Rapid stability assessment of barrier dams based on the extreme gradient boosting model

Barrier dams are major natural disasters that frequently occur in mountainous areas and have a high probability of destabilizing and failing within a short period. Catastrophic dam failures cause significant economic and ecological damage to downstream areas. Therefore, rapid and accurate assessment of barrier dam stability is crucial for effective emergency rescue efforts. A database was constructed with 1,738 barrier dam cases from 49 countries and regions. A machine learning algorithm was employed for barrier dam stability assessment, and the Sparrow Search Algorithm (SSA) was used to optimize the Gradient Boosting Decision Tree, Random forest, and Extreme Gradient Boosting (XGBoost) respectively. Then, 192 cases with detailed data were selected for training, validation, and comparison of model performance. Among the methods tested, the SSA-XGBoost model performed the best. In addition, the feature importance values of the input parameters were also analyzed. The dam height emerged as the most important factor, with a significantly higher importance value compared to other features. Based on this, 50 barrier dam cases with detailed information were randomly selected for stability assessment, and the results were compared with those of several classical methods. The results indicate that the SSA-XGBoost algorithm can more accurately judge the stability of barrier dams. The absolute accuracy, conservative accuracy, and misjudgment rates were 86%, 94%, and 6%, respectively. These findings demonstrate that the assessment method proposed in this study is reliable and can provide some practical references for the emergency rescue of barrier dams.