Does anybody have experience with the Finite Difference Element Method described in http://www.scc.kit.edu/scc/docs/FDEM/Literatur/FDEM-Survey-Feb09.pdf and would advocate for or against using it? The method works on an unstructured grid, the field quantities are represented by their values at the grid points, and spatial derivatives are evaluated at the grid points by means of fitting polynomials into the gridpoints in the neightborhood of the respective gridpoint.
To me, at first sight, the method seems to have a couple of advantages, compared to FEM:
- The meaning of the state variables of the discretized system is intuitive
- It can be applied as a "black box" to virtually every PDE
- There is no a priori pen and paper work necessary for calculating weak forms, quadrature formulas etc.
- It can readily be extanded to rather high orders
- The PDE can contain arbitrarily high derivatives, in principle
- The error of the derivatives can be estimated
Does anyone know any drawbacks of the method?