Imprecise moment-independent global sensitivity analysis of rock slope using Bayesian multi-model inference
Imprecise moment-independent global sensitivity analysis of rock slope using Bayesian multi-model inference
Traditional global sensitivity analysis (GSA) provides the relative importance of uncertain rock input properties towards rock structure response, given their best-fit probability distribution model and parameters. However, the availability of limited experimental data for the rock properties due to in-situ and laboratory testing complexities affects the model type and parameter accuracy. This study presents an augmented probabilistic methodology by coupling the multi-model and traditional Bayesian inference. The methodology incorporates the uncertainties associated with both the model type and parameter for estimating Borgonovos moment-independent sensitivity indices. The methodology is demonstrated for a rock slope having the potential of stress-controlled failure. It is concluded that the methodology results in imprecise sensitivity indices providing a measure of confidence in sensitivity estimates when limited data are available for properties. Further, the comparative analysis with traditional GSA concluded the superiority of the proposed methodology as it treats the uncertainties in both model type and parameters.
Akshay Kumar; G. Tiwari
10th European Conference on Numerical Methods in Geotechnical Engineering (NUMGE2023)
5. Probabilistic and inverse analysis
https://doi.org/10.53243/NUMGE2023-234
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