Flow through ribbed pipe-Periodic BC

genshih · November 20, 2015, 14:37

Hi !!!
This is my first post and I am a new user to FLUENT. I am working on a ribbed pipe with air as the fluid and am using a 2D geometry with axisymmetry & periodic BC. This BC offers only 2 input options & both give the following errors:
1.Pressure gradient: My experiments show I have a pressure loss of about 1500Pa in a 1.75m long pipe. My geometry is 19mm radius and only 10mm in the streamwise direction. I apply a pressure gradient of (1500/1.75) Pa/m (which is the input option in Fluent). I however get a simulation which keeps showing an increasing velocity with time steps. I do not know why..... I think it could be because
a) CFD would not be able to simulate the actual pressure losses and hence I would have to reduce the pressure gradient.
b) I am not changing the dimensions of length and area in the reference values section.......... The problem is I am not sure if I have to change them to the actual geometry values or whether I can proceed without making any changes to them.
However, I must mention that the simulation predicts flow phenomena similar to other papers on the same.

2. Mass Flow rate: I calculate Mass flow rate as: 30[lit/sec]*1.225(air density)*10^-3(lit/sec - kg/sec)=3.675*10^-3 [Kg/sec]
Now, the problem is when I give this as an input without any pressure gradient, Fluent seems to be overpredicting the pressure ( order of e58) and exits citing divergence in AMG solver: x-momentum
Why do you think this is happening?
I have also plotted the contours of the cell courant number and get values as high as 300 in some regions when I use time steps of 0.001. Could this be responsible. If so, why is the solution not giving problems when the pressure gradient is applied?
All suggestions are welcome as I am not able to get past these steps! Btw,The turbulence model is k-w SST.
Thank you for reading through!

LuckyTran · November 22, 2015, 11:30

First: 30 L/s @ 1.225 kg/m^3 is a massflow rate of 0.037 kg/s

I have run into some numerical issues using the axissymmetric boundary condition that I didn't have on a 2D planar simulation. It was prone to divergence which was eventually solved by lowering the urf's.

Reference values don't affect your solution, they only affect post-processed quantities.

I have performed several thousand simulations using 2D axissymmetric + periodic boundary conditions on circular and rectangular ribbed channel flow, so I know it works.

The pressure gradient approach is much more robust than specified mass-flow rate. Because the pressure gradient doesn't change each iteration, your boundary conditions remain fixed whereas with the mass-flow rate approach the boundary conditions are iterated in. Plus you are doing a transient simulation so your pressure gradient is changing each time-step.

1e58 is basically infinity, if you are over-predicting the pressure gradient that much it looks like you have a serious problem.

genshih · November 22, 2015, 12:37

Thank you for replying........Several thousands !?........wow!!!
Thank you for pointing out the mistake, I replaced the 10^-2 with 10^-3.
I have kind of given up on the mass flow rate because of the crazy pressure gradients issue.
Just by prescribing a pressure gradient, lets say 300Pa/m,the solution never seems to stop, i.e. with time, the velocity keeps increasing. Does this mean that:
1. The pressure gradient I imposed is very high ?
2. Something else I must have done wrong?

Shamoon Jamshed · October 3, 2017, 14:01

I have the same kind of problem, but I managed to solve with the flow but not heat transfer. Lucky Tran, you might be familiar with it.
I have a pipe 2m long, in the formula of computing heat transfer coefficient, I have to take the average of temperature values along the whole pipe length. So if I take a segment, and model periodic flow across it, the flow might be fully developed but hydraulically, not thermally.
For temperature, I tried to make a code that takes output to outlet at every set of (lets say 200) iteration and input this temperature to the inlet and repeat the procedure. At every iteration, I make Fluent to write Twall as area weighted average at wall. Then after several runs that makes the total length equals to 2m . E.g if 12.8mm is segment length, I need at least 157 runs to make it 2000m or 2m. At the end, I take the average of Twall that was written by Fluent for each segment.
This was for constant heat flux. I am not sure that approach is correct or not?

November 20, 2015, 14:37	Flow through ribbed pipe-Periodic BC	#1
genshih New Member CFDlover Join Date: Nov 2015 Posts: 5 Rep Power: 10	Hi !!! This is my first post and I am a new user to FLUENT. I am working on a ribbed pipe with air as the fluid and am using a 2D geometry with axisymmetry & periodic BC. This BC offers only 2 input options & both give the following errors: 1.Pressure gradient: My experiments show I have a pressure loss of about 1500Pa in a 1.75m long pipe. My geometry is 19mm radius and only 10mm in the streamwise direction. I apply a pressure gradient of (1500/1.75) Pa/m (which is the input option in Fluent). I however get a simulation which keeps showing an increasing velocity with time steps. I do not know why..... I think it could be because a) CFD would not be able to simulate the actual pressure losses and hence I would have to reduce the pressure gradient. b) I am not changing the dimensions of length and area in the reference values section.......... The problem is I am not sure if I have to change them to the actual geometry values or whether I can proceed without making any changes to them. However, I must mention that the simulation predicts flow phenomena similar to other papers on the same. 2. Mass Flow rate: I calculate Mass flow rate as: 30[lit/sec]1.225(air density)10^-3(lit/sec - kg/sec)=3.675*10^-3 [Kg/sec] Now, the problem is when I give this as an input without any pressure gradient, Fluent seems to be overpredicting the pressure ( order of e58) and exits citing divergence in AMG solver: x-momentum Why do you think this is happening? I have also plotted the contours of the cell courant number and get values as high as 300 in some regions when I use time steps of 0.001. Could this be responsible. If so, why is the solution not giving problems when the pressure gradient is applied? All suggestions are welcome as I am not able to get past these steps! Btw,The turbulence model is k-w SST. Thank you for reading through!

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Issues on the simulation of high-speed compressible flow within turbomachinery	dowlee	OpenFOAM Running, Solving & CFD	11	August 6, 2021 06:40
Periodic Pipe Flow LES; parabolic profile	jrrelx	FLUENT	34	August 13, 2015 17:58
[ICEM] periodic blocking - blade-to-blade turbine flow volume	Jonathan	ANSYS Meshing & Geometry	7	July 11, 2014 19:40
flow in perforated pipe distributor	pertupd	ANSYS	0	August 12, 2009 08:36
periodic fully developed flow in pipe	periodicbc	FLUENT	5	January 12, 2004 15:57

November 22, 2015, 11:30		#2
LuckyTran Senior Member Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,668 Rep Power: 65	First: 30 L/s @ 1.225 kg/m^3 is a massflow rate of 0.037 kg/s I have run into some numerical issues using the axissymmetric boundary condition that I didn't have on a 2D planar simulation. It was prone to divergence which was eventually solved by lowering the urf's. Reference values don't affect your solution, they only affect post-processed quantities. I have performed several thousand simulations using 2D axissymmetric + periodic boundary conditions on circular and rectangular ribbed channel flow, so I know it works. The pressure gradient approach is much more robust than specified mass-flow rate. Because the pressure gradient doesn't change each iteration, your boundary conditions remain fixed whereas with the mass-flow rate approach the boundary conditions are iterated in. Plus you are doing a transient simulation so your pressure gradient is changing each time-step. 1e58 is basically infinity, if you are over-predicting the pressure gradient that much it looks like you have a serious problem.

November 22, 2015, 12:37		#3
genshih New Member CFDlover Join Date: Nov 2015 Posts: 5 Rep Power: 10	Thank you for replying........Several thousands !?........wow!!! Thank you for pointing out the mistake, I replaced the 10^-2 with 10^-3. I have kind of given up on the mass flow rate because of the crazy pressure gradients issue. Just by prescribing a pressure gradient, lets say 300Pa/m,the solution never seems to stop, i.e. with time, the velocity keeps increasing. Does this mean that: 1. The pressure gradient I imposed is very high ? 2. Something else I must have done wrong?

October 3, 2017, 14:01		#4
Shamoon Jamshed Senior Member Shamoon Jamshed Join Date: Apr 2009 Location: Karachi Posts: 377 Rep Power: 17	I have the same kind of problem, but I managed to solve with the flow but not heat transfer. Lucky Tran, you might be familiar with it. I have a pipe 2m long, in the formula of computing heat transfer coefficient, I have to take the average of temperature values along the whole pipe length. So if I take a segment, and model periodic flow across it, the flow might be fully developed but hydraulically, not thermally. For temperature, I tried to make a code that takes output to outlet at every set of (lets say 200) iteration and input this temperature to the inlet and repeat the procedure. At every iteration, I make Fluent to write Twall as area weighted average at wall. Then after several runs that makes the total length equals to 2m . E.g if 12.8mm is segment length, I need at least 157 runs to make it 2000m or 2m. At the end, I take the average of Twall that was written by Fluent for each segment. This was for constant heat flux. I am not sure that approach is correct or not?