Analysis of LEAP Experimental Data and Validation of a Numerical Model: A Machine Learning Approach
Analysis of LEAP Experimental Data and Validation of a Numerical Model: A Machine Learning Approach
Twenty-three dynamic centrifuge tests of a saturated backfill deposit supported by a cantilever sheet-pile quay wall were conducted by the LEAP (Liquefaction Experiments and Analysis Projects) international collaboratory. In this article, a Gaussian process regression (GPR) is first used to assess the complex non-linear relationship between the backfill permanent lateral displacements and input motion and initial experimental conditions. The conducted analyses showed that the LEAP-2020 experimental results are consistent and revealed a number of salient characteristics and trends of the soil lateral displacements and interaction with the retaining wall during liquefaction. A GPR was thereafter used to analyze numerical simulations of the LEAP-2020 experiments conducted using the OpenSees-PDMY03 code. A probabilistic metric based on the GPRs of experimental and numerical results was used to assess the model performance and level of fitness of the simulations to the experimental results.
M. Zeghal; Alejandro Sepulveda
4th Asia-Pacific Conference on Physical Modelling in Geotechnics (ACPMG2024)
Keynotes
https://doi.org/10.53243/ACPMG2024-89
Contact Information
- Secretary General - Dr A M McNamara
- Email: [email protected]
- Phone: +44 20 7040 8154
- City, University of London
- Northampton Square
- London EC1V 0HB. United Kingdom