Study of Non-Intrusive Model Order Reduction of Neutron Transport Problems

Model order reduction is one important technique to speed up the solution of large-scale problems. Two model order reduction techniques, i.e., the proper orthogonal decomposition with interpolation and the Gaussian process regression, are investigated to compute the neutron flux and the effective multiplication factor. Two reduced order models are constructed and it is shown that a speedup of 3 similar to 4 orders of magnitude is achieved with guaranteed accuracy. The Gaussian process regression model can predict the mean value of keff and more importantly, the uncertainty of the predicted value can be analyzed as well. (C) 2021 Elsevier Ltd. All rights reserved.