// matlab
function [Ke, Me, TLG] = MITC4(Young, Poisson, thick, density, node, Vn)
nNPE = 4; % # of nodes per element
nDPN = 5; % # of DOFs per node
nDIM = 3; % Problem dimension
nGP = 2; % # of Gauss points
% Vn = Make_Vn(node);
[V1,V2] = Director(Vn(1,:)); % norn vec 구하기
TLG = [V1' V2' Vn(1,:)'];
Ke = Shell_Stiffness(Young, Poisson, thick, node, Vn);
Me = Shell_Mass(thick, density, node, Vn);
% nodewise2dofwise = reshape(1:nDPN*nNPE,nDPN,nNPE)';
% nodewise2dofwise = nodewise2dofwise(:)';
dofwise2nodewise = reshape(1:nDPN*nNPE,nNPE,nDPN)';
dofwise2nodewise = dofwise2nodewise(:)';
Ke = Ke(dofwise2nodewise,dofwise2nodewise);
Me = Me(dofwise2nodewise,dofwise2nodewise);
function Me = Shell_Mass(thick, density, node, Vn)
Me = zeros(nDPN*nNPE);
for i = 1:nGP
for j = 1:nGP
for k = 1:2
% ! Gauss integration point and weight factor
[r, s, t, weight] = Gauss22_Point(i, j, k);
% ! Jacobian
Jacob = Jacobian(r, s, t, thick, node, Vn);
% ! uvw: shape function matrix for displacment
uvw = Displacement(r, s, t, thick, Vn);
% Me = Me + weight * det(Jacob) * thick * transpose(uvw) * uvw * density;
Me = Me + weight * det(Jacob) * transpose(uvw) * uvw * density;
end
end
end
end
C++
#include <Eigen/Dense>
#include <vector>
using namespace Eigen;
void MITC4(double Young, double Poisson, double thick, double density, const MatrixXd& node, const MatrixXd& Vn, MatrixXd& Ke, MatrixXd& Me, MatrixXd& TLG) {
const int nNPE = 4; // # of nodes per element
const int nDPN = 5; // # of DOFs per node
const int nDIM = 3; // Problem dimension
const int nGP = 2; // # of Gauss points
Vector3d V1, V2;
std::tie(V1, V2) = Director(Vn.row(0)); // norn vec 구하기
TLG = (MatrixXd(3, 1) << V1, V2, Vn(0, 0)).finished();
Ke = Shell_Stiffness(Young, Poisson, thick, node, Vn);
Me = Shell_Mass(thick, density, node, Vn);
VectorXi dofwise2nodewise = VectorXi::LinSpaced(nDPN * nNPE, 1, nDPN * nNPE);
dofwise2nodewise = dofwise2nodewise.array() - 1; // Convert to 0-based indexing
Ke = Ke(dofwise2nodewise, dofwise2nodewise);
Me = Me(dofwise2nodewise, dofwise2nodewise);
}
MatrixXd Shell_Mass(double thick, double density, const MatrixXd& node, const MatrixXd& Vn) {
const int nDPN = 5; // # of DOFs per node
const int nNPE = 4; // # of nodes per element
const int nGP = 2; // # of Gauss points
MatrixXd Me = MatrixXd::Zero(nDPN * nNPE, nDPN * nNPE);
for (int i = 0; i < nGP; ++i) {
for (int j = 0; j < nGP; ++j) {
for (int k = 0; k < 2; ++k) {
// ! Gauss integration point and weight factor
double r, s, t, weight;
std::tie(r, s, t, weight) = Gauss22_Point(i, j, k);
// ! Jacobian
MatrixXd Jacob = Jacobian(r, s, t, thick, node, Vn);
// ! uvw: shape function matrix for displacement
MatrixXd uvw = Displacement(r, s, t, thick, Vn);
// Me = Me + weight * det(Jacob) * thick * transpose(uvw) * uvw * density;
Me += weight * det(Jacob) * uvw.transpose() * uvw * density;
}
}
}
return Me;
}
matlab
function Vn = Make_Avg_Vn(Node,etable)
NN = size(Node,1);
EN = size(etable,1);
nNPE = size(etable,2);
vn = zeros(nNPE,3);
Vn = zeros(NN,3);
for i = 1:EN
enode = Node(etable(i,:),:);
idx1 = [2:nNPE 1];
idx2 = [nNPE 1:nNPE-1];
for j = 1:nNPE
v1 = enode(idx1(j),:) - enode(j,:);
v2 = enode(idx2(j),:) - enode(j,:);
vc = cross(v1,v2);
vn(j,:) = vc/norm(vc);
end
Vn(etable(i,:),:) = Vn(etable(i,:),:) + vn;
end
Vn = Vn./vecnorm(Vn,2,2);
end
function [V1,V2] = Director(Vn)
e2 = [0 1 0];
if (Equal_Vector(Vn,e2))||(Equal_Vector(Vn,-e2))
V1 = [0 0 1];
else % for the special case Vn = e2
V1 = cross(e2,Vn);
V1 = Normalize(V1);
end
V2 = cross(Vn,V1);
end
function equal = Equal_Vector(Va, Vb)
cond = abs(Va(1) - Vb(1)) < eps & ...
abs(Va(2) - Vb(2)) < eps & ...
abs(Va(3) - Vb(3)) < eps;
if cond
equal = true;
else
equal = false;
end
end
function v1 = Normalize(v)
v1 = v / sqrt((v(1)^2 + v(2)^2 + v(3)^2));
end
function C = Material_Law(Young, Poisson)
C = zeros(6,6);
k = 1;
C(1,1) = Young / (1 - Poisson^2);
C(2,2) = C(1,1);
C(1,2) = Poisson * C(1,1);
C(2,1) = C(1,2);
C(4,4) = Young / (1 + Poisson) /2;
C(5,5) = k * C(4,4);
C(6,6) = C(5,5);
end
function [r, s, t, w] = Gauss22_Point(i, j, k)
gauss_point2 = [-0.577350269189626, 0.577350269189626];
w = 1.0;
r = gauss_point2(i);
s = gauss_point2(j);
t = gauss_point2(k);
end
function dHrs = dHrs_Matrix(r,s)
dHrs = 0.25*[1+s, -(1+s), -(1-s), 1-s; 1+r, 1-r, -(1-r), -(1+r)];
end
function H = Shape_Function(r,s)
H = 0.25*[(1+r)*(1+s) (1-r)*(1+s) (1-r)*(1-s) (1+r)*(1-s)];
end
function [e1,e2,e3] = Local_Basis(Jacob)
g2 = Jacob(2,:);
g3 = Jacob(3,:);
e3 = g3;
e3 = Normalize(e3);
e1 = cross(g2,g3);
e1 = Normalize(e1);
e2 = cross(g3,e1);
e2 = Normalize(e2);
end
function [cg1,cg2,cg3] = Contravariant_Basis(Jacob)
g1 = Jacob(1,:);
g2 = Jacob(2,:);
g3 = Jacob(3,:);
detJ = det(Jacob);
cg1 = cross(g2,g3) / detJ;
cg2 = cross(g3,g1) / detJ;
cg3 = cross(g1,g2) / detJ;
end
function [g1,g2,g3] = Covariant_Basis(Jacob)
g1 = Jacob(1,:);
g2 = Jacob(2,:);
g3 = Jacob(3,:);
end
function Jacob = Jacobian(r,s,t,thick,node,Vn)
H = Shape_Function(r,s);
dHrs = dHrs_Matrix(r,s);
nDIM = 3;
Jacob = zeros(nDIM,nDIM);
for i = 1:nDIM
Jacob(1,i) = dot(dHrs(1,:),node(:,i)) + 0.5*t*dot(dHrs(1,:),thick*Vn(:,i));
Jacob(2,i) = dot(dHrs(2,:),node(:,i)) + 0.5*t*dot(dHrs(2,:),thick*Vn(:,i));
Jacob(3,i) = 0.5*dot(thick*H,Vn(:,i));
end
end
C++
#include <Eigen/Dense>
#include <vector>
#include <cmath>
using namespace Eigen;
using namespace std;
MatrixXd Make_Avg_Vn(const MatrixXd& Node, const MatrixXi& etable) {
int NN = Node.rows();
int EN = etable.rows();
int nNPE = etable.cols();
MatrixXd vn = MatrixXd::Zero(nNPE, 3);
MatrixXd Vn = MatrixXd::Zero(NN, 3);
for (int i = 0; i < EN; i++) {
MatrixXd enode = Node.row(etable(i, 0)).transpose();
for (int j = 1; j < nNPE; j++) {
enode.conservativeResize(3, enode.cols() + 1);
enode.col(j) = Node.row(etable(i, j)).transpose();
}
VectorXi idx1(nNPE);
VectorXi idx2(nNPE);
for (int j = 0; j < nNPE; j++) {
idx1(j) = (j + 1) % nNPE;
idx2(j) = (j - 1 + nNPE) % nNPE;
}
for (int j = 0; j < nNPE; j++) {
Vector3d v1 = enode.row(idx1(j)) - enode.row(j);
Vector3d v2 = enode.row(idx2(j)) - enode.row(j);
Vector3d vc = v1.cross(v2);
vn.row(j) = vc.normalized();
}
Vn.row(etable.row(i)) += vn;
}
for (int i = 0; i < NN; i++) {
Vn.row(i) = Vn.row(i).normalized();
}
return Vn;
}
void Director(const Vector3d& Vn, Vector3d& V1, Vector3d& V2) {
Vector3d e2(0, 1, 0);
if (Equal_Vector(Vn, e2) || Equal_Vector(Vn, -e2)) {
V1 = Vector3d(0, 0, 1);
} else {
V1 = Vn.cross(e2).normalized();
}
V2 = Vn.cross(V1);
}
bool Equal_Vector(const Vector3d& Va, const Vector3d& Vb) {
return (abs(Va(0) - Vb(0)) < numeric_limits<double>::epsilon() &&
abs(Va(1) - Vb(1)) < numeric_limits<double>::epsilon() &&
abs(Va(2) - Vb(2)) < numeric_limits<double>::epsilon());
}
Vector3d Normalize(const Vector3d& v) {
return v.normalized();
}
MatrixXd Material_Law(double Young, double Poisson) {
MatrixXd C = MatrixXd::Zero(6, 6);
C(0, 0) = Young / (1 - Poisson * Poisson);
C(1, 1) = C(0, 0);
C(0, 1) = Poisson * C(0, 0);
C(1, 0) = C(0, 1);
C(3, 3) = Young / (1 + Poisson) / 2;
C(4, 4) = C(3, 3);
C(5, 5) = C(4, 4);
return C;
}
void Gauss22_Point(int i, int j, int k, double& r, double& s, double& t, double& w) {
double gauss_point2[2] = {-0.577350269189626, 0.577350269189626};
w = 1.0;
r = gauss_point2[i];
s = gauss_point2[j];
t = gauss_point2[k];
}
MatrixXd dHrs_Matrix(double r, double s) {
return 0.25 * (MatrixXd(2, 4) << 1 + s, -(1 + s), -(1 - s), 1 - s,
1 + r, 1 - r, -(1 - r), -(1 + r)).finished();
}
MatrixXd Shape_Function(double r, double s) {
return 0.25 * (MatrixXd(1, 4) << (1 + r) * (1 + s), (1 - r) * (1 + s), (1 - r) * (1 - s), (1 + r) * (1 - s)).finished();
}
void Local_Basis(const MatrixXd& Jacob, Vector3d& e1, Vector3d& e2, Vector3d& e3) {
Vector3d g2 = Jacob.row(1);
Vector3d g3 = Jacob.row(2);
e3 = Normalize(g3);
e1 = Normalize(g2.cross(g3));
e2 = Normalize(g3.cross(e1));
}
void Contravariant_Basis(const MatrixXd& Jacob, Vector3d& cg1, Vector3d& cg2, Vector3d& cg3) {
Vector3d g1 = Jacob.row(0);
Vector3d g2 = Jacob.row(1);
Vector3d g3 = Jacob.row(2);
double detJ = Jacob.determinant();
cg1 = g2.cross(g3) / detJ;
cg2 = g3.cross(g1) / detJ;
cg3 = g1.cross(g2) / detJ;
}
void Covariant_Basis(const MatrixXd& Jacob, Vector3d& g1, Vector3d& g2, Vector3d& g3) {
g1 = Jacob.row(0);
g2 = Jacob.row(1);
g3 = Jacob.row(2);
}
MatrixXd Jacobian(double r, double s, double t, const MatrixXd& thick, const MatrixXd& node, const MatrixXd& Vn) {
MatrixXd H = Shape_Function(r, s);
MatrixXd dHrs = dHrs_Matrix(r, s);
int nDIM = 3;
MatrixXd Jacob = MatrixXd::Zero(nDIM, nDIM);
for (int i = 0; i < nDIM; i++) {
Jacob(0, i) = dHrs.row(0).dot(node.col(i)) + 0.5 * t * dHrs.row(0).dot(thick * Vn.col(i));
Jacob(1, i) = dHrs.row(1).dot(node.col(i)) + 0.5 * t * dHrs.row(1).dot(thick * Vn.col(i));
Jacob(2, i) = 0.5 * (thick * H).dot(Vn.col(i));
}
return Jacob;
}
matlab
function uvw = Displacement(r,s,t,thick,Vn)
V1 = zeros(nNPE,nDIM);
V2 = zeros(nNPE,nDIM);
for i = 1:nNPE
[V1(i,:), V2(i,:)] = Director(Vn(i,:));
end
H = Shape_Function(r,s);
uvw = zeros(nDIM, nNPE*nDPN);
for j = 1:nNPE
uvw(1,nNPE*0+j) = H(j);
uvw(2,nNPE*1+j) = H(j);
uvw(3,nNPE*2+j) = H(j);
end
for i = 1:nDIM
for j = 1:nNPE
uvw(i,nNPE*3+j) = -0.5*t*thick*H(j)*V2(j,i);
uvw(i,nNPE*4+j) = 0.5*t*thick*H(j)*V1(j,i);
end
end
end
function [duvwdr, duvwds, duvwdt] = Derivative_Displacement(r,s,t,thick,node,Vn)
V1 = zeros(nNPE,nDIM);
V2 = zeros(nNPE,nDIM);
for i = 1:nNPE
[V1(i,:), V2(i,:)] = Director(Vn(i,:));
end
H = Shape_Function(r,s);
dHrs = dHrs_Matrix(r,s);
duvwdr = zeros(nDIM, nNPE*nDPN);
duvwds = zeros(nDIM, nNPE*nDPN);
duvwdt = zeros(nDIM, nNPE*nDPN);
for j = 1:nNPE
duvwdr(1,nNPE*0+j) = dHrs(1,j);
duvwdr(2,nNPE*1+j) = dHrs(1,j);
duvwdr(3,nNPE*2+j) = dHrs(1,j);
duvwds(1,nNPE*0+j) = dHrs(2,j);
duvwds(2,nNPE*1+j) = dHrs(2,j);
duvwds(3,nNPE*2+j) = dHrs(2,j);
end
for i = 1:nDIM
for j = 1:nNPE
duvwdr(i,nNPE*3+j) = -0.5*t*thick*dHrs(1,j)*V2(j,i);
duvwdr(i,nNPE*4+j) = 0.5*t*thick*dHrs(1,j)*V1(j,i);
duvwds(i,nNPE*3+j) = -0.5*t*thick*dHrs(2,j)*V2(j,i);
duvwds(i,nNPE*4+j) = 0.5*t*thick*dHrs(2,j)*V1(j,i);
duvwdt(i,nNPE*3+j) = -0.5*thick*H(j)*V2(j,i);
duvwdt(i,nNPE*4+j) = 0.5*thick*H(j)*V1(j,i);
end
end
end
function cov_B = Covariant_Strain(r,s,t,thick,node,Vn)
[duvwdr, duvwds, duvwdt] = Derivative_Displacement(r,s,t,thick,node,Vn);
Jacob = Jacobian(r, s, t, thick, node, Vn);
[g1, g2, g3] = Covariant_Basis(Jacob);
cov_B(1,:) = 0.5* (...
g1(1)*duvwdr(1,:) + g1(2)*duvwdr(2,:) + g1(3)*duvwdr(3,:) + ...
g1(1)*duvwdr(1,:) + g1(2)*duvwdr(2,:) + g1(3)*duvwdr(3,:) );
cov_B(2,:) = 0.5* (...
g2(1)*duvwds(1,:) + g2(2)*duvwds(2,:) + g2(3)*duvwds(3,:) + ...
g2(1)*duvwds(1,:) + g2(2)*duvwds(2,:) + g2(3)*duvwds(3,:) );
cov_B(3,:) = 0;
cov_B(4,:) = 0.5* (...
g2(1)*duvwdr(1,:) + g2(2)*duvwdr(2,:) + g2(3)*duvwdr(3,:) + ...
g1(1)*duvwds(1,:) + g1(2)*duvwds(2,:) + g1(3)*duvwds(3,:) );
cov_B(5,:) = 0.5* (...
g3(1)*duvwds(1,:) + g3(2)*duvwds(2,:) + g3(3)*duvwds(3,:) + ...
g2(1)*duvwdt(1,:) + g2(2)*duvwdt(2,:) + g2(3)*duvwdt(3,:) );
cov_B(6,:) = 0.5* (...
g1(1)*duvwdt(1,:) + g1(2)*duvwdt(2,:) + g1(3)*duvwdt(3,:) + ...
g3(1)*duvwdr(1,:) + g3(2)*duvwdr(2,:) + g3(3)*duvwdr(3,:) );
end
function ass_B = Assumed_Covariant_Strain(r, s, t, thick, node, Vn)
cov_B_st1 = Covariant_Strain( 1.0, 0.0, t, thick, node, Vn);
cov_B_st2 = Covariant_Strain(-1.0, 0.0, t, thick, node, Vn);
cov_B_tr1 = Covariant_Strain( 0.0, 1.0, t, thick, node, Vn);
cov_B_tr2 = Covariant_Strain( 0.0,-1.0, t, thick, node, Vn);
ass_B = Covariant_Strain(r, s, t, thick, node, Vn);
ass_B(3,:) = 0.0;
ass_B(5,:) = 0.5 * (1.0+r) * cov_B_st1(5,:) + 0.5*(1.0 - r) * cov_B_st2(5,:);
ass_B(6,:) = 0.5 * (1.0+s) * cov_B_tr1(6,:) + 0.5*(1.0 - s) * cov_B_tr2(6,:);
end
C++
#include <Eigen/Dense>
#include <vector>
using namespace Eigen;
using namespace std;
MatrixXd Displacement(double r, double s, double t, double thick, const MatrixXd& Vn) {
MatrixXd V1 = MatrixXd::Zero(nNPE, nDIM);
MatrixXd V2 = MatrixXd::Zero(nNPE, nDIM);
for (int i = 0; i < nNPE; i++) {
tie(V1.row(i), V2.row(i)) = Director(Vn.row(i));
}
VectorXd H = Shape_Function(r, s);
MatrixXd uvw = MatrixXd::Zero(nDIM, nNPE * nDPN);
for (int j = 0; j < nNPE; j++) {
uvw(0, nNPE * 0 + j) = H(j);
uvw(1, nNPE * 1 + j) = H(j);
uvw(2, nNPE * 2 + j) = H(j);
}
for (int i = 0; i < nDIM; i++) {
for (int j = 0; j < nNPE; j++) {
uvw(i, nNPE * 3 + j) = -0.5 * t * thick * H(j) * V2(j, i);
uvw(i, nNPE * 4 + j) = 0.5 * t * thick * H(j) * V1(j, i);
}
}
return uvw;
}
void Derivative_Displacement(double r, double s, double t, double thick, const MatrixXd& node, const MatrixXd& Vn,
MatrixXd& duvwdr, MatrixXd& duvwds, MatrixXd& duvwdt) {
MatrixXd V1 = MatrixXd::Zero(nNPE, nDIM);
MatrixXd V2 = MatrixXd::Zero(nNPE, nDIM);
for (int i = 0; i < nNPE; i++) {
tie(V1.row(i), V2.row(i)) = Director(Vn.row(i));
}
VectorXd H = Shape_Function(r, s);
MatrixXd dHrs = dHrs_Matrix(r, s);
duvwdr = MatrixXd::Zero(nDIM, nNPE * nDPN);
duvwds = MatrixXd::Zero(nDIM, nNPE * nDPN);
duvwdt = MatrixXd::Zero(nDIM, nNPE * nDPN);
for (int j = 0; j < nNPE; j++) {
duvwdr(0, nNPE * 0 + j) = dHrs(0, j);
duvwdr(1, nNPE * 1 + j) = dHrs(0, j);
duvwdr(2, nNPE * 2 + j) = dHrs(0, j);
duvwds(0, nNPE * 0 + j) = dHrs(1, j);
duvwds(1, nNPE * 1 + j) = dHrs(1, j);
duvwds(2, nNPE * 2 + j) = dHrs(1, j);
}
for (int i = 0; i < nDIM; i++) {
for (int j = 0; j < nNPE; j++) {
duvwdr(i, nNPE * 3 + j) = -0.5 * t * thick * dHrs(0, j) * V2(j, i);
duvwdr(i, nNPE * 4 + j) = 0.5 * t * thick * dHrs(0, j) * V1(j, i);
duvwds(i, nNPE * 3 + j) = -0.5 * t * thick * dHrs(1, j) * V2(j, i);
duvwds(i, nNPE * 4 + j) = 0.5 * t * thick * dHrs(1, j) * V1(j, i);
duvwdt(i, nNPE * 3 + j) = -0.5 * thick * H(j) * V2(j, i);
duvwdt(i, nNPE * 4 + j) = 0.5 * thick * H(j) * V1(j, i);
}
}
}
MatrixXd Covariant_Strain(double r, double s, double t, double thick, const MatrixXd& node, const MatrixXd& Vn) {
MatrixXd duvwdr, duvwds, duvwdt;
Derivative_Displacement(r, s, t, thick, node, Vn, duvwdr, duvwds, duvwdt);
MatrixXd Jacob = Jacobian(r, s, t, thick, node, Vn);
VectorXd g1, g2, g3;
tie(g1, g2, g3) = Covariant_Basis(Jacob);
MatrixXd cov_B = MatrixXd::Zero(6, duvwdr.cols());
cov_B.row(0) = 0.5 * (g1(0) * duvwdr.row(0) + g1(1) * duvwdr.row(1) + g1(2) * duvwdr.row(2));
cov_B.row(1) = 0.5 * (g2(0) * duvwds.row(0) + g2(1) * duvwds.row(1) + g2(2) * duvwds.row(2));
cov_B.row(2) = VectorXd::Zero(duvwdr.cols());
cov_B.row(3) = 0.5 * (g2(0) * duvwdr.row(0) + g2(1) * duvwdr.row(1) + g2(2) * duvwdr.row(2) +
g1(0) * duvwds.row(0) + g1(1) * duvwds.row(1) + g1(2) * duvwds.row(2));
cov_B.row(4) = 0.5 * (g3(0) * duvwds.row(0) + g3(1) * duvwds.row(1) + g3(2) * duvwds.row(2) +
g2(0) * duvwdt.row(0) + g2(1) * duvwdt.row(1) + g2(2) * duvwdt.row(2));
cov_B.row(5) = 0.5 * (g1(0) * duvwdt.row(0) + g1(1) * duvwdt.row(1) + g1(2) * duvwdt.row(2) +
g3(0) * duvwdr.row(0) + g3(1) * duvwdr.row(1) + g3(2) * duvwdr.row(2));
return cov_B;
}
MatrixXd Assumed_Covariant_Strain(double r, double s, double t, double thick, const MatrixXd& node, const MatrixXd& Vn) {
MatrixXd cov_B_st1 = Covariant_Strain(1.0, 0.0, t, thick, node, Vn);
MatrixXd cov_B_st2 = Covariant_Strain(-1.0, 0.0, t, thick, node, Vn);
MatrixXd cov_B_tr1 = Covariant_Strain(0.0, 1.0, t, thick, node, Vn);
MatrixXd cov_B_tr2 = Covariant_Strain(0.0, -1.0, t, thick, node, Vn);
MatrixXd ass_B = Covariant_Strain(r, s, t, thick, node, Vn);
ass_B.row(2) = VectorXd::Zero(ass_B.cols());
ass_B.row(4) = 0.5 * (1.0 + r) * cov_B_st1.row(4) + 0.5 * (1.0 - r) * cov_B_st2.row(4);
ass_B.row(5) = 0.5 * (1.0 + s) * cov_B_tr1.row(5) + 0.5 * (1.0 - s) * cov_B_tr2.row(5);
return ass_B;
}
matlab
function B = Strain_Displacement(r, s, t, thick, node, Vn)
MITC = true;
if MITC == true
cov_B = Assumed_Covariant_Strain(r, s, t, thick, node, Vn);
else
cov_B = Covariant_Strain(r, s, t, thick, node, Vn);
end
Jacob = Jacobian(r, s, t, thick, node, Vn);
[cg1, cg2, cg3] = Contravariant_Basis(Jacob);
[e1, e2, e3] = Local_Basis(Jacob);
B(1,:) = ...
(dot(cg1,e1)*dot(cg1,e1)) * cov_B(1,:) + ...
(dot(cg2,e1)*dot(cg2,e1)) * cov_B(2,:) + ...
(dot(cg1,e1)*dot(cg2,e1) + dot(cg2,e1)*dot(cg1,e1)) * cov_B(4,:) + ...
(dot(cg2,e1)*dot(cg3,e1) + dot(cg3,e1)*dot(cg2,e1)) * cov_B(5,:) + ...
(dot(cg3,e1)*dot(cg1,e1) + dot(cg1,e1)*dot(cg3,e1)) * cov_B(6,:);
B(2,:) = ...
(dot(cg1,e2)*dot(cg1,e2)) * cov_B(1,:) + ...
(dot(cg2,e2)*dot(cg2,e2)) * cov_B(2,:) + ...
(dot(cg1,e2)*dot(cg2,e2) + dot(cg2,e2)*dot(cg1,e2)) * cov_B(4,:) + ...
(dot(cg2,e2)*dot(cg3,e2) + dot(cg3,e2)*dot(cg2,e2)) * cov_B(5,:) + ...
(dot(cg3,e2)*dot(cg1,e2) + dot(cg1,e2)*dot(cg3,e2)) * cov_B(6,:);
B(3,:) = 0.0;
B(4,:) = ...
(dot(cg1,e1)*dot(cg1,e2)) * cov_B(1,:) + ...
(dot(cg2,e1)*dot(cg2,e2)) * cov_B(2,:) + ...
(dot(cg1,e1)*dot(cg2,e2) + dot(cg2,e1)*dot(cg1,e2)) * cov_B(4,:) + ...
(dot(cg2,e1)*dot(cg3,e2) + dot(cg3,e1)*dot(cg2,e2)) * cov_B(5,:) + ...
(dot(cg3,e1)*dot(cg1,e2) + dot(cg1,e1)*dot(cg3,e2)) * cov_B(6,:);
B(5,:) = ...
(dot(cg1,e2)*dot(cg1,e3)) * cov_B(1,:) + ...
(dot(cg2,e2)*dot(cg2,e3)) * cov_B(2,:) + ...
(dot(cg1,e2)*dot(cg2,e3) + dot(cg2,e2)*dot(cg1,e3)) * cov_B(4,:) + ...
(dot(cg2,e2)*dot(cg3,e3) + dot(cg3,e2)*dot(cg2,e3)) * cov_B(5,:) + ...
(dot(cg3,e2)*dot(cg1,e3) + dot(cg1,e2)*dot(cg3,e3)) * cov_B(6,:);
B(6,:) = ...
(dot(cg1,e3)*dot(cg1,e1)) * cov_B(1,:) + ...
(dot(cg2,e3)*dot(cg2,e1)) * cov_B(2,:) + ...
(dot(cg1,e3)*dot(cg2,e1) + dot(cg2,e3)*dot(cg1,e1)) * cov_B(4,:) + ...
(dot(cg2,e3)*dot(cg3,e1) + dot(cg3,e3)*dot(cg2,e1)) * cov_B(5,:) + ...
(dot(cg3,e3)*dot(cg1,e1) + dot(cg1,e3)*dot(cg3,e1)) * cov_B(6,:);
B(4:6,:) = 2.0 * B(4:6,:);
end
function Ke = Shell_Stiffness(Young, Poisson, thick, node, Vn)
C = Material_Law(Young, Poisson);
Ke = zeros(nNPE*nDPN, nNPE*nDPN);
for i = 1:nGP
for j = 1:nGP
for k = 1:2
[r, s, t, weight] = Gauss22_Point(i, j, k);
B = Strain_Displacement(r, s, t, thick, node, Vn);
Jacob = Jacobian(r, s, t, thick, node, Vn);
Ke = Ke + weight * abs(det(Jacob)) * B'*C*B;
end
end
end
end
C++
#include <Eigen/Dense>
#include <vector>
using namespace Eigen;
using namespace std;
MatrixXd Strain_Displacement(double r, double s, double t, double thick, const MatrixXd& node, const VectorXd& Vn) {
bool MITC = true;
MatrixXd cov_B;
if (MITC) {
cov_B = Assumed_Covariant_Strain(r, s, t, thick, node, Vn);
} else {
cov_B = Covariant_Strain(r, s, t, thick, node, Vn);
}
MatrixXd Jacob = Jacobian(r, s, t, thick, node, Vn);
VectorXd cg1, cg2, cg3;
tie(cg1, cg2, cg3) = Contravariant_Basis(Jacob);
VectorXd e1, e2, e3;
tie(e1, e2, e3) = Local_Basis(Jacob);
MatrixXd B(6, cov_B.cols());
B.row(0) = (cg1.dot(e1) * cg1.dot(e1)) * cov_B.row(0) +
(cg2.dot(e1) * cg2.dot(e1)) * cov_B.row(1) +
(cg1.dot(e1) * cg2.dot(e1) + cg2.dot(e1) * cg1.dot(e1)) * cov_B.row(3) +
(cg2.dot(e1) * cg3.dot(e1) + cg3.dot(e1) * cg2.dot(e1)) * cov_B.row(4) +
(cg3.dot(e1) * cg1.dot(e1) + cg1.dot(e1) * cg3.dot(e1)) * cov_B.row(5);
B.row(1) = (cg1.dot(e2) * cg1.dot(e2)) * cov_B.row(0) +
(cg2.dot(e2) * cg2.dot(e2)) * cov_B.row(1) +
(cg1.dot(e2) * cg2.dot(e2) + cg2.dot(e2) * cg1.dot(e2)) * cov_B.row(3) +
(cg2.dot(e2) * cg3.dot(e2) + cg3.dot(e2) * cg2.dot(e2)) * cov_B.row(4) +
(cg3.dot(e2) * cg1.dot(e2) + cg1.dot(e2) * cg3.dot(e2)) * cov_B.row(5);
B.row(2) = VectorXd::Zero(cov_B.cols());
B.row(3) = (cg1.dot(e1) * cg1.dot(e2)) * cov_B.row(0) +
(cg2.dot(e1) * cg2.dot(e2)) * cov_B.row(1) +
(cg1.dot(e1) * cg2.dot(e2) + cg2.dot(e1) * cg1.dot(e2)) * cov_B.row(3) +
(cg2.dot(e1) * cg3.dot(e2) + cg3.dot(e1) * cg2.dot(e2)) * cov_B.row(4) +
(cg3.dot(e1) * cg1.dot(e2) + cg1.dot(e1) * cg3.dot(e2)) * cov_B.row(5);
B.row(4) = (cg1.dot(e2) * cg1.dot(e3)) * cov_B.row(0) +
(cg2.dot(e2) * cg2.dot(e3)) * cov_B.row(1) +
(cg1.dot(e2) * cg2.dot(e3) + cg2.dot(e2) * cg1.dot(e3)) * cov_B.row(3) +
(cg2.dot(e2) * cg3.dot(e3) + cg3.dot(e2) * cg2.dot(e3)) * cov_B.row(4) +
(cg3.dot(e2) * cg1.dot(e3) + cg1.dot(e2) * cg3.dot(e3)) * cov_B.row(5);
B.row(5) = (cg1.dot(e3) * cg1.dot(e1)) * cov_B.row(0) +
(cg2.dot(e3) * cg2.dot(e1)) * cov_B.row(1) +
(cg1.dot(e3) * cg2.dot(e1) + cg2.dot(e3) * cg1.dot(e1)) * cov_B.row(3) +
(cg2.dot(e3) * cg3.dot(e1) + cg3.dot(e3) * cg2.dot(e1)) * cov_B.row(4) +
(cg3.dot(e3) * cg1.dot(e1) + cg1.dot(e3) * cg3.dot(e1)) * cov_B.row(5);
B.row(3) *= 2.0;
B.row(4) *= 2.0;
B.row(5) *= 2.0;
return B;
}
MatrixXd Shell_Stiffness(double Young, double Poisson, double thick, const MatrixXd& node, const VectorXd& Vn) {
MatrixXd C = Material_Law(Young, Poisson);
int nNPE = node.rows(); // Assuming nNPE is the number of nodes
int nDPN = node.cols(); // Assuming nDPN is the number of degrees of freedom per node
MatrixXd Ke = MatrixXd::Zero(nNPE * nDPN, nNPE * nDPN);
for (int i = 0; i < nGP; ++i) {
for (int j = 0; j < nGP; ++j) {
for (int k = 0; k < 2; ++k) {
double r, s, t, weight;
tie(r, s, t, weight) = Gauss22_Point(i, j, k);
MatrixXd B = Strain_Displacement(r, s, t, thick, node, Vn);
MatrixXd Jacob = Jacobian(r, s, t, thick, node, Vn);
Ke += weight * abs(Jacob.determinant()) * B.transpose() * C * B;
}
}
}
return Ke;
}
matlab
Nel = size(etable_4,1); % shell 수
K = sparse(tdof,tdof);
for i = 1:Nel
if nnz(etable_4(i,:))==4 % 사각형 이니까
nNPE = 4;
else
nNPE = 3;
end
eNode = node(etable_4(i,1:nNPE),:); % 노드 위치 정보 뽑기
eVn = Vn(etable_4(i,1:nNPE),:); % 노말 벡터
if nNPE==4
[ke, me] = MITC4(E, nu, thk, rho, eNode, eVn);
else
[ke, me] = MITC3(E, nu, thk, rho, eNode, eVn);
end
edofm = zeros(Ndof,nNPE);
for j = 1:Ndof % 5 자유도까지 있다.
edofm(j,1:nNPE) = Ndof*(etable_4(i,1:nNPE)-1)+j;
end
edof = edofm(:); % 위치 지정해주기
K(edof,edof) = K(edof,edof) + ke;% k 메트릭스 채우기
end
c++
#include <Eigen/Sparse>
#include <Eigen/Dense>
#include <vector>
using namespace Eigen;
using namespace std;
int Nel = etable_4.rows(); // shell 수
SparseMatrix<double> K(tdof, tdof);
for (int i = 0; i < Nel; i++) {
int nNPE;
if (etable_4.row(i).nonZeros() == 4) { // 사각형 이니까
nNPE = 4;
} else {
nNPE = 3;
}
MatrixXd eNode = node.row(etable_4(i, Eigen::seq(0, nNPE - 1))); // 노드 위치 정보 뽑기
MatrixXd eVn = Vn.row(etable_4(i, Eigen::seq(0, nNPE - 1))); // 노말 벡터
MatrixXd ke, me;
if (nNPE == 4) {
tie(ke, me) = MITC4(E, nu, thk, rho, eNode, eVn);
} else {
tie(ke, me) = MITC3(E, nu, thk, rho, eNode, eVn);
}
MatrixXd edofm = MatrixXd::Zero(Ndof, nNPE);
for (int j = 0; j < Ndof; j++) { // 5 자유도까지 있다.
edofm.row(j).head(nNPE) = Ndof * (etable_4.row(i).head(nNPE).array() - 1) + j + 1;
}
VectorXd edof = Map<VectorXd>(edofm.data(), Ndof * nNPE); // 위치 지정해주기
K.coeffRef(edof.cast<int>(), edof.cast<int>()) += ke; // k 메트릭스 채우기
}
MATLAB
clear, clc, close all
filename = 'ex1_MITC4_static.hob';
[MATERIAL, PSHELL, NODE, ~, SHELL_4, BC, FORCE] = InputReader(filename);
node = NODE(:, 2:4);
etable_4 = SHELL_4(:, 3:6);
fnode = FORCE(:,1);
fix_node = BC(:,1);
E = MATERIAL(1, 2);
nu = MATERIAL(1, 3);
rho = MATERIAL(1, 4);
thk = PSHELL(1, 3);
Nn = size(node,1);
Ndof = 5;
tdof = Ndof*Nn;
Vn = Make_Avg_Vn(node,etable_4);
load ex1_static.csv
load ex1_static_etable.txt
Nn_abaqus = size(ex1_static,1);
node_abaqus = ex1_static(:,2:4);
mag_abaqus = ex1_static(:,5);
sol_abaqus = ex1_static(:,6:8);
etable_abaqus = ex1_static_etable(:,2:5);
Nel = size(etable_4,1);
K = sparse(tdof,tdof);
for i = 1:Nel
if nnz(etable_4(i,:))==4
nNPE = 4;
else
nNPE = 3;
end
eNode = node(etable_4(i,1:nNPE),:);
eVn = Vn(etable_4(i,1:nNPE),:);
if nNPE==4
[ke, me] = MITC4(E, nu, thk, rho, eNode, eVn);
else
[ke, me] = MITC3(E, nu, thk, rho, eNode, eVn);
end
edofm = zeros(Ndof,nNPE);
for j = 1:Ndof
edofm(j,1:nNPE) = Ndof*(etable_4(i,1:nNPE)-1)+j;
end
edof = edofm(:);
K(edof,edof) = K(edof,edof) + ke;
end
%% Apply Boundary Condition
F = zeros(tdof,1);
F(Ndof*(fnode-1)+FORCE(1,2)) = FORCE(1,3);
d = zeros(tdof,1);
fix_node = fix_node';
fix_dofm = zeros(Ndof,length(fix_node));
for i = 1:Ndof
fix_dofm(i,:) = Ndof*(fix_node-1)+i;
end
fix_dof = fix_dofm(:);
%% Static Analysis
dof_a = fix_dof;
dof_b = setdiff(1:tdof,dof_a);
Ub = K(dof_b,dof_b)\F(dof_b);
d(dof_b) = Ub;
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function Vn = Make_Avg_Vn(Node,etable)
NN = size(Node,1); % NN: 전체 노드 수
EN = size(etable,1); % EN: 전체 요소 수
nNPE = size(etable,2); % nNPE: 각 요소의 노드 수
vn = zeros(nNPE,3); % 각 요소의 법선 벡터를 저장하는 행렬 (nNPE x 3 크기)
Vn = zeros(NN,3); % 각 노드의 평균 법선 벡터를 저장하는 행렬 (NN x 3 크기)
for i = 1:EN
enode = Node(etable(i,:),:); % 요소의 노드 좌표 가져오기
idx1 = [2:nNPE 1]; % 순환 인덱스를 위한 인덱스 배열 1
idx2 = [nNPE 1:nNPE-1]; % 순환 인덱스를 위한 인덱스 배열 2
for j = 1:nNPE
v1 = enode(idx1(j),:) - enode(j,:); % 첫 번째 벡터
v2 = enode(idx2(j),:) - enode(j,:); % 두 번째 벡터
vc = cross(v1,v2); % 두 벡터의 외적(법선 벡터)
vn(j,:) = vc/norm(vc); % 법선 벡터를 단위 벡터로 정규화하여 저장
end
Vn(etable(i,:),:) = Vn(etable(i,:),:) + vn; % 노드의 법선 벡터에 요소의 법선 벡터를 누적
end
Vn = Vn./vecnorm(Vn,2,2); % 모든 법선 벡터를 정규화
end
C++
#include <iostream>
#include <Eigen/Dense>
#include <Eigen/Sparse>
#include <vector>
#include <fstream>
#include <sstream>
using namespace Eigen;
using namespace std;
void Make_Avg_Vn(const MatrixXd& Node, const MatrixXi& etable, MatrixXd& Vn) {
int NN = Node.rows(); // NN: total number of nodes
int EN = etable.rows(); // EN: total number of elements
int nNPE = etable.cols(); // nNPE: number of nodes per element
MatrixXd vn(nNPE, 3); // matrix to store normal vectors for each element (nNPE x 3)
for (int i = 0; i < EN; ++i) {
MatrixXd enode = Node.row(etable.row(i).cast<int>()).eval(); // get node coordinates of the element
VectorXi idx1(nNPE);
VectorXi idx2(nNPE);
for (int j = 0; j < nNPE; ++j) {
idx1(j) = (j + 1) % nNPE; // cyclic index array 1
idx2(j) = (j == 0) ? nNPE - 1 : j - 1; // cyclic index array 2
}
for (int j = 0; j < nNPE; ++j) {
Vector3d v1 = enode.row(idx1(j)) - enode.row(j); // first vector
Vector3d v2 = enode.row(idx2(j)) - enode.row(j); // second vector
Vector3d vc = v1.cross(v2); // cross product (normal vector)
vn.row(j) = vc.normalized(); // normalize and store the normal vector
}
Vn.row(etable.row(i).cast<int>()) += vn; // accumulate the normal vectors to the nodes
}
Vn = Vn.array().rowwise().normalized(); // normalize all normal vectors
}
int main() {
string filename = "ex1_MITC4_static.hob";
// Assuming InputReader is defined elsewhere and returns the required matrices
MatrixXd MATERIAL, PSHELL, NODE, SHELL_4, BC, FORCE;
tie(MATERIAL, PSHELL, NODE, ignore, SHELL_4, BC, FORCE) = InputReader(filename);
MatrixXd node = NODE.rightCols(3);
MatrixXd etable_4 = SHELL_4.rightCols(4);
VectorXi fnode = FORCE.col(0).cast<int>();
VectorXi fix_node = BC.col(0).cast<int>();
double E = MATERIAL(0, 1);
double nu = MATERIAL(0, 2);
double rho = MATERIAL(0, 3);
double thk = PSHELL(0, 2);
int Nn = node.rows();
int Ndof = 5;
int tdof = Ndof * Nn;
MatrixXd Vn(Nn, 3);
Make_Avg_Vn(node, etable_4.cast<int>(), Vn);
MatrixXd ex1_static;
MatrixXd ex1_static_etable;
// Load ex1_static.csv and ex1_static_etable.txt into ex1_static and ex1_static_etable respectively
int Nn_abaqus = ex1_static.rows();
MatrixXd node_abaqus = ex1_static.rightCols(3);
VectorXd mag_abaqus = ex1_static.col(4);
MatrixXd sol_abaqus = ex1_static.rightCols(3);
MatrixXd etable_abaqus = ex1_static_etable.rightCols(4);
int Nel = etable_4.rows();
SparseMatrix<double> K(tdof, tdof);
for (int i = 0; i < Nel; ++i) {
int nNPE = (etable_4.row(i).array() != 0).count() == 4 ? 4 : 3;
MatrixXd eNode = node.row(etable_4.row(i).head(nNPE).cast<int>());
MatrixXd eVn = Vn.row(etable_4.row(i).head(nNPE).cast<int>());
MatrixXd ke, me;
if (nNPE == 4) {
tie(ke, me) = MITC4(E, nu, thk, rho, eNode, eVn);
} else {
tie(ke, me) = MITC3(E, nu, thk, rho, eNode, eVn);
}
MatrixXi edofm(Ndof, nNPE);
for (int j = 0; j < Ndof; ++j) {
edofm.row(j) = Ndof * (etable_4.row(i).head(nNPE).cast<int>() - 1).transpose() + j + 1;
}
VectorXi edof = Map<VectorXi>(edofm.data(), Ndof * nNPE);
for (int j = 0; j < edof.size(); ++j) {
K.coeffRef(edof(j), edof(j)) += ke(j);
}
}
// Apply Boundary Condition
VectorXd F = VectorXd::Zero(tdof);
F(Ndof * (fnode - 1).array() + FORCE(0, 1)) = FORCE(0, 2);
VectorXd d = VectorXd::Zero(tdof);
fix_node = fix_node.transpose();
MatrixXi fix_dofm(Ndof, fix_node.size());
for (int i = 0; i < Ndof; ++i) {
fix_dofm.row(i) = Ndof * (fix_node - 1).array() + i + 1;
}
VectorXi fix_dof = Map<VectorXi>(fix_dofm.data(), Ndof * fix_node.size());
// Static Analysis
VectorXi dof_a = fix_dof;
VectorXi dof_b = VectorXi::LinSpaced(tdof, 1, tdof).array().set_difference(dof_a).eval();
VectorXd Ub = K.block(dof_b, dof_b).ldlt().solve(F(dof_b));
d(dof_b) = Ub;
return 0;
}
Pshell=
1 1 0.001
Shetable=
0 1 1 41 42 21 20
1 2 1 62 63 42 41
398 399 1 400 401 380 379
399 400 1 421 422 401 400
bbns=
Materials=
1 2e+11 0.3 7850
shellcoord[0]=
-0.45 -0.45 -0.5 -0.5
0.45 0.5 0.5 0.45
0 0 0 0
shellcoord[399]=
0.5 0.5 0.45 0.45
-0.5 -0.45 -0.45 -0.5
0 0 0 0
scoord=
1 -0.5 -0.5 0
2 -0.5 -0.45 0
3 -0.5 -0.4 0
4 -0.5 -0.35 0
5 -0.5 -0.3 0
6 -0.5 -0.25 0
7 -0.5 -0.2 0
8 -0.5 -0.15 0
432 0.5 0.05 0
433 0.5 0.1 0
434 0.5 0.15 0
435 0.5 0.2 0
436 0.5 0.25 0
437 0.5 0.3 0
438 0.5 0.35 0
439 0.5 0.4 0
440 0.5 0.45 0
441 0.5 0.5 0
'FEM > HOB' 카테고리의 다른 글
Shell Beam Joint - HOB CAE (0) | 2024.10.16 |
---|---|
이 전에 진행한 Shell Beam Joint Concept (0) | 2024.10.12 |
Higher Order Beam CAE 제작 기록용 (0) | 2024.10.09 |
HOB - MITC4 Shell 계산 과정 (0) | 2024.08.14 |