SyFi - Symbolic Finite Elements

News

Release 0.3.3: Mainly a bug fix. A few examples described in the Para06 proceeding is also included.
Release 0.3: The Hermite and Nedelec elements have been added. Legendre polynomials are implemented. Examples demonstrating code generation and the use of Diffpack, Dolfin, PyCC, and Trilinos have been made.
Release 0.2: Added Bernstein polynomials. The Crouzeix-Raviart, Raviart-Thomas, and Taylor-Hood elements are implemented. Python support is added. Element matrices are computed for the Stokes problem and the mixed Poisson problem. Also, the Jacobian of a nonlinear convection diffusion problem is computed based on symbolic differentiation.
Release 0.1: Lagrangian elements and element matrices for Poisson problem are implemented.

Appetizer

The finite element method (FEM) package SyFi is a C++ library built on top of the symbolic math library GiNaC. The name SyFi stands for Symbolic Finite Elements. The package provides polygonal domains, polynomial spaces, and degrees of freedom as symbolic expressions that are easily manipulated. This makes it easy to define finite elements.

The following example shows the computation of the element matrix for the Poisson problem in C++,

void compute_element_matrix(Polygon& T, int order) { 
  std::map<std::pair<int,int>, ex> A;      // matrix of expression
  std::pair<int,int> index;                // index in matrix
  LagrangeFE fe;                           // Lagrangian element of any order 
  fe.set(order);                           // set the order 
  fe.set(T);                               // set the polygon 
  fe.compute_basis_functions();            // compute the basis functions
  for (int i=0; i< fe.nbf() ; i++) {
    index.first = i; 
    for (int j=0; j< fe.nbf() ; j++) {
      index.second = j; 
      ex nabla = inner(grad(fe.N(i)), grad(fe.N(j)));   // compute the integrands 
      ex Aij = T.integrate(nabla);                      // compute the weak form 
      A[index] = Aij;                                   // update element matrix
    }
  }
}

The following example demonstrates the computation of the Jacobian matrix for the nonlinear convection-diffusion equation in case of a power-law fluid:

polygon = ReferenceTriangle()
fe = VectorCrouzeixRaviart(polygon,1)
fe.compute_basis_functions()

# create sum u_i N_i
u, ujs = sum("u", fe) 
n = symbol("n") 

for i in range(0,fe.nbf()): 

  # nonlinear power-law diffusion term 
  mu = inner(grad(u), grad(u))
  fi_diffusion = pow(mu,n)*inner(grad(u), grad(fe.N(i))) 
  
  # nonlinear convection term 
  uxgradu = (u.transpose()*grad(u)).evalm()  
  fi_convection = inner(uxgradu, fe.N(i), True) 

  fi = fi_diffusion + fi_convection

  Fi = polygon.integrate(fi) 

  for j in range(0,fe.nbf()): 
    # differentiate to get the Jacobian 
    uj = ujs.op(j) 
    Jij = diff(Fi, uj) 
    # print out expression as C code 
    print "J[%d,%d]=%s\n"%(i,j, Jij.evalf().printc())

Requirement

SyFi relies on the symbolic math library GiNaC

Kent-Andre Mardal