[harryosu]

Hello Fluent folks,

As a Fluent newbie, I have a huge headache when trying to simulate turbulent flows in mini-channels. The research project i am now working on is related to mini-channel heat exchangers. To be brief, what I am simulating is a conjugate heat transfer problem involving hot and cold fluid and solid in-between. The channel is pretty small, with only 1.222 mm as hydraulic diameter. However, the inlet velocity is pretty large, in the range of around 20-40 m/s. This value is actually based on available experimental data, and some of the reasons for high-speed flow is because of large mass flow rate, extremely small density (helium as flowing medium) and extremely small cross-sectional area. Anyway, when I used the standard definition and suggested formulas to calculate turbulent parameters to input into Fluent, I found these calculated parameters are quite ridiculous. Specifically, for example, one cold channel inlet velocity is 21.73 m/s, density is 3.056 kg/m3, Reynolds number 3385.8, so I calculated and obtained turbulence intensity 5.793%, and then turbulent kinetic energy 2.38 m2/s2, dissipation rate of tke 7040.08 m2/s3!!!, and specific dissipation rate 2961.65 1/s!!!!!! This is extremely confusing since the values especially the velocity is from experiments, and the calculation method is completely based on Fluent manuals. But the calculated values are ridiculously high. What do you guys think? Is that normal? Btw when using this value for the input of setup in Fluent, the results showed these parameters are extremely high, which is unreasonable. For example, the turbulence intensity can go beyond 100%! So what is problem here? I really need help!

Thanks,
Harry

[LuckyTran]

Those numbers look pretty reasonable for initial guesses. Do you have knowledge from somewhere that suggests these are ridiculously high values?

Maybe you can describe what is confusing about the calculation procedure and we can help.

100% turbulence intensity does indeed sound very wrong. So look for the usual culprits, convergence issues, bad mesh, wrong b.c.'s, etc.

[harryosu]

Thanks LuckyTran.

The confusing part is that what I found the parameters range (I mean k, epsilon and omega) from some tutorials is way smaller than what I calculated, that is why I doubt whether it is a good guess for my application. And also I wonder the turbulent parameters for the inlets in the simulation results are not the same as what I input in the setup in the first place, are they? This is also confusing, because originally I thought like the velocity and temperature in velocity-inlet BCs, turbulent parameters will always be the same as I input in the setup. But in fact I have tried several turbulence intensities like 20% and 1% as input, the results show that the corresponding simulated turbulence intensities are around 400% and 20%, respectively. Obviously, the later results sounds reasonable, but it seems that the results do depend upon what you input initially, so then what is the appropriate method to guess the turbulent parameters for BCs? It turns out the manual-suggested method generates unreasonable results as I said before, but randomly selected turbulence intensity is not convincing.

Harry

[LuckyTran]

At your inlet, you must apply boundary conditions for all dynamic quantities in your simulation. This means you must completely specify velocity, temperature, k, epsilon/omega. Velocity & temperature is easily understood, however the turbulence at the inlet also must be specified by you and this is often overlooked.

In principle you should know what the turbulence at the inlet is from experiments or other and apply that at the inlet. For example, imagine you did not know your velocity/temperature at the inlet, which sounds ridiculous but it is nearly as ridiculous as not knowing your inlet turbulence. Your description of the problem is not complete until you have specified everything at the inlet. If you do not know, then you have to start guessing or modelling your inlet.

What Fluent suggests, some of the formulas given, are to simply use correlations for the average turbulence intensity, k, epsilon/omega, etc for fully developed pipe flow. Again these are only averages and not profiles. If you know better, then do better.

[harryosu]

Thanks again.

Okay, I got what you mean. The experiments do not provide any turbulence intensities (actually in heat exchangers they are difficult to measure). So I guess the only option left is to model the inlet turbulent parameters, or try several guesses to compare the results with experimental data in terms of friction factor and Nusselt number, etc., right? If they fit with data, then the simulation is reasonable. Otherwise, I will be stuck if no effective estimating method is available.

Harry