham.co.uk/phoenics/d_pol...mod/enc_tu.htm.

This is a question that has been addressed many times in the CFD literature, but in my opinion without any real consensus about which turbulence model is best to use for all types of AVA and indoor-air-flow applications.

I would aim to generate a first-cut converged solution using the standard two-equation k-e model, which is robust and numerically stable. Once you have a mesh that provides a converged solution of acceptable accuracy, you can then always restart from this solution trying more refined models such as the RNG, Chen-Kim or Realisable k-e variants, which basically offer improved performance in separated-flow regions. Energy-frequency models such as k-w-SST are favoured by some workers. Other factors such as stable stratification may warrant the inclusion of buoyancy terms in the turbulence transport equations, and in some situations low-Reynolds-number extensions may be needed, such as when near-wall processes are the focus, but this will involve greater computational expense, and k-w models are known to be numerically stable than k-e model for these applications. I hope this helps.

[TKSRichy]

Originally Posted by phoenics_cfd

Hi Richard,

This is a list of the turbulence models in PHOENICS: http://www.cham.co.uk/phoenics/d_pol...mod/enc_tu.htm.

This is a question that has been addressed many times in the CFD literature, but in my opinion without any real consensus about which turbulence model is best to use for all types of AVA and indoor-air-flow applications.

I would aim to generate a first-cut converged solution using the standard two-equation k-e model, which is robust and numerically stable. Once you have a mesh that provides a converged solution of acceptable accuracy, you can then always restart from this solution trying more refined models such as the RNG, Chen-Kim or Realisable k-e variants, which basically offer improved performance in separated-flow regions. Energy-frequency models such as k-w-SST are favoured by some workers. Other factors such as stable stratification may warrant the inclusion of buoyancy terms in the turbulence transport equations, and in some situations low-Reynolds-number extensions may be needed, such as when near-wall processes are the focus, but this will involve greater computational expense, and k-w models are known to be numerically stable than k-e model for these applications. I hope this helps.

Turns out i have been looking at the wrong aspect of the setting, for a long time, my simulation result have been getting ridiculous result, temperature over 40,000 degree Celsius in a model where the hottest object is just 60*C. Turns out it was the relaxation control defaulted a absurd value, making the simulation over-calculate and unable to converge.

[phoenics_cfd]

A temperature predicted that large is obviously a convergence issue and unrelated to turbulence-model selection. I would expect such a large temperature to initiate very rapid divergence of the solution, especially if the Buossinesq approximation hasn't been selected.

The default relaxation control in PHOENICS is "automatic convergence control", which by default provides linear relaxation of 0.25 on temperature, limits the maximum change in temperature to 50 degC per sweep, and sets an upper limit of 3000 degC on temperature. As a matter of course I always set the upper and lower limits on temperature to be a few degrees beyond the physically expected values for the given application.

[TKSRichy]

Indeed, I manged to solve the no.1 problem haunting my department, yet I wish they would have tell us about this important aspect of the software settings. When I choose Flair module, I expected the settings to be somewhat suitable for indoor simulations~~~

[phoenics_cfd]

The default settings should work for this case, as indicated by my last post, so please send us your Q1 input files together with any geometry files needed to run the case, and we will investigate and rectify the default settings as need be. The email address is: support@cham.co.uk.