Bayesian Inference for Constitutive Model Calibration
Bayesian Inference for Constitutive Model Calibration
This paper presents an automated method for calibrating several parameters in constitutive soil models using Bayesian inference. The probabilistic machine learning method combines triaxial test data with prior geotechnical knowledge to estimate constitutive model parameters and quantify their uncertainties using a Markov Chain Monte Carlo algorithm. The method has been tested and validated on two different constitutive models: the phenomenological non-linear Mohr-Coulomb model for sands and the critical state Modified Cam-Clay model for clays with Hvorslev surface. The known parameter values (from artificially generated data or previous projects) consistently reside within the 95% credible intervals, with the estimated means frequently approximating the known values. The approach can easily be extended to other constitutive models and experimental data types as it does not require knowing gradients with respect to model parameters.
C. Edmonds; B. Guilherme Saturnino; U. Cilingir
5th International Symposium on Frontiers in Offshore Geotechnics (ISFOG2025)
5 - Data Analytics and Machine Learning
https://doi.org/10.53243/ISFOG2025-164
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