LAUSR

Abstract Field data have shown the fact that soil spatial variability could aggravate the uncertainty of tunnel convergence ΔD (a key indicator for serviceability and safety of tunnels). This paper presents a detailed numerical analysis to investigate the probabilistic response of tunnel convergence in spatially varied soft soils. The soil Young's modulus Es is highlighted and modeled with isotropic and horizontally stratified anisotropic random fields, respectively. The influence of scale of fluctuation (SOF) δ of the Es on convergence ΔD is discussed in detail with respect to different directions, i.e., the vertical and horizontal directions both for δ and ΔD. It is observed that ignoring the spatial variability of Es, i.e., disregarding the possibility of unfavorable soft soil (low stiffness soil) locally around tunnel, can underestimate the mean value of ΔD. The horizontally stratified anisotropic random field is more appropriate than isotropic random field in the sense of an accurate prediction, especially when extreme tunnel convergence occurs. In horizontally stratified anisotropic random fields, the influence of horizontal and vertical SOF is different on tunnel convergence. The surrounding soils near tunnel crown and invert or across tunnel horizontal diameter are very critical to the tunnel convergence. In addition, the effect of horizontal SOF δx on failure probability of the calculated ΔD exceeding the specified allowable ΔDlim is limited when the δx is larger than 4.84 times of tunnel outer diameter.