Nuclear fuel lattice performance analysis by data mining techniques

In this paper a data mining analysis for BWR nuclear fuel lattice performance is shown. In a typical three-dimensional simulation of the reactor operation simulator gives the core performance for a fuel lattice configuration measured by thermal limits, shutdown margin and produced energy. Based on these results we can determine the number of fulfilled parameters of a fuel lattice configuration. It is interesting to establish a relationship between the fuel lattice properties and the number of fulfilled core parameters in steady state reactor operation. So, with this purpose data mining techniques were used. Results indicate that these techniques are able to predict with enough accuracy (greater than 75%) if a given fuel lattice configuration will have a either “good” or “bad” performance according to reactor core simulation. In this way, they could be coupled with an optimization process to discard fuel lattice configurations with poor performance and, in this way accelerates the optimization process. Data mining techniques apply some filter methods to discard those variables with lower influence in the number of core fulfilled parameter. From this situation, it was also possible to identify a set of variables to be used in new optimization codes with different objective functions than those normally used.