Multi-stage monitoring of abnormal situation based on complex event processing

This paper focuses on the monitoring of abnormal situation in workspace where complicate production activities are performed and possible abnormal situations vary in different stages. The monitoring application should track the production process, identifying the production stage and detecting anomaly in every stage as defined. With the development of ubiquitous computing technology and widespread of sensing equipment, context information pertaining to smart working environment is available for monitoring applications. Complex event processing (CEP) is usually introduced to process and correlate context information for its attractive feature of extracting composite event from a large amount of event data in real time according to user-defined event patterns. In this paper, we present context model and event model in which discrete event such as acquiring context value at a point of time is represented by context. The abnormal situation in every stage of production can be transformed into event expressions, called abnormal event patterns. Contexts in different time captured by sensors form data streams and processed by CEP engine to detect abnormal situation. We propose to use state transition to model each stage so that the normal transition period in the beginning and end of stage can be distinguished from abnormal situation. Once a stage is identified to be starting or ending, the application will change abnormal event patterns accordingly. Case study about metallographic examination proves that the approach we propose is effective and feasible for some multi-stage abnormal situation monitoring. (C) 2016 The Authors. Published by Elsevier B.V.