An efficient hybrid stochastic/deterministic coarse mesh neutron transport method

Abstract

A new incident flux response expansion method has been developed to significantly improve the accuracy of the hybrid stochastic/deterministic coarse mesh transport (COMET) method. Additionally, two acceleration techniques are introduced that significantly increase the computational efficiency of the method by several folds. The new expansion method removes singularities associated with the current method that degrade its accuracy and efficiency and ability to solve realistic problems with complexity and size that are inherent in operating commercial reactors. It also enables (paves the way for) the response method to be imbedded in low order transport methods (e.g., diffusion theory) for improving accuracy without degradation in efficiency. In general, the new expansion method also enables efficient and accurate coupling of different deterministic methods (e.g., characteristic to discrete ordinates and in general high order transport to high or low order transport). The new method improvements enable COMET to perform whole-core neutronics analysis in all light and heavy water operating reactors with Monte Carlo fidelity and efficiency that is several orders of magnitude faster than both direct Monte Carlo and fine mesh transport methods. A stylized CANDU-6 core benchmark problem with and without adjuster rods was used to test the accuracy and efficiency of the COMET method in whole (full) core configurations at two coolant states. The benchmark problem consisted of 4560 fuel bundles containing a total of 168,720 fuel pins and 21 adjuster rods. The COMET solutions were compared to direct Monte Carlo (MCNP) reference solutions. It was found that the core eigenvalue, bundle averaged and fuel pin power distributions predicated by COMET agree very well with the MCNP reference solution in all cases when the coarse mesh incident angular flux expansion in the two spatial and two angular (azimuthal and polar) variables is truncated at 4, 4, 2 and 2, respectively. These comparisons indicate that COMET can achieve accuracy comparable to that of the Monte Carlo method with a computational efficiency that is several orders of magnitude better.