PFV 3D convection matrix
From CFD-Wiki
Revision as of 12:20, 20 July 2011 by Jonas Holdeman (Talk | contribs)
Function CMat3D8W.m for pressure-free velocity convection matrix
function [Cm,RowNdx,ColNdx]=CMat3D8W(Xe, Elcon, nn2nft, Vdof)
% CMat3D8W - Returns the element convection matrix for the
% 3D linear-complete, normal-conforming, divergence-free, Hermite basis
% functions on 8-node rectangular hexahedral elements with 6 DOF per node
% using Gauss quadrature on the 2x2x2 reference cube.
% The 8 columns of the array V3dof must contain the six degree-of-freedom
% vectors in the nodal order (A,B,C,u,v,w).
% The assumed nodal numbering starts with 1 at the lower left corner (-1,-1,-1)
% of the element.
%
% Usage:
% [Cm,Rndx,Cndx] = CMat3D8W(Xe, Elcon, nn2nft,V3dof)
% [Cm,Rndx,Cndx,Rcm,RcNdx] = CMat3D8W(Xe, Elcon, nn2nft,V3dof)
% Xe(1,:) - x-coordinates of 8 corner nodes of element.
% Xe(2,:) - y-coordinates of 8 corner nodes of element.
% Xe(3,:) - z-coordinates of 8 corner nodes of element.
% Elcon(8) - connectivity matrix for this element, list of nodes.
% nn2nft(1,n) - global freedom number for node n.
% nn2nft(2,n) - global freedom type for node n.
% Vdof(6,8) - VP & velocity Dofs at 8 nodes.
%
% Calls:
% V8cW(nc,x,y,z), V8xyzcW(nc,x,y,z)
%
% Jonas Holdeman, July 2011
%
% Constants and fixed data
nc=[-1,-1,-1; 1,-1,-1; -1,1,-1; 1,1,-1; -1,-1,1; 1,-1,1; -1,1,1; 1,1,1]; % defines corner nodal order
ndfn=6; % number of degrees of freedom per node.
nne=8; % number of nodes per element
ndfe=ndfn*nne; % number of degrees of freedom per element
% Define 5-point quadrature data once, on first call.
% Gaussian weights and absissas to integrate 9th degree polynomials exactly.
global GQC5;
if (isempty(GQC5)) % Has 5-point quadrature data been defined? If not, define arguments & weights.
Aq=[-.906179845938664,-.538469310105683, .0, .538469310105683, .906179845938664];
Hq=[ .236926885056189, .478628670499366, .568888888888889, .478628670499366, .236926885056189];
GQC5.xa=zeros(125,1); GQC5.ya=zeros(125,1); GQC5.za=zeros(125,1); GQC5.wt=zeros(125,1);
nr=0;
for nz=1:5; for ny=1:5; for nx=1:5
nr=nr+1; GQC5.xa(nr)=Aq(nx); GQC5.ya(nr)=Aq(ny); GQC5.za(nr)=Aq(nz);
GQC5.wt(nr)=Hq(nx)*Hq(ny)*Hq(nz);
end; end; end
GQC5.size=nr;
end
xa=GQC5.xa; ya=GQC5.ya; za=GQC5.za; W=GQC5.wt; Nq=GQC5.size;
% ---------------------------------------------------
global Z3_V8c; global Z3_V8xc; global Z3_V8yc; global Z3_V8zc;
if (isempty(Z3_V8c) | isempty(Z3_V8xc) | size(Z3_V8xc,2)~=Nq)
Z3_V8c=cell(nne,Nq); Z3_V8xc=cell(nne,Nq);
Z3_V8yc=cell(nne,Nq); Z3_V8zc=cell(nne,Nq);
for k=1:Nq
for m=1:nne
ncm=nc(m,:);
Z3_V8c{m,k}=V8cW(ncm,xa(k),ya(k),za(k));
[Z3_V8xc{m,k},Z3_V8yc{m,k},Z3_V8zc{m,k}]=V8xyzcW(ncm,xa(k),ya(k),za(k));
end
end
end % if (isempty(*))
% ----------------- End fixed data ------------------
Ti=cell(nne);
for m=1:nne
Jt=Xe*GTL(nc(:,:),nc(m,1),nc(m,2),nc(m,3));
Det=det(Jt);
JtiD=inv(Jt)*Det;
J=Jt';
Ti{m}=blkdiag(J,JtiD);
end % loop m
Cm=zeros(ndfe,ndfe); S=zeros(3,ndfe); % Preallocate arrays
% Begin loop over Gauss-Legendre quadrature points.
for k=1:Nq
Jt=Xe*GTL(nc(:,:),xa(k),ya(k),za(k)); % transpose of Jacobian at (xa,ya,za)
Det=det(Jt);
JtbD=Jt/Det;
Jti=inv(Jt);
JtiD=Jti*Det;
Ji=Jti';
%
% Compute mapped element Si and the fluid velocity at the quadrature point (xa,ya,za).
Ua=[0;0;0];
for m=1:nne % velocity & derivatives at 8 corner nodes
mm=ndfn*(m-1); mm3=mm+1:mm+ndfn; ncm=nc(m,:);
S(:,mm3)= JtbD*Z3_V8c{m,k}*Ti{m};
Ua = Ua + S(:,mm+1:mm+ndfn)*Vdof(:,m); % A,B,C,u,v,w
end
Ub=Jti*Ua;
UgS=zeros(3,ndfe);
for m=1:nne % velocity & derivatives at 8 corner nodes
mm=ndfn*(m-1); mm3=mm+1:mm+ndfn;
UgS(:,mm3)=JtbD*(Ub(1)*Z3_V8xc{m,k}+Ub(2)*Z3_V8yc{m,k}+Ub(3)*Z3_V8zc{m,k})*Ti{m};
end
Cm = Cm + S'*UgS*W(k)*Det;
end % loop k over quadrature points
gf=zeros(ndfe,1);
m=0;
for k=1:nne
m=m+1; gf(m)=nn2nft(1,Elcon(k)); % get global freedom number
for k1=2:ndfn
m=m+1; gf(m)=gf(m-1)+1; % next
end % if
end % loop on k
RowNdx=repmat(gf,1,ndfe);
ColNdx=RowNdx';
RowNdx=reshape(RowNdx,ndfe*ndfe,1);
ColNdx=reshape(ColNdx,ndfe*ndfe,1);
Cm=reshape(Cm,ndfe*ndfe,1);
return;
% -----------------------------------------------------------------------
function G=GTL(ni,q,r,s)
% Transposed gradient (derivatives) of scalar trilinear mapping function.
% The parameter ni can be a vector of coordinate pairs.
G=[.125*ni(:,1).*(1+ni(:,2).*r).*(1+ni(:,3).*s), .125*ni(:,2).*(1+ni(:,1).*q).*(1+ni(:,3).*s), ...
.125*ni(:,3).*(1+ni(:,1).*q).*(1+ni(:,2).*r)];
return;
