Nuclear Reactor DBNR Calculation Using Deep Neural Networks

The patent replaces the traditional simplified thermo-hydraulic modeling approach with a neural network-based computational system. The neural network is trained offline using detailed 3D thermo-hydraulic reference code data, then deployed for rapid online DBNR calculations. This substitution maintains calculation speed while achieving precision comparable to the complex reference code, eliminating the need for large safety margins that were previously required due to simplified model inaccuracies.

The patent performs the complex computational work in advance by training a neural network offline using extensive 3D thermo-hydraulic reference code simulations. During actual reactor operation, the pre-trained neural network provides rapid DBNR calculations without requiring access to the complex reference code. This preliminary action transfers the computational burden from online operation to offline preparation, achieving high precision while maintaining simple real-time operations.

The patent implements a feedback mechanism where the neural network continuously monitors actual reactor conditions and adjusts DBNR calculations in real-time based on measured parameters such as neutron flux, coolant flow rate, and power distribution. This feedback loop enables precise determination of actual safety margins, allowing operators to optimize reactor control by avoiding excessive conservative margins while maintaining adequate safety levels through accurate, condition-specific calculations.