Study of dam safety and determination of risk scenarios using intelligent systems.

This project belongs to a research line that aims the definition of criteria for using intelligent systems for the safety analysis of earthfill dams and for serving as a tool for decision-making. In this project, the main goal was to stablish a methodology to reach that general purpose using neural networks.

Seprisis project demonstrated the applicability of neural networks to reach a more consistent interpretation of the data registered from different sensor families, and a series of interesting studies resulted from it: (I) displacement prediction (plumblines), (II) seepage prediction and (III) signal-noise analysis.

(I) For the analysis of radial displacements at different heights of the dam, increasing complexity models were posed. Most simple model considered only parameters related with reservoir water level and average temperatures. Other more complex models also considered previous values of the displacements analyzed (short-term models).

(II) The prediction of the seepage values required a deeper study on the relevance of different variables using neural networks.

(III) The influence of the noise (measuring errors) in test and validation sets in neural networks was studied using a model based on a gravity dam without joint injection.

Participants:

  This R&D project has been funded by: