Integrating Finite Element Method with ANN-SVM Techniques for Dam Deformation Prediction

Zouriq, Sara Mohamed; Elsalakawy, Tarik S.; Hamdy, Gehan; Khalil, Ehab; Elsayaad, Mohamed

doi:10.21608/erjsh.2025.375385.1406

Integrating Finite Element Method with ANN-SVM Techniques for Dam Deformation Prediction

Document Type : Research articles

Authors

¹ Civil Engineering Department, Faculty of Engineering at Shoubra, Benha University, Cairo, Egypt

² Construction Research Institute, National Water Research Center.

³ Arab Academy for Science Technology and Maritime Transport – Cairo Campus, Cairo, Egypt

10.21608/erjsh.2025.375385.1406

Abstract

Detection of structural damages is significant step in structural health monitoring process, and should be done as accurately as possible. Assessing the condition of a dam is one of the crucial steps in dam conditions assessment in the traditional method of Dam Health Index known as Dam Finite Element Analysis (DFA). In this paper, the author examines the methods of improving and extending the evaluation and scenario analysis based on analytical techniques of Machine Learning (ML). In particular, this paper examines the damage identification in the numerical simulation of the displacement of the dam from finite element analysis (FEA) using the classification techniques of support vector machine and artificial neural networks. Sizeable numerical nonlinear FEM simulation was carried out using ANSYS software on elements of water height changes with respect to upstream load, wave load and uplift forces database creation. This baseline FEA data provides the basis for more efficient and effective use of ML approaches that can then derive displacement performance within minutes under different operational rules or changed climate conditions.
The study shows that in the assessment of displacement performance of Koyna dam in India, the ANN model offers a better result than the SVM. This study shows that the application of ML technologies can be an indispensable addition that reduces the efforts, time, and calculations needed in comparison with pure FEA. The integration of ML techniques with FEA is shown to be a promising approach for supporting structural health monitoring in dam engineering.

Keywords