Novel Uncertainty Quantification Method for Computational Reactor Design Analysis of Nuclear Thermal Propulsion Cores
Abstract
Nuclear thermal propulsion (NTP) designs include large margins for manufacturing, thermal, and neutronic uncertainties. In the past these uncertainties could be better understood through rapid design and experimental measurements. In the current socio-political climate, rapid prototyping is not possible and work has shifted to computational designs. Larger computing power is now available such that uncertainties can be quantified using computational means. New NTP designs use Monte-Carlo (MC) analysis where well established deterministic uncertainty quantification techniques are not valid. This research develops a new method by extending the MC based Total Monte Carlo (TMC) and GRS uncertainty propagation methods such that they can be used for full-core design efforts that include calculating uncertainties in local quantities such as reaction rates and energy dependent flux and propagating uncertainties to the thermal hydraulics domain. mcACE is a software created in this research to implement the new method for practical applications using V&V’d codes, particularly MCNP. mcACE allows for sampling of any part of an MCNP input from random distributions to determine output uncertainties based on those inputs. Burn-up uncertainty propagation is achieved by coupling MCNP with ORIGEN (version in SCALE 6.1+) while using predictor/corrector methods with sub-steps for time-stepping. Manufacturing uncertainties are captured by sampling MCNP inputs through mcACE based on densities and geometry. Nuclear data uncertainties are captured by sampling continuous nuclear data cross-sections using ASAPy. ASAPy is a software created in this research to sample ACE data files using ENDF covariance information. Sensitivity analysis via variance decomposition for correlated inputs is used to create new insights on correlated sampling for MC neutronics. Calculating traditional sensitivity profiles proved to be out of reach for the selected uncertainty propagation methods though uncertainties can still be readily calculated. To verify uncertainty calculations and sampling methods, eigenvalue uncertainties are compared with iterated fission probability method. The MC uncertainty quantification is extended to the thermal domain by calculating uncertainties in power profiles and reaction rates from uncertain input sources. These uncertain power profiles are used in a hot-channel analysis to calculate performance uncertainties. The developed method is applied to two NTP cores, SNRE, a heritage HEU graphite fueled design and SCCTE, an LEU tungsten-cermet design are analyzed at the unit-cell and full core levels. Burn-up with uncertainty propagation is demonstrated on a typical NTP operation cycle.
Citation
Patel, Vishal K (2019). Novel Uncertainty Quantification Method for Computational Reactor Design Analysis of Nuclear Thermal Propulsion Cores. Doctoral dissertation, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /189154.