Parallel approach of a Galerkin-based methodology for predicting the compressive strength of the lightweight aggregate concrete

Abstract

A methodology based on the Galerkin formulation of the finite element method has been analyzed for predicting the compressive strength of the lightweight aggregate concrete using ultrasonic pulse velocity. Due to both the memory requirements and the computational cost of this technique, its parallelization becomes necessary for solving this problem. For this purpose a mixed MPI/OpenMP parallel algorithm has been designed and different approaches and data distributions analyzed. On the other hand, this Galerkin methodology has been compared with multiple linear regression models, regression trees and artificial neural networks. Based on different measures of goodness of fit, the effectiveness of the Galerkin methodology, compared with these statistical techniques for data mining, is shown.