This paper deals with the problem of evaluating the system performance of general manufacturing systems for which an analytical method is available to estimate the main performance indicators such as throughput, buffer levels, lead times, etc. The main assumptions are that the analytical method is approximate (i.e., it is biased in estimating the system performance) and rapid (or cheap) in execution. A further assumption is that a higher accuracy method is available, such as an exact Markov chain or a discrete event simulation model, to correctly predict the system performance. Respect to approximate analytical model this model is exact, except for the biasness introduced by the finite sample size in simulation runs, but slow or expensive. The paper proposes a method to combine experiments made using both analytical and simulation models. The results from few simulation experiments are used to adjust the estimate of analytical method by using a kernel based regression predictor. Numerical results show the proposed approach decreases the bias committed by analytical methods and also provide some insights about their improvement.

Operational Learning of Approximate Analytical Methods for Performance Evaluation of Manufacturing Systems

MATTA, ANDREA;
2015-01-01

Abstract

This paper deals with the problem of evaluating the system performance of general manufacturing systems for which an analytical method is available to estimate the main performance indicators such as throughput, buffer levels, lead times, etc. The main assumptions are that the analytical method is approximate (i.e., it is biased in estimating the system performance) and rapid (or cheap) in execution. A further assumption is that a higher accuracy method is available, such as an exact Markov chain or a discrete event simulation model, to correctly predict the system performance. Respect to approximate analytical model this model is exact, except for the biasness introduced by the finite sample size in simulation runs, but slow or expensive. The paper proposes a method to combine experiments made using both analytical and simulation models. The results from few simulation experiments are used to adjust the estimate of analytical method by using a kernel based regression predictor. Numerical results show the proposed approach decreases the bias committed by analytical methods and also provide some insights about their improvement.
2015
Proceedings of the 10th Conference on Stochastic Models of Manufacturing and Service Operations SMMSO 2013
978-960-9439-33-6
approximate analytical methods, simulation, multifidelity, kernel regression
File in questo prodotto:
File Dimensione Formato  
smmso01_30_03_2015.pdf

Accesso riservato

Descrizione: Paper definitivo
: Publisher’s version
Dimensione 283.42 kB
Formato Adobe PDF
283.42 kB Adobe PDF   Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11311/986821
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact