ML21277A093
| ML21277A093 | |
| Person / Time | |
|---|---|
| Issue date: | 08/18/2021 |
| From: | Jason Carneal, Lisa Regner, Rebecca Sigmon, Chris Speer, Julie Winslow NRC/NRR/DRO/IOEB |
| To: | NRC/RES/DSA |
| Dennis M | |
| References | |
| Download: ML21277A093 (7) | |
Text
NRC Operating Experience Artificial Intelligence Workshop An Overview Jason Carneal Chris Speer Team Julie Winslow Rebecca Sigmon Lisa Regner Focus
- Machine learning / natural language processing applications for operating experience
- Progress to date
- Impacts to the reactor oversight process
OpE Artificial Intelligence Projects The operating experience branch is developing machine learning / natural language processing algorithms to make our processes more efficient.
Automation of operating experience processing Extending existing search tools to allow association of reports to inspection procedure, system, and available risk information
OpE Hub - Deployed Products Consolidation of Deployed Products
- Website portal for NRC users
- One stop shop for all OpE products
- Easy to navigate
- Facilitates user interaction and support
ROP Machine Learning Case 1 Classification of Operating Experience Documents by Technical Review Group Objectives ML / NLP
- Build a classifier that can sort Training Set Historical OpE Classification incoming OpE documents by Clearinghouse Data (SVC / multinomial Naive Bayes) technical review group
- Automate certain aspects of operating experience workflow Incoming OpE Documents Progress (e.g., licensee
- Working classifier for event event reports) notifications (ranges from 60%-90% accuracy depending Automated on technical area) Technical Review
- Exploring extension to Group Classification licensee event reports
ROP Machine Learning Case 2 Classifying Operating Experience Documents by Inspection Procedure Objectives
- Build a classifier that maps OpE documents to applicable Training Set ML / NLP inspection procedures Historical mapping of Classification
- Expand to include different generic communications to (Naive Bayes, Support Vector Machine) forms of classification inspection procedure (equipment types, failure modes, etc.)
Incoming OpE
- Deploy advanced search Documents capability (e.g., licensee event reports)
Progress Inspection
- Generic Communication Procedure Specific classifier built (ranges from 65-Operating 70% accuracy) Experience
- Exploring extension to licensee event reports and other operating experience documents
Offsetting redundant tasks
- Eliminating repetitive manual reports on popular topics
- Tools for inspectors / NRC staff / management to review data of interest
- Lowering bar of access to data for both internal and external users Improving data-driven decision-making
- Consolidating and democratizing access to sparse and difficult to access data
- Deploying tools that allow users to explore data on their own
- Sharing insights previously difficult to ascertain
Operating Reactor Analytics Public Site Public Site for Reactor Oversight Process Information
- Action Matrix
- Performance Indicators
- Findings Lead
Contact:
Reed Anzalone NRR/EMBARK https://www.nrc.gov/reactors/operating/oversight/analytics.html