In this master thesis a validation and evaluation of the recently released ADVANTG code, which combines a well-known Monte Carlo (MC) transport code MCNP with a deterministic transport code Denovo, is presented. The aim of ADVANTG is to automate the process of generating the variance reduction parameters (weight-windows) for the fixed source MCNP calculations, which consequently accelerate the simulations in terms of the required CPU time.
Reliability and consistent performance of the ADVANTG code were tested on two computationally demanding benchmarks from the ICSBEP handbook: the ''Labyrinth'' experiment where neutron fluxes were measured and calculated in a three-section concrete labyrinth for a \ce{^{252}Cf} neutron source and the ''Skyshine'' experiment where neutron and photon scattering in the air above an open operating reactor are simulated. The efficiency of each simulation has been obtained by the statistical test FOM. Compared to the analog MCNP simulations, the speed-up factors or the increases in relative efficiency of up to 1400 (neutrons), for the ''Labyrinth'' case, and up to 30000 (neutrons) and 1400 (photons), for the ''Skyshine'' case, were achieved using the ADVANTG-generated variance reduction parameters. As the mean values obtained by the ADVANTG-accelerated simulation sit within the statistical uncertainties of the analog simulation for both cases, it was shown that no additional bias is introduced by the ADVANTG code.
Because of its reliability and consistent performance the ADVANTG code was used to analyse $\gamma$ dose rates from the future Slovenian silo type low and intermediate level waste (LILW) repository. Using ADVANTGs variance reduction technique different detailed calculations of $\gamma$ dose rates were once again achievable in reasonable time. Contrary, all previous calculations used rough geometric models or other different kind of approximation, which did not take into account the effect of backscattering in the air properly. For the optimized ADVANTG settings the maximum speed-ups up to 10000 were achieved compared to the analog MCNP simulation. Within the framework of the study we analysed the influence of different repository configurations on the values of the annual photon dose rates for different measurement positions on the surface. For all configurations of the repository, with the exception of using LILW from NPP Krško, the annual photon dose rates, due to $\gamma$ rays, were for all measurement positions lower than the prescribed limit values (worker: 20 mSv/year). For the latter, an additional analysis was made, where the concrete N2b disposals, in which the LILW is stored, were modelled in more detailed. To ensure that the photon dose rates are lower than the prescribed limit values, the required thickness of the additional grout was determined by varying the additional quantity of the grout.
The ADVANTG code has proven to be a powerful and reliable tool for generating effective variance reduction parameters and to greatly accelerate analog MCNP simulations. Therefore, in even more complex cases, it will be necessary or the only option for simulations with the statistically relevant results.
|
