Context: The amount of gravity data and the resolution of gravity field models has increased continuously over the last decades. This results in large dense linear systems with N unknowns where N is typically on the order of 100,000. The set-up of the linear system requires O(N2) operations and the solution is an order O(N3) process if classical direct solvers are used. Therefore, there is a strong need to develop methods that can solve the problem in order O(N,logs,N) for some non-negative small number s.
Aim: Study and implementation of the panel clustering technique for the numerical solution of geodetic boundary value problems using boundary element methods.
Methods: The focus is on the approximation of the kernel function by degenerate kernels in the farfield using Taylor series expansions (panel clustering).
Present topics: The panel clustering technique. |
