Simulation Results on Buffer Allocation in a Continuous Flow Transfer Line with Three Unreliable Machines

Engineers designing a continuous flow production line consisting of machines and buffers in series, have to determine the optimal sizes of the intermediate buffers. When the machines operating on the product are unreliable, this dimensioning decision becomes even more difficult. This paper aims to provide some insight into the optimal sizes of intermediate buffers in a continuous flow transfer line with three machines and two buffers. To this end, an extensive series of simulation experiments are performed, and noted the performance (measured as the availability, i.e., the fraction of time that the system as a whole is producing output) under a large variety of different input conditions. The simulation model for this purpose is based on Petri-nets and has been extensively validated prior to this research. Using these simulation results, we try to find rules to determine where to allocate which amount of buffer space in order to obtain a sufficiently high availability of the system. Different scenarios are simulated: all machines have equal reliability, one machine has significantly higher reliability and one machine has significantly lower reliability. For each scenario, a large number of settings is tested. A careful study of the availability of the system as a function of the sizes of the two buffers reveals some interesting results. We find that the optimal buffer sizes depend strongly on the (relative) reliability of the different machines and that–in general–the performance of the entire system can be considerably increased by correctly choosing the sizes of the two buffers. When all machines are equally reliable, both buffers should be equally large. When the first or the last machine is less reliable, the first respectively the second buffer should be used to dampen the impact of this unreliable machine. Another finding is that if the middle machine is considerably less reliable than the other two, the buffers have far less effect on the performance of the overall system. The results reported in this paper can be used by practitioners in the field of design of continuous flow transfer line systems to gain insight into the optimal dimensioning of buffers in production systems. Future research in this area will consider simulation models of larger linear systems or more complex networks of machines and buffers. Also, the effects of relaxing one or more of the assumptions made in the simulation model (e.g., on the distribution of upand down-times of the machines) should be investigated.