A practical bottleneck detection method

This paper describes a novel method for detecting the bottleneck in a discrete event system by examining the average duration of a machine being active for all machines. The machine with the longest average uninterrupted active period is considered the bottleneck. The method is widely applicable and also capable of analyzing complex and sophisticated systems. The results are highly accurate, distinguishing between bottleneck machines and non-bottleneck machines with a high level of confidence. This approach is very easy to use and can be implemented into existing simulation tools with little effort, requiring only an analysis of the log file which is readily available by almost all simulation tools. This method satisfies not only academic requirements with respect to accuracy but also industry requirements with respect to usability.