Multi-objective probabilistic back analysis for selecting the optimal updating strategy based on multi-source observations

Accurate prediction of geotechnical performance is essential to safety control and risk mitigation in an uncertain environment. In situ observation could be incorporated into the Bayesian inference to improve the reliability of prediction for a rational treatment of various geotechnical-related uncertainties. However, due to technical limitations and cost-efficiency, previous studies generally adopted one or two types of observations in Bayesian updating and focused only on the prediction fidelity. Thus, a fixed updating strategy is usually prescribed ac-cording to the engineer's experience, which may lead to the unreliability of prediction. This paper proposes a novel multi-objective probabilistic back analysis approach for determining the optimal updating strategy, based on multi-source observations, to improve the reliability of subsequent predictions. The features of this approach include three aspects: (1) more than two types of observations are incorporated into the Bayesian updating, (2) prediction fidelity and prediction variability are both considered for determining the optimal strategy, and (3) multi-objective optimization results in a set of non-dominated optimal strategies in the pool of candidate updating strategies. A diaphragm wall-supported excavation in clay is analyzed to demonstrate the effectiveness of the proposed approach.