A system monitoring model by examining entity dynamics

Abstract

Monitoring plays an essential role in the manufacturing processes, as it bridges the gap between system conditions and the necessary corrective activities. Generally, for a Computer Integrated Manufacturing (CIM) system, monitoring is about collecting various facility signals, which are fed into specific models to interpret the signals. However, from the managerial viewpoint, the healthiness of the whole system is more desirable than the operation detail. In effect, the monitoring operation of a CIM system is analogous to diagnosing a system by checking the functioning of some targeted domains. Inspired by the distinctive philosophy that a proper system should be working in complete harmony, a novel method for presenting a holistic picture of a manufacturing system by examining the flowing entities is presented in this research. This research was conducted in three stages. First, a manufacturing system was modelled as the integration of a set of Regions of Interest (ROIs) in a high level manner. Second, analogous to the concept of checking blood pulses in the human body, several features were extracted from a system to constitute the "pulses" of an ROI; these features include the Regional Inconsistency (RI), the Inter-component Arrival Time (IAT), the Inter-component Leaving Time (ILT), and the Instant Work-In-Progress (IWIP). A reasoning scheme was then established to detect two types of popular abnormalities (blockings and slowdowns) in an ROI. Third, an ROI segmentation technique was developed to assist in the design of the monitoring framework by taking into consideration the tolerable system response time. At the outset of the study, it was anticipated that, based on analyzing the "pulses" tones of all ROIs involved, the healthiness of the holistic system could be reflected. Simulation experiments were conducted to validate the effectiveness of the monitoring approach proposed in this research. It was found that, in terms of the hardware requirements, only simple counter devices with time-stamp functions are needed, and this highly enhances the portability of the proposed approach.