I recently stumbled over the following problem:
Consider a simple tube, closed at both ends. In my case it has a diameter of 4 mm and a length of about 0.1 m. If you introduce a small pressure impulse at one end the impulse will travel through the tube. The speed, at which the pressure disturbance advances, should equal the speed of sound of the medium. In my case the medium is air at a pressure of 1 bar (1E5 Pa) with a sonic velocity of about 330 m/s.
I set up a transient simulation with a local initial condition at one end, wich creates the initial pressure impulse of about 0.2 bar. After tweaking the time steps the impulse travels through the tube and is reflected at the opposing end, just as one would expect. At a sonic velocity of about 330 m/s the pressure impulse should travel the distance of 0.1 m in about 2.9E-4 seconds.
The speed, at which the impulse travels, by far exceeds the sonic velocity by a factor of about 100. This does clearly not make sense from a physical point of view. The speed, at which the impulse moves can be influenced by the time-step and mesh setting to a small degree, but I never managed to come even close to the actual sonic velocity.
However, if I enable the "High Mach Number Flow" option, the pressure impulse travels at just the right speed. However, since the pressure impulse is small there does not occur supersonic flow, which basically means that the accuracy of the solution obtained this way should be pretty low.
For my current project I need to investigate a resonance phenomenon within a cavity. I utterly failed to reproduce experimental results with Flow Simulation. In order to investigate this issue I set up the very simple transient problem detailed above, which should basically yield the cavitie's resonance frquency, which can also be calculated analytically with good accuracy. Until today I have employed steady-state FlowSim simulations for a number of projects with good results. This is my first real transient analysis.
Has anyone ever experienced similar issues? Maybe I'm overlooking something fundamental here. Or maybe the FlowSim solver is not designed for operating at these small timescales? Any advice on this topic is greatly appreciated.