Ting Zhu

Date of Award


Degree Type


Degree Name

Master of Applied Science (MASc)


Engineering Physics


Adriaan Buijs




In 1984, the ZED-2 research reactor was used to study five (Th, Pu)O2 fuel bundles with the goal to provide both benchmark tests for future reactor code validation and experimental measurements for a possible thorium fuel cycle in CANDU. In this work, the neutronic models of these critical experiments were investigated by TSUNAMI, a sensitivity and uncertainty (S/U) analysis tool, part of the SCALE6 reactor physics package from the Oak Ridge National Laboratory.

TSUNAMI consists of different modules that are capable of calculating the values of keff and the uncertainties in keff due to uncertainties in the nuclear data. It generates energy-dependent sensitivity coefficients from which the percentage change in keff due to perturbations in nuclear data values can be determined. The calculated keff has a bias which is the difference between calculation and measurement. Several sources of uncertainties are responsible for the observed keff bias. But the most dominant and universal contributor is the uncertainties in the nuclear data. Because of the shared nature of nuclear data in all simulations, correlations among the calculated keff values exist and can be measured in terms of the correlations in the nuclear data uncertainties. TSUNAMI provides a consistent and systematic approach to examine these correlations and their effect on keff biases. It also offers an interesting application of the Generalized Linear Least Squares (GLLS) methodology which applies adjustments to the original nuclear data so that keff biases are reduced.

Using TSUNAMI, the computed keff values agree with the experimental keff up to a bias of 2.3mk or lower. It is also found that the list of top uncertainty contributors in keff is identical among the simulations, confirming a high degree of correlation in the nuclear data uncertainties. An illustrative example demonstrated the application of TSUNAMI to coolant void reactivity. In this particular simulation, the coolant density was reduced to mimic the loss of coolant condition. Positive reactivity was induced as the simulation predicted. Through S/U analysis, the uncertainty in the computed reactivity was determined in a similar fashion as the keff uncertainty. Top reactivity uncertainty contributors were also identified. Finally the GLLS procedure was explored using the coolant void example along with the experiments.

Even though the results are not conclusive on the applicability of these (Th, Pu)O2 ZED-2 benchmark experiments to future thorium CANDU reactors, they have called attention to the need for a larger number and better designs of relevant benchmark experiments. It is proposed that they should be designed to maximize the sensitivity information of thorium and plutonium in order to offer more meaningful S/U analysis and establish good confidence in the future investigation of (Th, Pu)O2 fuel in CANDU reactors.

McMaster University Library

Files over 3MB may be slow to open. For best results, right-click and select "save as..."