Abstract: The traditional two-dimensional design method for the side and front slopes of tunnel portal projects face some challenges, such as difficulties in quality control, low parameterization and automation, and a lack of spatial relationship. To address these problems, a design scheme for tunnel portal slopes based on parametric, three-dimensional, and intelligent concepts is proposed. Initially, a database of tunnel portal slope design cases is established; further, a deep learning algorithm is employed to intelligently determine the portal boundary mileage and excavation parameters of the tunnel portal slope. Subsequently, a slope transition mode based on a curved surface is introduced, a swarm intelligence algorithm is applied to optimize the selection of controlling transition sections, and excavation control points are estimated to stitch them into a mesh, thereby establishing a slope excavation model. Finally, the integrated slope excavation model and terrain model are subjected to Boolean operations to determine the boundary and quantities of excavation accurately and automatically. A case study conducted on the Changsha-Ganzhou railway using the proposed method reveals that compared with the traditional method, the design scheme obtained via intelligent recommendations exhibits high reliability, a reduction in the excavation boundaries with notable distortion by 28.0%, and an increase in the average accuracy rate of excavation quantities by 8.7%. These results confirm that the proposed method improves the efficiency and quality of the tunnel portal slope design.

Key words: tunnel portal design, side and front slopes, parametric design, intelligentization, deep learning