compression
Hi all
I am trying to simulate a couple of floating bodies in a water tank that I allow to fall over time. So here comes the problem. I always observe a loss in height (i.e. height of the water level decreases over time). This looks quite extreme in a 2d case. Then I tried to benchmark h, speed of sound time steps... all parameters that might affect the compressibility of the fluid but it still is compressing to a certain extent. I am kind of happy with the flow speeds however looking at that continious loss in hight seems somewhat obscure. I am not sure if this is an effect of the dynamic boundaries and if yes, what would be a best practice guide?
- Should one create a void around each floating that would be fixed over some time so the fluid could equilibrate?
- Should one wait for some time - "warm up" and then allow the floating’s to fall? But, after some time those strange fluctuations occur again - Stillwater case problem…
- Should one generate the floating’s out of the fluid and let the drop into it?
I appreciate any help !!!
I am trying to simulate a couple of floating bodies in a water tank that I allow to fall over time. So here comes the problem. I always observe a loss in height (i.e. height of the water level decreases over time). This looks quite extreme in a 2d case. Then I tried to benchmark h, speed of sound time steps... all parameters that might affect the compressibility of the fluid but it still is compressing to a certain extent. I am kind of happy with the flow speeds however looking at that continious loss in hight seems somewhat obscure. I am not sure if this is an effect of the dynamic boundaries and if yes, what would be a best practice guide?
- Should one create a void around each floating that would be fixed over some time so the fluid could equilibrate?
- Should one wait for some time - "warm up" and then allow the floating’s to fall? But, after some time those strange fluctuations occur again - Stillwater case problem…
- Should one generate the floating’s out of the fluid and let the drop into it?
I appreciate any help !!!
Comments
1.) Are you losing any particles during the simulation?
2.) How much height (absolute and relative to dp) are you losing?
3.) Are you comparing your results with CaseFloatingSphereVal2D?
4.) If you use the FtPause parameter (for, say, 1.0 seconds or more), does the water height decrease during the time that floatings are paused?
5.) Are your constants (coefsound, cflnumber, etc.) the same as those in CaseFloating_Def.xml or CaseFloatingSphereVal2D_Def? If not, how different are they?
6.) Are your floating bodies rising in the water tank as the height of the water level decreases over time? If so, then perhaps your floating bodies are submerged in the tank deeper than their density allows and they are rebounding. You could try submerging them higher up, but not above the water, unless you are okay with splashing. I apologize if this question is missing the point of your issue, but I ask it for the sake of thoroughness.
Thank you so much for your willing to help me out.
1) No, I am not loosing any particles.
2) I loose over 5 seconds, 3 cm of water height (the initial height of the model is 24m) which is a dploss of 20.
3) Yes, I am doing that. However, the difference is that I generate the floatings within the fluid. I was thinking to set a void around them so the boundary can develop...
4) Yes, the water height is decreasing by 20 dp. My FtPause is set to 5 seconds, but the water height keeps decreasing over time.
5) My dp is 0.0015, the speed of sound is actually set to 120 but due to the model height it becomes set by gencase to 153.393. Which, is a reasonable value when comparing it to the other cases, I guess. The one in case floating is 138.664. In addition to that I tries a very high resolution case. But it did not really help.
6) well that’s kind of funny. I had rising floating bodies in combination with "boiling / overcooking" water before, I even was able to expand them. However, after carefully adapting the h value, the speed of sound, the "timestepping" i.e. setting the CoefDtMin, I kind of got rid of these effects, despite the fact that I observe always this strange compression. Actually I try to avoid any splashing.
I hope that my answers give you a better understanding of what`s causing this compression.
Thank you so much I appreciate any help!!!
Can I ask how long is your simulation? I mean how long does it take to "lose" 20 dx of height in your simulation.
cheers
G.
PS: what is you δ-SPH value?
I have a 2D case. It is 24 m in Z and 12 m in X. I am soing some benchmarking and increase the model lenght. It seams to get better, but I never get such a ntabile solution as in the floating case. I simulate 30 s and it takes around 5 secounds to loose 20 dp. Obviesly in dependence to the speed of sound it somewhat varies. The delta-sph value is set to 0.1. Thank you very very much.
Best H
First of all thanks to the ones that intended to help me on this. I kind of got rid of that strange compression by increasing the length of model i.e. fluid and fixed boundaries. It seems that the model is unhappy when the Z scale is way larger than X. Actually I ended up with two cases.
One case is 30 m in length and the other one is 32 m in length. The slightly longer case surprisingly behaved quiet well and I observed a similar pattern as in the floatingval2d case. The slightly shorter case did strange things, again. Despite different domain sizes and slightly different x lengths, both cases have exactly the same parameters e.g. same dp, same mass, or same cs0 158 that was automatically calculated over the height.
In the case of the shorter 30 m run the height in the model was more or less maintained over time, but a flow towards the model bottom started to develop, right at the surfaces of the floating bodies, which I found super strange. Then I decreased the coefh value from 1.2 incrementally to 0.7, which influenced the dtini value. By doing that these strange flow patterns around the floating bodies disappeared, and I got a more or less acceptable pattern again. In addition to that I decreased the cflnumber from 0.2 to 0.1 which also had an effect on the time stepping, and somehow seemed to affect control the flow behavior.
Anyway, I find it very strange that two almost similar cases (only 2 meters difference in length, and different domain sizes), require a large difference in the coefh to obtain comparable results. I am not sure but I remember from one of the forum entries that the cell size has an effect on the performance of the simulations in dependence of the coefh. Anyway, I am somewhat irritated, which simulations I can trust on, if only minor differences in the parameters cause such differences.
May be some of you could explain that to me … Best H
On the other hand, if you resolution is poor you can define by hand the mass or density of the object.
These two tips should help.
Regards
thank you very much for your kind input. I already had the draw mode "full" activated. And I used the relative weight so it computes the mass of the floating with respect to the mass of the mass bound value, which I guess is the best option also in low resolution cases. Anyway the issue with the velocity increase at the surface of the floating remains. I could track it down to the relative density.
In order to make thinks a bit clearer. I did the following. I Copy pasted the case FloatingVal2D in a new folder. Then I generated the cylinder into the centre of domain at Z = 0.05, instead of at the top of the water column. In addition I set the Ftpause value to 5 seconds to see what happens if I keep the floating fixed. When I run the case with the present rel. dens of 1.6 everything is fine. But when I increase that value to 2.6 than the velocity at the boundaries of the floating starts to build up and some artificial velocity plume, appears to develop. In reducing the coefvalue from 1.2 to 0.8 it’s getting better, also lowering the cflnumer to obtain a smaller time step is also helpful, but I could not get a similar pattern as for the 1.6 rel. density case. Since I want to apply a density of 2650 to replicate an experimental case, it would be extreamly helpful if you could comment on that. Thank you and enjoy your weekend. Best H
Alex
I want to believe… but I am sorry, maybe I am missing something. But to the best of my knowledge in the CaseFloatingSphereVal2D there is no stl that`s getting loaded, right ???
I only modified the cases that comes in DualSphysics by changing the relative weight to 2.6 and setting the FtPause to 5 seconds. The problem I have now doing that is that there will be a artificial flow developing, which I do not understand. I attached to screenshots for clarity.
https://drive.google.com/drive/folders/0B3ajAgMnxT6kNXUzaXhnaEd1UEk
One screenshot is from the case with a rel. weight of 1.6 and one with a relative weight of 2.6. The same also happens if I use the other options such as rhopbody="2600".
Thank you very much for your help, I appreciate it a lot.
// *********************************************
<======
<setmkbound mk="51" />
<===== changed only that to 2.6