To investigate the mechanical properties of fractured dolomite, this study analyzed the fracture characteristics (dip angle, length, position, quantity) using the Pearson coefficient and MIC coefficient. Subsequently, the data pertaining to fracture characteristics is preprocessed using a third-degree polynomial, and a three-classification strategy is implemented to improve the logistic regression algorithm to establish the strength prediction model of fractured dolomite. Furthermore, the significance order of the impact of fracture characteristics on rock strength was determined using the numerical simulation software PFC3D, and the dip angle effect was explained from the perspective of internal fracture propagation within the rock. The results show that: (1) When the regularization coefficient λ = 10,000, the algorithm has the highest prediction accuracy and the strongest model generalization ability. (2) The numerical simulation analysis software PFC3D can accurately invert rock failure process and characteristics, and the order of influence of fracture characteristics on rock strength is dip angle > length > position.
© 2023. Springer Nature Limited.