On Wed, Feb 15, 2012 at 1:13 AM, Daniel Wheeler
<[email protected]>wrote:

>
>
> On Mon, Feb 13, 2012 at 7:13 PM, Michael Brown <
> [email protected]> wrote:
>
>>
>>  Thanks for your help. Even if I don't end up using FiPy for my current
>> project, it looks like a great package and I'll definitely keep it in mind
>> for the future/recommend it.
>>
>
> Cool. Thanks.
>

No problem.


>  I don't think you need to split between the x and u parts of f as long as
> you can construct a g such that g_x = f. g can be arbitrary.
>
>

Okay, but the problem is f depends on u, so to solve g_x = f(u(x, t), x) I
need to know u, the unknown! That makes g a functional of u, not a local
function of u(x). That's why I tried to split off the u dependence, but it
doesn't look like it's going to help. Either way I get an
integro-differential equation to solve.  Either what I had before or:

u_t - v_x = 0
v_t - u_x = g = \int f(u(x, t), x) dx

It's starting to look like a nonlocal problem.


>> Also, how would this method extend to two or three dimensions?
>>
>
> You could use
>
>   u_t - v_x - w_y = 0
>
>   v_t - u_x = g
>   w_t - u_y = h
>
> where g and h are arbitrary as long as
>
>   g_x + h_y = f
>
> Either one could be zero for instance. It might matter in some numerical
> way, but not for the maths.
>

I get it now. That's a neat trick.


>
> Good luck!
>  --
> Daniel Wheeler
>
Thanks!
_______________________________________________
fipy mailing list
[email protected]
http://www.ctcms.nist.gov/fipy
  [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ]

Reply via email to