[Deepaksha]

I am doing a analysis on ceiling fan which is placed in a standard Air Delivery chamber ( as per IS 374-1979 - details attached). Two domains are created one is the rotating domain in which the ceiling fan enclosed with cylinder and the remaining portion of the test chamber is taken as Stationary domain. Both domains are connected through interface. The result required is air velocity at a plane 1.5 m below the fan along 4 perpendicular axis at an increment of 80mm. Actually the Air delivery of ceiling fan is finding in such a testing setup and my intention is to replicate the test in CFD so that we can virtually test new models/designs at design stage itself.


In the results I have obtained the velocity values at the initial points near the origin (Near the center) is not matching with test readings can anyone suggest some better method to improve the results and co relation also can anyone suggest good time step for transient analysis and physical time scale for steady state analysis (now I have used 2/omega as physical time step in steady state analysis and time for 30 degree rotation as time step for transient analysis with 5 rotation total time - Rotational speed of the model is 400RPM)


ANSYS CFX
Analysis - Steady state
Physical time scale -2/omega
max no of iterations - 500
Created monitor points at 1.5m below the fan along 4 axis to measure velocity


details of Fan
no of blades -3
rotational speed 400 RPM
direction - Anti clockwise



It will be very help full if someone give some insights in Meshing also since the rotating domain (Fan ) is very small as compared to the test chamber ie, 1mm blade thickness and 7 meter test room width

[ghorrocks]

You need to do a sensitivity analysis on all tunable variables to check you are getting accurate results. You should do this before you start interpreting the results as before you check it who knows how accurate the results are? If the results are very inaccurate then there is no point in interpreting the results.

To do a sensitivity study on your convergence tolerance:
* Do a simulation at a convergence tolerance (you appear to have already done this)
* Do another simulation where you converge to 10 time tighter residuals tolerance
* Compare the results of the two runs. If they are the same then your simulation is not convergence tolerance sensitive. If the results are different by an amount you are concerned about then do another simulation with the residual tolerance 10 times tighter again and repeat until you do get insensitive results.

Then do this same process for time step size (your time step in a transient simulation should be set by sensitivity analysis, not by a simple estimate) and mesh size. For the time step size one halve the time step size each step, and for the mesh size one halve the element edge length (which will result in very big meshes very quickly).

Once you are confident your convergence tolerance, mesh size and time step size is OK only then can you assume your simulation is accurate and only then you start to interpret the results.

The same applies to a steady state simulation, the only difference is the time step size does not need a sensitivity analysis. Just mesh size and convergence tolerance.

[Deepaksha]

Thank You Sir.
In this case the rotating fan blade thickness is 1mm only and the chamber size is almost 7*7*4 meters. So it is very difficult to mesh. I am using ANSYS mesh. can you give some inputs for best practices in meshing

[AtoHM]

As you will mesh the rotating and stationary parts separately anyway, you can use a quite fine mesh for the rotor and determine from your independence study how far you can go with the stationary part in terms of mesh size. I expect you won't need an extremely fine mesh here.

[Deepaksha]

If we kept fine mesh for rotor and coarse mesh for stationary part, the mesh concentration on both side of interface will be very different . Now I am taking the domain in to Spaceclaime as single multi body object and giving "Group'" option at the contact surface (Interface surfaces) and also meshing as single multi body geometry.Giving general element size which will be applicable for both Stator and rotor. In further for refining the rotor blade with face sizing of small element size. So finally i will get a refined mesh region at rotor portion and equal mesh size at both side of the Interface. Can you suggest a better meshing practice. I am using ANSYS mesh