Re: Getting good results with NDSolve?

*To*: mathgroup at smc.vnet.net*Subject*: [mg62938] Re: Getting good results with NDSolve?*From*: dh <dh at metrohm.ch>*Date*: Fri, 9 Dec 2005 05:10:17 -0500 (EST)*References*: <dn8g3t$bqd$1@smc.vnet.net>*Sender*: owner-wri-mathgroup at wolfram.com

Hi Christopher, something is fishy with your expression: dg/dt = Norm(Grad g) * Laplacian(Grad g / Norm (Grad g)) On the left side you have a scalar. On the right side you have Laplacian(Grad g ...) what is a vector. Maybe you mean Div( Grad g / Norm (Grad g))????? Daniel pantagruel123 at aol.com wrote: > > Hi, I'm experimenting with NDSolve, not being an expert with differential equations. I'm getting some results, but it seems like I'm getting too much precision where I don't need it, and not enough where I do. I'm wondering if playing with the options would be something worth trying. Also it takes a really long time to compute what I've been told should take seconds with a C++ routine. > > If I want to solve a PDE of three variables (t,x,y) on the region {0,1}^3, how can I tell NDSolve to have, say a resolution of N x N points for (x,y) and a resolution of M for t? I find that if I just reduce the max points for t, then NDSolve quits before gets to the end of the interval. Maybe there is a way to tell NDSolve to trash some intermediate results and just keep every Mth point? > > The equation i'm using is dg/dt = Norm(Grad g) * Laplacian(Grad g / Norm (Grad g)) > where we're looking for g(t,x,y), and the right-hand derivatives are taken only with respect to x and y. My initial value g(0,x,y) had been generated from a ListInterpolation--I'm guessing this is the best way to use a bitmap as the inital value. > > Regards, > > Christopher Arthur > >