Optimization of subchannel simulations by using genetics algorithms

The channel of a nuclear power reactor contains the fuel bundles which are made up of fuel elements distributed in a manner that creates a series of interconnected subchannels through which the coolant flows. Therefore, the 'Subchannel' modeling is a quite well technique used to predict local flow variables in nuclear fuel channels. In general, these models require the use of correlations for calculating mixing effect across the lateral interconnected subchannels. However, there is no unique set of correlations that can be used to predict a complete rage of experimental conditions. In order to avoid this major drawback, in this paper the coupling of a Genetic Algorithm (GA) to a subchannel model is presented. The use of a GA in conjunction with an appropriate objective function allows the subchannel model to internally determine the optimal values of the coefficients required by the lateral mixing correlations, without user intervention. The predictions obtained by using the coupled model were compared with experimental data collected in two interconnected subchannels. The model was able to predict the experimental trends without using external tuning coefficients. (author)