[katty17]

Hi Glenn,
Many thanks for your prompt attention and help.


I need some time to implement your suggestions into the simulation. I will inform you about the result.


About the tiny subdomain; may be “thin” was a better term rather than “tiny”. Sorry for not clarifying enough. Attached is a brief description of the problem. According to this image, in reality when the outlet pressure gets higher than the inlet one (t=t0), the valve (the two moving walls) blocks the flow path and prevents backward flow in the duct. However we are not interested in flow distortion around the valve, and hence as is shown in the attached figure, we only used a subdomain to obstruct the flow path at t=t0. The difference between the pressure values at two sides of this thin subdomain is very high.

In my experience, the faster we increase in the value of general source term coefficient, the bigger pressure pulse we will see in our results.
About the pipe deformation; this error is occurs even if the walls are modeled flexible.


Regarding the convergence criteria; I was not able to bring the values of residuals below 0.0003 after that the valve is fully closed. Is it acceptable?


Best

[ghorrocks]

Is the opening action of the valve important? It is easier to assume the valve opens instantly, because you can then just start with an initial condition with a pressure difference and the flow progresses from there. You don't need to model any action to do this at all.

In my PhD thesis I used modelling shock wave flow very similar to this as a benchmark case for CFX4. This may be useful for you: https://opus.lib.uts.edu.au/handle/10453/20133

[katty17]

If I got it correctly, you mean it is better to start the simulation from t=t0 till t=T+t0. This way we can provide the simulation with better initial conditions.
Seems to be a good solution because the opening action of the valve is not important for me at all.

Many thanks for sharing you PhD thesis. I downloaded it and need to read it. It should be replete with very useful information not only for this simulation but also for other cases.

Best

[ghorrocks]

Have a look how I modelled the shock tube in my thesis. Simple, nothing moves (except the gas, of course) and you should have no problems getting it to converge. My thesis also shows you how you should validate the model with checking convergence criteria, time step size, mesh size and other factors.