Variational methods and speed-up of Monte Carlo perturbation computations for optimal design in nuclear systems

Seeking optimal material distribution in a nuclear system to maximize a response function of interest has been a subject of considerable interest in nuclear engineering. Examples are the optimal fuel distribution in a nuclear reactor core to achieve uniform burnup using minimum critical mass and the use of composite materials with an optimal mix of constituent elements in detection systems and radiation shielding. For such studies, variational methods have been found to be useful but, they have been used for standalone analyses often restricted to idealized models, while more elaborate design studies have required computationally expensive Monte Carlo simulations ill-suited to iterative schemes for optimization. Such an inherent disadvantage of Monte Carlo methods changed with the development of perturbation algorithms but, their efficiency is still dependent on the reference configuration for which a hit-and-trial approach is often used. In the first illustrative example, this paper explores the computational speedup for a bare cylindrical reactor core, achievable by using a variational result to enhance the computational efficiency of Monte Carlo design optimization simulation. In the second example, the effect of non-uniform material density in a fixed-source problem, applicable to optimal moderator and radiation shielding, is presented. While applications of this work are numerous, the objective of this paper is to present preliminary variational results as inputs to elaborate stochastic optimization by Monte Carlo simulation for large and realistic systems.