Services allocation in a metrology laboratory using genetic algorithms / Distribuição de serviços em laboratório de metrologia utilizando algoritmos genéticos

AUTOR(ES)
DATA DE PUBLICAÇÃO

2009

RESUMO

As simple as it might look, the problem of allocating a group of tasks to be executed with a limited number of available machines will always present a number of details and restrictions that in certain cases can result in a solution of great complexity. Some of the variables involved are: service priority, available raw materials, machinery or equipment capacities, qualified technical personnel existence, production historical records, service execution time, machinery idle times, additional services that could call for specialized personnel. Currently, its easy to find some software solutions which provide this type of logistics to the industrial sector. The main goal of these solutions is to minimize the problems created when dealing with services distribution. Generally, integrated management systems, known as ERP (Enterprise Resource Planning) are provided with an entire module to deal with services distribution; unfortunately its not uncommon for this module to discard some or many of the factors involved. The specificities of each different case and the complexity to introduce these variables into the algorithm flux have a crucial role to play when choosing the best solution; the best alternative will depend heavily on its flexibility and versatility. When a machine has variable capacity, as it is in many cases, the problems complexity greatly increases and the traditional solution techniques like mathematic programming fall short of the solution. More successful results have been observed when using artificial intelligence techniques, particularly using genetic algorithms to solve the problem of services allocation (job scheduling) as there are written reports about it. This work introduces the modeling and solution for cases of services allocation using metrology laboratories and the use of a genetic algorithm tailored specifically to each case. It takes into account, for each case, all the variables and restrictions that the specialized personnel will face in a metrology laboratory. The solution modeling to optimize the services allocation among a limited number of resources using Genetic Algorithms, allows a better understanding about important aspects in real case situations, e.g. service type performed by the resource, item characteristics, resources working range, items metrology range, delivery times attainable, items and available resources quantities, services execution times. Worthwhile to mention and additional benefits obtained using the implemented algorithms are a better understanding of the problem when dealing with cloned items and the development of an associated Learning Genetic Algorithm with which a parameter definition automation for the main Genetic Algorithm Implementation is intended. These parameters are: crossover type, selection type, adequate mutation rate values, elitism and tournament. Preliminary results obtained using the proposed algorithm are showing the viability of the solution and its applicability in practical cases

ASSUNTO(S)

service order analise de algoritmos e complexidade de computacao algoritmo genético metrologia ordem de serviço metrology genetic algorithm

ACESSO AO ARTIGO

Documentos Relacionados