Martin J. Gander
Université de Genève
Section de Mathématiques
Rue du Conseil-Général 9, CP 64
1211 Genève 4, Suisse
martin.gander(at)unige.ch

Home Research Teaching Consulting Personal  
   
Seminars
Genève:
- Analyse numérique
McGill:
- CSE
- Applied Mathematics
Conference
Ascona 2013

Available projects
Postdoctoral:
- upon request
Doctoral degree:
- upon request
Master thesis:
- upon request

:: Fast non-conforming interface projections
An example of a fast advancing front technique to compute L2 projections from an arbitrary triangular grid onto another arbitrary one for non-conforming (e.g. mortar) methods. To obtain the Matlab code, see below. Check also out the 3d case, and the case for contact problems, where the interfaces two dimensional, but in 3d, and not at precisely the same geometrical location.



Below is the short Matlab code which does the projection: it is explained in the paper "An algorithm with optimal complexity for non-matching grid projections", M.J. Gander and C. Japhet, Proceedings of the 18th international conference of Domain Decomposition methods, 2008. The complete demo code in Matlab with an example can be downloaded here, and also the slides of the corresponding presentation with several helpful illustrations.
        
function M=InterfaceMatrix(Na,Ta,Nb,Tb);
% INTERFACEMATRIX projection matrix for nonmatching triangular grids 
%   M=InterfaceMatrix(Na,Ta,Nb,Tb); takes two triangular meshes Ta
%   and Tb with associated nodal coordinates in Na and Nb and
%   computes the interface projection matrix M

bl=[1];                        % bl: list of triangles of Tb to treat
bil=[1];                       % bil: list of triangles Ta to start with
bd=zeros(size(Tb,1)+1,1);      % bd: flag for triangles in Tb treated 
bd(end)=1;                     % guard, to treat boundaries
bd(1)=1;                       % mark first triangle in b list.
M=sparse(size(Nb,2),size(Na,2));
while length(bl)>0
  bc=bl(1); bl=bl(2:end);      % bc: current triangle of Tb 
  al=bil(1); bil=bil(2:end);   % triangle of Ta to start with
  ad=zeros(size(Ta,1)+1,1);    % same as for bd
  ad(end)=1;
  ad(al)=1; 
  n=[0 0 0];                   % triangles intersecting with neighbors
  while length(al)>0
    ac=al(1); al=al(2:end);    % take next candidate
    [P,nc,Mc]=Intersect(Nb(:,Tb(bc,1:3)),Na(:,Ta(ac,1:3)));
    if ~isempty(P)             % intersection found
      M(Tb(bc,1:3),Ta(ac,1:3))=M(Tb(bc,1:3),Ta(ac,1:3))+Mc;
      t=Ta(ac,3+find(ad(Ta(ac,4:6))==0)); 
      al=[al t];               % add neighbors 
      ad(t)=1;
      n(find(nc>0))=ac;        % ac is starting candidate for neighbor  
    end
  end
  tmp=find(bd(Tb(bc,4:6))==0); % find non-treated neighbors
  idx=find(n(tmp)>0);          % take those which intersect
  t=Tb(bc,3+tmp(idx));
  bl=[bl t];                   % and add them
  bil=[bil n(tmp(idx))];       % with starting candidates Ta
  bd(t)=1;
end
       
        
   
Home Research Teaching Consulting Personal