Parallel Computing Techniques

Abstract

Meshless methods are numerical methods for solving partial differential equations. The connectivity relationships between the discretized nodes do not need to be explicitly established. Instead, a collection of nodes is distributed over the problem domain. This represents an advantage when it is necessary to generate a new mesh several times, as in the case of moving structures. However, this advantage has its price, as this class of methods consumes more computing time than methods such as finite element methods. Thus, for more complex problems that require “remesh” such a moving electrical machine models, this work aims to apply parallel programming techniques to accelerate the execution of computational codes that solve mathematical models of electromagnetism based on meshless methods. Programming with threads, in this case open multi-processing, was used to parallelize the most time-consuming parts of the code of an electromagnetic model of an induction machine based on the Element-Free Galerkin meshless method.