Selecting time series models for predicting concrete dam foundation uplift pressure

  • Sheila R. Oro Programa de Pós-Graduação em Engenharia de Automação e Sistemas, Universidade Federal de Santa Catarina, SC, Programa de Pós-Graduação em Engenharia Ambiental: Análise e Tecnologia Ambiental, Universidade Tecnológica Federal do Paraná, PR
  • Matheus H. D. M. Ribeiro Programa de Pós-Graduação em Engenharia de Produção e Sistemas, Universidade Tecnológica Federal do Paraná, PR
  • Ubirajara F. Moreno Departamento de Automação e Sistemas, Universidade Federal de Santa Catarina, SC
  • Nestor Roqueiro Departamento de Automação e Sistemas, Universidade Federal de Santa Catarina, SC
  • Edivaldo J. Silva Junior Estrutura de Barragens, Parque Tecnológico Itaipu, BR
Keywords: Autoregressive integrated moving average, Exponential smoothing state space, Feed-forward neural network autoregression, Vector autoregression, Dam monitoring, Piezometer

Abstract

The level of piezometric pressure at the concrete-rock interface and at discontinuities in the foundation of a concrete dam has a great influence on overturning movements. The objective of this study was the selection of models for the prediction of uplift pressures in the foundation of a concrete dam on the basis of historical time series values only. The models were automatically modeled using the fests package in the R software. The models were evaluated on the basis of RMSE. All of them satisfactorily represented the time series profile of the monthly averages of the measurements of six piezometers of the Itaipu Dam and were able to predict the long-term underpressures with reasonable accuracy. The results obtained indicate that the automation of the modeling can assist in the monitoring and follow-up of the evolution of the behavior of the dam in its current phase.
Published
2023-10-18