BEYOND THE PREDICTION OF CRITICAL FLOOD LEVELS: USING ARTIFICIAL NEURAL NETWORKS FOR FAILURE MECHANISMS BELOW CREST LEVEL

摘要

Dike failure and subsequent flooding may not only be caused by overtopping, but by other mechanisms as well, like e.g. instability of the inner slope and piping. For these mechanisms, ANNs have been devised to obtain a quick and efficient safety evaluation.Flood forecasting and flood warning systems traditionally do not look beyond the crest level of flood defences. However, overtopping is not necessarily the main (potential) cause of flooding. In large part of the Netherlands for example, the traditional focus on sufficient height of the dikes and levees has led to a situation where piping (seepage erosion) and the stability of the embankment, especially under uplift conditions, turn out to be the dominant failure mechanisms, as recent studies by the Dutch government have shown.