LN cooling on mechanical properties and fracture characteristics of hot dry granites involving ANN prediction

Abstract

Hot dry rock (HDR) development is essential to reduce carbon emission and achieve renewable energy applications. Liquid nitrogen (LN) fracturing can significantly create complex fractures in HDR. Understanding the damage characteristics of heated granites after LN cooling is essential. This study systematically investigates the mechanical properties of heated granites subjected to LN cooling, followed by a two-scale analysis for fracture morphology variation. Increasing rock temperature and cooling cycle cause more complex microcracks and macro fracture networks, leading to higher fragmentation degrees. The degradation of mechanical properties mainly occurs within the first 10 min and sixth cycles. After the six cycles, the failure model transforms from shear failure to a complex composite failure. The feldspar-to-quartz ratio was first proposed to quantitatively describe the relationship between the mechanical strength and granite types. The linear regression method developed two accurate models with four factors to predict the uniaxial compressive strength (UCS) and tensile strength (UTS). Artificial neural network models can precisely predict the UCS and UTS of various granites under different LN cooling parameters based on a comprehensive data set. It provides a novel method to optimize HDR stimulation based on artificial intelligence.