How to identify earth pressures on in-service tunnel linings: A Bayesian learning perspective

The identification of earth pressures acting on in-service transportation tunnel linings is essential for their health monitoring and performance prediction, especially for those exhibiting poor structural performance. Since pressure gauges incur substantial costs, the inversion of pressures based on easily observed structural responses, such as deformations, is desirable. The inherent challenge in this inverse problem lies in the non-uniqueness of solutions, which arises from the fact that various pressures can yield structural responses fitting equally well with the observed data. However, existing approaches for pressure inversion predominantly rely on a deterministic framework, often neglecting a detailed discussion on this non-uniqueness. In addressing this gap, this study introduces a Bayesian approach. The proposed statistical framework enables the quantification of uncertainty induced by non-uniqueness in inversion results. The analysis identifies the uniform component in distributed pressures as the primary source of non-uniqueness. The mitigation of solution non-uniqueness can be achieved by increasing the quantity of deformation data or incorporating an observation of internal normal force in a tunnel lining – the latter proving to be notably more effective. The practical application in a numerical case demonstrates the effectiveness of this approach and the associated findings. In addition, our investigation recommends maintaining deformation measurement accuracy within the range of [-1, 1] mm to ensure satisfactory outcomes. Finally, deficiencies and potential future extensions of this approach are discussed.