Numeric mapping of geological stratum topology for shield tunnel

• Published in: IEEE transactions on geoscience and remote sensing Vol. 61; p. 1
• Main Authors: Zhang, Jinyu, Zhang, Yijie, Shen, Mengru, Zhang, Dingli
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.01.2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Arches; back propagation algorithm; Back propagation networks; Boreholes; Dredging; Earth; Excavation; Faces; Friction; Friction resistance; Geological mapping; geological recognition; Geology; Lateral stress; Mathematical models; Mechanics; Model testing; neural network; Neural networks; Polynomials; Shape; Shear strength; Slope stability; Soil; Soil stability; Soils; Stress; Stress ratio; Terzaghi's theory; Topology; tunnel face; Tunneling shields; Tunnels
• ISSN: 0196-2892; 1558-0644
• DOI: 10.1109/TGRS.2023.3325444

The geological condition above tunnel face is essential to excavation and support. In order to solve the shortcoming of ground borehole method that cannot obtain continuous information while destroying soil stability, this paper presents a numeric algorithm to continuously recover the topology of each soil stratum. Specifically, firstly, a Silo-Wedge tunnel face model and its topology& mechanics parameter algorithms were proposed based on Terzaghi theory and Mohr stress circle. Through considering the friction effect of cutter-head and designing suitable parabolic arch element contour, the calculated shape and lateral stress ratio closely fits model tests. Secondly, a inter-stratum load transferring algorithm was designed based on differential element force equilibrium model. Through involving arch effect, shearing resistance and friction force, the result become more practical. Thirdly, a bottom-to-top two virtual stratum model was proposed to recursively calculate the thickness change of soil stratum corresponding to sensed pressure deviation. Through designing the load contribution weight of stratum, complex polynomial solving problem is avoided. Finally, a back propagation (BP) neural network was improved based on proposed model characteristics, calculation error and learning time are reduced. Finally, effectiveness was tested.