Study on the release pattern of radon exhalation in the overburden soil of uranium tailings under arid climate and prediction based on Fully Connected Neural Network (FCNN)-based deep learning radon prediction model.

Predicting radon exhalation rate under high-temperature and sun-exposure conditions has always been a challenging issue for uranium tailings management units, involving the coupled effects of three important factors: temperature, humidity, and fractures rate. Under arid climate conditions, a series of indoor simulation experiments with different temperatures were conducted on the covering soil of uranium tailings. The Fully Connected Neural Network (FCNN)-based deep learning radon prediction model was proposed, and through error comparison with the Long Short-Term Memory (LSTM) model, the FCNN-based deep learning radon prediction model model demonstrated a better ability to reflect the laws of radon gas release and could more accurately express the relationships between temperature, soil moisture content, the overburden fractures rate, and radon exhalation rate. This paper provides a feasible prediction method for radon control and prevention in uranium tailings.