Bayesian updating for progressive excavation of high rock slopes using multi-type monitoring data

Systems for monitoring the deformation and stress conditions of excavated high rock slopes are usually implemented for safety reasons and to predict the stability of future works. This paper adopts Bayesian methods for updating important geomechanical parameters namely: EURO, E(III2), E(IV2), E(f(a)), c(III1), c(III2), c(V-1) and c(f(a)) in these types of cases. The proposed method utilizes parametric sensitivity analysis, the BP neural network (back propagation neural network), and Bayesian updating to effectively reduce the number of variables, improve the computational efficiency and gradually update the random variables by using progressive monitoring information. The high rock slope excavation on the left bank at the Lianghekou Hydropower Station in China is illustrated as a detailed case study. Initially, only one type of measurement is first used for Bayesian updating (measured displacements or anchorage forces), and then both types of measurements are used. Compared to using only one type of measurement, the parameter uncertainty is reduced and the model accuracy is improved when both types of measurements are employed.