The behavior of transient period of nonterminating simulations: an experimental analysis

Abstract

The design and control of many industrial and service systems require the analysts to account for uncertainty. Computer simulation is a frequently used technique for analyzing uncertain (or stochastic) systems. One disadvantage of simulation modeling is that simulation results are only estimates of model performance measures. Therefore, to obtain better estimates, the outputs of a simulation run should undergo a careful statistical analysis. Simulation studies can be classified as terminating and nonterminating according to the output analysis techniques used. One of the major problems in the output analysis of nonterminating simulations is the problem of initial transient. This problem arises due to initializing simulation runs in an unrepresentative state of the steady-state conditions. Many techniques have been proposed in the literature to deal with the problem of initial transient. However, existing studies try to improve the efficiency and effectiveness of currently proposed techniques. No research has been encountered that analyzes the behavior of the transient period. In this thesis, we investigate the factors affecting the length of the transient period for nonterminating manufacturing simulations, particularly for serial production lines and job-shop production systems. Factors such as variability of processing times, system size, existence of bottleneck, reliability of system, system load level, and buffer capacity are investigated.