[kveki]

I apologize in advance for my poor English, I use a translator.

Dear Friends, I ask for help in the following question.
There is a simple task: the pipe section is considered under axisymmetric conditions. Input and output - translational periodic boundaries. I want to get a steady turbulent flow of water in the pipe.
I work in Fluent and there are no problems. For periodic boundaries, I set either the mass flow rate or the pressure gradient. In all cases, I get a convergent solution and physically correct values.
Now I decided to do the same on CFX 19.2. Similarly, I set translational periodic boundaries, but as a result I get incomprehensible things.
1. When I set the mass flow rate, I get sinusoidal fluctuations in flow and pressure drop along a pipe with a wide amplitude. The values are clearly non-physical and strangely periodic.
2. When I set the pressure gradient, the situation in CFX is noticeably better - there are no fluctuations, the solution almost reaches the required value. At the same time, in order for the velocity and flow rate in the channel to reach stationary values, the Residual Target must be set no higher than 1e-10!!!, otherwise (for example, with 1e-5) the flow rate in the channel is very far from the real value. There is no such thing on Fluent and everything converges normally with residuals =1e-5.

I made a test case for fluent and CFX:
R_pipe= 0.1 [m]
L_pipe= 0.2 [m]
rho= 998.2 [kg/m3], mu= 0.001003 [kg/(m*s)]
mass flow rate= 78.398 [kg/s] -variant 1 (corresponds to the average velocity in the pipe 2.5 [m/s])
press gradient= -207.1115 [Pa/m] -variant 2 (corresponds to the average velocity in the pipe 2.5 [m/s])

initialization:
u= 0.1 [m/s] (I specifically took a value very different from 2.5 [m/s] so that the convergence of the solution was clearly visible)
k= 0.01 [m2/s2]
eps= 1 [1/s]

Characteristics of the established flow in the pipe:
average velocity 2.5 [m/s]
mass flow rate= 78.398 [kg/s]
press gradient= -207.1115 [Pa/m]

In fig.1 and fig.2 for comparison, the graphs of the average velocity in the pipe obtained on fluent and CFX are shown for the cases of assignment at the periodic boundaries of mass flow rate and press gradient

fig1.jpg, fig2.jpg

Why are there such differences in solutions on Fluent and CFX? What are these strange sinusoidal oscillations when setting the mass flow rate at periodic boundaries in CFX?

Since I started working with CFX recently, it is possible that this is my mistake, but I will not understand where it is. I apologize in advance if the answer is corny simple!

[Opaque]

In Ansys CFX, did you set the Domain Interface Target in the Solver Control panel?

If you monitor the domain interface imbalance, is it converging slowly?

For the mass flow option, it seems you need some under-relaxation for the updates. Looking into the installation files, you can include a

Pressure Update Multiplier = <real number>

along with the Mass Flow Rate specification.

It seems the default is 0.25, perhaps needs a smaller value, say 0.1

[kveki]

Thanks for the quick response!

Yes, I set Domain Interface Target to values 1 e-2 and 1e-5. There was no difference in the results.

I also experimented with Pressure Update Multiplier. Smaller values reduce the frequency of oscillations, but do not eliminate them in any way.

The strange thing is that this is not just a dissimilarity of the task, but clearly expressed periodic fluctuations. I can't find an explanation for this yet.

[ghorrocks]

If Pressure Update Multiplier operates like an under relaxation factor then there will be a threshold value above which you get oscillations and below you get convergence. So keep lowering the Pressure Update Multiplier until you do not get oscillations.

[kveki]

I set the multiplier = 0.0001 for the experiment, and the oscillations began to damped. The task converged to the established mass flow rate. Looks like you're right!
fig1.jpg
Now I'm experimenting with different values.

Thanks!

[kveki]

Additionally, I wanted to share some information. I made calculations with the values of Pressure Update Multiplier 0.25, 0.01 and 0.0001. Graphs of changes in mass flow rate are shown in the figures.
fig2.jpg fig3.jpg
It can be seen that at PUM = 0.25, the flow fluctuations do not fade and have a constant amplitude. At PUM= 0.01, there are damped oscillations with a much lower frequency, but when the mass flow rate converges to the set value, the oscillations still occur (albeit with a very small amplitude). And finally, at P= 0.0001, the oscillation frequency is very low (although the amplitude has increased compared to the case of PUM = 0.01) and the solution exactly converges to the set value. At the same time, convergence is achieved approximately by the 16000th iteration, which is 2 orders of magnitude higher than on Fluent, where the solution converged already on the 150th iteration.
I have a question, what is the difference between CFX and Fluent periodic boundary techniques? Why are there such big differences in the solution? What is the Pressure Update Multiplier parameter and why is it not in Fluent?