Estimation of geological strength index through a Bayesian sequential updating approach integrating multi-source information

Explicit determination of geotechnical parameters is an important but difficult task in rock engineering. Paradoxically, data are often limited even though data from multiple sources (e.g., different testing procedures, testing positions, and estimation models) are commonly available, and the integration of multi-source information for the determination of geotechnical parameters in a probabilistic context remains a great challenge. We have modified an existing Bayesian sequential updating approach and used it for the first time to estimate the geological strength index (GSI) of rock masses by integrating mull-source information in a way that considers prior information, multiple estimation models and probabilistic properties of model uncertainties. The argillaceous siltstone in Zigui County in the Three Gorges Reservoir region of China was used to quantitatively illustrate this method. Three data sets using rock mass rating (RMR)- , tunneling quality index (Q)-, and rock mass index (RMi)-based estimation models were sequentially incorporated into the Bayesian sequential updating framework, through which the information on the mean value and standard deviation of the GSI can be updated. Then a large number of equivalent GSI samples were generated through Markov chain Monte Carlo (MCMC) simulation for further statistical analysis. The results showed that the proposed Bayesian sequential updating approach can provide reasonable probabilistic estimates of the GSI, which compared well with observed data using the standard GSI chart. The Bayesian sequential updating approach is a promising tool for geotechnical parameter estimation and can combine large amounts of information, thereby effectively depicting the probabilistic characteristics of GSI.