Prediction of Sandstone Dilatancy Point in Different Water Contents Using Infrared Radiation Characteristic: Experimental and Machine Learning Approaches

Abstract In rock mechanics, the dilatancy point is always occurring before failure during loading process. Water content plays a significant role in physiomechanical properties, which also impact under This significantly warning engineering structures stability. Therefore, it essential to predict different water contents get an early for effective monitoring of projects. study investigates effects on sandstone presence infrared radiation (IR). Furthermore, this IR was used first time as input parameter artificial intelligence (AI) techniques stress-strain curve. The experimental findings show that stress range curve stages (crack closure and unstable crack propagation) increases with content. However, deformation stable propagation decreases stress, initiation elastic modulus are negatively linearly correlated, while peak level quadraticaly correlated high (R2). absolute strain energy rate, gives sudden increase at dilatancy, index. 0.86 σmax dry index predicted from data using three computing techniques: neural network (ANN), random forest regression (RFR), k-nearest neighbor (KNN). performance all evaluated R2 root-means-square error (RMSE). results models satisfactory performances all, but KNN remarkable. research will be helpful provide guidelines about underground project stability evaluation environments.