Risk Assessment and Reliability Analysis on Long-term Settlement of Soft Soils

Publication Type:
Thesis
Issue Date:
2023
Full metadata record
Nowadays, there is a growing interest in applying reliability analysis in geotechnical design approaches as a method of managing and quantifying geotechnical risks with respect to uncertain geotechnical input parameters. Since creep settlement occurs in an extremely long period of time, prediction of creep settlement is a challenging task for geotechnical engineers to utilize soft grounds. Among several methods to evaluate the long-term behaviour of soft soils, the elastic visco-plastic model could be an effective model. However, the difficulties and uncertainties in determining the model parameters is one of the most important limitation of this method. As a result, the aim of this study is to investigate the influence of model parameters uncertainties on predicting the time dependent behaviour of soft soils. In this research, an elastic visco-plastic creep model was combined with the Monte-Carlo probabilistic method to investigate the effects of uncertainties in the elastic visco-plastic model parameters on time-dependent behaviour in soft soils. By adopting monitoring data from the case study of Väsby test fill, the most appropriate cross correlation coefficient between an elastic-plastic model parameter (λ/V) and the creep coefficient (ψ_0/V) was introduced. Moreover, the time-dependent behaviour of soft soils was analysed incorporating spatial variability of elastic visco-plastic model parameters. Standard Gaussian random fields for the adopted random variables were generated adopting Karhunen-Loeve expansion method. The probability of failure was calculated adopting random field (RF) and single random variable (SRV) analysis to determine the critical spatial correlation length, resulted in a maximum probability of failure. In this study, Bayesian updating method of identifying the model parameters was used to update the elastic visco-plastic model parameters using field and oedometer test data by applying the transitional Markov Chain Monte Carlo (TMCMC) method for two case studies. The results confirm that adopting even 20% of total monitored data has a considerable impact in predicting more realistic post-construction settlements. This study provides an insight into selecting the most suitable cross correlation coefficient and the critical spatial correlation length while adopting elastic visco-plastic model parameters as random variables. Therefore, the risks in predicting long-term settlement of soft soils reduces and the reliability of the design in construction increases. Moreover, adopting field monitoring data at early stages to update model parameters has significant impact in predicting more realistic long-term settlement which affects the risks associated with time and cost when adopting low embankment strategy for design of transport infrastructure.
Please use this identifier to cite or link to this item: