SOLVING THE MIXTURE DESIGN PROBLEM ON A SHARED MEMORY PARALLEL MACHINE
Author(s):
J.a. Martínez,
L.g. Casado,
I. García,
Eligius M.t. Hendrix
Paper abstract: A computational model which explores the power of the distributed computing paradigm is the well known Branch and Bound scheme. One of the characteristics of such a scheme is the unpredictable behavior of the model in relation to the set of sub-problems dynamically generated, because this is application dependent. Therefore, specific load balance and task partition strategies have to be applied. This paper deals with a Branch and Bound approach for the semi-continuous quadratic mixture design (SCQMD) problem from a parallel computing perspective. SCQMD is a multi-objective global optimization problem with linear, quadratic and semi-continuity constraints. The Branch and Bound approach for the SCQMD problem is extremely complex from a computational point of view. Practical cases coming from the industry show a high computational complexity that make the problem hard to be solved on a uniprocessor platform. In this work, two parallel approaches are proposed and evaluated on a shared memory distributed computer
