Methods for optimization of neutron detector performance in nuclear power plants

Abstract

Safety of nuclear power plants (NPPs) requires shutdown systems to mitigate design-basis accidents. Shutdown is based on measurements from in-core flux detectors. Thus, this thesis proposes a methodology for managing neutron-detector aging. Detector performance improvement for prompt fraction along with genetic algorithm optimization, risk estimation, and the fault semantic network (FSN) are studied. Detector models have been developed using circuit models to represent detector behavior in aged conditions. For cases with preprocessed measured data, initial conditions are set to the initial sample value, and an inverse of the transform is applied to reconstruct the signal at the detector. The model’s response is compared to the plant data to validate output. Proposed gains have been implemented in the model with the old gain values to compare with FUELPIN code predictions. A genetic algorithm has been selected to solve optimization and search problems where the Dynamic Signal Compensator (DSC) function is the objective function and the detector function is the fitness function. Two new sets of gain constants have been used to solve the power ramp-up case, allowing reduction of maximum margin-to-trip (MTT) loss to 0.8%. The risk model has been established in the FSN using historical data, and the risk factor for all channels has been identified. The initial channel table and risk prioritization map are updated as information becomes available. This is demonstrated in three case studies where assumed component failure rates were used in a qualitative/quantitative manner for risk estimation. The results suggest the simulated ramp rate with the optimized DSC gain constants is conservative compared to actual fueling and signal oscillation rates. Further gains in detector prompt fraction were obtained by optimization of the amplifier gain settings through genetic algorithm with no adverse impact on operating margins.