Problems with Simulation of a Valve with Cavitation
 

Hi everyone,

I am working on a simulation of a valve connected to an discharge pipe, where I try to find the mass flow and velocity profile at the exit of this discharge pipe.
I keep the valve at a certain fixed opening (30%). At the inlet I set up a total pressure of 40 bar and at the outlet I set an opening pressure of 1 bar (absolute).

I initially did a run (Steady State - no Cavitation) with a smaller valve and everything seemed reasonable (reasonable expected massflow and the velocity profile was similar to what I expected and to literature).

With a larger valve, however, I noticed an unrealistic exit velocity profile and large zones with negative (absolute) pressure. As a result I turned on cavitation (water + water vapor) with steady state. With cavitation I get however:

- At the outlet of the discharge pipe (opening type, that allows only inflow of water) an inflow/backflow of water. There is no mass flow going out of the discharge pipe, hence large mass imbalances;
- Very large zones of water vapor (cavitation), especially near the outlet. The amount of water vapor does seem excessive.

I tried turning on total energy, but that did not change anything.

I searched online and in this forum and often I read that the backflow with cavitation can be solved by:
- Setting a lower opening pressure, or;
- Performing a transient simulation;
- I read that putting the outlet further downstream often does not work / resolve the problem.

The problem that I see with a transient simulation is that my model is quite large (discharge tube has a diameter of 2 meters and 8 meters long. The valve is relatively small. Element size 8-12 mm. In total I have 100 million elements. The flow velocity is locally very high, especially at the valve opening speeds can go above 50 m/s) and I read that for cavitation you need small elements and very small timesteps.

So to summarize:
- Does my backflow problem stem indeed from the fact that for accurate simulation of the cavitation you need a transient simulation?
- Using transient flow, I presume I should use adaptive time step. What is a suggested minimum time step or at least starting point for my time step for cavitation (I read someone posting 1E-6 sec, that seems very low)? The total flow-through time of the valve + discharge tube is around 3 seconds.
- Does it make sense to increase the element size (thus reducing the total amount of elements) to be able to run a simulation within a reasonable time? As increase the element size, reduces the accuracy of capturing caviation. Also the valve opening is relatively small, thus I need at least local refinement to capture the flow accurately.

I ran the 100 million steady state model with 100 iterations in 15 hours, so my computer is quite capable of handling a bit of calculations.

I hope I gave enough information. In case something is missing or needs more explanation, please let me know!

Why do you say increasing the length to the outlet does not work? That is the normal way to respond to these issues and it usually does work (or at least help).

Only run a transient simulation if you are sure it is transient as they are very expensive for simulations like this.

Your mesh size should be determined by a sensitivity study where you look at the accuracy you require. You should not determine mesh size based on your opinion of what the run time should be.

Thank you for your response.

- I read many posts / threats and almost all stated that simply increasing the length of the outlet did not help for cases with cavitation. That is why I dismissed it. For normal cases indeed, the first step for stabilizing the outlet condition is increasing the outlet length. Furthermore, due to the large outlet diameter, it is very computational expensive to make the outlet longer (although larger elements can be used towards the end of course).

- Exactly, I fear that in the limited time I have a transient run is not possible.

- I maybe posted it a bit crude, but initially I started with a coarser mesh, 1-10 million elements, to find the right opening grade of the valve. Then I used a finer mesh to see the sensitivity of the study and see how much the outlet mass flow changed (that is my most important variable). However, also with the cavitation model I keep having large mass imbalances (as a result of the backflow) Moreover due to very high local speeds (small first inflation layer), small openings and large size of the model I ended up with a lot of elements.

I also feared that using too large elements would not be able to capture the cavitation. I will try at least start a simulation with less elements and with a longer outlet to see what happens and then start to refine.

Well I can say if you have large sections of cavitated flow on your outlet boundary you are going to have a hard time converging. Before dismissing it out of hand you really should try it.

You will find that obtaining mesh convergence for a cavitation model will be challenging. If you have obtained an adequate mesh for a single phase model then that is a very good start for a cavitation model.

Thank you again for your suggestion. Tomorrow I'll do some trials, it can't hurt!