Background Traditional rock mass engineering design largely relies on empirical formulas or the comparison and selection of a limited number of numerical simulation schemes, making it difficult to effectively balance multiple constraints such as safety, economy, and resource recovery rate. The finite-discrete element method (FDEM) based on the intrinsic cohesive zone model (ICZM) can effectively simulate the transition of quasi-brittle materials from continuum to discontinuum and is particularly suitable for highly discontinuous rock engineering problems; it has therefore become an important numerical analysis tool in the field of rock mass engineering. However, there is an urgent need for a high-precision, integrated, and automated parameter optimization method that combines rock mass engineering design with the finite-discrete element approach.

Methodology  To address the technical challenge of selectively inserting zero-thickness cohesive elements at complex geometric interfaces in Abaqus, a two-dimensional zero-thickness cohesive element insertion plugin (ACE) was developed and open-sourced using the Abaqus Python API. This plugin enables global or local selective insertion along any specified mesh edges, thereby establishing an Abaqus-FDEM numerical simulation framework based on the intrinsic cohesive zone model (ICZM). To tackle the low degree of automation in the parameter optimization process for rock mass engineering, a multi-platform integrated, fully automated FDEM parameter optimization approach is proposed. Based on the Isight optimization platform, this method integrates Rhino-Grasshopper parametric modeling, Gmsh adaptive meshing, Abaqus numerical solving, and automated result extraction modules. Driven by Python scripts, it constructs a fully integrated automated optimization workflow covering geometric modeling, mesh generation, cohesive element embedding, finite-discrete element numerical simulation, and macroscopic response extraction.

Results Taking a room-and-pillar mining stope as the engineering background, with the goaf span fixed at 5 m and the pillar width selected as the single design variable, the Nelder-Mead optimization algorithm was employed. With the constraint of satisfying the bearing capacity requirement and the objective of minimizing the pillar width, a single-objective automated optimization was performed. The results show that when the goaf span is 5 m, the minimum pillar width that still ensures the bearing capacity of the pillar is 3.5 m.

Conclusions The reliability of the fully automated parameter optimization method was verified. The research outcomes provide the rock mechanics community with an open-source, free, and easily accessible cohesive element insertion tool for conducting FDEM studies on the Abaqus platform. At the same time, this work offers a new approach for simulation-based parameter optimization in rock mass engineering.