Analyzing the impact of natural radioactivity and hazards from granite samples using modified XceptionNet architecture

Granite is a common stone used for construction purposes and it contains diverse amounts of radioactive substances. The radiological hazard and dose parameters associated with granite samples include the measurement of radionuclide concentrations such as potassium 40 ( 40 K), uranium 238( 238 U), and thorium-232 ( 232 Th). The proposed modified XceptionNet architecture analyzes the radioactive levels in the granite samples which affect the health of humans. The modified XceptionNet model is formed by altering the XceptionNet model with different layers and activation functions. The XceptionNet is modified using a dropout regularization parameter which helps the proposed model to perform well even with minimal samples. The experiments are conducted by taking 30 granite samples from south India and evaluated using a gamma ray spectrometer equipped with a high-purity Germanium (HPGe) Radiation Detector. The proposed model accesses different key parameters associated with the safety of granite such as Radium Equivalent Activity (REA), absorbed dose rate, annual effective dose, external and internal hazard rate, and excess lifetime cancer risk. This paper analyzes the potential of deep learning in radioactive hazard prediction and its effect on different organs such as ovaries, testes, bone marrow, and the entire body. This work serves as a reliable deep-learning prediction model to predict the radiological hazards of granite used in construction sites.