Erland Hejn Nielsen

In managing production systems, a strong emphasis is placed on the reduction of variance in specific transformation processes (e.g. for quality control purposes) and on controlling the level of variability in general, as for instance with the Bullwhip effect. However, the possible disturbing interference from autocorrelation (variability dependence over time) in otherwise level stable event streams has received little attention. From the literature it is known that autocorrelation in queuing systems can lead to a large deterioration in performance compared to an event independent scenario. Thus, either production managers are missing important aspects in production planning, or the 'realistic' autocorrelation patterns inherent in actual production setups are not like those considered in the literature. In this paper, relevant and 'realistic' types of autocorrelation schemes are characterised and their levels discussed. The paper puts previous works on the impact of autocorrelation in queuing networks in perspective for production systems.