Determining an approximate finite element mesh density using neural network techniques

Abstract

A system is presented which uses a neural network to predetermine the mesh density for modeling a magnetic device with finite elements. The system 'learns' how to mesh from examples of ideal meshes. Once trained, the system computes the mesh density given the geometric and material descriptions of a device. A mesh based on this density information can be used as the initial mesh for an adaptive solver.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>