Automated Analysis for the Joyo Remote Monitoring System

The quantity and nature of monitoring system data makes creates a need for automated analysis. The Knowledge Generation (KG) software provides a tool for automated analysis of monitoring system data. KG uses cascaded state machines to model processes and compare monitoring system data to the declared operations at the facility. As part of an on going cooperation between the Japan Nuclear Cycle Development Institute and DOE, the KG system was recently applied to the Joyo Remote Monitoring System. KG provides a generic framework for developing models to analyze remote monitoring system data. Since KG doesn’t know about any particular facility, a key part of applying KG to a facility is developing a model of the facility and its processes. This requires an understanding of the remote monitoring system and the facility operations. The JOYO KG analysis models spent fuel pond transfer machine movements. Analysis of this sensor data is a difficult task for a human even with the help of images and other visual aids and is a prime candidate for automation. Modeling the transfer machine provided a challenge since the remote monitoring system provides only simple motion detectors that were intended to detect the cask as it moved in to and out of the facility. The KG model exploits motion detector response parallax that indicates the transfer machine direction. Analysis of existing event sequences identified key criteria for determining the direction of the transfer machine. These criteria were coded into state machine models that determine the transfer machine direction. These tracking state machines were combined with higher level state machines to create a robust and reliable facility model that identifies keys steps in a spent fuel transfer operation. The KG model successfully tracks the spent fuel pond transfer machine and identifies expected and abnormal motion sequences. More importantly, this result shows that an analysis program like KG can automatically synthesize raw data from a marginally adequate set of sensors into information and knowledge about the processes occurring in a monitored facility.