Scholarlink

<p>Proactive slope stability management in open-pit mines requires real-time, accurate forecasting of failure risk. Traditional methods lack the adaptive capability to process continuous IoT sensor data for predictive early warning. The primary contribution of this study is the development and validation of a novel IoT-enabled hybrid machine learning digital twin (DT) model, formalised as DT = ⟨M, D, U, P⟩, for real-time prediction of the factor of safety (FoS) and total displacement at the Pulang Copper Mine, Yunnan Province, China, which serves as a geotechnically complex validation testbed. A digital twin integrating IoT sensor networks with a hybrid ensemble model (Random Forest+XGBoost) was deployed across five monitoring stations at the Pulang Copper Mine. The model utilized 43 engineered temporal and statistical features. Hyperparameters were optimized via Bayesian Optimization, and model interpretability was achieved using SHAP analysis. For FoS, the model attained an average test R² (0.975), with station-specific ranging (0.968–0.986) and MAE (0.0023–0.0080&#xa0;mm). Total displacement prediction yielded test R² (0.978–0.990) for moderate deformations. Multiclass FoS safety classification achieved testing accuracy (97.5–100%) and macro F1 scores (0.908–1.000). SHAP analysis revealed adaptive feature importance, with FoS lag-1 dominant in stable regimes and longer-term rolling means critical in near-failure conditions. The model successfully transforms slope stability monitoring from reactive observation to proactive risk management. The study validates a practical, scalable solution for real-time geotechnical risk assessment in critical infrastructure.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Factor of safety prediction for open-pit slope using IoT-enabled hybrid machine learning digital twin with real-time adaptive updating

  • 期刊论文

  • 发表日期：

    2026年7月27日

查看全文
  • Ibrahim Haruna Umar,
  • Hang Lin,
  • Chaoyi Yang,
  • Abdullah Alzouba

摘要

Proactive slope stability management in open-pit mines requires real-time, accurate forecasting of failure risk. Traditional methods lack the adaptive capability to process continuous IoT sensor data for predictive early warning. The primary contribution of this study is the development and validation of a novel IoT-enabled hybrid machine learning digital twin (DT) model, formalised as DT = ⟨M, D, U, P⟩, for real-time prediction of the factor of safety (FoS) and total displacement at the Pulang Copper Mine, Yunnan Province, China, which serves as a geotechnically complex validation testbed. A digital twin integrating IoT sensor networks with a hybrid ensemble model (Random Forest+XGBoost) was deployed across five monitoring stations at the Pulang Copper Mine. The model utilized 43 engineered temporal and statistical features. Hyperparameters were optimized via Bayesian Optimization, and model interpretability was achieved using SHAP analysis. For FoS, the model attained an average test R² (0.975), with station-specific ranging (0.968–0.986) and MAE (0.0023–0.0080 mm). Total displacement prediction yielded test R² (0.978–0.990) for moderate deformations. Multiclass FoS safety classification achieved testing accuracy (97.5–100%) and macro F1 scores (0.908–1.000). SHAP analysis revealed adaptive feature importance, with FoS lag-1 dominant in stable regimes and longer-term rolling means critical in near-failure conditions. The model successfully transforms slope stability monitoring from reactive observation to proactive risk management. The study validates a practical, scalable solution for real-time geotechnical risk assessment in critical infrastructure.