Machine learning algorithms for the prediction of pile impedance in nonhomogeneous soils
The following data driven and physics informed machine learning models are used in predicting the impedance of pile foundations subject to vertical and lateral dynamic loads: Gaussian Processes, Neural Networks, Decision Trees, and Gradient Boosted Trees. The models were trained on finite element analyses data and predicted the impedance of piles reliably when the input was similar to training data. When predicting impedance for data not used in training, the Gaussian Process and Neural Networks models performed better. Neural Networks was the best performer overall and was selected for further applications. The application of Neural Networks in predicting pile responses against sinusoidal and earthquake loading was examined and compared against different methods. Trained machine learning algorithms can reliably predict the impedance of piles, which would help reduce the time and computational power needed for analyses of piles subjected to dynamic loads and can reduce the number of iterations for finite element analyses.
Authors
- Ghassan Hassan Sutaih (ORCID: https://orcid.org/0000-0003-1804-0214)
- Mohamed Aggour (ORCID: https://orcid.org/0000-0001-5786-7279)
Institutions
- King Abdulaziz University (SA)
- University of Maryland, College Park (US)
Publication Details
- Journal
- International Journal of Geotechnical Engineering
- Published
- 2026-11-16
- DOI
- https://doi.org/10.1080/19386362.2026.2757123
- Primary Topic
- Geotechnical Engineering and Soil Mechanics
- Type
- article
- Field-Weighted Citation Impact
- 0.00