A multi-scale 3D geological modeling method for mountain tunnels: Integrating geological genesis constraints with multi-source data

The construction of complex mountain tunnels is challenged by geological data scarcity, highly heterogeneous rock mass conditions, and scale variations in multi-source data. Conventional modeling relies on mathematical interpolation or pure data-driven methods, resulting in models that may not conform to geological formation mechanisms. This study presents a novel multi-scale 3D geological modeling framework that integrates geological genesis constraints with multi-source data fusion. At the regional scale, stratigraphic interfaces are reconstructed using a zonal interpolation strategy combining Kriging with variable-order B-spline surface interpolation. An anisotropic ellipsoidal search algorithm preserves the continuity of steeply dipping strata, while complex fault systems are restored using the balanced section method. At the engineering scale, a structured voxel grid is generated using volume-proportion weighting and cubic convolution interpolation. At the outcrop scale, high-resolution Light Detection and Ranging (LiDAR) point cloud is integrated with tunnel face mapping, employing an improved Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm to accurately extract joint networks. Model integration is achieved through spatial datum unification, information priority coordination mechanism, and hierarchical topology fusion that resolves cross-scale conflicts. Application to a complex plateau railway tunnel demonstrates that the regional scale model achieves a geometric accuracy of 87.0% and a Kappa coefficient of 0.82, significantly outperforming conventional methods. Additionally, the model was validated against excavation-revealed tunnel-face records, achieving an average lithology prediction accuracy of 92.68%. This framework generates reliable geological visualizations to support risk assessment and construction optimization.