Prediction of critical heat flux using Gaussian process regression and ant colony optimization

Abstract

Critical heat flux(CHF) is a significant parameter that determines the heat transfer capability of nuclear reactors, and therefore prediction of CHF with accuracy is of great importance for the design and safety analysis of nuclear power plants (NPPs). This paper presents a novel hybrid model based on Gaussian process regression (GPR) and ant colony optimization (ACO) for the prediction of CHF. In this model, the ACO algorithm is employed to optimize the hyper-parameters of GPR based on a training set derived from two published literature sources. Prediction of CHF is performed under three conditions: fixed inlet condition, local condition, and fixed outlet condition. The predicted results of the hybrid model are compared to those of support vector regression (SVR). It is shown that this hybrid model is superior to SVR in terms of lower prediction errors. The parametric trends of CHF are also analyzed to evaluate the prediction performance of both models. The predicted values of the hybrid model have a closer agreement with experimental values than SVR. It is observed as well that the presented model can favorably improve the prediction precision of CHF and has excellent potential for other related applications in nuclear engineering.