Abstract In this work, a preconditioner is developed based on the linear coarse mesh finite difference (CMFD) formulation for the flexible generalized minimal residual (FGMRES) algorithm, and applied to accelerate… Click to show full abstract
Abstract In this work, a preconditioner is developed based on the linear coarse mesh finite difference (CMFD) formulation for the flexible generalized minimal residual (FGMRES) algorithm, and applied to accelerate within-group Krylov iterations based on the two-dimensional (2-D) method of characteristics (MOC). The conventional CMFD method is linearized by replacing the multiplicative updating operator with an additive correction operator in the prolongation step. The effectiveness of the linearized CMFD preconditioner for problems featuring steep flux gradients and high scattering ratios can be demonstrated by the numerical results for the IAEA LWR pool reactor problem. The total FGMRES iterations and computing time were decreased by 52.7% and 41.8%, respectively. However, only modest efficiency improvements were achieved for the 2-D C5G7 and the KAIST-2A benchmark problems, revealing the degraded performance of the linearized CMFD preconditioner for problems with strong local heterogeneities.
               
Click one of the above tabs to view related content.