Hi,

I'm interested in using CFX to model a linear air bearing (aerostatic bearing); but I'm having quite a lot of trouble.

Air bearing dimensions: Outer dims: 60x60 mm Inlet dia: 2 mm Orifice dia: 0.4 mm

The bearing is supposed to float on a cushion of air above a flat and smooth surface at anywhere between 5 and 50 microns.

I've modelled one quadrant of the fluid volume. The orifice is located at the centre of the bearing (i.e. at 0,0) and the air bearing is parallel to the smooth surface. I'm currently investigating the bearing properties at a 50 micron separation.

I'm modelling the fluid domain as an ideal gas at 300K and using the standard turbulence model.

Here are my questions:

1. Ideally, I'd like to specify an inlet pressure of 5 bar and an outlet pressure of 0 bar; this has been problematic; so instead, I've been applying a mass flow rate to the inlet (whilst keeping a 0 bar outlet); if I determine the pressure at the inlet (for a given mass flow rate), would I have got the same mass flow rate if I'd specified that inlet pressure in the first place?

2. In order to achieve a model with a reasonable number of elements, I've split the model into three parts and I'm meshing them individually. This allows me to use a very fine mesh for the air gap, whilst using larger elements for the inlet and outlet regions. Is this the best strategy because I've read that "knitting" different meshes together can cause inaccuracies?

3. My main problem is that the vertical thrust predicted by CFX acting on the bearing is very low. I think this is due to large negative pressures existing around the orifice; from what I've read this does not occur in the real world. What could cause it?

We have test results for a very similar bearing and some antique bearing analysis software (that works) that we're comparing ANSYS against.

Thanks for your help.

Hi,

Questions: 1) As is discussed in the manual, it is usually easier to get convergence when the flow is specified as a mass flow rate inlet and pressure outlet. Pressure inlet and outlet is significantly harder to converge. Providing there is no pressure gradient across the inlet (and there should not be for an accurate simulations) the pressure/pressure specification and the mass flow/pressure specification are equivalent. You can check this by using a massflow/pressure simulation as an initial guess to a pressure/pressure simulation. It should converge to effectively the same answer.

2) Depends how you "knit". If you use a conformal mesh then the important issue is to keep mesh quality under control. If you use GGIs then the issue is whether the GGI causes problems. Both approaches are OK providing they are done correctly. Both approaches can make the simulation rubbish if done poorly.

3) Have you checked the overall accuracy of your simulation? You need to check the convergence tolerance, mesh size, time step size (if transient) by sensitivity analysis and check the model physics is correct.

Glenn Horrocks