ML23200A230
| ML23200A230 | |
| Person / Time | |
|---|---|
| Issue date: | 07/19/2023 |
| From: | Tanya Smith Office of Nuclear Security and Incident Response |
| To: | |
| References | |
| Download: ML23200A230 (1) | |
Text
Putting our Vision into Focus:
NRC Collaboration to Enhance Emergency Preparedness The NRCs vision is to demonstrate the Principles of Good Regulation (independence, openness, efficiency, clarity, and reliability) in performing our safety and security mission. The agency is also working on a vision and strategy for the use of technologies like artificial intelligence and digital twins in meeting our mission. Consistent with this vision, the NRC is embracing innovative approaches, new and diverse ideas, and an environment of collaboration. The Division of Preparedness and Response within the Office of Nuclear Security and Incident Response provides design challenges for Nuclear Engineering students at Purdue University to learn about Emergency Preparedness and Health Physics and develop creative solutions for the use of new technologies in these fields. These design projects benefit the students and the NRC and bring our vision into focus on ways to enhance emergency preparedness.
Design of a Novel Protective Action Decision-Making Tool Optimization of Mask Designs for Dose Reduction The student team combined dose projection outputs from RASCAL with state-of-the art models of shelter The student team optimized a mask designed to reduce inhalation dose. The team also used release data from protection efficiency and evacuation into a user-friendly rapid decision-making aid for protective actions (PRISEM) Fukushima and made use of available NRC RAMP radiation protection codes to quantify the benefit of wearing a mask for a radiological emergency.
RASCAL Dose Projection Outputs Dose Projection Data and Shelter Models Mask efficiency can be calculated based on One-Story Building Cloudshine Protection Factors Example PRISEM (Dickson et. al) output showing known distribution of radioactive particle sizes Vinyl Brick No basement First Floor 0.084 ln(x) + 0.7698 No basement First Floor 0.1167 ln(x) + 0.5897 benefit of sheltering Weighted Average 0.084 ln(x) + 0.7698 Weighted Average 0.1167 ln(x) + 0.5897 over time Basement Basement First Floor 0.0836 ln(x) + 0.7604 First Floor 0.1187 ln(x) + 0.5897 https://ramp.nrc-gateway.gov Basement 0.0871 ln(x) + 0.4442 Basement 0.076 ln(x) + 0.2937 Weighted Average 0.0875 ln(x) + 0.6019 Weighted Average 0.0973 ln(x) + 0.4382 Shelter Ventilation Model Post-Radiological 2 2 2
Incident Shelter-in-Place = 0.16 + 0.25 + 0.4 +
0.0263 +
Wearing a mask
=
+ 1 vs. Evacuation Model 2 2 2 provides a net
= + 10.9 17 2 + 0.25 + 0.4 + 0.0263 +
(PRISEM) benefit to dose
+1 = +
= 1 exp 4 reduction 1
Use of Machine Learning for Predictive Emergency Response Design and Application of Digital Twins to Emergency Response The student team combined probabilistic risk assessment (PRA) with artificial intelligence (AI) to develop a The student team designed a Digital Twin (DT) capable of operating in a fully-automated, real-time, remotely machine learning tool to accurately predict accident release timing based on developing plant conditions. accessible manner. The DT is a modern update to the Response Technical Manual and Response Technical Tool Such a tool could provide decision-makers with advanced warning of a release in time to inform action. which are used for manually estimating core damage states during an emergency based on plant data.
What if we combined PRA with machine learning? Can we achieve real-time accident The student team advanced the state-of-the-art for tools to aid decision-making prediction to inform response?
In this scenario, the AI was able to accurately predict time of release into containment with Digital twins (DT) are virtual good accuracy for the representations of an object entire event time or system that use simulation Z. Dahm, S. Kannappel, W. Kelly. Design of a Machine Learning Tool for Predictive and machine learning to aid Emergency Response, Transactions of the ANS Winter Meeting 2022, Phoenix, AZ, pp. 56-59. decision-making doi.org/10.13182/T127-39694 Response Technical Manual (RTM) Response Technical Tool (RTT) Digital Twin Application of the RTM/RTT (circa 1980s-1990s) (circa 2010s) The DT uses Fuzzy Logic to combine damage state predictions into a best estimate Contact information Todd Smith, PhD Senior Level Advisor (301) 287-3744 todd.smith@nrc.gov