Bayesian Back Analysis for Settlement Prediction of Embankments Built on Soft Soils Incorporating Monitoring Data—A Case Study

A novel method for back analysis was used for an embankment over deep soft soil along a major highway upgrade between Woolgoolga and Ballina, NSW. Bayesian back analysis was undertaken using monitored settlement data. The key parameters of interest were the compression ratio, recompression ratio, creep strain rate and coefficient of vertical and horizontal consolidation. Posterior distributions were sampled using a multi-chain Monte Carlo algorithm through a likelihood function to estimate the updated model parameters and subsequent settlement prediction. The simplified geotechnical model, incorporating parameter ratios, can be shown to reduce the amount of computational time required. The predictions were shown to converge to the field measurements regardless of some assumptions made about measurement error and aided in providing a more consistent prediction based on the available data. The intent of the study was to demonstrate that key geotechnical parameters can be updated, and settlement predictions revised and verified from limited site investigation data using Bayesian back analysis incorporating monitored surface settlement data.