From: Roy Stogner <roystgnr@ic...>  20061017 14:01:25

On Tue, 17 Oct 2006, li pan wrote: > I have a question of boundary condition. I want to set > Dirichlet boundary condition. The boundary surface in > any kinds of plane. The nodes on the plane can only > slid but not move out of the plane. Can I still use > penalty method? How can I do it? Yes: Instead of penalizing the vector between each boundary node and its desired point, penalize the normal distance between each boundary node and its desired plane. Instead of weighting that penalty residual against a full vector of test functions, weight it against the inner product of that vector and the normal derivative to the plane. If the boundary nodes need to be constrained not to move past the edges of a bounded region of a plane, things can get a little harder, but depending on what the rest of your PDE is (it sounds like you're doing mesh smoothing, right?) this may not be a problem.  Roy 