Agrégation des données pour l'évaluation des performances de systèmes flexibles de production par modèles à réseaux de files d'attente

Abstract

This paper deals with the use of Queueing Network (QN) models for evaluating/predicting quantitatively the performances of Flexible Manufacturing Systems (FMS) at the strategic and tactical decision levels. One of the main problems in such studies consists in obtaining objectively the characteristics of a computationally tractable QN model from the data provided by industrial experts (process plans, production ratios). Indeed, when exploiting directly these data, one is usually led to considering QN models that cannot be analytically or numerically solved because they exhibit too many customer chains and classes. Thus, we propose a formalization of the workload of an FMS that is further exploited for defining a systematic workload data aggregation process. This process enables a transparent obtention of the exact characteristics of QN model which can be analytically and/or numerically solved. The application and the quality of our results are shown on several examples, one of which being an industrial FMS.