[enr_venkat]

Am currently solving a steady state 3D Centrifugal pump problem using ANSYS CFX.

Objective:

To validate the performance curve of pump based on experimental inputs

Domain decomposition:

Stationary domain: Inlet, Outlet and Casing (1 Body)

Rotating domain: Impeller fluid volume surrounding the solid impeller. But solid impeller trace is left within the impeller fluid volume

Domain interface: Between Stationary domain and rotating domain.

Method: SRF and Frozen Rotor with no pitch change

Mesh:

Fine On curvature with skewness of 0.92 and Aspect ratio of 60

Boundary conditions used:

Domain ref pressure = 1 bar.

Inlet: Total pressure = 0 bar

Outlet: Mass flow rate = 0.065 kg/s or 3.9 lpm.

Turbulence intensity: Medium

Heat transfer: none

Rotating domain: 3100 RPM about Global Y

Issues:

Convergence beyond 1E-3 seem to be impossible with lot of wiggles Number of iterations: 700; Criteria: Want to set 1E-6 but I set it to default for first run.

If I consider the heat transfer, I get similar behavior. My idea is to consider the temperature later when I get the desired pressure rise.

I tried various combinations of BCs. Gone through previous threads.

It was so tricky to get the solution converged even if I use thumb rule of 1/2w , 1/4w....as physical timescale. Checked with automatic timescale in the beginning.

After 290 iterations, flow is getting blocked at inlet and outlet with changes in block % ranging from 0.5 % to 67%

Any suggestions or discussions related to this matter is highly appreciated.

Please find attached the current status as images:

[ghorrocks]

FAQ: http://www.cfd-online.com/Wiki/Ansys...gence_criteria

[surajkashyap]

Hi, you have mentioned a thumb-rule to determine the physical timescale. Can you please elaborate? How do I estimate a starting physical timescale given the speed and number of blades in the impeller?
I am analysing a centrifugal pump with 6 guide vanes running at 1440 rpm.

[ghorrocks]

The general rule of thumb is a starting point for the time step size is the fluid residence time in the simulation domain (if the flow goes through without gross recirculations).

From there you should adjust it higher or lower as described in the FAQ.

You can also start with a guessed time step size and adjust it from there. This works fine as well in many applications.

[highorder_cfd]

Hi

First, I would try with a big time step, something like auto timescale with a factor of 10, or physical time scale 1 or 2 sec, to try to remove some unsteady behaviour. Second, I would try to increase (ramp up) the rpm slowly, maybe starting from 1000 rpm.
Finally, I would try it is to rotate the impeller of few degrees. As you are solving with a MRF model, the impeller position will impact on the final solution. A slight change in the configuration might help the convergence.

[surajkashyap]

@highorder_cfd and @ghorrocks Thanks for the tips. I achieved a much faster convergence this time

[Gape]

How did you manage it? with which Timescale?

[ghorrocks]

What worked for his simulation is unlikely to work for yours as they are all different. Read the FAQ: http://www.cfd-online.com/Wiki/Ansys...gence_criteria