COMPARING DIFFERENTE FEED-FORWARD PERCEPTRON ARCHITECTURES IN RADIOACTIVE PARTICLE TRACKING TECHNIQUE

Resumen

Nuclear techniques based on attenuation of gamma radiation are widley used in theindustry to calculate volume fractions, to predict fluid density and track radioactive particle to evaluate industrial units. This work presents a method based on the principles of the radioactive particle tracking (RPT) technique where counts obtained by an array of detectors properly positioned around the unit will be correlated to predict the instantaneous positions occupied by the radioactive particle by means of an appropriate mathematical search location algorithm. Detection geometry developed employs an array of eight NaI(Tl) scintillator detectors, a 137Cs point source with isotropic emission of gamma-rays and a polyvinyl chloride test section filled with air that represents an industrial mixer. The modeling of the detection system is performed by MCNPX code. The aim of this work is to compare multilayer perceptron neural network as a location algorithm in the RPT technique to predict the position of a radioactive particle to evaluate a concrete mixer.