Intelligent prediction of grouting in fractured rock masses

Curtain grouting projects are characterized by their large scale and complexity, presenting significant challenges for real-time prediction of grout penetration using traditional methods. This study introduces an intelligent prediction method for grouting in fractured rock masses based on three core principles: integration of multi-source input features, fracture voxel modeling, and shortest path in sequential grouting. Three categories of data (geological structure data, grouting environmental data, and grouting operation data in the concept of a grouting geological model) are integrated and served as multi-source structured data in the intelligent prediction of grouting. A voxelization model quantifies the spatial characteristics of fractures, with voxel size optimized for capturing grouting paths. A shortest path algorithm based on a hierarchical solution is then developed to calculate grout penetration distances in the process of sequential grouting. A complete analysis framework is established, from the voxelization of the fracture network model to precise voxel classification, ultimately achieving an accurate prediction of grout penetration. The method demonstrates excellent performance on the test set, with validation against numerical methods in single-fracture and sequential grouting scenarios confirming its accuracy and prediction efficiency as hundreds of times faster than numerical methods. Application to the Dongzhuang hydraulic project’s grouting test area further validates its effectiveness in multi-hole grouting scenarios. • An intelligent prediction method for grouting in a 3D fracture network is established. • Three principles of intelligent analysis of grouting engineering are proposed. • The excellent performance of the proposed method on the sequential grouting scenarios is demonstrated.