![]() |
|
|
|
Journal of Engineering, Project, and Production Management, 2026, 16(7), 2026-0101
Numerical Simulation and Risk Assessment for Geological Engineering Stability in Construction Projects Using Machine Learning
1 Ph.D.
candidate, China University of Petroleum (East China), China
Engineering Management
Received June 15, 2026; revised August 4, 2026; accepted August 10, 2026
Available online November 5, 2026
Abstract: Geological engineering stability is a primary determinant of safety, schedule, and cost performance in large construction projects, yet conventional limit-equilibrium and finite-element procedures struggle to quantify risk under parameter uncertainty and time-varying loading. This study proposes a hybrid Numerical-Simulation and Machine-Learning (NS-ML) framework that couples physics-based Finite-Element Modeling (FEM) of slopes, tunnels, and foundations with a stacked ensemble learner for quantitative stability classification and probabilistic risk assessment. A balanced dataset of 4,812 instances is assembled by augmenting 1,684 published field case histories with 3,128 calibrated FEM realizations spanning slope, tunnel, and shallow-foundation geometries. Six geomechanical descriptors feed the ensemble: unit weight, cohesion, friction angle, slope/overburden ratio, pore‑pressure ratio and seismic coefficient. The base learners are Extreme Gradient Boosting (XGBoost), Random Forest (RF), Light Gradient Boosting Machine (LightGBM), and a Bidirectional Long Short‑Term Memory (BiLSTM) network. These base learners are combined into a stacked ensemble using a logistic regression meta-learner. The proposed NS-ML model attains 95.7% classification accuracy, an F1-score of 0.954, and an AUC of 0.978 on a held-out test set, outperforming the strongest single learner (XGBoost, 92.4%) by 3.3 percentage points. SHAP-based interpretation identifies the friction angle and pore-pressure ratio as dominant risk drivers, while an ablation study quantifies the marginal value of each base learner and the FEM-augmented data partition. A Monte Carlo risk assessment module then converts the ensemble’s probabilistic output into project-level expected loss, providing practical decision support for construction risk management.
Keywords: Geological engineering stability; numerical simulation; machine learning; ensemble learning; construction risk assessment; finite-element method. Copyright © Journal of Engineering, Project, and Production Management (EPPM-Journal). This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License. Requests for reprints and permissions at eppm.journal@gmail.com. Citation: Pan, H. and Faming, G. (2026). Numerical Simulation and Risk Assessment for Geological Engineering Stability in Construction Projects Using Machine Learning. Journal of Engineering, Project, and Production Management, 16(7), 2026-0101.
|