Thanks for your reply.
I probably found the reason of the problem. As said before I'm trying to solve an eq of this type: A(d phi/d t) = div (D grad phi) + q where A and D are costant coefficient and q is a spatially varying heat source. The problem is in how I create the cylindrical 1D mesh (origin of the mesh is not in 0.). It doesn't matter if the discretization is logarithmic or uniform but how I declare it: (using a uniform spacing:) ** mesh = CylindricalGrid1D(dx=dr, nx=(len(DR))) + (r_int,) ** SHOULD BE EQUAL TO: ** mesh = CylindricalGrid1D(dx=DR) + (r_int,) ** [dr = 5e-04, nr=58, r_int=0.00125, DR is a list which contains the various dx(58 elements of value dr for uniform grid)] BUT It is NOT. The mesh (cell centers, facecenters) is ok but the results are NOT. The results are right only if I create the mesh using dx=dr and nx=.. . And this is why before I thought the problem was the logaritmic discretization (must use DR=[...]). As said before with Cylindrical 2D mesh results are instead correct. Is this a bug? > Date: Thu, 6 May 2010 11:11:10 -0400 > From: [email protected] > To: [email protected] > Subject: Re: Problem in solving Poisson equation with 1D Cylindrical mesh > > > Edward, This may be to do with having a very small volume (or area or > line or point) for the inner most element of the domain. It should be > the same whether one is using a 1D or 2D mesh. Since you are getting > differences in the 1D and 2D case, it should be relatively easy to > debug and figure out what's going on. It could also be that the > boundary condition on the inner boundary as zero area and this is > causing issues. Try shifting the grid by a small value away from the > zero point and see if things are improved. I have always had this > issue with cylindrical grids and have never really had a satisfactory > solution (other than shifting away from the zero point). If you > discover a better way to handle this, let me know. Cheers. > > > If you can't debug it, then send me the most minimalist scripts that > show the issue and I'll give it a shot. > > On Wed, May 5, 2010 at 9:53 PM, Eduard Manley <[email protected]> wrote: > > Problem partially solved: > > > > I'm using a logarithmic discretization (first dr= 5e-04 and next > > dr increasing as 1.05)(with internal radius= 0.00125, external radius=0.03) > > and, > > for some unknown reason this create problems and wrong result with > > cylindrical 1D mesh. I tried using a uniform discretization (dr=5e-04 nx=58) > > and now the result is correct. > > > > However I need to use the logarithmic discr. so after some hours of sleep > > I'll think about the reason..... > > > > ________________________________ > > Hotmail: Trusted email with powerful SPAM protection. Sign up now. > > > > -- > Daniel Wheeler > > _________________________________________________________________ Your E-mail and More On-the-Go. Get Windows Live Hotmail Free. https://signup.live.com/signup.aspx?id=60969
