Currently, cyclic data cannot be input directly into AutoSimOA. The cycle would need to be taken out of the data by averaging over the cyclic periods. Also, the Warm-up Calculator cannot currently detect the initial bias in transient data. Warm-up methods assume steady-state and the question of initial bias in transient data has not yet been adequately addressed in the research literature.This work has concentrated on estimating the mean and variance of output data. There are however other measures of performance that may be of interest to the user 37
e.g. proportions, quantiles… It would be useful to extend AutoSimOA to provide decisions for these measures.8CONCLUSIONAn automated simulation output analyser, AutoSimOA, has been created in order to estimate the mean and variance of output data. The main components of AutoSimOA analyse and advise on the warm-up period and number of replications or run-length that should be used. The completed analyser has been demonstrated with case studies using transient and steady-state data, and has been shown to run smoothly and robustly. 38
39NOMODELTransient Steady StateDeterministicStochasticWarm-upReplications / One long runWarm-up Batch CyclesWHAT?TransientSteady StateOne long runWARM-UP?CYCLE?WARM-UP?CYCLE?YESNOYESNOYESNOYESNOYESYESNOWARM-UP?WARM-UP?YESYESNOWARM-UP?WARM-UP?ReplicationsBatch CyclesWarm-upWarm-upWarm-up Batch CyclesBatch CyclesWarm-upHOW?Figure 20: The output analysis decision process.NO
ACKNOWLEDGEMENTSThis work was part of the Automating Simulation Output Analysis (AutoSimOA) project () that was funded by the UK Engineering and Physical Sciences Research Council (EP/D033640/1). The work was carried out in collaboration with SIMUL8 Corporation, who also provided sponsorship for the project.REFERENCESAlexopoulos, C., Fishman, G.S. and Seila, A.F. 1997 Computational Experience withthe Batch Means Method?" Proceedings of the Winter Simulation Conference, p194-201Alexopoulos, C. (2006) A comprehensive review of methods for simulation output analysis. Proceedings of the 2006 Winter Simulation Conference, p168-178Banks, J., Carson II, J.S., Nelson, B.L., Nicol, D.M. (2005). Discrete-Event SystemSimulation. 4thEdition. Prentice Hall, Upper Saddle River, NJ.Bischak, D.P., Kelton, W.D. and Pollock, S.M. (1993) Weighted Batch Means forConfidence Intervals in Steady-State Simulations Management Science, Vol. 39, No.8, p1002-1019.Chen, E.J. and Kelton, W.D. (2007) A Procedure for Generating Batch-Means Confidence Intervals for Simulation: Checking Independence and NormalityChow YS, Robbins H (1965). On the asymptotic theory of fixed-width sequentialconfidence intervals for the mean. The Annals of Mathematical Statistics; 36:457-462.Conway, R.W. (1963). Some technical problems in digital simulation. ManagementScience, 10 (1), 47-61.40
Delaney, P.J. 1995. Control of initialisation bias in queuing simulations using queuingapproximations. M.S. thesis, Department of Systems Engineering, University of Virginia.