Modern scientific simulations generate large datasets at remote sites with appropriate resources (supercomputers and clusters). Bringing these large datasets to the computers of all members of a distributed team of collaborators is often impractical or even impossible: there might not be enough bandwidth, storage capacity or appropriate data analysis and visualization tools locally available. To address the need to access remote data, avoid heavy Internet traffic and unnecessary data replication, Tech-X Corporation developed a tool, which allows running remote data visualization collaboratively and sharing the visualization objects as they get generated. The size of these objects is typically much smaller than the size of the original data. For marketing reasons, we renamed the product CoReViz. The detailed information on this product can be found at http://www.txcorp.com/products/CoReViz/. We installed and tested this tool at multiple machines at Tech-X and on seaborg at NERSC. In what follows, we give a detailed description of this tool.

The objective of the performed research is to develop an early anomaly detection methodology so as to enhance safety, availability, and operational flexibility of Boiling Water Reactor (BWR) nuclear power plants. The technical approach relies on suppression of potential power oscillations in BWRs by detecting small anomalies at an early stage and taking appropriate prognostic actions based on an anticipated operation schedule. The research utilizes a model of coupled (two-phase) thermal-hydraulic and neutron flux dynamics, which is used as a generator of time series data for anomaly detection at an early stage. The model captures critical nonlinear features of coupled thermal-hydraulic and nuclear reactor dynamics and (slow time-scale) evolution of the anomalies as non-stationary parameters. The time series data derived from this nonlinear non-stationary model serves as the source of information for generating the symbolic dynamics for characterization of model parameter changes that quantitatively represent small anomalies. The major focus of the presented research activity was on developing and qualifying algorithms of pattern recognition for power instability based on anomaly detection from time series data, which later can be used to formulate real-time decision and control algorithms for suppression of power oscillations for a variety of anticipated operating conditions. The …