Bayesian network safety risk analysis for the dam–foundation system using Monte Carlo simulation

Abstract

Gravity dams are mostly located in alpine and gorge regions with complex terrain and geological conditions, facing challenges of large peak flows, developed weak structural planes, high-stress levels, etc. The determination of failure risks of different failure modes in gravity dams is closely related to structural characteristics, materials properties, and loading conditions. However, the failure risks in the traditional risk analysis methods are usually calculated based on the expert evaluation methods and statistical data, ignoring the uniqueness of dam characteristics, and leading to an unreasonable risk assessment. Therefore, an improved probabilistic risk analysis method for the dam–foundation system using Bayesian network (BN) and Monte Carlo simulation is proposed in this study, which considers the combined action of overtopping, foundation stability, and structural failure. Based on the Monte-Carlo method, the risk analysis of overtopping is proposed by considering the influences of peak flows and maximum wind speeds, and the dam–foundation system is developed by establishing response surface equations of performance functions. Then, the BN safety risk analysis method for gravity dams is formed. This improved method is applied in the GD gravity dam project and finds that the system risk is 2.10 × 10−3. The most critical failure factor is the foundation instability, and the importance and sensitivity analysis also proved it, implying the validity and reasonability of the proposed method. Thus, the BN safety risk analysis method for gravity dams enables an improved and more reliable estimate of dam risk.