Modelling radium equivalent activity from 226Ra, 232Th, and 40K series of recycled waste materials : analytical and artificial intelligence approaches

Abstract

Primordial radionuclides in the decay sequence beginning with 238U, 232Th, and 40K, as well as cosmic radiation, account for most of the natural radiation in environments and humans. Construction and building materials contain primordial radionuclides. This research predicts the radium equivalent activity (Raeq) from the 226Ra, 232Th, and 40K concentrations of recycled waste materials using the deep neural networks of artificial intelligence. The Levenberg-Marquardt backpropagation technique was used to train the network, which had a three-hidden layer structure and 5–30 neurons in each layer. Predicting the Raeq of recycled waste materials was achieved with high precision using all network architectures. The best performance metrics for training, validation, and testing were demonstrated by a 3-15-15-15-1 network architecture. Furthermore, using untrained data, a robust correlation of 0.9996 R2 was obtained from the model’s confirmation.