Calculation of the roll moment of an oscillating U tank

Dear all,

I am trying to get the roll moment of a 2D U-shaped tank in forced oscillating motion. The tank is partially filled with water.

Since the computeforces tool does not provide such a quantity I am trying a different approach.

I measure pressures on several probes over the tank boundaries (I took the sloshing simulation as a reference starting point). The idea would be to compute the transverse and vertical forces from the pressure distribution. To compare these forces (e.g. the phase averaged time histories) with those resulting from te compute forces. To eventually scale the forces from pressures by the founded ratio. To compute the force moment (force x arm) by using the corrected pressures.

This reasoning would imply to get an almost correct pressure distribution at least in terms of 'shape' since the absolute value would then be corrected.
However I get very fluctuating pressures that lead to un useful results.

I have also checked the pressure distribution with the steady tank (no motion at all). At the very first time step the distribution seems correct and the described process results in the correct hydrostatic force. Unfortunately, from the second time step even those pressures on the steady tank have fluctuations that do not converge anywhere.


is there a chance to get rid in some ways of those fluctuations (I read in other discussions this is related to the boundary condition. BTW I am using delta-sph=0.1)?

or alternatively,

has someone another idea to compute the force moment I am looking for?

Thanks a lot


  • Hi I dont kown if it helps ... I did not really find a exact solution to get rid of the fluctuations but a combo of all these steps somewhat helped a bit...

    - Put everything on auto in the xml! -hswl -speedsystem - speedsound
    - Play with the coefsound. So far I used 20 for most cases.
    - The soothing lengh needs also some care. I replaced the line coefh value with hdp - -> hdp value="1.4" comment="coef to calculate the smoothing length (hdp=h/dp)".
    I did that because I endup with the same values for the coehf as usd in the example cases e.g. 0.8..... I found the best values between for hdp of 1.2 -1.5... but I think Alex knows that way better :=) ???
    - Reduce the cfl number ->> smaller the value a bit
    - Double check the particle spacing -> dp and your domain dimensions sould be "devidable". (If you case is 10 m long use a dp that results in a equal number)
    - Ensure that the fluid particles don't "sit" ontop of the boundaries and doublecheck again that the dimenions you use are devideable by the dp again.
    - If you accelerate the flow by any force such as shaking the box as in your case dont do it to fast. Give the system some time. I got fluctuations just becaus i was impatient and accelerated to fast. Try to accelerate the box slowly in small increments until you reach your target frequency.
    - May be this one also helps, but I am not 100% sure. Doublecheck the aspect ratio of your case. I think I had some troubles with that too.
    Best H
  • Hi, based on a paper "A multi-phase particle shifting algorithm for SPH simulations of violent hydrodynamics with a large number of particles" increase speed of sound can reduce pressure fluctuations. I also use several coefsound(ver4.0) for sloshing case and it shows pressure fluctuations can be reduced using an appropriate speed of sound.

  • Thanks. I am trying new simulations following your suggestions.
    I have processed the results from a previous one with a higher number of particles (around 1 ML). Attached there is a plot of the both the transverse (Fx) and vertical (Fz) forces on the whole tank. Blue ones are those computed by the 'computeForces'. Red ones are those integrated from pressures. There is a very good agreement on the Fx while the Fz have an increasing trend (almost linear, as shown in the last subplot) that is quite strange. If such a global trend is removed the two signals are again in good agreement.

    So, is there a numerical reason for this 'linearly' increasing trend in the vertical forces? and eventually do you known how to handle this?

    Thanks again for your help.

Sign In or Register to comment.