Análise do período transiente da simulação de sistemas não terminais

Abstract

The selection of a suitable warm-up period for a discrete-event simulation is still a difficult choice to determine. Some methods have been developed and will be briefly reviewed here. There are still no definitive criteria for choosing a particular method among existing ones. We discuss some existing methods and compared with a novelmethodology implemented. A performance analysis for the methods will be exposed by an evaluation procedure for comparison. The methods are applied to the output generated by a simple work model with different levels of utilization. Some metrics will be used to investigate the efficiency of each method compared to the novel proposal implemented. Results are presented which can be divided into two groups. Initially, a firstmodel, M/M/1 queue, used as a validation technique for the proposal. A study was conducted for different intensity levels. We investigated the efficiency of the proposed method and also the quality evolution of the estimates presented as the traffic intensity in the model test is increasing. The second model evaluated, M/M/C queues, with dif-ferent values c for each of the procedures provided in series was used for comparison procedure between the methods for the detection of the appropriate warm-up period. The individual results of each method and also the comparison between different techniques were evaluated. These results, ensure that the novel proposal can be effectively considered as a valid technique for procedures that require the estimation of warm-upperiods.