A regression-based approach to the prediction of crest settlement of embankment dams under earthquake shaking

Abstract

The settlement of embankment dams is among the many major damages caused by earthquakes that, eventually, leads to dam instability. Therefore, an accurate assessment of the seismic settlement of embankment dams is of particular concern. This study aims to evaluate the settlement of embankment dams subjected to earthquake loads using regression-based methods. wide-ranging cases of real data on crest settlement of embankment dams caused by earthquakes were analyzed. Yield acceleration of dam (ay), maximum horizontal earthquake acceleration (amax), fundamental period of dam body (Td), predominant period of earthquake (Tp), and earthquake magnitude (Mw) were considered as the most inuential parameters that a_ect the seismic crest settlement of embankment dams. By applying Support Vector Regression (SVR) and Multiple Linear Regression (MLR) methods, two models were developed to estimate the earthquake-induced settlement of embankment dams. Subsequently, sensitivity analysis was conducted in order to assess the behavior of the proposed models under di_erent conditions. Finally, the accuracy of the proposed models was compared with the existing relationship for the estimation of earthquake-induced crest settlement of embankment dams. Although both MLR- and SVR-based models enjoy acceptable accuracy in the estimation of the crest settlement of embankment dams under earthquake loading, the SVR-based model has higher accuracy.