Matrix effects corrections in prompt gamma-ray spectra of a PGNAA online analyzer system using artificial neural network

Abstract

One of the well-known online monitoring techniques used for quality control of bulk samples is Prompt Gamma Neutron Activation Analysis (PGNAA). PGNAA suffers from the so-called matrix effect problems such as density, thickness and moisture content of the sample under study. In this work, an Artificial Neural Network (ANN) model is introduced to deal with these effects. The required spectra for training and testing the proposed ANN model are obtained by Monte Carlo simulation of the gamma-ray spectra recorded in a PGNAA online analyzer system used in cement factories. The gamma-ray spectra related to given set of density, thickness and moisture content are corrected channel-to-channel using the proposed ANN model. The corrected spectra are then compared against the standard spectra obtained for predefined standard density, thickness and moisture content. The improvement in Theil coefficient is larger than 70% for the tested spectra.