Impending failure detection for a discrete process

Abstract

Signals from a discrete process contain a strong modulation as a result of the discrete events in the process, such as paper passage in a recirculating document feeder (RDF). This paper presents a study of the methodology of process monitoring for a RDF system. A fault tree has been established that shows the cause-and-effect relationship regarding possible malfunctions of a RDF system. Critical components of the RDF system have been identified for condition monitoring. The signature from the measurements of position, vibration, vacuum pressure, and drive motor current have been analysed. A data separation scheme was used in signal processing to demodulate the strong signal component associated with paper passage. Unique index extraction algorithms based on time series analysis and modeling have been developed to detect failures of these components. A decision-making scheme based on multiple voting has been implemented.