Georgia Institute of Technology
Real-Time Neutron Multiplicity and Source Localization for Reactivity Monitoring
Abstract
dc:description.abstractNeutron-emitting systems containing fissile material produce statistical signatures that encode both neutron multiplication and spatial configuration. Temporal correlations between detections reflect fission chain dynamics and reactivity, while spatial count gradients across segmented detectors encode source position. Existing methods treat these domains independently: time-correlation techniques estimate multiplication retrospectively under static assumptions, and localization approaches infer position from spatial distributions without incorporating temporal correlation. An integrated real-time framework has not previously been demonstrated. This dissertation presents a proof-of-concept architecture for unified real-time neutron monitoring that combines streaming multiplicity estimation with machine-learning-based spatial localization under explicit statistical control. The system operates on time-tagged list-mode neutron data and is designed to remain bounded, interpretable, and detector-agnostic. The temporal component adapts the Hage–Cifarelli compound Poisson formalism to a sliding-window streaming implementation. Reduced factorial moments are computed continuously across multiple gate widths, and adaptive window control regulates integration length based on statistical uncertainty. Covariance-aware uncertainty propagation and a finite-state controller ensure that low-rate, degraded, or nonstationary conditions are declared explicitly rather than producing unstable estimates. In parallel, a supervised learning pipeline estimates neutron source position from spatial count patterns measured by a 15-tube He-3 detector array. Random forest regression models trained on simulated detector responses are evaluated experimentally. The localization subsystem incorporates count thresholds, hypothesis routing between single- and dual-source models, classifier confidence gating, and out-of-distribution detection. The primary contribution is an episode-based decision architecture in which real-time multiplicity analysis operates continuously and localization is invoked selectively under sustained statistical ambiguity. Experimental validation with the NoMAD detector demonstrates agreement with conventional batch multiplicity calculations within uncertainty and median localization errors below 5 cm under controlled conditions, establishing feasibility of integrated, uncertainty-aware real-time neutron monitoring.
Degree
thesis:*- Name thesis:degree_name
- Nuclear Engineering, PhD
- Department thesis:degree_department
- Mechanical Engineering
- Grantor
- Georgia Institute of Technology
- Year dc:date.issued
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Cannon, Natalie Lauren
- Advisors dc:contributor.advisor
-
- Biegalski, Steven R.
- Erickson, Anna S.
- Committee members dc:contributor.committeemember
-
- Dewji, Shaheen
- Tsvetkov, Pavel
- Thompson, Nicholas
- Hutchinson, Jesson
- ORCID person:identifier.orcid
Rights
- Language dc:language.iso
- English
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/1853/81686