Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Timestep too small?
  • Hello,

    I'm currently trying to simulate a free decay test of a floating structure and I'm experiencing some problems after changing the timestep. The first simulation, which worked fine and had reasonable results, was run with the standard values:
    CFL = 0.2
    CoefDtMin = 0.05
    #DtIni = 0.0001
    #DtMin = 0.00001

    After this I just wanted to check, if the results are changed by using a different timestep. To do so, I changed the values to the following:
    CFL = 0.0002 (factor 1/1000)
    CoefDtMin = 0.05 (factor 1/1)
    #DtIni = 0.00000001 (factor 1/10000)
    #DtMin = 0.000000001 (factor 1/10000)
    As you can see, the factor we used in comparison to the stable simulation is not the same for the four variables.

    Simulating with these parameters led to errors that looked like numerical instabilities. After ca. 5 s there was a big increase in the fluid velocity and the floating body was moved to the right side by this accelerated fluid particles.

    Has anyone ever experienced something similar? Or is there something wrong in the definition, mainly because I didn't use the same factor for all four variables?

    Best regards,
    Jannik
  • You have to note what is the meaning of each value...
    As you know the time step is computed automatically be default.

    If you use a different CFL number (factor of 1/1000) then you will have 1000 times smaller time step... why do you want that?
    Changing Dtini and DtMin without removing "#" is not going to work... the use of "#" means that parameter is not going to be recognized so it is not being used... so default values are being used

    However if variable time step is computed automatically, why do you want to change it?

    Alex
  • Hello Alex,

    thank your for your quick response.
    I understand, that my changes in DtIni und DtMin did not have any influence because of the "#", but I still have a 1000 times smaller timestep by changing CFL, which leads (as I suppose) to these numerical instabilities.

    And I understand, that the timestep is variable and computed automatically, but I wanted to change it, because the match of my results and the experimental data isn't that good. Currently I'm checking the influence of some parameters to improve the results, and one of them is the timestep.

    Regards
    Jannik
  • "because the match of my results and the experimental data isn't that good. Currently I'm checking the influence of some parameters to improve the results, and one of them is the timestep..."

    I believe that maybe the velocities in your system are too large... that is why you need a smaller DT, however DT is computed startinf from your Speed of Sound... so what you need to change is the speed of sound or speed of system according to your problem...
    By default the maximum speed of the system is defined as the celerity of propagation of a dam-break

    Alex