The numerical performance of wavelets for PDEs: The multi-scale finite element (Q1570310)

The research summarized in this paper is part of a multi-year effort focused on evaluating the viability of wavelet bases for the solution of partial differential equations. The primary objective for this work is to establish a foundation for hierarchical/wavelet simulation methods based upon numerical performance, computational efficiency, and the ability to exploit the hierarchical adaptive nature of wavelets. This work demonstrates that hierarchical bases can be effective for problems with a dominant elliptic character. The author use orthogonality in energy norm with a nodal change-of-basis to develop multi-scale finite elements that make use of a hierarchical Schauder basis. The use of the wavelet transform in conjuction with the assembly of the mass and stiffness operators for the hierarchical basis is presented in the framework of 1-D and 2-D multi-scale finite elements. In addition, a comparison between the hierarchical Schauder basis and the linear finite element basis is presented. The application of an ad-hoc row-column lumping procedure for the hierarchical basis is outlined, and its effectiveness for solving elliptic boundary value problems in one and two dimensions is demonstrated.