Determination of uranium and thorium concentrations in thorium ore sample using artificial neural network and comparison with net area peak method

Abstract

Abstract In order to determine favorable and different elements in soil, the rapid and accurate methods are required. In this research, simultaneous prediction of thorium and uranium in soil samples was performed via gamma spectrometry. Then, the acquired data were analyzed with artificial neural network (ANN) and net area peak (NAP). Natural soil samples obtained from thorium ore consisting of thorium and uranium were used to train models (ANN and NAP). The techniques were evaluated with respect to prediction ability of uranium and thorium concentrations and robustness. Using proposed ANN and NAP methods, the thorium concentration was predicted with mean relative error percentage less than 8.27% and 9.30%, respectively. Also, uranium concentration just was determined with ANN because the NAP method cannot measure uranium concentration. The performance of the neural network model and NAP technique were compared with the acquired empirical data. The results showed that the neural network can more accurately predict the thorium and uranium concentrations in soil samples.