3DM
Data-Driven Decision Making (3DM)
Welcome to the Data-Driven Decision Making (3DM) team wiki.
Applies data science, advanced analytics, and artificial intelligence / machine learning (AI/ML) to improve operational insight and decision-making. This work supports development of common methods, tools, and data practices to enable consistent, data-informed outcomes and scalable analytics applications across the nuclear enterprise.
Additional information is available on the Operating Plant Initiatives (OPI) Program website.
What We Stand For
The 3DM team is committed to advancing the real-world application of AI in data science through innovation, collaboration, and knowledge dissemination. These principles guide how we approach research and support the nuclear sector.
What We Do
The 3DM team provides the capabilities that support AI and data-driven innovation across the nuclear sector. These capabilities span applied research, technology development, technical collaboration and knowledge sharing
| Capability | Description |
|---|---|
| Applied Research | Investigate emerging AI, ML, and data science methods through research projects and proof-of-concept applications. |
| Technology Development | Develop prototypes, tools, and technical guidance that demonstrate practical applications of AI and advanced analytics. |
| Technical Collaboration | Partner with Nuclear Sector programs, researchers, and engineers to address real-world technical and operational challenges. |
| Knowledge Sharing | Create technical resources, best practices, and educational materials that support the adoption of AI and data science across the industry. |
Research Areas
The 3DM team applies data science, artificial intelligence (AI), and machine learning (ML) across several key research areas. These focus areas represent the primary domains where the team develops new methods, evaluates emerging technologies, and supports Nuclear Sector research initiatives.
- Insights-Transform data into meaningful information for decision-making
- Prognostics- Predict the condition of equipment and performance
- Automation- Improve efficiency with AI-enabled workflows.
- Optimization- Improve engineering and operational performance
Research Goals
The 3DM program's research goals are guided by the priorities outlined in the proposed Five-Year Plan (2025–2029). These priorities focus on strengthening AI and data science capabilities, supporting collaboration across the Nuclear Sector, and advancing research that enables the practical application of AI technologies.
Research Projects
The following projects highlight how the 3DM team applies AI, machine learning and data science to address rural world engineering, operational and research challenges across the nuclear sector
TroubleBot
TroubleBot is an troubleshooting assistant that helps engineers investigate equipment and system issues. By combining Retrieval-Augmented Generation (RAG) with EPRI's technical knowledge base, it assists users with identifying potential failure modes, gathering evidence, and generating troubleshooting reports.
Operating Experience Review Enhancements
Operating Experience Review Enhancements is an AI-assisted platform that streamlines the review of operating experience (OE) documents. It uses natural language processing, semantic search, and automated summarization to help engineers analyze historical events, identify trends, and locate relevant technical information more efficiently.
Relief Request Index
The Relief Request Index is a searchable web application that organizes NRC relief request documentation into a centralized knowledge base. AI-generated summaries and enhanced search capabilities allow users to quickly locate historical requests, supporting documents, and regulatory information.
AI-Enhanced Fuel Failure Detection
AI-Enhanced Fuel Failure Detection is a machine learning research project that analyzes operational plant data to identify indicators of potential fuel failures. The project evaluates predictive analytics techniques that can support earlier detection and improve situational awareness for plant personnel.
AI for Non-Destructive Examination (NDE)
AI for Non-Destructive Examination (NDE) investigates the use of artificial intelligence to assist with the analysis of inspection data collected during non-destructive examinations. The project explores methods for improving defect detection, data interpretation, and engineering evaluations.
Featured Publications & Technical Resources
The 3DM program develops publications, technical guidance, and educational resources that help researchers, engineers, and member utilities understand and apply artificial intelligence (AI), machine learning (ML), and data science within the nuclear industry. These resources share research findings, implementation guidance, best practices, and lessons learned from ongoing projects and emerging technologies.
Machine Learning Assisted Engineering Analyses
A collection of applied machine learning research projects that investigate AI-driven solutions for engineering challenges, including predictive modeling, uncertainty quantification, transformer monitoring, and acoustic emission analysis. These efforts demonstrate how machine learning can improve engineering analyses and support technical decision-making.
Generative AI and ML Enabling Technology Guides
A series of technical guidance documents that introduce emerging AI technologies, implementation considerations, and best practices for nuclear applications. Topics include recent advancements in large language models (LLMs), AI design control and verification and validation (V&V) methods, and integrating LLMs with existing EPRI software systems.
Generative AI Assisted Text Analyses POCs
A collection of proof-of-concept (POC) studies evaluating the use of generative AI, natural language processing (NLP), and large language models (LLMs) for technical document analysis. Current efforts include design requirement identification, probabilistic risk assessment (PRA) document retrieval, ALARA planning support, and information retrieval for nuclear decommissioning.
How to Run an LLM Project: The Story of EPRI's STELLA Series
A technical resource that documents EPRI's experience planning, developing and implementing large language model (LLM) applications through the STELLA project. The publication shares practical lessons learned, implementation strategies, and recommendations for developing enterprise LLM solutions.
Keyword Glossary
The following glossary defines common terms and acronyms used throughout the 3DM wiki and related research projects.
- Artificial Intelligence (AI)
- Computer systems designed to perform tasks that typically require human intelligence, such as reasoning, learning, language processing, and decision-making.
- Machine Learning (ML)
- A branch of artificial intelligence that uses data to identify patterns and make predictions or decisions.
- Data Science
- An interdisciplinary field that combines statistics, computing, and subject-matter knowledge to analyze data and produce useful insights.
- Advanced Analytics
- Analytical methods that use statistics, machine learning, optimization, and other computational techniques to support decision-making.
- Large Language Model (LLM)
- An artificial intelligence model trained on large amounts of text that can understand, generate, summarize, and analyze natural language.
- Natural Language Processing (NLP)
- A field of artificial intelligence focused on enabling computers to understand, interpret and generate human language.
- Retrieval-Augmented Generation (RAG)
- An artificial intelligence approach that combines document retrieval with a large language model so responses can be generated using relevant source information.
- Proof of Concept (POC)
- An early implementation used to determine whether a proposed technology, method or application is feasible and useful.
- Operating Experience (OE)
- Information and lessons learned from past operating events that can be used to improve plant performance, reliability and safety.
- Non-Destructive Examination (NDE)
- Inspection methods used to evaluate materials, components or structures without damaging them.
- Probabilistic Risk Assessment (PRA)
- A method used for evaluating the likelihood and potential consequences of failures, events or accidents.
- Nuclear Regulatory Commission (NRC)
- The federal agency responsible for regulating the civilian use of nuclear materials and nuclear power in the United States.
- As Low As Reasonably Achievable (ALARA)
- A radiation safety principle focused on keeping exposure as far below regulatory limits as achievable within possible reason.