Reliability analysis of PVD-improved soft soil settlement prediction using Bayesian back analysis framework and simplified Hypothesis B method

Abstract The time-dependent settlement of soft soils is one of the key problems in geotechnical engineering. Using Bayesian back analysis, this study examined the reliability of settlement of the Ballina embankment in Australia. As random variables, the primary compression index (Cc), swelling index (Ce), and secondary compression index (Cαe) were examined for their influence on the settlement probability distribution. To generate compression index samples, Monte Carlo Markov chain simulations （MCMCS） were used, and predicted settlement samples were derived from the compression index samples. Consequently, the predicted settlement samples can be used for the reliability analysis. Based on a comparison between the field settlement data and the predicted settlement, the results indicate that the 90% confidence interval of the predicted settlement is in reasonable agreement with the field settlement monitoring data. With the incorporation of more monitored settlement data into the Bayesian framework, the distribution of predicted settlement shifts from the Weibull distribution to the normal distribution. In addition, the degree of uncertainty in the prediction of settlement decreases with the amount of data incorporated into the model. Additionally, the small amount of data used in the Bayesian framework can also lead to underestimations of failure probability.