The structural responses of segmental lining to the construction loads during shield tunneling are not well understood yet. To this end, a hybrid sensing method is proposed to understand the mechanical responses of tunnel lining to the construction loads via a wireless sensor network (WSN) and machine learning algorithms. Firstly, the displacement of tunnel lining is monitored in real-time during construction by using the WSN. The rotation of the whole segmental ring is found for the first time based on the in-situ data. A refined three-dimensional numerical model is then applied to investigate the additional stress of segmental rings under such a new deformation mode. Subsequently, a scenario-based construction database is established, including 130 numerical cases of load conditions. For the dynamic control of shield tunneling, an efficient multi-layer perception surrogate model is proposed with the help of the established database. The additional stress state of segmental rings is sensed on-site in a hybrid mode efficiently, both from real structural responses via WSN and the construction parameters via surrogate model. Meanwhile, the construction parameters can be adjusted timely if the additional stress exceeds its limit. For the case study of this paper, with the measured maximum ring rotation at 0.25° and the maximum arc length of circumferential misalignment between two adjacent rings at 7.1 mm, an additional rotation-induced stress of 19.34*103 kN/m2, i.e., 39% of the concrete compressive strength, is found for the first time at longitudinal bolt holes of segmental rings, which has not been considered but should be emphasized in the future design practice.
Read full abstract