Intelligent Design and Evaluation of Roadway Support Based on Unsupervised Machine Learning: A Case Study of the 171106 Roadway in LiuZhuang Mine

The design of support for mining roadways often heavily relies on the expertise of engineering technicians, which can result in inadequate control over the surrounding rock, overly conservative support parameters. Using the 171106 panel of LiuZhuang Mine as the study context, this study optimized an unsupervised machine‐learning algorithm, built a machine‐learning–based evaluation model for gateroad surrounding rock, developed an intelligent decision‐support system for roadway support, and optimized support design parameters; the feasibility of the scheme was cross‐validated through numerical simulation and mine pressure monitoring. The results show that, after optimization, the unsupervised algorithm achieved 92% accuracy in surrounding‐rock classification. Applied at the working face, the dynamic classification indicated Class III surrounding rock; leveraging the intelligent decision system, a support cross‐section was generated, and an optimized support design was produced. Compared with the original scheme in numerical simulations, the optimized design reduced left‐rib displacement by 21%, surrounding‐rock stress by 9.1%, roof displacement by 10%, and shrank the plastic zone by 1 m. Field measurements further showed that roof bolt stress decreased by 8.98%, rib bolt stress by 7.7%, roof separation by 16%, roof displacement by 25%, and rib displacement by 21%. These findings verify the feasibility and scientific soundness of the optimized support scheme and provided a valuable reference model for the optimization of support systems in similar mining environments.