Nuclear reactors safety core parameters prediction using Artificial Neural Networks

The present work investigates an appropriate algorithm based on Multilayer Perceptron Neural Network (MPNN), Apriori association rules and Particle Swarm Optimization (PSO) models for predicting two significant core safety parameters; the multiplication factor Keff and the power peaking factor Pmax of the benchmark 10 MW IAEA LEU research reactor. It provides a comprehensive analytic method for establishing an Artificial Neural Network (ANN) with self-organizing architecture by finding an optimal number of hidden layers and their neurons, a less number of effective features of data set and the most appropriate topology for internal connections. The performance of the proposed algorithm is evaluated using the 2-Dimensional neutronic diffusion code MUDICO-2D to obtain the data required for the training of the neural networks. Experimental results demonstrate the effectiveness and the notability of the proposed algorithm comparing with Trainlm-LM, quasi-Newton (Trainbfg-BFGS), and Resilient Propagation (trainrp-RPROP) algorithms.