Difference between revisions of "Point contact"

From Medusa: Coordinate Free Mehless Method implementation
Jump to: navigation, search
(Results)
(Numerical solution with FreeFem++)
 
(46 intermediate revisions by 5 users not shown)
Line 3: Line 3:
 
=Point contact on a 2D half-plane=
 
=Point contact on a 2D half-plane=
  
A starting point to solve problems in contact mechanics is to understand the effect of a point-load applied to a homogeneous, linear elastic, isotropic half-plane. This problem may be defined either as plane stress or plain strain (for solution with FreeFem++ we have choosen the latter). The traction boundary conditions for this problem are:
+
A starting point to solve problems in contact mechanics is to understand the effect of a point-load applied to a homogeneous, linear elastic, isotropic half-plane. This problem may be defined either as plane stress or plain strain (for solution with FreeFem++ we have chosen the latter). The traction boundary conditions for this problem are:
 
\begin{equation}\label{eq:bc}
 
\begin{equation}\label{eq:bc}
 
\sigma_{xy}(x,0) = 0, \quad \sigma_{yy}(x,y) = -P\delta(x,y)
 
\sigma_{xy}(x,0) = 0, \quad \sigma_{yy}(x,y) = -P\delta(x,y)
Line 9: Line 9:
 
where $\delta(x,y)$ is the Dirac delta function. Together these boundary conditions state that there is a singular normal force $P$ applied at $(x,y) = (0,0)$ and there are no shear stresses on the surface of the elastic half-plane.
 
where $\delta(x,y)$ is the Dirac delta function. Together these boundary conditions state that there is a singular normal force $P$ applied at $(x,y) = (0,0)$ and there are no shear stresses on the surface of the elastic half-plane.
  
The analytical relations for the stresses can be found from the [https://en.wikipedia.org/wiki/Flamant_solution Flamant solution] (stress distributions in a linear elastic wedge loaded by point forces a the tip. When the "wedge" is flat we get a half-plane. The derivation uses polar coordinates.) and are given as:
+
== Closed form solution ==
 +
 
 +
The analytical relations for the stresses can be found from the [https://en.wikipedia.org/wiki/Flamant_solution Flamant solution] (stress distributions in a linear elastic wedge loaded by point forces a the tip. When the "wedge" is flat we get a half-plane). For a point $(x, y)$ denote its square distance to the origin by $r^2 = x^2 + y^2$, and the stresses in point $(x, y)$ are given by:
 
\begin{equation}
 
\begin{equation}
\sigma_{xx} = -\frac{2P}{\pi} \frac{x^2y}{\left(x^2+y^2\right)^2},
+
\sigma_{xx} = -\frac{2P}{\pi} \frac{x^2y}{r^4},
 
\end{equation}
 
\end{equation}
 
\begin{equation}
 
\begin{equation}
\sigma_{yy} = -\frac{2P}{\pi} \frac{y^3}{\left(x^2+y^2\right)^2},
+
\sigma_{yy} = -\frac{2P}{\pi} \frac{y^3}{r^4},
 
\end{equation}
 
\end{equation}
 
\begin{equation}
 
\begin{equation}
\sigma_{xy} = -\frac{2P}{\pi} \frac{xy^2}{\left(x^2+y^2\right)^2},
+
\sigma_{xy} = -\frac{2P}{\pi} \frac{xy^2}{r^4},
 
\end{equation}
 
\end{equation}
for some point $(x,y)$ in the half-plane. From this stress field the strain components and thus the displacements $(u_x,u_y)$ can be determined. The displacements are given by
+
for some point $(x,y)$ in the lower half-plane. From this stress field the strain components and thus the displacements $(u_x,u_y)$ can be determined. The displacements are given by
 
\begin{align}
 
\begin{align}
u_x &= -\frac{P}{4\pi\mu}\left((\kappa-1)\theta - \frac{2xy}{r^2}\right), \label{eq:dispx}\\
+
u_x &= -\frac{P}{4\pi\mu}\left(\frac{2xy}{r^2}+\frac{2\mu}{\lambda + \mu} \arctan(y/x)\right), \label{eq:dispx}\\
u_y &= -\frac{P}{4\pi\mu}\left((\kappa+1)\log r + \frac{2x^2}{r^2}\right), \label{eq:dispy}
+
u_y &= -\frac{P}{4\pi\mu}\left(\frac{y^2-x^2}{r^2}-\frac{\lambda+2\mu}{\lambda + \mu} \log(r^2)\right), \label{eq:dispy}
 
\end{align}
 
\end{align}
where $$r = \sqrt{x^2+y^2}$$ and $$\tan \theta = \frac{x}{y}.$$ The symbol $\kappa$ is known as Dundars constant and is defined as
+
where again $r^2 = x^2+y^2$ and $\lambda$ and $\mu$ are Lame parameters. For plane stress, do the substitution
\[
+
$\lambda \gets \frac{2\mu\lambda}{\lambda+2\mu}$ in the solution and in the Navier equation.
\kappa = \begin{cases} 3 - 4\nu & \quad \text{(Plane strain)}, \\
+
 
                      \cfrac{3 - \nu}{1+\nu} & \quad \text{(Plane stress)}. \end{cases}
+
The proof of correctness and plots of the solution are available in this notebook: [[:File:point_contact.nb]]
\]
+
 
The last remaining symbol is $\mu$ which represents the shear modulus (sometimes also denoted with $G$).
+
Copyable C++ code for closed form solution:
 +
<syntaxhighlight lang="c++" line>
 +
std::function<Vec2d(Vec2d)> analytical = [=](const Vec2d& p) {
 +
    double x = p[0], y = p[1], r2 = x*x + y*y, factor = -force / (4.0*M_PI*mu);
 +
    double ux = 2*x*y/r2 + 2*mu/(lam + mu)*std::atan2(y, x);
 +
    double uy = (y*y-x*x)/r2 - (lam + 2*mu)/(lam + mu)*std::log(r2);
 +
    return Vec2d(factor*ux, factor*uy);
 +
};
 +
</syntaxhighlight>
 +
 
 +
Copyable Matlab code for closed form solution:
 +
<syntaxhighlight lang="matlab" line>
 +
function [ux, uy] = point_contact_analytical(x, y, lam, mu, P)
 +
% Evaluates closed form solution for point contact.
 +
% lam and mu are Lame parameters, P is the point force magnitude
 +
r2 = x.^2 + y.^2; factor = -P / (4.0*pi*mu);
 +
ux = factor*(2*x.*y./r2 + 2*mu/(lam + mu)*atan2(y, x));
 +
uy = factor*((y.^2-x.^2)./r2 - (lam + 2*mu)/(lam + mu)*log(r2));
 +
end
 +
</syntaxhighlight>
  
 
==Numerical solution with [http://www.freefem.org/ FreeFem++]==
 
==Numerical solution with [http://www.freefem.org/ FreeFem++]==
Line 56: Line 77:
 
Equation (\ref{eq:weak}) can be handed to FreeFem++ with the help of [https://en.wikipedia.org/wiki/Voigt_notation#Mandel_notation Voigt or Mandel notation], that reduces the symmetric tensors $\boldsymbol{\sigma}$ and $\boldsymbol{\varepsilon}$ to vectors. The benefit of [https://en.wikipedia.org/wiki/Voigt_notation#Mandel_notation Mandel notation] is that it allows the tensor scalar product to be performed as a scalar product of two vectors.
 
Equation (\ref{eq:weak}) can be handed to FreeFem++ with the help of [https://en.wikipedia.org/wiki/Voigt_notation#Mandel_notation Voigt or Mandel notation], that reduces the symmetric tensors $\boldsymbol{\sigma}$ and $\boldsymbol{\varepsilon}$ to vectors. The benefit of [https://en.wikipedia.org/wiki/Voigt_notation#Mandel_notation Mandel notation] is that it allows the tensor scalar product to be performed as a scalar product of two vectors.
 
For this reason we create the following macros:
 
For this reason we create the following macros:
 +
<syntaxhighlight lang="c++">
 
  macro u [ux,uy] // displacements
 
  macro u [ux,uy] // displacements
 
  macro v [vx,vy] // test function
 
  macro v [vx,vy] // test function
Line 62: Line 84:
 
  macro em(u) [dx(u[0]),dy(u[1]),sqrt(2)*(dx(u[1])+dy(u[0]))/2] // strain in Mandel notation
 
  macro em(u) [dx(u[0]),dy(u[1]),sqrt(2)*(dx(u[1])+dy(u[0]))/2] // strain in Mandel notation
 
  macro A [[2*mu+lambda,mu,0],[mu,2*mu+lambda,0],[0,0,2*mu]] // stress-strain matrix
 
  macro A [[2*mu+lambda,mu,0],[mu,2*mu+lambda,0],[0,0,2*mu]] // stress-strain matrix
 +
</syntaxhighlight>
  
The weak form (\ref{eq:weak}) can then be expressed naturally in FreeFem++ syntax as
+
The weak form (\ref{eq:weak}) can then be expressed naturally in FreeFem++ syntax as <syntaxhighlight lang="c++" inline>int2d(Th)((A*em(u))'*em(v)) - int2d(Th)(b'*v)</syntaxhighlight>
int2d(Th)((A*em(u))'*em(v)) - int2d(Th)(b'*v)
 
  
  
==Numerical solution with [[Meshless Local Strong Form Method (MLSM)]]==
+
==Explicit solution==
  
 
<syntaxhighlight lang="c++" line>
 
<syntaxhighlight lang="c++" line>
Line 75: Line 97:
 
     domain.fillUniformBoundaryWithStep(O.d_space);
 
     domain.fillUniformBoundaryWithStep(O.d_space);
 
     domain.findSupport(O.n, internal);
 
     domain.findSupport(O.n, internal);
     domain.findSupport(O.n, boundary, internal, true); //search only among internal nodes + itself
+
     domain.findSupport(O.n, boundary, internal, true);
  
 
     EngineMLS<vec_t, Gaussians, Gaussians> mls(
 
     EngineMLS<vec_t, Gaussians, Gaussians> mls(
         {pow(domain.characteristicDistance()*O.sigmaB,2), O.m},   //basis functionsa
+
         {pow(domain.characteristicDistance()*O.sigmaB,2), O.m},
 
         domain.positions[domain.support[0]],
 
         domain.positions[domain.support[0]],
         pow(domain.characteristicDistance()*O.sigmaW,2));      //weight function    */
+
         pow(domain.characteristicDistance()*O.sigmaW,2));       
 
 
 
     auto mlsm = make_mlsm(domain, mls, internal);
 
     auto mlsm = make_mlsm(domain, mls, internal);
 
     for (auto& i : boundary)    u_3[i]  = u_anal(i);
 
     for (auto& i : boundary)    u_3[i]  = u_anal(i);
Line 91: Line 112:
 
         for (i=0;i<internal.size();++i){
 
         for (i=0;i<internal.size();++i){
 
             u_3[i] = O.dt * O.dt / O.rho * (
 
             u_3[i] = O.dt * O.dt / O.rho * (
                     O.mu * mlsm.lap(u_2,i) + O.E/(2-2*O.v ) * mlsm.graddiv(u_2,i) +
+
                     O.mu * mlsm.lap(u_2,i) + O.E/(2-2*O.v )  
 +
                    * mlsm.graddiv(u_2,i) +
 
                     force[i] - O.dampCoef * (u_2[i] - u_1[i])/O.dt
 
                     force[i] - O.dampCoef * (u_2[i] - u_1[i])/O.dt
 
                 ) /// navier part
 
                 ) /// navier part
Line 100: Line 122:
 
         u_2 = u_3;
 
         u_2 = u_3;
 
}
 
}
 +
</syntaxhighlight>
 +
 +
Note that an operator grad(div(u)) that requires also mixed derivatives is used. To obtain stable solution a minimal second order monomial basis is required.
 +
 +
==Implicit solution==
 +
<syntaxhighlight lang="c++" line>
 +
    /// [Field declarations]
 +
    Eigen::SparseMatrix<double, Eigen::RowMajor> M(2 * N, 2 * N);
 +
 +
    M.reserve(Range<int>(2 * N, 2*O.n));
 +
    Eigen::VectorXd rhs(2 * N);
 +
 +
    // Set equation on interior
 +
    for (int i : internal) {
 +
        op.lapvec(M, i,  O.E / (2 + 2 * O.v));
 +
        op.graddiv(M, i, O.E / (2 - 2 * O.v));  // graddiv + laplace in interior
 +
        rhs(i) = 0;
 +
        rhs(i+N) = 0;
 +
    }
 +
    // Set boundary conditions
 +
    for (int i : boundary) {
 +
        M.coeffRef(i, i) = 1;  // boundary conditions on the boundary
 +
        M.coeffRef(i+N, i+N) = 1;  // boundary conditions on the boundary
 +
        Vec2d bc = analytical(i);
 +
        rhs(i) = bc[0];
 +
        rhs(i+N) = bc[1];
 +
    }
 +
 +
    // Sparse solve
 +
    Eigen::BiCGSTAB<Eigen::SparseMatrix<double, Eigen::RowMajor>> solver;
 +
    solver.setTolerance(1e-15);
 +
    solver.compute(M);
 +
    Eigen::VectorXd sol = solver.solve(rhs);
 
</syntaxhighlight>
 
</syntaxhighlight>
  
 
== Results ==
 
== Results ==
 +
=== Purely Dirichlet case ===
 +
For starters we check the solution of the problem with Dirichlet BC. The conditions are obtained from closed form solution [[#Closed form solution]].
  
Comparison of MLS based MLSM, FEM and analytical solution of displacements at different cross-sections.  
+
[[File:image_1b19ssjsn1e4t7qvq0ffnb1pft9.png|600px]]
  
[[File:v_for_y_05.png|400px]][[File:v_for_x_0.png|400px]][[File:u_for_y_05.png|400px]]
+
The $L^2$ error norm is used to quantify the error. Since the displacements are the variable that obtain from FreeFem++, we use the displacement magnitude $\|\boldsymbol{u}\| = \sqrt{u_x^2+u_y^2}$ to define our $L^2$-error norm. We first define the relative error field:
  
<figure id="fig:point_contact_convergence">
+
\begin{equation}
 +
L^2(\b{x}_i) = \frac{\|\b{u}(\b{x}_i)-\b{u}_\text{analytical}(\b{x}_i)\|}{\mathrm{max}\left(\| \b{u}_\text{analytical}(\b{x}_i)\| ,10^{-10}\right)},
 +
\end{equation}
  
 +
and then take the maximum $L^2$ error as our indicator:
  
For the convergence between analytical and numerical solution we vary the number of nodes by increasing the grid size in both $x$- and $y$- directions simultaneously by powers of two from $2^2$ (16 nodes all together) to $2^7$ (16384 nodes all together).
 
The $L^2$ error norm is used to measure the "difference" between solutions. Since the displacements are the variable that obtain from FreeFem++, we use the displacement magnitude $|\boldsymbol{u}| = \sqrt{u_x^2+u_y^2}$ to define our $L^2$-error norm. The exact equation we have used is
 
 
\begin{equation}
 
\begin{equation}
L^2\text{-norm} = \sqrt{\frac{\int_\Omega (|\boldsymbol{u_{\text{numerical}}}|-|\boldsymbol{u_{\text{analytical}}}|)^2d\Omega}{\int_\Omega|\boldsymbol{u_{\text{analytical}}}|^2d\Omega}}. 
+
L^2 = \mathrm{max}_i\left(L^2(\b{x}_i)\right)
 
\end{equation}
 
\end{equation}
Results are shown in <xr id="fig:point_contact_convergence"/>
+
 
[[File:Convergence.png|400px|<caption>FEM Convergence results for the point contact problem. The colours blue, red and green represent linear, quadratic and cubic finite elements, respectively.</caption>]]
+
For the FreeFem++ convergence test we vary the number of nodes by increasing the square grid size in both $x$- and $y$- directions simultaneously by powers of two from $2^2$ (16 nodes all together) to $2^7$ (16384 nodes all together).
</figure>
+
 
 +
Comparison of MLS based explicit and implicit MLSM, FEM and analytical solution of displacements at different cross-sections.
 +
 
 +
[[File:v_for_y_05.png|400px]][[File:v_for_x_0.png|400px]][[File:u_for_y_05.png|400px]]
 +
 
 +
[[File:crossY.png|400px]][[File:crossX0.png|400px]][[File:crossX.png|400px]]

Latest revision as of 18:39, 23 November 2022

Click here to return back to Solid Mechanics

Point contact on a 2D half-plane

A starting point to solve problems in contact mechanics is to understand the effect of a point-load applied to a homogeneous, linear elastic, isotropic half-plane. This problem may be defined either as plane stress or plain strain (for solution with FreeFem++ we have chosen the latter). The traction boundary conditions for this problem are: \begin{equation}\label{eq:bc} \sigma_{xy}(x,0) = 0, \quad \sigma_{yy}(x,y) = -P\delta(x,y) \end{equation} where $\delta(x,y)$ is the Dirac delta function. Together these boundary conditions state that there is a singular normal force $P$ applied at $(x,y) = (0,0)$ and there are no shear stresses on the surface of the elastic half-plane.

Closed form solution

The analytical relations for the stresses can be found from the Flamant solution (stress distributions in a linear elastic wedge loaded by point forces a the tip. When the "wedge" is flat we get a half-plane). For a point $(x, y)$ denote its square distance to the origin by $r^2 = x^2 + y^2$, and the stresses in point $(x, y)$ are given by: \begin{equation} \sigma_{xx} = -\frac{2P}{\pi} \frac{x^2y}{r^4}, \end{equation} \begin{equation} \sigma_{yy} = -\frac{2P}{\pi} \frac{y^3}{r^4}, \end{equation} \begin{equation} \sigma_{xy} = -\frac{2P}{\pi} \frac{xy^2}{r^4}, \end{equation} for some point $(x,y)$ in the lower half-plane. From this stress field the strain components and thus the displacements $(u_x,u_y)$ can be determined. The displacements are given by \begin{align} u_x &= -\frac{P}{4\pi\mu}\left(\frac{2xy}{r^2}+\frac{2\mu}{\lambda + \mu} \arctan(y/x)\right), \label{eq:dispx}\\ u_y &= -\frac{P}{4\pi\mu}\left(\frac{y^2-x^2}{r^2}-\frac{\lambda+2\mu}{\lambda + \mu} \log(r^2)\right), \label{eq:dispy} \end{align} where again $r^2 = x^2+y^2$ and $\lambda$ and $\mu$ are Lame parameters. For plane stress, do the substitution $\lambda \gets \frac{2\mu\lambda}{\lambda+2\mu}$ in the solution and in the Navier equation.

The proof of correctness and plots of the solution are available in this notebook: File:point_contact.nb

Copyable C++ code for closed form solution:

1 std::function<Vec2d(Vec2d)> analytical = [=](const Vec2d& p) {
2     double x = p[0], y = p[1], r2 = x*x + y*y, factor = -force / (4.0*M_PI*mu);
3     double ux = 2*x*y/r2 + 2*mu/(lam + mu)*std::atan2(y, x);
4     double uy = (y*y-x*x)/r2 - (lam + 2*mu)/(lam + mu)*std::log(r2);
5     return Vec2d(factor*ux, factor*uy);
6 };

Copyable Matlab code for closed form solution:

1 function [ux, uy] = point_contact_analytical(x, y, lam, mu, P)
2 % Evaluates closed form solution for point contact.
3 % lam and mu are Lame parameters, P is the point force magnitude
4 r2 = x.^2 + y.^2; factor = -P / (4.0*pi*mu);
5 ux = factor*(2*x.*y./r2 + 2*mu/(lam + mu)*atan2(y, x));
6 uy = factor*((y.^2-x.^2)./r2 - (lam + 2*mu)/(lam + mu)*log(r2));
7 end

Numerical solution with FreeFem++

Due to the known analytical solution the point-contact problem can be used for benchmarking numerical PDE solvers in terms of accuracy (as well as computational efficiency). The purpose of this section is to compare the numerical solution obtained by FreeFem++ with the analytical solution, as well as provide a reference numerical solution for the C++ library developed in our laboratory.

For purposes of simplicity we limit ourselves to the domain $(x,y) \in \Omega = [-1,1] \times[-1,-0.1]$ and prescribe Dirichlet displacement on the boundaries $\Gamma_D$ from the known analytical solution (\ref{eq:dispx}, \ref{eq:dispy}). This way we avoid having to deal with the Dirac delta traction boundary condition (\ref{eq:bc}). The problem can be described as find $\boldsymbol{u(\boldsymbol{x})}$ that satisfies \begin{equation} \boldsymbol{\nabla}\cdot\boldsymbol{\sigma}= 0 \qquad \text{on }\Omega \end{equation} and \begin{equation} \boldsymbol{u} = \boldsymbol{u}_{\text{analytical}} \qquad \text{on }\Gamma_D \end{equation} where $\boldsymbol{u}_\text{analytical}$ is given in equations (\ref{eq:dispx}) and (\ref{eq:dispy}).

To solve the point-contact problem in FreeFem++ we must first provide the weak form of the balance equation: \begin{equation*} \boldsymbol{\nabla}\cdot\boldsymbol{\sigma} + \boldsymbol{b} = 0. \end{equation*} The corresponding weak formulation is \begin{equation}\label{eq:weak} \int_\Omega \boldsymbol{\sigma} : \boldsymbol{\varepsilon}(\boldsymbol{v}) \, d\Omega - \int_\Omega \boldsymbol{b}\cdot\boldsymbol{v}\,d\Omega = 0 \end{equation} where $:$ denotes the tensor scalar product (tensor contraction), i.e. $\boldsymbol{A}:\boldsymbol{B} =\sum_{i,j} A_{ij}B_{ij}$. The vector $\boldsymbol{v}$ is the test function or so-called "virtual displacement".

Equation (\ref{eq:weak}) can be handed to FreeFem++ with the help of Voigt or Mandel notation, that reduces the symmetric tensors $\boldsymbol{\sigma}$ and $\boldsymbol{\varepsilon}$ to vectors. The benefit of Mandel notation is that it allows the tensor scalar product to be performed as a scalar product of two vectors. For this reason we create the following macros:

 macro u [ux,uy] // displacements
 macro v [vx,vy] // test function
 macro b [bx,by] // body forces
 macro e(u) [dx(u[0]),dy(u[1]),(dx(u[1])+dy(u[0]))/2] // strain (for post-processing)
 macro em(u) [dx(u[0]),dy(u[1]),sqrt(2)*(dx(u[1])+dy(u[0]))/2] // strain in Mandel notation
 macro A [[2*mu+lambda,mu,0],[mu,2*mu+lambda,0],[0,0,2*mu]] // stress-strain matrix

The weak form (\ref{eq:weak}) can then be expressed naturally in FreeFem++ syntax as int2d(Th)((A*em(u))'*em(v)) - int2d(Th)(b'*v)


Explicit solution

 1     RectangleDomain<vec_t> domain(O.domain_lo, O.domain_hi);
 2     domain.fillUniformInteriorWithStep(O.d_space);
 3     domain.fillUniformBoundaryWithStep(O.d_space);
 4     domain.findSupport(O.n, internal);
 5     domain.findSupport(O.n, boundary, internal, true);
 6 
 7     EngineMLS<vec_t, Gaussians, Gaussians> mls(
 8         {pow(domain.characteristicDistance()*O.sigmaB,2), O.m},  
 9         domain.positions[domain.support[0]],
10         pow(domain.characteristicDistance()*O.sigmaW,2));       
11     auto mlsm = make_mlsm(domain, mls, internal);
12     for (auto& i : boundary)    u_3[i]  = u_anal(i);
13     /// [MAIN TEMPORAL LOOP]
14     for (size_t step = 0; step * O.dt < O.time; step++) {
15         int i;  
16         ///[NAVIER-CAUCHY EQUATION :: explicit Plane Stress]
17         #pragma omp parallel for private(i) schedule(static)
18         for (i=0;i<internal.size();++i){
19             u_3[i] = O.dt * O.dt / O.rho * (
20                     O.mu * mlsm.lap(u_2,i) + O.E/(2-2*O.v ) 
21                     * mlsm.graddiv(u_2,i) +
22                     force[i] - O.dampCoef * (u_2[i] - u_1[i])/O.dt
23                 ) /// navier part
24                 + 2 * u_2[i] - u_1[i];   
25         }
26         ///[STEP FORWARD]
27         u_1 = u_2;
28         u_2 = u_3;
29 }

Note that an operator grad(div(u)) that requires also mixed derivatives is used. To obtain stable solution a minimal second order monomial basis is required.

Implicit solution

 1     /// [Field declarations]
 2     Eigen::SparseMatrix<double, Eigen::RowMajor> M(2 * N, 2 * N);
 3 
 4     M.reserve(Range<int>(2 * N, 2*O.n));
 5     Eigen::VectorXd rhs(2 * N);
 6 
 7     // Set equation on interior
 8     for (int i : internal) {
 9         op.lapvec(M, i,  O.E / (2 + 2 * O.v));
10         op.graddiv(M, i, O.E / (2 - 2 * O.v));  // graddiv + laplace in interior
11         rhs(i) = 0;
12         rhs(i+N) = 0;
13     }
14     // Set boundary conditions
15     for (int i : boundary) {
16         M.coeffRef(i, i) = 1;  // boundary conditions on the boundary
17         M.coeffRef(i+N, i+N) = 1;  // boundary conditions on the boundary
18         Vec2d bc = analytical(i);
19         rhs(i) = bc[0];
20         rhs(i+N) = bc[1];
21     }
22 
23     // Sparse solve
24     Eigen::BiCGSTAB<Eigen::SparseMatrix<double, Eigen::RowMajor>> solver;
25     solver.setTolerance(1e-15);
26     solver.compute(M);
27     Eigen::VectorXd sol = solver.solve(rhs);

Results

Purely Dirichlet case

For starters we check the solution of the problem with Dirichlet BC. The conditions are obtained from closed form solution #Closed form solution.

Image 1b19ssjsn1e4t7qvq0ffnb1pft9.png

The $L^2$ error norm is used to quantify the error. Since the displacements are the variable that obtain from FreeFem++, we use the displacement magnitude $\|\boldsymbol{u}\| = \sqrt{u_x^2+u_y^2}$ to define our $L^2$-error norm. We first define the relative error field:

\begin{equation} L^2(\b{x}_i) = \frac{\|\b{u}(\b{x}_i)-\b{u}_\text{analytical}(\b{x}_i)\|}{\mathrm{max}\left(\| \b{u}_\text{analytical}(\b{x}_i)\| ,10^{-10}\right)}, \end{equation}

and then take the maximum $L^2$ error as our indicator:

\begin{equation} L^2 = \mathrm{max}_i\left(L^2(\b{x}_i)\right) \end{equation}

For the FreeFem++ convergence test we vary the number of nodes by increasing the square grid size in both $x$- and $y$- directions simultaneously by powers of two from $2^2$ (16 nodes all together) to $2^7$ (16384 nodes all together).

Comparison of MLS based explicit and implicit MLSM, FEM and analytical solution of displacements at different cross-sections.

V for y 05.pngV for x 0.pngU for y 05.png

CrossY.pngCrossX0.pngCrossX.png