Modeling nonlinear fractional-order subdiffusive dynamics in nuclear reactor with artificial neural networks

Area of Science:

• Nuclear Engineering
• Computational Physics
• Artificial Intelligence

Background:

• Nuclear reactor dynamics are often modeled using complex fractional-order (FO) equations.
• Solving these nonlinear FO models analytically or numerically presents significant challenges.
• Neutron transport is frequently simplified as a subdiffusion process in reactor modeling.

Purpose of the Study:

• To develop and analyze artificial neural network (ANN) models for various nonlinear fractional-order nuclear reactor models.
• To provide a robust method for capturing the transient and steady-state dynamics of these complex systems.
• To demonstrate the efficacy of ANNs in representing nonlinear subdiffusive neutron transport.

Main Methods:

• Development of ANN models trained on data generated from nonlinear fractional-order point reactor kinetics, FO Nordheim-Fuchs, inverse FO point reactor kinetics, and FO constant delayed neutron production rate approximation models.
• Iterative ANN learning process involving adjustments to network layers and neuron counts.
• Extensive simulation studies to validate model performance.

Main Results:

• The developed ANN models demonstrated a faithful representation of both transient and steady-state dynamics.
• ANNs effectively captured the nonlinear behavior inherent in the fractional-order models.
• Satisfactory representation of the nonlinear subdiffusive process of neutron transport was achieved.

Conclusions:

• ANNs offer a powerful and accurate tool for modeling complex nuclear reactor dynamics described by nonlinear fractional-order equations.
• The developed ANN models provide a viable alternative for simulating challenging reactor physics scenarios.
• This approach facilitates a better understanding and prediction of reactor behavior under subdiffusive conditions.