Thesis title: SEISMIC RESPONSE OF A TUNNEL EXCAVATED IN SATURATED SANDY SOILS
Subway tunnels are major transportation infrastructures accommodating heavy traffic in modern cities. Until not long ago, underground structures were generally believed to be unaffected by seismic actions; however, numerous cases of tunnels suffering severe damage following seismic events have been reported (Iida et al., 1996; Wang et al., 2009; Callisto and Ricci, 2019; Sudevan et al., 2020).
In particular, at sites characterised by liquefiable sandy saturated soils, significant damage has been observed, as highlighted by post-earthquake damages following the Loma Prieta earthquake, USA (1989) (Schmidt et al., 1998), Kushiro-Oki, Japan (1993) (Koseki et al., 2000), Kocaeli, Turkey (1999) (Erdik, 2001), Maule, Chile (2010) (Lew et al., 2010), Tohoku, Japan (2011) (Tokimatsu et al., 2012), and Christchurch, New Zealand (2011) (Kaiser et al., 2012). In these cases, underground structures were primarily damaged by significant buoyant forces generated by the liquefaction of the surrounding saturated soil, which locally exceeded the self-weight of the underground structures (Shen et al., 2022).
Often, the dynamic response of tunnels to seismic actions is investigated using a simplified decoupled approach, in which the free-field propagation of seismic waves in the soil is first evaluated and the forces induced in the tunnel lining are subsequently computed by adopting analytical solutions based on simplified linear assumptions. However, owing to recent technological advances that have significantly reduced computational times, the use of a fully coupled approach, which allows the soil–structure interaction to be explicitly taken into account in the context of dynamic finite element simulations, is becoming increasingly common.
The latter schematisation of the problem may lead to different results depending on the constitutive model adopted to describe the seismic response of the soil. In this PhD thesis, the objective is to investigate the performance of different constitutive laws in predicting the transverse seismic response of a circular tunnel embedded in saturated sandy soil. In particular, finite-element-based numerical predictions obtained using the advanced SANISAND model, here adopted as a reference behaviour, are first compared with those obtained using a visco-elastic formulation, whose parameters are calibrated against the results of an equivalent linear visco-elastic analysis, and subsequently with those obtained by adopting the Hardening Soil model with small-strain stiffness, a constitutive law available in a well-known commercial code. A seismic input motion (Arcelik) was scaled to three different peak ground acceleration (PGA) levels in order to assess the range of validity of the different constitutive models. The results show that, with increasing PGA, only the SANISAND constitutive model is able to describe the soil behaviour in a more realistic manner. Indeed, thanks to the deviatoric–volumetric coupling, this model is able to capture the triggering of liquefaction, which, in the case of higher PGA values, develops in zones close to the piezometric surface. The limitations associated with the use of simpler constitutive assumptions are thus highlighted with respect to the considered seismic intensity, thus providing useful insight into the complex problem under investigation. In the tunnel lining, the effects of early irreversibility are reflected in the distributions of the hoop force and bending moment at the end of the seismic event and at the end of consolidation. These effects arise from the different deformations of the lining, which are a consequence of the accumulation of plastic strains that can only be captured through the use of advanced constitutive models, i.e. by using the SANISAND constitutive model and, to a lower extent, the HSsmall one.
The analyses are performed using both a fully undrained approach and a dynamic consolidation (u–p) formulation, in which different values of soil permeability are adopted in order to investigate the role of this parameter in the seismic response of the tunnel. The results of the u–p analyses show the importance of the permeability on the seismic response of the soil. For instance, by observing the time history of the excess pore pressure ratio (Ru) at a point located 4.25 m below ground level, defined as the ratio between the excess pore pressures generated during the seismic event and the lithostatic effective vertical stress, it can be seen that Ru increases rapidly over the time interval corresponding to the most significant part of the input accelerogram. After the peak, it reduces with a rate that depends on the permeability. In contrast, in the case of lower permeability, as well as in the fully undrained analyses, Ru increases monotonically with time until it reaches a unit value, which is indicative of the occurrence of liquefaction. For higher PGA levels, the importance of adopting constitutive models capable of correctly representing the hydro-mechanical coupling of the soil therefore becomes evident, as this is essential for a comprehensive and realistic assessment of the seismic response of tunnels in saturated sandy soils.