Data Fusion and Feature Selection to Inform the State of a Nuclear Reactor

• Activity Number: 62 - Data Fusion: An Exploration of Practical Aspects
• Type: Topic Contributed
• Date/Time: Sunday, July 28, 2019 : 4:00 PM to 5:50 PM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): Nidhi Parikh* and Garrison Flynn and Adin Egid and Emily Casleton
• Companies: Los Alamos National Laboratory and Los Alamos National Laboratory and Los Alamos National Laboratory and Los Alamos National Laboratory
• Keywords: Data fusion; Feature selection; Classification; Missing data

Abstract

Validating that a nuclear reactor is operating at the declared power level is of specific interest to the nonproliferation community. Under the Multi-Informatics for Nuclear Operations Scenarios (MINOS) venture, a collocated research nuclear reactor and reprocessing facility at Oak Ridge National Laboratory have been instrumented with various sensors recording data. In this work, we focus on predicting the power level of the reactor using data from four modalities: seismic, acoustic, radiation, and electro-magnetic. However, combining data from multiple modalities poses two challenges: 1) Long-term monitoring of sensors is prone to outages of different sensors at different times leading to missing data from one or more modalities. 2) Each modality can be used to compute several features (potentially 100s of features) some of which may not be useful or redundant. We will perform feature selection to extract informative features from various modalities and combine these features through a classifier that is robust against missing features to predict the reactor power level.