Long-Term Settlement Induced by EPB Tunnelling Studied with Numerical Simulations

10.2478/sggw-2018-0012

Annals of Warsaw University of Life Sciences – SGGW Land Reclamation No 50 (2), 2018: 139–157 (Ann. Warsaw Univ. of Life Sci. – SGGW, Land Reclam. 50 (2), 2018)

Long-term settlement induced by EPB tunnelling studied with numerical simulations

MACIEJ OCHMAŃSKI Faculty of Civil Engineering, Silesian University of Technology, Gliwice, Poland

Abstract: Long-term settlement induced by EPB 2002, Mair 2008, Wu et al. 2011, Ng tunnelling studied with numerical simulations. et al. 2013). This aspect of tunnelling Long-term settlement induced by tunnelling in becomes particularly important when soft cohesive subsoil is rarely considered as a sub- considering that long-term settlement, ject for research despite substantial evidence that it increases considerably after construction. The pa- which is generally caused by consolida- per aims to investigate the time effects of tunnel- tion, may constitute 30–90% of total ling with an earth pressure balance (EPB) shield settlement (Shirlaw 1995). There are two by means of analysing subsoil deformation with factors leading to tunnelling-induced the most up-to-date numerical tool. A three-dimen- consolidation, with the first one related to sional numerical model with a detailed description the technological processes significantly of the geometrical and mechanical characteristics disturbing subsoil around the cavity. of the tunnelling process and the various materials involved was built. The computational model vali- Mechanized tunnelling in cohesive soils dated on a real-case was then used to simulate tun- is mainly carried out using earth pressure nelling in different groundwater conditions prior balance (EPB) shields that excavate soil to construction. That was followed by a simulation with a rotary cutterhead and are pushed of groundwater drawdown conditions after tunnel- forward against previously installed seg- ling with impermeable and permeable tunnel lin- mental lining by a set of hydraulic jacks. ings. The calculations show the development of During this process, a grout backfilling is the settlement over time in different groundwater conditions together with the significant influence injected at the tail of the shield to fill the of the permeability of the tunnel. circumferential void between the lining extrados and cavity boundary. Immedi- Key words: numerical simulations, EPB shield, ately after injection, grout behaves as a long-term settlement liquid and with the progress of hydration turns into a relatively strong and stiff ma- INTRODUCTION terial preventing further deformation of the cavity. Subsoil consolidation caused There is substantial evidence that the by the imbalance between stress state in settlement continues to increase up to the soil and the pressure of the grout is several years after the construction of limited to relatively small portions of tunnels in soft cohesive subsoil (O’Reilly soil around the cavity and the intensity et al. 1991, Bowers et al. 1996, Harris of the induced effect is determined by 140 M. Ochmański the pressure difference between these more significant effects may be induced two materials. The second factor lead- by a groundwater drawdown after tunneling to the consolidation settlement is ling, this effect is manifested not only by that tunnels act as drains introducing additional long-term settlement but also new boundary conditions at the cavity by the rotation of the lining segments contour. This results in a negative excess and the change in the structural forces in pore water pressure induced in the sur- that elements. rounding soil, thus an increase in the The analyses of the tunnelling proc- effective stress in the soil leads to the ess mainly refer to the almost immedi- consolidation process (Ward and Thomas ate short-term effects and are mainly 1965, Palmer and Belshaw 1980, Harris carried out using empirical approaches 2002). Additionally, the existence of the (Peck 1969) derived from a wide data- thin grouting layer around the segmental base of evidence. On the other hand, lining has a rather limited effect on its analytical closed-form solutions are not water-tightness, especially over the frequently used in view of their rather long-term (Mair 2008). Water flows into limited capabilities which are reduced to the tunnel through the grouting layer and elasticity and rarely plasticity with ques- the joints between the lining segments. It tionable assumptions that ignore much is often observed that the internal surface of the modern knowledge concerning of the lining is visibly wet even though soil behaviour. The most sophisticated auxiliary measures are applied to make tools are numerical simulations that have it watertight. The final long-term set- recently gained popularity due to their tlement resulting from the combination impressive capabilities which allow for of both factors generally occurs over the introduction of complex geometry of a much larger area and the settlement the problem and include various contact trough profile is much deeper and wider behaviours between different elements with little curvature compared to the and sophisticated constitutive models short-term one (Mair and Taylor 1997). for natural and artificial materials. Fur- In densely urbanized cities, cyclic and thermore, the progressive development transient seasonal effects are drastically of contemporary IT technology allows, limited by the presence of buildings especially thanks to parallel comput- and concrete pavements, roads which ing, coupled problems to be solved with efficiently reduce the migration of water precisely described geometry discretized from rainfall and prevent the evaporation with hundreds of thousands of finite of moisture from the subsurface. There- elements. Moreover, according to the fore, the level of the groundwater table provisions contained in the current codes is influenced by the fluctuation of river of practice (e.g. in the European Union levels, leakage from water mains, deep- CEN, 2004), geotechnical design should -well pumping and the presence of nearby always consider the effects induced by tunnels. The change in the groundwater the changes in the groundwater table level prior to tunnelling may also lead as it leads to time-related effects, one to significant changes in the ground set- of the consequence is increased earth tlement as noted by Mair (2008). Even pressure acting on the lining (ASCE Long-term settlement induced by EPB tunnelling studied... 141

1984). Despite that, most of the analy- description of the computational model, ses of the tunnelling process even using a validation based on a real case study of sophisticated numerical tools still refer the Metropolitan Rapid Transit Authority to the immediate response and neglect to (MRTA) project in Bangkok (Suwansa- analyse the long-term effects related to wat 2002, Surarak 2010, Sirivachiraporn consolidation which are assumed to be and Phienwej 2012, Likitlersuang et al. rather negligible. 2013) is presented. Finally, the long- The aim of the paper is to analyse the -term effects related to the consolidation effects related to long-term consolida- of soil induced by the tunnelling process tion as a result of shield tunnelling and are analysed together with the effects of the groundwater table drawdown. The the groundwater drawdown by means of paper demonstrates the capabilities of the ground surface settlement. Analyses the most up-to-date and realistic repro- are performed introducing impermeable duction of the tunnelling process with and fully-permeable tunnel linings. EPB technology by means of numerical simulations. An automated numerical COMPUTATIONAL MODEL model introduced with a script prepared in python objected-oriented language The computational model is introduced for commercial AbaqusTM FEM code in the commercial AbaqusTM (Hibbitt developed by Ochmański et al. (2018) is and Sorensen 2001) FEM code as a used here to allow for fast solutions with three-dimensional symmetrical problem a strong theoretical basis. After a brief as depicted in Figure 1. Simulations are

FIGURE 1. A three-dimensional symmetrical FEM model of tunnel drilled with the EPB shield 142 M. Ochmański fully automatized and controlled by a is calculated assuming hydraulic heads at script written in the Python program- the boundaries of the model and known ming language. The model introduced value of the permeability coefficient of presents an attempt to describe in as pre- each layer. The nonlinear, irreversible cise as possible all of the components of stress-displacement response of the sub- the EPB technology which allows for the soil in the presented model is described elimination of many subjective assump- by means of the hypoplastic constitutive tions which are typical of the simulation models (Gudehus et al. 2008) freely of this process, such as the fictitious available as a FORTRAN implementa- boundary conditions of prescribed tion at the website of Soilmodels project displacements at the cavity boundary to (www.soilmodels.com). Specifically, the reproduce volume loss. For the sake of von Wolffersdorff (1996) hypoplastic brevity, only a brief description of the model was used for the coarse-grained computational model characteristics is materials and the Mašín (2005) model given here, in order to learn about the for the fine-grained materials. To cali- fine details the reader is referred to the brate the former one, 13 parameters need work of Ochmański et al. (2018). to be assigned, while for the later one, 12 parameters are necessary. Further- Geometry of the model more, to include increased stiffness in the The introduced model produces the small-strain range, an extension called possibility of reproducing any possible the intergranular strain concept (ISC) geometrical setup of a tunnel and subsoil (Niemunis and Herle 1997) has been including e.g. the stratification of the introduced. Subsoil layers are assumed subsoil, irregular surface shapes, curved to be fully saturated with water flow de- tunnel alignment etc. The model bounda- scribed by the well-known Darcy’s law. ries are fixed at the distance which pre- serves their minimum influence on the Tunnel heading with the EPB shield region of stress concentration according An essential element of the tunnel head- to Gunn’s (1993) recommendations. ing machine is a conical shield tapering Depending on the model complexity, off towards the end as depicted in its geometry may be discretized with Figure 2 with other elements forming an eight-node linear bricks or with 10-node altogether realistic reproduction of the quadratic tetrahedral elements. tunnelling process with the EPB shield. The shield’s deformations are determined Subsoil by a linear-elastic behaviour, whereas The role played by the subsoil behaviour its interaction with the surrounding soil due to the fluctuations of the groundwater is simulated as a “hard contact” in the table on the overall response of the geo- normal direction and with Coulomb technical model is undeniably important. friction in the tangential direction. The The initial distribution of the effective tunnel face is supported by a hydrostati- stress in the case of horizontal subsoil cally distributed pressure applied to an layers is computed using the K0 proce- impermeable membrane at the front dure, while the initial pore water pressure surface (Müeller-Kirchenbauer 1977) Long-term settlement induced by EPB tunnelling studied... 143

FIGURE 2. Definition of the computational model for tunnelling with the EPB shield (Ochmański et al. 2018) transferring pressure to both the soil cavity has been simulated by a layer of skeleton and pore fluid. The self-weight finite elements with the initial stress state of the shield is applied to the bottom part hydrostatically varying corresponding of the shield’s internal surface as a re- to the injection pressure. The mechani- sultant pressure from all devices located cal response of this material has been within the shield. assumed to be isotropic linear-elastic A tunnel lining is introduced as a cyl- with a time-dependent stiffness based on inder with a specific diameter and thick- the relationships defined by Meschke et ness. Segmentation of the lining has been al. (1996). The initial nil volume change considered only along the tunnel axis with corresponding to the liquid state of the a frictional contact between the adjacent grout has been accounted for by fixing rings, thus omitting stiffness reduction the value of the Poisson’s ratio to 0.499 due to the radial joints. Whereas, the (to overcome singularity for 0.5) with material of lining has been assumed to a progressive drop in time up to a value be linear-elastic with parameters typical of 0.2 specific for the solid material. for mature material of precast concrete The presence of the backup trailer fol- elements. Face supporting pressure, lowing the shield has been simulated as forces generated by the hydraulic jacks a set of concentrated forces applied to necessary to push the shield forward the single cubic elements which repre- and the frictional force generated at the sent the weight of the gantry cranes. soil–shield interface have been applied as a resultant pressure to the transversal Computational steps outer surface of the last installed ring. The tunnelling process is schematized The grout backfilling of the circum- by a repetitive sequence of two discrete ferential void formed between an extra- steps, specifically for the shield advance- dos of the lining and the boundary of the ment and lining installation. Each step is 144 M. Ochmański computed by assuming a transient cou- VALIDATION OF THE MODEL pled stress–displacement analysis with – CASE STUDY pore fluid diffusion. In the step of shield advancement, displacements are confined A project called Blue Line was undertaken to the back of the shield, pushing it ahead by the Metropolitan Rapid Transit Au- of the current face position, together with thority (MRTA) in the area of Bangkok, it the corresponding shift of all of the ap- served as a reference case for the validation plied loads, i.e. the shield’s self-weight of the computational model. It consisted and backup-trailer loads transmitted to of 18 cut-and-cover stations connected by the lining. In the next step, the excavation two twin tunnels (Northbound and South- of the soil is simulated by the deactivation bound) each about 20-kilometer long and of the finite elements accompanied by the drilled with EPB shields. Validation was hydrostatic pressure applied to the newly performed within a short section of the exposed face and the activation of the line characterized by a sequence of the lining with tail void backfilling elements. following subsoil layers: weathered crust After the simulation of the tunnel and backfill material (BM), Bangkok soft construction, an additional two steps of clay (BSC), medium stiff to stiff clay a long-term consolidation analysis are (MSC), and medium dense to very dense introduced. During the first step, the excess sand (DS). The mechanical behaviour of pore pressure generated during tunnelling each layer from small to large strain range is dissipated, whereas the second step has been calibrated with reference to the refers to conditions of the groundwater rich database of field and laboratory tests drawdown simulated by applying the new reported e.g. by Balasubramaniam et al. boundary conditions of the pore pressure (1976), Balasubramaniam and Hwang at the bottom surface of the soil block. (1980) or Seah and Koslanant (2003). The initial shear stiffness profile through depth Computational unit has been calibrated from the above tests, A computationally expensive coupled while the stiffness degradation curves stress–displacement analysis simulating for each layer have been fitted to pass the seepage flow problem with geometry through a strain thresholds – γ0.7 (ratio at discretized using around 400 thousands which G/Gmax is equal to 0.7) which were finite elements (~1.5 million DOF) was taken from studies of Likitlersuang et al. carried out on two nodes of the Prometh- (2013). A detailed explanation of the con- eus supercomputer being a part of the stitutive model calibration along with its Polish grid infrastructure. The message- performance is presented in the work of -passing library (MPI) parallelization Ochmański (2016). For the sake of brevity, of the element operations and equation the results of the calibration are presented solution with the direct sparse solver here only for the medium stiff to stiff clay resulted in a significant reduction of the (MSC) as depicted in Figure 3. The state simulation time (230 vs 53 h) compared parameters for all the Bangkok subsoil to modern personal computers equipped layers with the calibrated parameters of with the Intel i7 CPU (12 cores) and the hypoplastic constitutive models are 64 GB of memory. presented in Table 1. TABLE 1. State properties and calibrated parameters of the constitutive models for Bangkok subsoil layers (Ochmański et al. 2018)

State properties Hypoplasticity for fine-grained materials + ISC cl Layer γd en OCR k φc λ* κ*N νpp αG R βr Χ Ag ng mrat × (kN/m3) (-) (-) (m/s) (°) (-) (-) (-) (-) (-) (-) (-) (-) (kPa) (-) (-) BM 5.69 3.86 1.3 1e-9 23 0.2 0.03 2.2 0.21 1.2 1e-4 0.023 1.0 1 500 0.5 0.7 × SBC 6.18 3.3 1.3 3e-9 28.8 0.17 0.02 1.85 0.14 1.2 1e-4 0.003 1.0 1 500 0.5 0.7 × MSC 11.80 1.2 1.65 1e-9 30 0.16 0.01 1.45 0.25 1.2 1e-4 0.001 1.0 1 500 0.75 0.7 × Hypoplasticity for coarse-grained materials + ISC

γd en OCR k φc hs Ned0 ec0 ei0 ΑβR βr Χ mR mT Layer (kN/m3) (-) (-) (m/s) (°) (GPa) (-) (-) (-) (-) (-) (-) (-) (-) (-) (-) (-) DS 16.86 0.6 1.5 1e-7 36 1.5 0.28 0.55 0.95 1.05 0.25 1.5 1e-4 0.5 6 5 2

FIGURE 3. Mechanical response of the medium stiff to stiff clay reproduced by the hypoplastic model (Ochmański et al. 2018) 146 M. Ochmański

Simulation of the EPB tunnelling has to construct a tunnel of 6.3 m in a diam- been validated on the Southbound tunnel eter (external lining diameter without drilled before the Northbound one. In the including tail void). The parameters analysed section, a shield with a diameter characterizing the computational model of 6.43 m at a depth of 18 m was used summarized in Table 2, were taken from

TABLE 2. Parameters of the model adopted to simulate tunnelling with the EPB shield of the MRTA project Parameter Symbol Unit Value Parameter Symbol Unit Value Subsoil Tail void backfilling Ratio of initial Width W m60 E /E – 0.6 stiffness* 1 28 Stiffness after 28 Length L m 120 E GPa 5 days* 28

Height H m 60 Poisson’s ratio v28 – 0.49 3 Number of soil layers n – 4 Grout density* pgr kg/m 1 800 Distance to tunnel Grouting pressure z m18 G kPa 170 springline 0 (axis) P Shield Tunnel lining

Diameter at the front Dch1 m 6.43 Outer diameter Dl m 6.3

Diameter at the back Dch2 m 6.43 Thickness Tl m 0.3

Length Ls m 8.35 Ring width Wl m 1.2

Conicity δ m 0 Young’s modulus El GPa 31.5 Front shield T m 0.08 Poisson’s ratio v – 0.2 thickness* fs l 3 Tail shield thickness* Tts m 0.04 Bulk density pl kN/m 25

Front shield length Lfs m 4.0 Backup trailer

Tail shield length Lts m 4.35 Total length* Ltr m72

Young’s modulus Esh GPa 200 Total weight* Ftr_tot kN 3980

Poisson’s ratio vsh – 0.3 Transversal spacing* Str m 3.0

Overcut Dexc m 0.02 Number of forces* Ftr_n –9 Gap at the shield 3, 7.5, 7.5, 3, g m 0.065 Longitudinal spacing* Lf m back t tr 15, 3, 15, 3, 15 230, 230, 230, Total self-weight of F kN 3 000 Forces* F kN 280, 280, 270, the shield* s tr 270, 100, 100 Coulomb friction μ – 0.20 Construction steps coefficient*

Frictional force WM kN 8 700 Penetration rate texc m/h 1.5 Face pressure at the F kPa 120 Stand-still phase t h 0.5 axis P std Gradient of the face F kPa/m 15.7 × pressure P_gr * Values assumed based on similar studies (Kasper and Meschke 2004). Long-term settlement induced by EPB tunnelling studied... 147 a

FIGURE 4. Surface settlement profiles obtained from simulations in the transversal (a) and longitudinal section (b) compared with observations the literature describing the case study are slightly different from those obtained (Suwansawat 2002, Suwansawat and from evidences, i.e. 13.5 mm and 1.0%. Einstein 2007) and partly assumed based The settlement profile is narrower on similar studies (Kasper and Meschke compared to the mean characteristic of 2004). observations, however, it still fits into The surface settlement generated the cloud of points. On the other hand, during tunnelling obtained from numeri- the longitudinal profile fits well with the cal simulations is presented in Figure 3 observations. together with a cloud of points representing observations from the real case. SETTLEMENT INDUCED The results obtained for the longitudinal BY TUNNELLING IN DIFFERENT section are presented up to the moment GROUNDWATER CONDITIONS when the shield reached a distance from the reference section equal to double the In order to investigate the influence depth of the tunnel (36 m). The settle- of the groundwater conditions on the ment trough profile (Fig. 4a) obtained long-term settlement induced by tun- from a simulation with the maximum nelling, the following scenarios have value of 19 mm and volume loss of 1.2% been analysed: groundwater drawdown 148 M. Ochmański conditions prior to tunnelling; and the horizontal stress is reduced. A positive conditions of hydrostatically distributed outcome of tunnel excavation in condi- pore water pressure followed by ground- tions of groundwater drawdown is that water drawdown after tunnelling. The a lower face pressure is required to bal- latter analysis has been extended by the ance the horizontal stress from the soil simulation of long-term consolidation ahead compared to the situation before with and without drainage conditions at the drawdown with the hydrostatic pore the tunnel boundaries representing im- water pressure. permeable and fully permeable lining. In the reference case used for the vali- Groundwater drawdown results in dation, a simulation has been carried out a change of stress transfer between the in conditions of groundwater drawdown soil skeleton and pore water. The verti- prior tunnelling with a face pressure cal component of the pore water pres- equal to 120 kPa (measured at the axis). sure is transferred to the horizontal one This value means that the face pressure by a ratio of lateral pressure equal to being lower than horizontal total stress 1.0, while the horizontal component of (Fig. 5a) presents an abnormal situation, the effective stress is computed from the therefore, for the purpose of analysis face vertical one with the use of the K0 coef- pressure has been increased to the value ficient which is usually lower than unity. of the horizontal total stress (195 kPa). In this way, an obvious increase of the Considering the previously described vertical effective stress due to the reduc- effect of the horizontal total stress reduction of pore water pressure results in an tion due to the groundwater drawdown, increase of horizontal effective stress but supporting face pressure in the conditions to a lesser extent than the reduction of of tunnelling prior to groundwater draw- pore water pressure; thus, in the condi- down has been increased by the value of tions of groundwater drawdown, the total Δu · (1 – K0) = 33 kPa as depicted in the a b

FIGURE 5. Face pressure applied in a simulation of groundwater drawdown conditions (a) prior to, and (b) after tunnelling Long-term settlement induced by EPB tunnelling studied... 149

Figure 5b. In this way, face pressure is reached a distance from the reference equal to the horizontal total stress com- section equal to three times the tunnel ponent and a comparison between both depth (54 m), while results in the trans- cases may be carried out. versal section correspond to the situation The short-term settlement of the when the reference section is behind the subsoil induced during tunnelling in face at the same distance of 3 · z0. The both groundwater conditions plotted as latter ones are vertically and horizon- longitudinal and transversal settlement tally scaled with reference to their maxi- profiles are depicted in Figure 6. Addi- mum values and depth to the springline, tionally, settlement profiles from the respectively. The results in the short-term simulation in groundwater drawdown present a typical deformation pattern conditions with face pressure as in the induced by tunnel excavation depicted reference case study (Fp = 120 kPa) are at the ground surface as a settlement also given. The results presented in the trough, with some differences related to form of longitudinal profiles refer to the the different initial water-bearing condi- deformations induced until the shield tions. Simulations of tunnelling in initial a

FIGURE 6. Short-term surface settlement presented in the (a) longitudinal section and (b) transversal section 150 M. Ochmański groundwater drawdown conditions show The load transferred from the soil to the influence of the face pressure on the the lining is more uniform with a better reduction of pre-convergence, thus limit- lateral confinement, which results in ing surface settlement ahead of the tunnel a lower lining deformation followed by face. At the same time, after passing the a reduction of the subsoil settlement. On reference section (y/z0 ≥ 0) by the shield, the other hand, an increase in the set- increased face pressure barely influ- tlement trough width in the transversal ences the settlement profile and none section is observed. In Figure 7 a 3D is observed in the transversal section. contour plot of settlements is presented, Comparing the results of the simulation but for the sake of brevity, only for the in groundwater drawdown conditions, simulation of tunnelling in groundwater with an increase in face pressure com- drawdown conditions. pared to a value similar to the horizontal Simulation of the subsoil consolida- stress component (Fp = 195 kPa), with tion after tunnel construction reveals those from the simulation in hydrostatic further differences between the studied groundwater conditions with the cor- cases. The longitudinal and transversal responding face pressure, significant settlement profiles are plotted together differences are observed. In the latter with the contour plots of excess pore conditions, surface settlement recorded water pressure generated during tun- in the longitudinal profile is almost nelling in drawdown and hydrostatic twice smaller (11 vs 19 mm) as that from groundwater conditions in Figure 8. the former simulation in groundwater Relative settlement induced after drawdown conditions. It is connected tunnelling in drawdown conditions with an increased horizontal total stress (Fig. 8a) reaches 13 mm, which consti- component at the tunnel level, while the tute about 65% of the short-term settle- vertical component remains unchanged. ment, while consolidation for tunnelling

FIGURE 7. Contour plot of the settlement from numerical analysis with groundwater drawdown conditions prior to tunnelling with increased face pressure (Fp ≈ σh) a b

c d

FIGURE 8. Surface settlement profiles generated during consolidation in: a – longitudinal section, b – transversal section; excess pore water pressure generated during tunnelling in: c – drawdown conditions, d – hydrostatic conditions 152 M. Ochmański in hydrostatic groundwater conditions is the steady-state pore water pressure in slightly more intensive and settlement the case of fully-permeable lining are increases by 7 mm, which is 80% of the presented in the Figure 9. The observa- initial value. Looking at the transversal tions indicate a progressive increase of section of the trough profile (Fig. 8b) for settlement for the fully permeable lining tunnelling in hydrostatic groundwater by 0.65 m for the tunnel in hydrostatic conditions, an increase in its width by groundwater conditions and by 0.22 m almost 100% is observed resulting in for the tunnel in drawdown conditions, a significant reduction of its curvature. which is a result of water flow to the The reason for this behaviour is that tunnel from surrounding subsoil. In a more intensive consolidation process both cases additional settlement is at is induced in the soil around the tunnel least one order of magnitude larger than in groundwater drawdown conditions that in the case of the tunnel with an as confirmed by excess pore pressure impermeable lining. The difference of plotted for both cases in Figures 8c and settlement between both cases is related 8d. In both cases, the maximum excess again to the intensity of the consolida- pore water pressure is generated ahead tion which, in drawdown conditions, is of the face and around the lining in the lower due to less negative excess pore upper half, whereas a higher value is water pressure build-up after the intro- observed for the initial drawdown condi- duction of a drainage boundary at the tions. On the other hand, minimal excess tunnel cavity. The contour plots of the pore pressure appears under the tunnel pore water pressure presented in the at a distance approximately equivalent Figure 9c and d indicates a lowering to the tunnel radius with a concentra- of the free-field pressure and its drop tion located directly under the shield, around the tunnel. Besides that, in both and similar to the simulation with ini- cases similar nonnegligible widening of tial drawdown conditions indicates ex- the transversal profiles of the settlement tremum. This may be explained by the trough quantified by a 20% increase of fact that face pressure at the crown and the distance to the inflection point is invert is respectively lower and higher observed when a permeable lining is than the horizontal component of the total introduced. stress by around 30 kPa due to their dif- Effects of the groundwater table low- ferent gradients as depicted in Figure 5. ering on the ground surface settlement Subsoil consolidation is influenced by are depicted in Figure 10. In both ana- the permeability of the lining introduced lysed cases, a progressive increase of in the presented simulations by means of the free-field settlement after the intro- the pore pressure boundary conditions duction of new boundary conditions of applied at the tunnel perimeter. The pre- pore pressure at the bottom of the soil sented results refer to the simulation of block may be observed. In the case of tunnelling with hydrostatic and draw- tunnelling with an impermeable lining, down groundwater conditions. The set- settlement increases by almost 1.3 m, tlement profiles in the longitudinal and while for a permeable lining, additional transversal sections, and distribution of settlement (0.9 m) is lower than in the a b

c d

FIGURE 9. Surface settlement profiles generated during consolidation: a – longitudinal profile, b – transversal profile; and steady-state pore water pressure for the tunnel with fully-permeable lining in: c – drawdown conditions, d – hydrostatic conditions 154 M. Ochmański a

FIGURE 10. Surface settlement profiles resulting from groundwater drawdown: a – longitudinal profiles and b – transversal profiles previous case due to the earlier consoli- smoother transversal settlement profile dation of the soil in close proximity to with less curvature. the tunnel. After filtering out the free- The transversal profiles expressed -field settlement, it may be noticed that as a total settlement show that in both the tunnel with an impermeable lining impermeable and fully-permeable lin- leads to an unfavourable heave of the ings a similar amount of the free-field ground surface due to the existence of settlement equal to around 1.4 m occurs. rather stiff lining comparing to the soft This significant increase of settlement subsoil, which undergoes the consolida- observed during the lowering of the pore tion process. The intensity of this effect water pressure is a consequence of the depends on the relative stiffness of the very compressible soil layer (λ = 0.4 lining and will increase with a lower and e = 3.3) with a large thickness. overburden depth and in the case of These results are in accordance with the deep tunnels, it may even be negligible. estimation completed by Nutalaya et al. On the other hand, a fully-permeable (1989) indicating 1.6 m of long-term set- lining induces additional settlement, tlement due to deep well pumping from which nevertheless results in a much the Bangkok subsoil. Long-term settlement induced by EPB tunnelling studied... 155

CONCLUSIONS permeability of the lining is not without interest especially when lowering of the The paper presents the most up to date groundwater table is induced after tunnel and realistic reproduction of tunnelling construction. A few centimetres of dif- process with EPB technology by means ferential settlement of the ground sur- of automated FEM simulations applied face induced during groundwater table to investigate effects related to different lowering has a much larger impact on the groundwater conditions on the induced response of the existing structures than ground deformations. The computational even an order of magnitude larger free- model, successfully validated with a real field settlement. The former, in the case case study, has been used to simulate of fully-permeable lining, is smaller, long-term settlement induced by EPB resulting in a smooth transversal profile tunnelling in hydrostatic and ground- with little curvature, while in the case water drawdown conditions with imper- of the impermeable lining, the relative meable and fully-permeable lining. heave of the ground surface being for the The presented analyses have shown existing structures much more destruc- that both short- and long-term settle- tive than the settlement is observed. This ments are significantly influenced by dif- effect will depend on the lining stiff- ferent groundwater conditions and dem- ness and at the same time will become onstrated the importance of the lining more important for shallow tunnels with permeability. During tunnelling in the a lower overburden, while for the deep groundwater drawdown conditions, the ones it may not even occur. positive outcome of a lower face supporting pressure becomes unimportant due to Acknowledgements the possible increase of short- as well as This research was supported in part by long-term settlements. This is a result of PL-Grid Infrastructure. the difference between the face pressure gradient and horizontal stress component gradient, leading to insufficient support REFERENCES at the crown and excess pore water pres- ASCE 1984: Guidelines for tunnel lining design. sure build-up around the cavity. In the ASCE, New York. long-term, the concentration of excess BALASUBRAMANIAM A.S., CHAUDHRY pore water pressure around the tunnel, A.R., HWANG M., UDDIN W., LI Y.G. 1976: induces more intensive consolidation State boundary surface for weathered and soft of the soil in its close proximity, thus Bangkok Clay. Austral. Geomech. J. 6 (1): 43–50. generating further ground deformations. BALASUBRAMANIAM A.S., HWANG Z.-M. On the other hand, when the fully-per- 1980: Yielding of weathered Bangkok clay. meable lining is introduced, the situation Soils Found. 20 (2): 1–15. is reversed, generating lower long-term BOWERS K.H., MILLER D.M., NEW B.M. settlement as an effect of tunnelling (1996): Ground movement over three years at the Heathrow Express Trial Tunnel. In: R.J. in already consolidated subsoil due to Mair, R.N. Taylor (Eds.), Geotechnical aspects groundwater drawdown than that which of underground construction in soft ground. occurs under hydrostatic conditions. The Balkema, Rotterdam: 647–652. 156 M. Ochmański

EN 1997-1. Eurocode 7. Geotechnical design. NG C.W.W., LIU G.B., LI Q. 2013: Investigation GUDEHUS G., AMOROSI A., GENS A., HER- of the long-term tunnel settlement mechanisms LE I., KOLYMBAS D., MAŠIN D., WOOD of the first metro line in Shanghai. Can. Geo- D.M., NIEMUNIS A., NOVA R., PASTOR tech. J. 50 (6): 674–684 M., TAMAGNINI C., VIGGIANI G. 2008: NIEMUNIS A., HERLE I. 1997: Hypoplastic The soilmodels.info project. Int. J. Num. Anal. model for cohesionless soils with elastic strain Meth. Geomech. 32 (12): 1571–1572. range. Mech. Cohesive-Frictional Mat. 2 (4): GUNN M.J. 1993: The prediction of surface set- 279–299. tlement profiles due to tunnelling. In: G.T. NUTALAYA P., YONG R.N., CHUMNAN- Houlsby, A.N. Schofield (Eds.), Predictive KIT T., BUAPENG S. 1989: Land subsidence soil mechanics. In: Proceedings of the Wroth in Bangkok during 1978–1988. In: Workshop memorial symposium, 27–29.07.1992, St. on Bangkok Land Subsidence – What is Next? Catherine’s College, Oxford. Thomas Telford: Bangkok: 1–48. 304–316. O’REILLY M.P., MAIR R.J., ALDERMAN G.H. HARRIS D.I. 2002: Long term settlement fol- 1991: Long-term settlements over tunnels: an lowing tunnelling in overconsolidated London eleven-year study at Grimsby. In: Proceedings Clay. In: R. Kastner, F. Emeriault, D. Dias, of Conference Tunnelling ’91, London: 55–64. A. Guilloux (Eds.), Geotechnical aspects of OCHMAŃSKI M. 2016: Numerical analyses of underground construction in soft ground. the effects of tunnels construction. PhD thesis, Spécifique, Lyon: 393–398. Silesian University of Technology, Gliwice HIBBITT K., SORENSEN 2001: ABAQUS/ [typescript]. /Standard User’s Manual. ABAQUS/Standard OCHMAŃSKI M., MODONI G., BZÓWKA J. User’s Manual. Hibbitt, Karlsson & Sorensen. 2018: Automated numerical modelling for the KASPER T., MESCHKE G. 2004: A 3D finite control of EPB technology. Tunn. Undergr. Sp. element simulation model for TBM tunnelling Tech. 75: 117–128. in soft ground. Int. J. Num. Anal. Meth. Geo- PALMER J.H.L., BELSHAW D.J. 1980: Defor- mech. 28 (14): 1441–1460. mations and pore pressure in the vicinity of LIKITLERSUANG S., TEACHAVORASIN- a precast, segmented, concrete-lined tunnel in SKUN S., SURARAK C., OH E., BALASUB- clay. Can. Geotech. J. 17 (2): 174–184. RAMANIAM A. 2013: Small strain stiffness PECK R.B. 1969: Deep excavations and tunnel- and stiffness degradation curve of Bangkok ling in soft ground. In: Proceedings of the 7th Clays. Soils Found. 53 (4): 498–509. International Conference on Soil Mechanics MAIR R.J. 2008: Tunnelling and geotechnics: and Foundation Engineering: 225–290. new horizons. Géotechnique 58 (9): 695–736. SEAH T.H., KOSLANANT S. 2003: Anisotropic MAIR R.J., TAYLOR R.N. 1997: Bored tunnel- consolidation behavior of soft Bangkok clay. ling in the urban environment: State-of-the-art Geotech. Test. J. 26 (3): 266–276. report and theme lecture. In: Proceedings of SHIRLAW J.N. 1995: Observed and calculated 14th International Conference on Soil Mechan- pore pressures and deformation induced by ics and Foundation Engineering, Hamburg. earth pressure balance shield. Can. Geotech. J. Vol. 4: 2353–2385. 32: 181–189. MAŠÍN D. 2005: A hypoplastic constitutive mod- SIRIVACHIRAPORN A., PHIENWEJ N. 2012: el for clays. Int. J. Num. Anal. Meth. Geomech. Ground movements in EPB shield tunneling of 29 (4): 311–336. Bangkok subway project and impacts on adja- MESCHKE G., KROPIK C., MANG H.A. 1996: cent buildings. Tunn. Undergr. Sp. Tech. 30: Numerical analyses of tunnel lining by means 10–24. of a viscoplastic material model for shotcrete. SURARAK C. 2010: Geotechnical Aspects of the Int. J. Num. Meth. Eng. 39 (18): 3145–3162. Bangkok MRT Blue Line Project. PhD thesis, MÜELLER-KIRCHENBAUER H. 1977: Stabil- Griffith School of Engineering, Brisbane [type- ity of slurry trenches in inhomogeneous sub- script]. soil. In: Proceedings of the 9th International SUWANSAWAT S. 2002: Earth Pressure Balance Conference on Soil Mechanics and Foundation (EPB) Shield Tunneling in Bangkok: Ground Engineering. Tokyo. Vol. 2: 125–132. Response and Prediction of Surface Settle- Long-term settlement induced by EPB tunnelling studied... 157

ments Using Artificial Neural Networks. PhD pressure balance – EPB) przez przeprowadzenie thesis, Massachusetts Institute of Technology analiz numerycznych deformacji podłoża grun- [typescript]. towego. W tym celu zbudowano trójwymiarowy SUWANSAWAT S., EINSTEIN H.H. 2007: De- model numeryczny szczegółowo odzwierciedlają- scribing settlement troughs over twin tunnels cy proces tunelowania, geometrię analizowanego using a superposition technique. J. Geotech. problemu oraz odpowiedź mechaniczną różnych Geoenviron. Eng. 133 (4): 445–468. materiałów. Model obliczeniowy został zweryfi- WARD W.H., THOMAS H.S.H. 1965: The de- kowany w odniesieniu do rzeczywistego tunelu velopment of earth loading and deformation wydrążonego w Bangkoku, a następnie wykorzy- in tunnel linings in London Clay. In: Proceed- stany do symulacji procesu tunelowania w różnych ings of 6th International Conference on Soil warunkach gruntowo-wodnych. Przeprowadzone Mechanics Foundation Engineering, Toronto. symulacje przedstawiają wpływ obniżenia zwier- Vol. 2: 432–436. ciadła wody gruntowej po wydrążeniu tunelu przy von WOLFFERSDORFF P.A. 1996: A hypoplas- założeniu w pełni przepuszczalnej i nieprzepusz- tic relation for granular materials with a prede- czalnej segmentowej obudowy tunelu na rozwój fined limit state surface. Mech. Cohesive-Fric- osiadań podłoża gruntowego w czasie. tional Mat. 1 (3): 251–271. WU H., XU Y., SHEN S., CHAI J. 2011: Long- Słowa kluczowe: symulacje numeryczne, tarcza -term settlement behavior of ground around EPB, konsolidacja shield tunnel due to leakage of water in soft deposit of Shanghai. Front. Arch. Civil Eng. China 5 (2): 194–198. MS received 23.01.2018 MS accepted 04.06.2018 Streszczenie: Analizy numeryczne długotermino- wego osiadanie gruntu wywołanego drążeniem tunelu tarczą wyrównanych ciśnień gruntowych. Author’s address: Mimo istnienia wielu dowodów przedstawiają- Maciej Ochmański cych rozwój w czasie osiadań wywołanych proce- Katedra Geotechniki i Dróg sem tunelowania w spoistym podłożu gruntowym, Wydział Budownictwa tematyka ta rzadko jest przedmiotem badań nauko- Politechnika Śląska wych. Celem artykułu jest zbadanie zmiany w cza- ul. Akademicka 5, 44-100 Gliwice sie efektów wywołanych drążeniem tunelu tarczą Poland wyrównanych ciśnień gruntowych (ang. earth e-mail: [email protected]