Dear All,

I am currently using Fluent 5.5 to test if a straight pipe can be modelled using the 2D axisymmetric solver. For each tests I run the flow simulation twice. From the first test I take a profile of x,y,z velocity and k and eps from the cross section when the flow is fully developed and use this as an input for the second test.

My first aim was to get a grid independent solution. To do this I fixed the boundary layer thickness to a y+ of 30 and then increase the number of cells in the domain until the increase showed no change in the parameters of interest, such as a velocity profile. For low discharges this worked well, but I have been having trouble with the higher discharges. When I read the profile in the second time the discharge, which should have been 5kg/s, changes to another, larger number. For a coarse grid the problem is not so noticeably, but with finer grids it is much worse (mass flow rate from 5 kg/s to 5.11 kg/s). The grid is not very fine i.e. the cells in the main domain are not smaller than the boundary layer, and I am using square cells.

Do you know what the problem is and is there a way to work around it? If not how is it possible to tell when grid independence is reached because the flow rate is changing and therefore the parameters of interest, such as velocity profile, are changing?

Thanks in advance, John

Are you sure that you calculation is properly converged for the finer grids?. If you are specifying the mass flow at the duct inlet to be 5 kg/s, you should get 5 kg/s or an extremely close value to 5 kg/s at whaterver section of the duct. Remember to check the mass flow in report/fluxes rather than report/surface integrals, since the values may differ. Can you give an idea on how far the velocity profiles are for finer grids in comparison to coarser grids? you should get very similar profiles in shape and in mean value.

Cheers

Thanks for your help Peter. I hope I have answered your questions as you would like.

Q)Are you sure that you calculation is properly converged for the finer grids?.

A)I have never been happy with the 1e-3 convergence criteria so I always let the simulations continue until the residuals are horizontal for a considerable period of time (This is usually around 1e-7). I do this for both coarse and fine grids.

Q)If you are specifying the mass flow at the duct inlet to be 5 kg/s, you should get 5 kg/s or an extremely close value to 5 kg/s at whatever section of the duct.

A)I do not directly specify the mass flow rate at the inlet. Instead as the pipe has a known area I select a velocity to match the discharge I require (Q=VA). When I run the first flow simulation the mass flowrate is OK when I take slices through the pipe. Even from the slice I take the profile from the mass flow rate is reported to be what I require. But when I read this in to the new simulation the flow rate differs.

Q)Remember to check the mass flow in report/fluxes rather than report/surface integrals, since the values may differ.

A)No the values are the same in this case

Q)Can you give an idea on how far the velocity profiles are for finer grids in comparison to coarser grids? you should get very similar profiles in shape and in mean value.

A)The profiles are similar in shape, but due to the increase in flow rate for the fine meshes they are always above the coarse meshes. The mean value of velocity also differs

Thanks once again for your help, its much appreciated.

Regards, John

In my first message I said that I record a profile of x, y and z velocity. This is incorrect, I actually record a profile of axial and radial velocity. Sorry

John, because you are using the same velocity profile for both the first and the second runs, and the duct area is the same, you should get the same mass flow rate in both cases providing that the density is the same for both. Now it is interesting to see if you assume that the flow is incompressible (you should specify a value for density, equal in the two runs) or compressible (in that case, you should specify the right temperature boundary condition in both cases. It will be another profile rather than a constant value). So, if velocity, density and area are the same in the two runs you should get the same mass flow. Cheers

Q) John, because you are using the same velocity profile for both the first and the second runs, and the duct area is the same, you should get the same mass flow rate in both cases providing that the density is the same for both.

A) The density is the same so I was expecting the same result, which is why I am thinking there may be a problem with Fluent

Q) Now it is interesting to see if you assume that the flow is incompressible (you should specify a value for density, equal in the two runs) or compressible (in that case, you should specify the right temperature boundary condition in both cases. It will be another profile rather than a constant value). So, if velocity, density and area are the same in the two runs you should get the same mass flow. Cheers

A)The flow is incompressible so the density is the same in both runs. This is why I was expecting the same result

Has anyone had this problem before or does anyone have any more ideas?