search

LudwigBoess / SPHKernels.jl

Julia implementation of some common SPH kernels
MIT License
10 stars 0 forks source link

Question about x/h and h_inv #8

Closed AhmedSalih3d closed 3 years ago

AhmedSalih3d commented 3 years ago

Hello!

Thank you very much for making this package, I am trying to use it for some calculations.

This way of defining smoothing kernels is a bit new for me, in the sense that I have only seen kernels previously where "2h" (2 times the smoothing length) is the defining factor.

Therefore I am a bit confused of what I actually need to insert as value for x/h and h_inv.

I am trying to use WendlandC4 in 2D, where I have:

h = 0.04 m.

According to the notation used in Dehnen and Aly's paper, I suppose I should use;

x/u = x/(2*h)

h_inv = 1/(2*h)

But when I test this I get wrong results.

Would you mind clearing up my confusion?

Kind regards, Ahmed

LudwigBoess commented 3 years ago

Hi Ahmed,

please excuse the late reply, I neglected this package a little and only just saw the issue.

I checked and there was unfortunately quite a severe (and stupid) bug in the C4 kernel.

Can you please update to version 1.4 and check if you still get a wrong result?

Cheers, Ludwig

LudwigBoess commented 3 years ago

closing this for now

AhmedSalih3d commented 3 years ago

Hi again

I am looking into this again, sorry for not having answered earlier.

Based on your first comment, my understanding is that now everything should follow "conventional" notion of having q = r/h, where r is the Euclidean distance, h is smoothing length?

I am not that familiar with the kernels you have implemented so I will try to extend with one I know well as mentioned in your docs. I think it could be very interesting to have this kind of SPHKernel library you already have been working hard on.

Kind regards