An Extreme Weather-Related Risk Analysis Model for Embankment Dam: Causal Inference in Historic Data Statistics

The process of dam failure is affected by various uncertain and interrelated risk factors. In order to quantitatively analyze the risk of dam failure, a reasonable method should be used for risk analysis. The Bayesian networks (BNs) has an excellent ability to inferencing event-related potentials. Most previous studies show this method has an inefficient analysis process and a subjective result due to the limitations of relying on domain knowledge. Therefore, this paper develops a Bayesian risk analysis model based on historical data statistics. The network structure was established through causal loop analysis, and the probability parameters were obtained by Bayesian learning, which can effectively improve the reliability of the model. Herein, a risk analysis model has been constructed based on the calculation of correlation factors in historical dam failure events from 1954 to 2021. Based on this model, the probability parameters of different dam failure modes caused by extreme weather have been deduced. According to the results, overtopping and structural instability are highly affected by extreme flood factors.