An iterative a posteriori grid optimization procedure is presented. Grid quality is measured by efficient non-orthogonality, non-smoothness, and clustering functionals that are normalized using geometric considerations to provide for the combination of mutually incompatible characteristics in the final grid. Proper formulation and normalization of these functionals yields a scheme which is capable of creating non-overlapping grids in concave to convex regions. An additional functional which aids in untangling already overlapped grids is presented. Applications to solution adaptive grids are presented and the cost effectiveness of the procedure versus an elliptic grid generator is evaluated.