An approximation method to analyse polling models of pull-type production systems

In this paper, a polling model with finite queues to analyse a production system that is operating under a pull-type control mechanism is studied. The polling model is analysed under three well-known service policies namely, exhaustive, gated and g-limited. These policies determine when and how machines switch from one part to another. These polling models can be solved by an exact approach that requires the complete characterisation of the system as a Markov chain. However, the state space increases dramatically with queue capacity and the number of parts manufactured. Hence an approximation approach that uses the regenerative property of the system is devised. Our numerical study shows that the approximation approach is not only very efficient in solving the problem but also very effective in finding accurate estimations of the system performance measures. [Received 13 December 2006, Revised 21 Febrauary 2007, Accepted 7 March 2007].