Fast nuclide identification method based on hybrid dynamic Bayesian network

Area of Science:

• Nuclear Physics and Engineering
• Radiation Detection and Measurement
• Data Analysis and Modeling

Background:

• Accurate and efficient nuclide identification is crucial for various applications.
• Traditional methods can be limited in speed and precision, especially in noisy environments.
• Detector output pulse events offer rich information for nuclide characterization.

Purpose of the Study:

• To develop an efficient and precise nuclide identification method.
• To enhance the speed and accuracy of nuclide identification in radiation detection scenarios.
• To validate the proposed method's performance under various noise and nuclide mixture conditions.

Main Methods:

• Hybrid dynamic Bayesian network modeling of detector output pulse events.
• Utilizing Monte Carlo simulations to assess pulse event-to-monoenergetic ray energy correspondence.
• Implementing a probabilistic propagation algorithm for continuous model refinement and a sequential test for enhanced identification.

Main Results:

• Achieved over 91.3% identification accuracy in single-nuclide scenarios with a 1:7 background-to-nuclide intensity ratio.
• Demonstrated high detection rates for specific nuclides (e.g., >81% for 137Cs, ~100% for 60Co) in dual-nuclide scenarios.
• In multi-nuclide scenarios, identification rates reached up to 99.6% with a false alarm rate below 0.8%.

Conclusions:

• The proposed hybrid dynamic Bayesian network method is feasible and effective for rapid nuclide identification.
• The method shows significant potential for improving radiation detection and analysis.
• Continuous refinement and sequential testing enhance performance in realistic detection environments.