Classification of group structures for a multigroup collision probability model using machine learning

Multigroup neutron transport models can be significantly faster than continuous energy methods. While minimizing the number of groups improves runtime, this causes accuracy a calculation to become strongly dependent on selected group boundaries. Machine learning exceed human performance range classification tasks. In work, potential for supervised machine classify few-group structures is evaluated. Artificial neural network and random forest classifiers were trained determine whether given 20-group structure enables multigroup collision probability model calculate accurate multiplication factors in light water reactor lattice simulation. The training data consisted 20,000 their associated factors. Five-fold cross-validation was used optimizing hyperparameters both algorithms. could input with up 95.3% 95.5% respectively.