Optimization and bearing capacity analysis of truss support structures in deep coal mine roadways

Abstract Large rock deformations and support structure failures are common in deep coal mine excavations. These issues are often caused by high in situ stress, strong disturbances, and complex geological conditions. This study addresses these challenges by optimizing the structure and analyzing the bearing characteristics of the Truss Support System (TSS). A three-dimensional numerical model is developed based on nonlinear rock mechanics theory to simulate the interaction between the surrounding rock and the truss support. A coupled simulation method, combining Finite Element Analysis (FEA) and the Discrete Element Method (DEM), is used to accurately capture the progressive failure of jointed rock masses at depths exceeding one kilometer, as well as the dynamic response of the support system. Topology optimization is then applied to reconfigure the truss’s node layout and member cross-sections for reduced weight. Using a Multi-Objective Genetic Algorithm (MOGA), Pareto-optimal support configurations are identified, balancing three objectives: minimizing material usage, maximizing structural stiffness, and achieving optimal stress distribution. In addition, an intelligent prediction model based on Artificial Neural Network (ANN) is developed to map geomechanical parameters, support geometry, and system bearing capacity, enabling rapid scheme selection and early warning of failure risks. The results show that topology optimization reduces material usage by 18.3%, maximum stress by 26.6%, and roof displacement by 31.6%, while increasing the buckling factor by 45.2%. The medium material usage range yields the best cost-performance ratio. The area under high stress is reduced by 41.2%, and the load-bearing process exhibits a three-stage stiffness pattern: degradation, transition, and recovery. During the local yielding phase, energy dissipation accounts for 62.3%. Roof nodes retain a safety margin of 32.5 mm. The ANN-based model maintains prediction errors below 12% under various working conditions. Overall, this study improves the lightweight design and bearing performance of truss supports in deep coal mines, enhances adaptability to complex geological environments, and provides a theoretical and methodological basis for safe, efficient mining operations.