Annals of Operations Research 93 (2000) 145–176 145 Production systems with interruptions, arbitrary topology and finite buffers Surendra M. Gupta ∗ and Ayse Kavusturucu Laboratory for Responsible Manufacturing, 334 SN, Department of MIME, Northeastern University, 360 Huntington Avenue, Boston, MA 02115, USA E-mail: gupta@neu.edu We consider a production system with finite buffers and arbitrary topology where service time is subject to interruptions in one of three ways, viz. machine breakdown, machine vacations or N -policy. We develop a unified approximation (analytical) methodology to calculate the throughput of the system using queueing networks together with decomposition, isolation and expansion techniques. The methodology is rigorously tested covering a large experimental region. Orthogonal arrays are used to design the experiments in order to keep the number of experiments manageable. The results obtained using the approximation methodology are compared to the simulation results. The t-tests carried out to investigate the differences between the two results show that they are statistically insignificant. Finally, we test the methodology by applying it to several arbitrary topology networks. The results show that the performance of the approximation methodology is consistent, robust and produces excellent results in a variety of experimental conditions. Keywords: production systems, expansion methodology, machine breakdown, machine va- cations, N -policy, split-merge topology, finite buffers 1. Introduction Modern production systems have become very complex and consequently, their analyses are equally complicated. A production system can be considered as a collec- tion of various service areas where jobs arrive at different rates and demand services with unequal processing times. Due to their nondeterministic behavior, researchers de- velop stochastic models (using queueing networks) to analyze such systems. Queueing networks are efficient tools for modeling systems of nondeterministic nature. They even provide product form solutions for some restrictive assumptions (for example, there should be infinite buffers at each service unit). However, when the system be- comes more complicated (such as, when the buffer capacities are finite), the solution rapidly becomes intractable as the size of the network and the number of nodes with limited buffer capacities grow. This is a serious drawback, since many real life sit- uations impose even more restrictions on the system. In this paper we focus on two * Corresponding author.  J.C. Baltzer AG, Science Publishers