Bi-objective optimization of thermal conductivity and thermal stress of UO2-Mo-Nb composite through Gaussian process regression and NSGA-II method

As a promising candidate for accident tolerant fuels, the UO2-Mo composite with continuous Mo channel gains a distinct improvement in the thermal conductivity (TC) of UO2 fuels to enhance the safety and economy of reactors. However, due to a large difference between the coefficient of thermal expansion (CTE) of UO2 matrix and Mo channel, microcracks induced by thermal stress usually occur in the matrix of the UO2-Mo composite with a high Mo content. In this paper, it is proved that adding Nb into Mo channel could significantly modify the CTE, hence reduce the thermal stress of matrix (MTS). However, under the constant proportion of additive materials, increasing Nb simultaneously decreases the TC of the whole composite due to its lower TC than that of Mo. To balance this contradiction and search an optimizing composition range, a bi-objective optimization method for material properties was proposed. Two models for predicting TC and MTS of the UO2-Mo-Nb composite were constructed using Gaussian Process Regression, based on the dataset with uncertainty built by finite element simulation. Leveraging the two models and NSGA-II method, an optimizing composition range of the UO2-Mo-Nb composite was obtained. According to the range, a UO2 composite containing 3 vol% Mo and 1 vol % Nb was fabricated, which exhibited a TC around 30% higher than that of UO2 at 1000 degrees C and no microcracks due to the moderate MTS.