Hi everybody! I am having problems specifying mass flow boundaries at the inlet. Can anyone help me with this? Have a nice day, everyone

Hi,

Can you explain you problem further?

Jon

One phase turbulent flow entering a 2D bend (nice and long straight channels on either side of bend - inlet at the bottom channel, outlet at the top). I am using CFX4.4.

What I really want to obtain in flow separation right after the bend. I have run this specifying the inlet as an inlet, using the default turbulent intesity. I have tried running it both as a steady state and transient case. Neither gave separation. (Actually, the steady state case did before it had converged good. When the convergence got better = after more iterations, the separation effect dissapeared).

CFX Support suggested I use a mass flow boundary as the inflow condition, but the program won't let me do this, so I'm probably doing it wrong.

So my question(s) is: How do I specify mass flow boundaries at the inlet, and how do I obtain flow separation after the bend?

You will need to use a low-Reynolds number turbulence model to achieve seperation around the bend, if indeed it exists. Therefore you will need your mesh such that the y+ is less than 1.0, or perhaps 3.0, if you are using Wilcox-type models.

Jon, I am using the Low Re number model, with y+ less than 1.0.

(1). Flow separation is a function of the geometry, and Reynolds number. We need to know these two parameters first. (2). A bend in a channel does not guarantee the existence of flow separation. (3). Try a 90 degree sharp turn first (no rounded corner). (4). What is your mesh size and the mesh distribution around the bend? (5). If the axial (streamwise) mesh is coarse around the bend, you will not get accurate solution. (6). What was the turbulence model used? The two-equation k-epsilon model has the tendency to give more diffusive solution. (7). What was the numerical scheme used? Upwind scheme is also very diffusive. (8). All of these can give you a smoother or better solution. (unfortunately, you are after the flow separation) (9). If you try to hide information from the readers, it will only take longer to get the right help. (a message to all readers seeking for help)

John, thank you for your reply, didn't mean to hide anything from you

1) Reynolds number: 750 000

Geometry: 180 degree 2D bend, width 50mm, inner radius is 25mm, outer radius is 75mm. 150mm straight channel before bend,300mm after bend.

4) Around the bend I have 80 mesh seeds uniformly distributed. In the boundary layer I have used the tabular option, and distributed 27 mesh seeds, also within y+ = 1. In the rest of the cross section (y-dir) I have distributed 20 mesh seeds using two way bias, L2/L1 =2. The lower and upper straight channels have 80 and 160 uniform mesh seeds, respectively.

6) Low Re number

7) I have only used the default, hybrid