USE OF GAMMA RADIATION AND DEEP NEURAL NETWORKS TO MONITOR PETROLEUM TRANSPORT BY PRODUCTS THROUGH PIPELINES

Resumo

This report presents a methodology for identifying the interface region of petroleum by-products during transportation in polyducts using an artificial neural network (ANN) to improve the accuracy of material purity determination. The setup includes a 662 keV gamma-ray source and two NaI(Tl) detectors to measure transmitted and scattered beams. Steel ducts with radii from 4 to 10 inches were investigated. Data from the detectors were used for ANN training, with the entire pulse height distribution (PHD) as input. Models for a stratified flow regime were developed using the MCNPX code, generating datasets for training and test for ANN with different purity levels for the biphasic oil-gasoline system. The mean relative error was 2.22%, demonstrating the ANN's capability to accurately predict purity levels up to 99%.