# Troubleshooting: different results for different mesh on exact same geometry (Rectangular channel)

Member Posts: 4

Hello All! I have been running some simulations on a simple rectangular (square) channel on ANSYS Fluent, and analytically verified the results.

I am facing two problems: (1) When I lower the mesh size (add more elements), keeping dimensions and other parameters fixed, the pressure drop somehow keeps increasing, even when the centerline velocity seems to converge.

(2) To analytically verify the ANSYS results, I used Navier Stokes for a thin rectangular channel. I am aware that for a square channel, the error should be ~10%. Using the following equations I obtained the analytical centerline velocity and pressure drop values which differ greatly from the ANSYS calculations. While I understand that the centerline velocity of fully developed flow will not be the same as the inlet velocity, I am confused as to how the analytical centerline velocity can be an order of magnitude over the inlet velocity.

Does anyone have some insight on how to resolve these two issues?

Tagged:

• GermanyPosts: 6,847Forum Coordinator

Entrance length point?

• GermanyPosts: 6,847Forum Coordinator

Here it is not plane HGP flow and not in pipe. This is a square duct.

• GermanyPosts: 6,847Forum Coordinator

Is your channel long enough to allow flow to develop?

• Posts: 21Member

My channel width and thickness are both 1 mm, while the length is 12 mm

• UKPosts: 8,855Forum Coordinator

What material properties are you using in the model?

• GermanyPosts: 6,847Forum Coordinator

There is a phenomenon called centerline velocity overshoot.

Also increase length of channel to ensure fully developed laminar flow. Laminar flows require longer entrance length than turbulent flows.

• Posts: 21Member

I am using Water as a fluid under ambient conditions, at laminar flow

• UKPosts: 8,855Forum Coordinator

Can you double check? If you didn't set water in the cell zone it'll still be using air in the model.

• Posts: 21Member

This is what shows up when I first pull down cell zone

• Posts: 21Member

The ANSYS calculated centerline velocity was an order of magnitude smaller than what I had calculated analytically though.

• Posts: 21Member

For the entrance length, point I basically got what I input as inlet velocity (0.02 m/s):

The centerline velocities I showed in the table are mid-channel (6mm in a 12 mm channel).

• GermanyPosts: 6,847Forum Coordinator

Again have you estimated the entrance length for your flow?

• UKPosts: 8,855Forum Coordinator

Put a line down the centre of the channel and plot velocity on that: that'll give the centreline velocity with position and answer whether the profile has established.

• Posts: 21Member

Sorry, misunderstood before. Thanks for the clarification Rob.

I would estimated the length for the flow to fully develop to be approximately 2.7 mm out of the total length of 12 mm

• GermanyPosts: 6,847Forum Coordinator

Re check the way you are calculating streamwise velocity: hint check h.

• Posts: 21Member

I had the velocity in the lengthwise direction calculate at the exact center line of the thickness. The yellow line shows where the calculations were taken

• GermanyPosts: 6,847Forum Coordinator

I meant your "analytic" derived velocity.

• UKPosts: 8,855Forum Coordinator

The CFD result looks to be about right, peak velocity is around twice the inlet speed and in around half the area is covered by the "high" speed region.

• Posts: 21Member

Yes you are right. The calculation for general rectilinear channels turned out to be 0.6% from the ANSYS calculation

• GermanyPosts: 6,847Forum Coordinator