Blind nonnegative matrix factorization algorithms to estimate the neutron flux of fission chamber detector: Application to neutron-gamma discrimination

In this study, we deal with the neutron flux monitoring inside the TRIGA MARK II reactor as one of the nonnegative matrix factorization problems. The fact that these methods of separating blind sources does not require any assumptions on the way the signal and noise were mixed has encouraged us to apply them to online neutron flux monitoring through neutron-gamma discrimination. We synthesize fission chambers output signals using the Python-based Fission Chamber simulator (pyFC). These signals are then considered as time series mixtures that are analyzed through nonnegative algorithms known as the most popular blind source separation techniques to extract original sources signals. The obtained results show that the Regularized Lee-Sung algorithm is the most appropriate method for discriminating neutrons and gammas. Furthermore, the neutron-gamma discrimination is highlighted through the computation of the cross-correlation functions between the separated source signals and pure neutron and gamma signals. The presented method provides a better characterization of the neutron and gamma signals with very high accuracy.