Efficient random sampling method based on the control variates method and sensitivities of multiple mockup parameters

The CV-S method is one of efficient sampling-based uncertainty quantification methods. In order to enhance the efficiency of the CV-S method, we propose to use multiple mockup parameters instead of a single parameter. Before doing that, we address an issue that the efficiency of the CV-S method is case-dependent, and introduce a concept of dimensionless parameters. By doing this, the case-dependent efficiency of the original CV-S method disappears. One fictitious mockup parameter which combines multiple mockup parameters is prepared. Proper weights in this combination are dynamically determined from the finite number of samples. Test calculations are carried out for k∞ of a BWR assembly model during burnup using k∞ of several pincell models composing the target assembly as mockup parameters. It is clearly demonstrated that the efficiency is enhanced by using multiple mockup parameters as much as possible.