Key Impact Factors on Dam Break Fatalities

Dam failures can lead to catastrophes on human society. However, there is a lack of research about dam break fatalities, especially on the key factors that affect fatalities. Based on the analysis of historical dam break cases, most studies have used the regression analysis to explore the correlation between those factors and fatalities, but without implementing optimization to find the dominating factors. In order to understand and reduce the risk of fatalities, this study has proposed a new method to select the impact factors on the fatality. It employs an improved ANN (Artificial Neural Network) combined with LOOCV (Leave-one-out cross-validation) and SFS (Stepwise Forward Selection) approach to explore the nonlinear relationship between impact factors and life losses. It not only considers the factors that have been widely used in the literature but also introduces new factors closely involved with fatalities. Dam break cases occurred in China from 1954 to 2013 are summarized, within which twenty-five cases are selected with a comprehensive coverage of geographic position and temporal variation. Twelve impact factors are taken into account as the inputs, i.e., severity of dam break flood (SF), population at risk (PR), public understanding of dam break (UB), warning time (TW), evacuation condition (EC), weather condition during dam break (WB), dam break mode (MB), water storage (SW), building vulnerability (VB), dam break time (TB), average distance from the affected area to the dam (DD) and preventive measures by government (PG).From those, three key factors of SF, MB and TB are chosen. The proposed method is able to extract the key factors, and the derived fatality model performs well in various types of dam break conditions.