Prediction of Earthquake-Induced Liquefaction State in Embankment Dams Using Back-Propagation Neural Network

One of the most seismic damaging events in earth dams is a phenomenon called Liquefaction. In this research, a Feed-forward Back-propagation Neural Network (FBNN) model as a classification tool, with one output parameter which represents the occurrence and non-occurrence of earthquake-induced liquefaction state in embankment dams, was developed. A total of 152 embankment dam datasets were considered: 42 dams that liquefied and 110 dams that did not liquefy. The accuracy of the (7-10-1) developed FBNN model is more than 94% for this problem. Finally, the developed model has an adequate accuracy in prediction of the liquefaction state of embankment dams under earthquake loading, and this might prove to be a useful alternative, with distinct advantages over the conventional methods.