Real-time data-driven average active period method for bottleneck detection

Abstract

Prioritising improvement and maintenance activities is an important part of the production management and development process. Companies need to direct their efforts to the production constraints (bottlenecks) to achieve higher productivity. The first step is to identify the bottlenecks in the production system. A majority of the current bottleneck detection techniques can be classified into two categories, based on the methods used to develop the techniques: Analytical and simulation based. Analytical methods are difficult to use in more complex multi-stepped production systems, and simulation-based approaches are time-consuming and less flexible with regard to changes in the production system. This research paper introduces a real-Time, data-driven algorithm, which examines the average active period of the machines (the time when the machine is not waiting) to identify the bottlenecks based on real-Time shop floor data captured by Manufacturing Execution Systems (MES). The method utilises machine state information and the corresponding time stamps of those states as recorded by MES. The algorithm has been tested on a real-Time MES data set from a manufacturing company. The advantage of this algorithm is that it works for all kinds of production systems, including flow-oriented layouts and parallel-systems, and does not require a simulation model of the production system.