Combination of polynomial chaos and Kriging for reduced-order model of reacting flow applications

The combination of Proper Orthogonal Decomposition (POD) with Kriging has been shown to be a reliable choice for the development Reduced-Order Models (ROMs) prediction combustion data at unexplored operating conditions. In this study, POD is combined Polynomial Chaos Expansion (PCE), PCE and (PC-Kriging) Artificial Neural Networks (ANN) ROM that can predict 2D Non-negative Matrix Factorization (NMF) instead as compression method also investigated. This chosen because it intrinsically guarantee non-violation physical constraints such positivity chemical species mass fractions, although POD's reconstruction errors are lower. performances NMF in proposed supervised methods compared, normalized root mean squared (NRMSE) being less than 10% spatial fields temperature, CH4 O2 all approaches.