Length of Wave Simulation When using Jonswap Piston

Hello,

I'm wondering if anyone has any advice on my issue.

I'm using a piston wavemaker with irregular waves in a nuumerical flume. From some literature reviews I've found that the length of a numerical simulation with a jonswap spectrum should be your peak period, Tp*150, and that the dt or "time out" should be your Tp/800, in order to properly capture a full jonswap spectrum. I'm running cases with a Tp between 3 and 7 seconds. This is creating some very long simulations even with a dpi of say .02m. Its also taking up a ton of space on my hard-drive to run these. Does anyone have any different guidance on this that could lessen my computational time? I haven't been able to find an SPH-specific paper that talks about simulation length for irregular waves.

Thanks,

Liam

Comments

  • Hello!

    From a practical perspective you could look into how long it takes for your simulation to produce a stable signal. If your simulation produces stability after 0.5T, why let it finish to T, even if litterature requests it?

    Also what litterate recommends it? Perhaps they use/work with different methods, where it is more feasible.

    It could also be that what they require is correct but that the hardware you are using insufficient.

    Personally I think that practically it is always a trade-off between simulation length, convergence and stability.

    Kind regards

  • edited February 2023

    Hi,


    The theoretical surface elevation is created in the WavePaddle_mkxxxx.csv file when you begin running a DualSPhysics simulation.


    1. I would recommend you evaluate the spectrum from the time series with different lengths and timeout. Identify the minimum length and timestep where you can recreate the input spectrum to whatever standard you deem acceptable.
    2. For the storage issue, you can run the simulation in parts. Once a part is done, you can complete the postprocessing and delete the out files.
    3. If your simulation is symmetric about y axis, I would say try running a 2D simulation.
    4. Using Inlet/Outlet will reduce domain size. But I am not so sure if that will necessarily speed up the computation. See https://forums.dual.sphysics.org/discussion/2115/why-does-inlet-outlet-take-a-significantly-longer-time-to-simulate-than-piston#latest asked by me.


  • Hi,

    I would recommend to create a domain that is long at least as 1 wavelength (in deep water, this corresponds to 1.56Tp^2).

    About the output time, there is no rule, since your calculation timestep has nothing to do with that. I would use 20Hz, should be already enough to measure water surface elevation and reconstruct your spectrum.

    Regards

Sign In or Register to comment.