Desarrollo de un algoritmo evolutivo híbrido para la optimización de una cadena de suministro de dos empresas de ensamblaje

Abstract

A fundamental goal in the world is to obtain an optimal state, or rather, an accurate, precise and perfect solution for a specific problem. These optimal states can be found in different areas such as medicine, engineering or architecture. For example, Industrial Engineering has as one of its objectives to improve or optimize the processes of a company in order to obtain more benefits with lower costs. There are a lot of optimization algorithms, such as genetic, particle swarm optimization, micro-algorithms or memetic. Therefore, an optimal solution would be to take advantage of their benefits and then build a hybrid algorithm that incorporates their best features. In this manner, it is possible to find solutions to the problem much faster, in terms of runtime and convergence warranty. In this context, this research aims to find an optimization algorithm for a supply chain, which is a network that has different entities such as manufacturers, suppliers, distributors, retailers, transporters and customers or end-users. To this end, this work presents two case studies, a manufacturing, and an assembly company. The first company assembles furniture; while the second focuses on the assembly of televisions and motorcycles. In the first case study, the objective is to minimize the cost of the supply chain that is subject to several variables, such as transportation, distribution or manufacturing costs; and it is also desired to minimize the cost of product cutting. Meanwhile, in the second case, in addition to minimizing the cost of the chain, the objective is to maximize customer satisfaction. To optimize these statements, it was necessary to build functions, known as objective functions; likewise, there were several restrictions, such as the existing demand or the storage capacity of a product that has a distributor. In order to determine if this new hybrid algorithm finds the optimal solution to the problem, it was necessary to make a comparison among all the algorithms. This comparison is based on the runtime required for the algorithm to perform its work and the quality of the algorithm´s convergence. These results were found thanks to the implementation and execution of the algorithms using the same execution environment and the same general characteristics. In the end, a discussion and conclusion are made where the strongest points of the hybrid algorithm are determined, as well as, a comparison with the other algorithms and with hybrid algorithms proposed by other authors.