next up previous contents
Next: Compressible Navier-Stokes Simulations Up: Numerical Formulation Previous: Numerical Formulation

Discontinuous Galerkin for Diffusion

  We consider as a model problem the parabolic equation with variable coefficient $\nu$ to demonstrate the treatment of the viscous contributions:

\begin{displaymath}
u_t = \nabla \cdot (\nu \nabla u) + f, \quad \mbox{in $\Omega$}, \quad u \in L^2(\Omega) \end{displaymath}

\begin{displaymath}
u = g({\bf x}, t), \quad \mbox{on $\partial \Omega$} \end{displaymath}

We then introduce the flux variable

\begin{displaymath}
{\bf q} = -\nu \nabla u \end{displaymath}

and re-write the parabolic equation

\begin{displaymath}
u_t = - \nabla \cdot {\bf q} + f, \quad \mbox{in $\Omega$}\end{displaymath}

\begin{displaymath}
\frac{1}{\nu} {\bf q} = -\nabla u, \quad \mbox{in $\Omega$}\end{displaymath}

\begin{displaymath}
u = g({\bf x}, t), \quad \mbox{on $\partial \Omega$},\end{displaymath}

The variational formulation of the problem is then as follows: Find $({\bf q},u) \in H({\em div};\Omega)\times L^2(\Omega)$ such that

\begin{displaymath}
(u_t, w)_{E_i} = ({\bf q}, \nabla w)_{E_i} - <w,{\bf q^B} \c...
 ..._{\partial E_i} + (f,w)_{E_i}, 
\quad \forall w \in L^2(\Omega)\end{displaymath}

\begin{displaymath}
\frac{1}{\nu} ({\bf q}, v) = (u, \nabla v)_{E_i} - <u^B, v {\bf n}\gt _{\partial E_i},\quad \forall v \in H({\em grad};\Omega)\end{displaymath}

\begin{displaymath}
u = g({\bf x}, t), \quad \mbox{on $\partial \Omega$} \end{displaymath}

where the parentheses denote standard inner product in an element (Ei) and the angle brackets denote boundary terms on each element, with ${\bf n}$ denoting the unit outwards normal and:

The surface terms contain weighted boundary values of vB, qB, which can be chosen as the arithmetic mean of values from the two sides of the boundary, i.e. $v^B = \frac{( v^E + v^I)}{2}$, and $q^B = \frac{( q^E + q^I)}{2}$.

Integrating by parts once more, we obtain an equivalent formulation which is easier to implement and it is actually used in the computer code. The new variational problem is

\begin{displaymath}
(u_t, w)_{E_i} = (-\nabla \cdot {\bf q}, w)_{E_i} - <w,({\bf...
 ...bf n}\gt _{E_i} + 
(f,w)_{E_i}, \quad \forall w \in L^2(\Omega)\end{displaymath}

\begin{displaymath}
\frac{1}{\nu} ({\bf q}, v) = (-\nabla u, v)_{E_i} - 
<(u^B-u...
 ..., v\gt _{\partial E_i},\quad \forall v \in H({\em grad};\Omega)\end{displaymath}

\begin{displaymath}
u = g({\bf x}, t), \quad \mbox{on $\partial \Omega$} \end{displaymath}



T. Warburton
10/24/1998