Comparison of acceleration schemes for parallel S{sub N} algorithm solution of shielding problems

Journal Article · Tue Dec 31 04:00:00 EST 1996 · Transactions of the American Nuclear Society

OSTI ID:436905

Hunter, M A; Haghighat, A ^[1]

---

1. Pennsylvania State Univ., University Park, PA (United States)

In recent years, parallel discrete ordinates algorithms based on domain decomposition have been developed to decrease the running time of neutron transport calculations. Previously, acceleration techniques such as diffusion synthetic acceleration (DSA) and coarse-mesh rebalance (CMR) were used to accelerate the convergence of these calculations and thus decrease the CPU time. Both methods are based on using lower order equations to find improved fluxes to use in the next transport iteration. Diffusion synthetic acceleration was originally developed to accelerate convergence of thermal neutron groups (with high c = {sigma}{sub s}/{sigma}{sub t}) and eigenvalue problems; however, the method has been used as an acceleration technique for all types of problems. In the past, we have always used CMR acceleration with our parallel S{sub N} algorithm development because the CPU time spent performing the CMR calculations is insignificant compared to the time spent performing the transport calculations. In this study, we compare these two acceleration schemes based on problem convergence and CPU time.

• OSTI ID: 436905
• Report Number(s): CONF-9606116--
• Journal Information: Transactions of the American Nuclear Society, Journal Name: Transactions of the American Nuclear Society Vol. 74; ISSN 0003-018X; ISSN TANSAO
• Country of Publication: United States
• Language: English

Related Subjects

66 PHYSICS
99 GENERAL AND MISCELLANEOUS
ALGORITHMS
COMPUTER CALCULATIONS
CONVERGENCE
DISCRETE ORDINATE METHOD
EFFICIENCY
NEUTRON TRANSPORT
SHIELDING