Torque Prediction
 

Dear All,
I am simulating a flow over a ceiling fan in an enclosed room. the fan is rotating at speeds in the range of 300-400 RPM. I am defining a rotating domain around the fan and following the frozen rotor approach.

The fan blade is only 1 mm thick. I have tried to predict the torque over the fan, the error varies from 15-60% from the actual value for various cases. However the mass flow variation is well within 2% of experiments.

I recently approached ANSYS support for it and they suggested going for y plus value for first cell as 1 as I am using SST model. But because of that the error increased further by 10% in torque and to 20% in mass flow ( y did this happen ? )

Can anyone suggest how to improve my accuracy and consistency of such a simulation.

1. Firstly, you need to carry out a numerical parameter analysis to guage the uncertainity of the CFD results. More information regarding this can be found in "ANSYS CFX best practices guide for numerical accuracy".
2. After establishing uncertainity, I hope you will be in a position to understand the effect of Turbulence model, y-plus & etc..

All the best :)

Have you read this FAQ? http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

Hi Glenn,
Yes I have read that a number of times. Still I couldnt find a solution to my problem. One additional thing, by reducing yplus to below 1, my solution is not converging properly, my residuals are stable ( although high ) but my torque changes with every iteration it has already been around 1400 iterations.
How should I get a stable and accurate solution for my problem.

Does a stable and accurate solution exist? You are assuming a steady state solution (well, frozen rotor to be exact, but same thing), but if you have large separations in the flow then it is highly likely that a steady state solution does not exist as the flow is inherently transient. You might need to move to transient models, and possibly DES/SAS/LES.

Hi ghorrocks,
My simulation is quite a low rpm and low blade loading simulation. It does not involve flow separation.

Then you should be able to get an accurate result for this model.

Can you post some images and your CCL? Maybe the output file too.

Hi,

I also have the same issue in terms of torque prediction. The CFM predictions are quite good. But the torque predictions are varying from 0 to 50%. Did you get any breakthrough?We are trying several methods and didnt get any success. Please reply to my email id: trilok99@gmail.com

What you need to do for an accurate simulation is described in the link I posted in post #3.