$\gamma$ spectrum-dose conversion algorithm using artificial neural networks for spectroscopic EPD

Abstract

This thesis study is about an algorithm for the direct calculation of the ambient dose equivalent H$\ast$(10) from the gamma spectra of a spectroscopic electronic personal dosimeter (EPD) using an artificial neural network (ANN). Previous conventional algorithms of spectrum-to-dose conversion have limitations in terms of several aspects including low accuracy in energy range below ~ 300 keV. It leads to need a new approach for spectrum-to-ambient dose equivalent conversion. Therefore, the ANN based dose conversion model is proposed. The proposed ANN model was trained with the simulated spectra of a CsI(Tl)-PIN diode based EPD using Monte Carlo N-particle transport code 6. The results showed that H$\ast$(10) calculated using the ANN is highly consistent with the theoretical H$\ast$(10) in simulation test and is comparable performance with G(E) function method in experimental test.