Advanced simulation of CANDU Reactors at Low-Cost with Parallel Processing

Abstract

The present work continues a research program on the speeding-up simulation of a Candu reactor by techniques involving parallel computers. In the first phase (reported at the 16th CNS Simulation symposium), the parallelization of NUCIRC had been executed on a first generation parallel machine. It has been pursued with more up-to-date parallel architectures in the phase reported now. More specifically, the thermalhydraulic simulation corresponds to the ITYPE = 4 configuration of the NUCIRC code and involves a representation of the full CANDU loop with only one pass. After a brief analysis of the parallelism inherent to NUCIRC and a technical description of the parallel hardware (set of Transputers from Transtech and a Volvox system) on which the parallelized code is implemented, the improvement in performance is shown and commented upon. Finally, an extension of this approach to more sophisticated parallelism (simulation of transient flows) presently under progress, is outlined.