Dissertação
Modelos matemáticos para os problemas de dimensionamento e programação de bateladas em máquina única e máquinas paralelas
Fecha
2014-03-19Registro en:
TRINDADE, Renan Spencer. MATHEMATICAL MODELS FOR SCHEDULING A SINGLE AND PARALLEL IDENTICALS BATCH PROCESSING MACHINES WITH NON-IDENTICAL JOB SIZES. 2014. 100 f. Dissertação (Mestrado em Ciência da Computação) - Universidade Federal de Santa Maria, Santa Maria, 2014.
Autor
Trindade, Renan Spencer
Institución
Resumen
Problems of scheduling on batch processing machines to minimize makespan are widely
exploited by academic literature, mainly motivated by reliability testing in the semiconductor
industry. These problems consist in grouping jobs as a batch and scheduling the processing in
single or parallel machines. The jobs have non-identical processing times and non-identical
sizes and the total size of the batch cannot exceed the machine capacity. The processing time
of a batch is given by the longest processing time of any job in the batch. Jobs with nonidentical
release times can also be considered, and in this case a batch can only be processed
after the job with the longest release time in the batch is available. We consider four different
problems of scheduling on batch processing machines with non-identical job size and
different characteristics: single batch processing machine (1|sj,B|Cmax), single batch
processing machine with non-identical job release times (1|rj,sj,B|Cmax), identical parallel
batch processing machines (Pm|sj,B|Cmax), and identical parallel batch processing machines
with non-identical job release times (Pm|rj,sj,B|Cmax). New mathematical models are proposed
with formulations that exploit characteristics of each problem. The mathematical models are
solved using CPLEX and the computational results show that the proposed models performed
better than other models from literature. The new models for 1|sj,B|Cmax and 1|rj,sj,B|Cmax are
compared with previously published meta-heuristics and the results show that the models
provide better solutions than meta-heuristics methods with competitive computational times.