Using neural networks in earthfill dams emergency planning

Abstract

Safety control of dams is made during the normal exploitation phase, with the support of monitoring data from the observation system and also from the visual inspections information and data, by comparing the real and actual measurements with the values predicted by models of the expected dam behavior. The analysis of abnormal situations obliges to an intervention performed by a dam safety specialist who, facing the data from the observation system and the dam behavior model, will define the correspondent emergency level. This traditional approach, used on a daily basis for assessing dam safety, is adequate, but sometimes, it may delay significantly the actions to restore dam safety standards. In fact, a important time period can occur, between the identification of an abnormal situation in the dam and the definition of the level of seriousness associated, as well as all the subsequent actions. The use of new technologies to help decision support and emergency planning can contribute to mitigate the effects of this disadvantage. The current paper presents a case study concerned with the use of an artificial neural network (ANN) in order to evaluate the behavior of an earthfill dam, Valtorno-Mourão Dam in Portugal. The developed model allowed the identification of both normal and abnormal situations, establishing the correspondent dam alert levels.