An AI ‐ MCDA Framework for Mitigating Geohazard Risks in Energy Infrastructure: Integrating BeiDou Geospatial Data for Sustainable Trenchless Pipe Jacking

This study develops an integrated AI‐MCDA framework for mitigating geohazard risks in sustainable trenchless pipe‐jacking planning for energy infrastructure. The framework combines BeiDou‐enabled geospatial intelligence, explainable machine learning, and fuzzy multi‐criteria decision analysis to support route‐mitigation selection under subsurface uncertainty. First, an XGBoost model is used to estimate segment‐level geohazard probability from geological, hydrogeological, terrain, monitoring, and engineering variables, while SHAP is applied to interpret the main drivers of risk. The model achieved an AUC of 0.807, with recall of 0.708 and an F1‐score of 0.607, suggesting usable performance for preliminary geohazard screening. Second, fuzzy DEMATEL and DANP are employed to identify causal interdependencies among geohazard, engineering, sustainability, and resilience criteria and to derive influential weights. Third, fuzzy VIKOR is used to rank competing route‐mitigation bundles, with fuzzy TOPSIS and scenario analysis applied for robustness checks. The findings show that groundwater pressure, slope instability, monitorability, and jacking‐force demand are decisive factors. A1, the Northern Bypass + Enhanced Lubrication bundle, emerged as the preferred alternative with the lowest VIKOR compromise index (( = 0.0300)) and the highest TOPSIS closeness coefficient (0.762), while remaining stable under the tested scenario conditions. The study offers a transparent decision‐support framework for climate‐aware underground energy infrastructure planning.