Article ID: | iaor1998124 |
Country: | United Kingdom |
Volume: | 24 |
Issue: | 5 |
Start Page Number: | 387 |
End Page Number: | 398 |
Publication Date: | May 1997 |
Journal: | Computers and Operations Research |
Authors: | Narahari Y., Hemachandra N., Gaur M. S. |
Keywords: | markov processes |
In this article we examine the transient performance of a flexible machine centre that processes several classes of jobs with significant setup times and with priority scheduling. Our results show that the transient performance of the system over typical observation intervals can be significantly different from the performance predicted by steady-state analysis. The machine centre served three classes of jobs – class 1, class 2 and class 3; class 3 jobs have non-preemptive priority over jobs of class 1 and class 2. The results show that the throughput and cycle time of class 1 and class 2 jobs are affected quite dramatically by the arrival of class 3 jobs. However, over a typical observation period in the initial evolution of the system, steady-state results overestimate this effect of the high priority jobs whereas the more credible estimates are given by transient analysis. The transient analysis is carried out by solving explicitly the corresponding Markov chains, using a higher level stochastic Petri net model to generate the Markov chains.