ACCELERATION OF NEUTRON TRANSPORT CALCULATIONS USING KRYLOV SUBSPACE ITERATIONS AND OPTIMALLY DIFFUSIVE COARSE MESH FINITE DIFFERENCE METHOD

Method of characteristics (MOC) is one of the most efficient deterministic technique for high fidelity neutronic analysis of complex and heterogeneous reactor problems. However, the poor convergence speeds of MOC source iteration scheme, especially for problems with large scattering to transport cross-section ratio are well known. This creates a serious hindrance for its effective application to realistic problems. Initially, the optimally diffusive coarse mesh finite difference (odCMFD) method was employed for improving the performance of the transport solver in code DIAMOND, which is an assembly level neutronic analysis code based on the MOC inner-outer formulation and unstructured meshing. It was found that the odCMFD method is highly effective in accelerating the flux convergence by coupling accurate yet computationally intensive high order transport solutions with fast but approximate low order diffusion solutions. Although, the odCMFD method drastically reduces the number of high order MOC sweeps, the use of Gauss-Seidel scheme for solving the odCMFD source iteration substantially piles up the number of low order diffusion sweeps. To overcome this issue, the use of Bi-Conjugate Gradient Stabilized (BiCGSTAB) method, a Krylov subspace iteration scheme, has been examined to further improve the efficiency of odCMFD acceleration. This paper presents implementation of the BiCGSTAB-odCMFD acceleration scheme in code DIAMOND, benchmarking results and its performance analysis.