A moving mesh strategy for solving high dimensional PDEs is presented
along the lines of the moving mesh PDE (MMPDE) approach recently devel
oped in one dimension by the authors and their collaborators, With thi
s strategy, a moving mesh PDE is formulated from the gradient flow equ
ation for a suitable functional, and the underlying physical PDE is re
placed with an extended system for computing the physical solution and
the mesh. The method allows simultaneously for adaptation and mesh sk
ewness controls. In certain cases the mesh function, being a harmonic
map, is guaranteed to exist when the boundary of the computational dom
ain is convex. Efficiency of the method is predicated upon being able
to solve the moving mesh PDE in an efficient manner. Numerical results
for 2D are given to demonstrate the ability of the MMPDE to concentra
te mesh points and to control the mesh skewness. A spatial-eigenvalue
approximate factorization scheme is used for solving the MMPDE. (C) 19
98 Elsevier Science B.V.