Speeding-up Finite Element analyses by replacing the linear equation solver with a Boundary Element code. Part 1: 2D linear elasticity

A substantial computational advantage can often be obtained by replacing the solver for the governing linear system of equations, inside a linear elastic Finite Element program, with a Boundary Element code implemented so as to calculate displacements (and, if required, stresses) in the same internal points (nodes and Gauss points) where the original Finite Element code would provide an output. The analysis of several test problems shows that, even in the unfavorable case of 2D geometries, this modification of the Finite Element method becomes rapidly convenient as the density of the starting Finite Element mesh increases.