Estimation of percolation of earthen dam using the combined model of deep learning and wavelet transform

Document Type : Research Article

Authors

Faculty of Civil Engineering, Tabriz University, Tabriz, Iran.

Abstract

Seepage prediction is one of the important tools in preventing erosion and destruction earth fill dams.In recent years, due to the uncertainty, complexity, and nonlinearity of seepage relationships, the use of artificial intelligence methods for estimation and prediction of this phenomenon has gained attention. The objective of this research is to estimate seepage in the Starakhan earth fill dam located in northwest Iran. To achieve this objective, in this research, the long-short-term memory network and the wavelet-deep network hybrid model have been used in two different scenarios, and the results obtained from these models have been compared with the feed forward neural network. The results obtained indicated that deep recurrent networks, in the modeling of the seepage phenomenon, outperformed the forward neural networks in terms of estimation accuracy. This can be attributed to their recursive connection between the output and input at each time step, as well as their ability to learn dependencies from previous time sequences. The modeling accuracy was improved by up to 6% as a result. Furthermore, the combined wavelet-deep network model demonstrated superior performance compared to other models, resulting in a 10% increase in modeling accuracy. In conclusion, the utilization of deep recurrent networks and the combined wavelet-deep network model in seepage modeling holds the potential to enhance estimation accuracy when predicting this phenomenon.

Keywords

Main Subjects

Amirkabir Journal of Civil Engineering

Articles in Press, Accepted Manuscript
Available Online from 03 October 2024

Files

Share

How to cite

Statistics