Development of optimal framework SARAX/DAKOTA for multiple objective optimization of fast reactor and its application

The neutronic design of fast reactor (FR) considers complex objectives and constraints but the operating experience of FRs is much lower than that of commercial reactors. It is complicated and time-consuming to perform optimization design manually. This study applies Genetic Algorithm (GA) and modifies the search method in coding strategy. Distinguished from previous research, a new conception of coding that maps continuous decimal integers into a loading pattern and searches using uniform sampling is developed. This coding strategy realizes the coverage of the multidimensional search space, so that it is possible to deal with the problems of continuous and discrete variables. With this coding strategy, initial designs are generated uniformly and automatically, which leads to better global convergence. Multiple design variables, objectives, and constraints are considered to cater to the requirements of practical application. With the above coding strategy and considerations, a framework that couples the SARAX neutronics analysis code and the GA module in the DAKOTA toolkit is constructed and tested on several cases for verification. It is easy to add any of the design variables, optimization objects, objectives and constraints into the framework and set the optimal range. Both the neutronic software and optimization have been fully validated and performed as a black box. These features are beneficial to the design of FRs, which is lacking of experience. Last but not the least, is a demonstration of this framework in which Pu238 production is optimized based on CEFR. The searching result showed the excellent properties of this work.