A bulk analysis system using the prompt gamma neutron activation method and neural network
Nuclear Science and Technology Research Institute (NSTRI), Tehran, Iran
* e-mail: email@example.com
Accepted: 9 May 2017
Published online: 22 June 2017
In this work, an on-line and bulk analysis system based on the prompt gamma neutron activation method and neural network is introduced. Using a setup that includes a 252Cf source and a BGO scintillator detector, a set of semi-experimental data obtained from cement raw materials is produced to train an optimized neural network. The neural network is trained based on a back-propagation algorithm with 100 experimental prompt gamma-ray spectra. The elements existing in the different cement samples are specified. With a good precision compared to the least square analysis, the ANN (Artificial Neural Network) could identify elements. One of the key points in this work is that more than 100 different prompt gamma spectra of neutron activated samples were produced without the need for different cement samples or Monte Carlo simulations.
© Società Italiana di Fisica and Springer-Verlag GmbH Germany, 2017