Slope rockbolting using key block theory: Force transfer and artificial intelligence-assisted multi-objective optimisation

This paper presents a novel artificial intelligence (AI)-assisted two-stage method for optimising rock slope stability by integrating advanced 3D modelling with rock support design, aiming at minimising risks, material usage, and costs. In the first stage, an extended key block analysis identifies key blocks and key block groups, accounting for progressive failure and force interactions. The second stage uses AI algorithms to optimise rockbolting design, balancing stability, cost, and material use. The most efficient algorithms include the multi-objective tree-structured Parzen estimator (MOTPE) and non-dominated sorting genetic algorithms (NSGA-II and NSGA-III). Applied to the Larvik rock slope, the optimised solution uses 18 pre-tensioned cablebolts, providing 13.2 MN of active force and achieving a factor of safety of 1.31 while reducing the average anchorage length by approximately 16% compared to traditional design. The AI-assisted approach also reduces computation time by over 90% compared to Quasi-Monte Carlo (QMC) methods, demonstrating its efficiency for small-scale civil engineering projects and large-scale mining operations. The developed tool is practical, compatible with Building Information Modelling (BIM), and ready for engineering implementation, supporting sustainable and cost-effective rock slope stabilisation. While the method is largely automated, professional judgement remains crucial for verifying ground conditions and selecting the final solution. Future work will focus on integrating data uncertainties, addressing complex block deformation mechanisms, refining optimisation objectives, and improving the performance of multi-objective optimisation for slope rockboling applications to further enhance the method's versatility.