Abstract
Modular Pebble-bed HTGR (High-Temperature Gascooled Reactor) employs online HPGe (High Purity Germanium) detector systems to measure the nuclide inventory in spherical fuel elements. The determination of the detector’s DE (Detection Efficiency) curve involves Monte Carlo modeling of the detection system. However, the actual dimension parameters of the HPGe detectors often deviate from the nominal values provided by the manufacturer due to manufacturing limitations and accumulated dead layer growth. Therefore, the calculated DE is subject to systematic errors when the detector model is constructed according to nominal values in Monte Carlo simulations.
Focusing on the HTR-10 online burnup measurement system, this work first performs a sensitivity analysis of the DE curve to relevant dimension parameters using Monte Carlo method and analyzes the effect of each variable on the DE profile between 344 and 1408 keV. Then four meta-heuristic algorithms including GA (Genetic Algorithm), DE (Differential Evolution), PSO (Particle Swarm Optimization) and SA (Simulated Annealing) are incorporated to find the optimal modeling parameter combination that minimizes the discrepancy between the measured and calculated DEs. In addition, different algorithms are compared in terms of accuracy and convergence rates to assess the applicability of meta-heuristic algorithms to such detector parameter optimization problems.
Sensitivity analysis reveals that the side dead layer thickness, germanium crystal radius, crystal height, copper radius, and source-detector distance are five key parameters in the determination of DE profiles with different energy selectivity characteristics. After the optimization, the difference between the calculated and measured DEs is reduced to less than or equal to the measurement uncertainty of the measured DEs. Therefore the optimized model can be applied to the efficiency calibration of the above-mentioned detector system to improve its measurement accuracy. Finally, three group intelligence algorithms present satisfactory optimization capability and convergence rate, and the method developed in this paper can be extended to other HPGe systems.
