Deep Excavation of the Gate of the Orient in Suzhou Stiff Clay: Composite Earth-Retaining Systems and Dewatering Plans

Fig. 7(a) depicts variations in phreatic levels at SW1 to SW20, approximately 2 m behind the CBP walls. During excavation of the uppermost 9.25 m of soils in Stages S1 to S2, phreatic levels behind the north wall (SW17 to SW20), west wall (SW12 to SW16), and south wall (SW7 to SW9) had 4 to 11 m drawdown; i.e., groundwater levels were satisfactorily lowered below the open-cut surface. In contrast, there were only small decreases, no more than 2.0 m, in phreatic levels behind the east wall (SW1 to SW6). Because no light well point was installed along the east pit side (Fig. 6), the monitored data indicated that the waterproof curtains (Figs. 1 and 3) successfully cut off potential seepage flow during dewatering for removal of the uppermost 9.25 m of soils.

Fig. 7(b) shows variations in the artesian levels at CY1 to CY4, approximately 25 m behind the CBP walls. In Stages S3 to S4, the monitored artesian levels at CY1 to CY4 had 10–14 m drawdown and then maintained these levels in subsequent Stages S4 to S7. Obviously, pumping of the confined aquifer water inside the pit caused significant drawdown in the artesian levels outside the pit. Fig. 8(a) illustrates the locations of the observation wells behind the pit and the pumping wells inside the pit relative to the CBP wall and waterproof curtain along the east pit side. As outlined in this figure, the waterproof curtain had not extended deeply into the confined aquifer layer at 38–44 m BGS; i.e., the water flow path between the pit and the ground outside had not been cut off. Consequently, pumping of the confined aquifer by relief wells inside the pit led to dramatic drawdown in the artesian level outside the pit. As shown in Fig. 7(a), phreatic levels did not show substantial variations in Stages S3 to S7; i.e., dewatering of the confined aquifer water did not cause apparent leaking or seepage flow in the upper clayey strata (aquitard) featuring low permeability (Fig. 2). Following completion of the base slab, all pumping wells were shut off upon the end of Stage S7, and artesian levels recovered rapidly to the original levels within approximately 40 days.

Fig. 9(a) presents typical lateral CBP wall displacements at P2, P6, P11, and P13 at different construction stages. Apparently, P6, P11, and P13 developed smaller displacements than P2 on the east pit side; this discrepancy should result from the open cut, which reduced lateral earth pressures behind the CBP walls along the other three pit sides. Fig. 9(b) shows corresponding lateral ground displacements behind the CBP walls, i.e., T2 at 1 m behind P2, T5 at 13.5 m behind P6, T8 at 23.5 m behind P11, and T10 at 23.5 m behind P13. Both the CBP walls and ground developed typical deep-seated displacements in Stages S2 to S5, and their displacement developments synchronized with each other. Benefitting from the SNW and jet-grouting piles, excavation of the inner pits just incurred very limited additional wall and ground displacements in Stage S6. During the subsequent 31 days for pouring the base slab (Stage S7), neither the CBP walls nor the ground underwent apparent time-dependent displacement; this differed from those excavations in Shanghai soft clay (Tan et al. 2015b) also constructed by bottom-up (BU) method. It was evident that stiff clay hardly featured creep behavior. However, both the CBP walls and ground underwent noticeable postexcavation lateral displacements up to 20 mm in 1 year (Stage S8). Fig. 10 plots typical time histories of settlements for the buried communication cable at D1 to D13 and bridge abutments at X1 to X4 behind the pit (Fig. 1). The shallowly buried pipelines and bridge abutments experienced remarkable settlements during excavation (Stages S1 to S7); however, all of them developed very limited settlements in Stage S8. This was clearly indicative of limited consolidation-induced ground settlement postexcavation. Moreover, as shown in Fig. 7, the confined aquifer levels recovered rapidly to their original levels after completion of excavation, which would push up the overlying impermeable clayey strata (aquitard) somewhat. In light of these, it can be reasonably deduced that the significant postexcavation lateral wall and ground displacements should be largely associated with demolishing the rigid concrete struts in Stage S8 for casting permanent underground structure elements (e.g., inner side walls and floor slabs) rather than dissipation of excess pore pressure (consolidation) in the ground. Dismantling of the concrete struts caused an immediate reduction in supporting system stiffness; moreover, it would take poured underground structural elements at least 28 days to get fully cured (gain sufficient strength) for resisting lateral wall displacements. As a result, remarkable postexcavation lateral wall displacements occurred in Stage S8, followed by lateral ground displacement directly behind.

By comparing the lateral wall and ground displacements, it can be seen that the lateral ground displacements of T2 immediately behind the east pit side were comparable to the corresponding lateral wall displacements of P2. Even at a distance of 23.5 m, the lateral ground displacements of T8 and T10 behind the open cuts were only slightly smaller than the lateral wall displacements of P11 and P13. These observations reveal that the excavation-induced lateral ground displacement zone extended far beyond $1.1H_e$ behind the pit, where $H_e$ denotes the final excavation depth at 21.5 m BGS. This is demonstrated by the measured ground settlement to be presented later.

Fig. 11(a) plots the relationship between the maximum lateral displacement ${\delta }_{hm}$ of CBP walls and excavation level $H$; Fig. 11(b) plots the relationship between $H$ and depth $H_m$ with ${\delta }_{hm}$. Because lateral displacements of CBP walls were not measured during removal of the uppermost 7.85 m soils, $H$ in Fig. 11 was the excavation depth depicted in Table 1 subtracted by 7.85 m. To better understand excavation performance in Suzhou stiff clay, field data from another 21 excavations in Suzhou summarized in Table 2 were also included. Suzhou is a city located on the alluvial deposit—Yangtze River Delta plain. Similar geological and hydrological conditions feature across the entire city; i.e., all the cases in Table 2 are comparable. Except for Case 16 (Tan et al. 2015a) and Case 20 following zoned-excavation procedure, as well as Case 17 following the top-down (TD) excavation procedure, all these excavations adopted a typical bottom-up (BU) procedure. Moreover, empirical envelopes in the literature were included in Fig. 11 for comparison, i.e., Clough and O’Rourke (1990) and Moormann (2004) for stiff clays worldwide, Hashash et al. (2008) for the Central Artery/Tunnel (CA/T) project in medium stiff Boston Blue clay (BBC), and Tan and Wang (2013a, b) for BU and TD excavations in Shanghai soft clay. As shown in Fig. 11(a), the east CBP wall with a SNW had slightly greater ${\delta }_{hm}$ than the other CBP walls with open cut in this project, and their ${\delta }_{hm}$ ranged between ${\delta }_{hm}=0.10\%H$ and ${\delta }_{hm}=0.50\%H$, which agreed with those Suzhou excavations retained by multipropped CBP walls but was greater than those supported by multipropped diaphragm walls (DWs). Compared with the envelopes in literature, the upper boundary for Suzhou excavations (${\delta }_{hm}=0.50\%H$) matched with that of Clough and O’Rourke (1990), but was only half of that (${\delta }_{hm}=1.0\%H$) of Moormann (2004). Compared with those excavations in Shanghai soft clay with ${\delta }_{hm}\le 1.0–1.2\%H$ (Tan and Wang 2013a, b), ${\delta }_{hm}$ in Suzhou stiff clay were much smaller. Despite BBC being weaker than Suzhou clay, ${\delta }_{hm}$ reported in Hashash et al. (2008) was considerably smaller than those in Suzhou. This should mainly arise from the narrow excavation widths (17.7–32.9 m) of the CA/T, whose braced struts were much shorter than those in this case. From the viewpoint of material mechanics, a short rod (strut) is much stronger than a long rod (strut) against buckling when subjected to axial loading because of its smaller slenderness ratio.

Distinct from excavations in Shanghai in which ${\delta }_{hm}$ occurred at depths within 7 m below and above excavation surfaces ($H_m=H-7m$ to $H_m=H+7m$), most ${\delta }_{hm}$ in Suzhou occurred above the excavation surface ($H_m=H-6m$ to $H_m=H$ for those retained by propped CBP walls and $H_m=H-12m$ to $H_m=H$ for those by propped DWs). As summarized in Table 2, those cases in Suzhou had final excavation depths $H_e$ of 5.15–26 m and retaining wall lengths $L_w$ of 10–46 m. With the exception of Case 11, retaining walls of all the cases were completely embedded in the upper clayey strata and wall toes did not penetrate into the underlying silty sand layer. As for Case 11, it had an excavation depth of 26 m and wall embedment length (wall length below excavation base) $L_e$ of 20 m; i.e., approximately 12 m of $L_e$ was in the upper clayey strata and 8 m of $L_e$ was in the underlying silty sand. Both the clayey strata and silty sand layer in Suzhou were stronger than those of Shanghai. Therefore, the discrepancy for $H_m$ in Suzhou and Shanghai should be associated with the much stiffer clay strata or denser silty sand in Suzhou; i.e., the soils below excavation level in Suzhou could provide larger resistance against inward lateral wall deflection than those in Shanghai. Consequently, ${\delta }_{hm}$ in Suzhou had a much greater possibility of occurring above excavation surfaces than ${\delta }_{hm}$ in Shanghai.

Fig. 14 presents distribution of ${\delta }_{fv}$ along distance $d$ from the pit, in which both ${\delta }_{fv}$ and $d$ were normalized by $H_e$. Furthermore, data available from Cases 11, 12, 16, and 17 and Moormann (2004), as well as the empirical envelopes in literature (Peck 1969; O’Rourke 1976; Clough and O’Rourke 1990; Hashash et al. 2008), were included. Obviously, ${\delta }_{fv}$ behind this multipropped excavation had spandrel-type profiles typical for excavations retained by cantilever wall (e.g., Peck 1969; Clough and O’Rourke 1990) instead of the concave type typical for multipropped excavations (e.g., Ou et al. 1993; Tan and Wang 2013a, b). This inconsistency should derive from excavation of the uppermost 7.85 m soils being achieved by open cut or retained by SNWs. Evidently, the composite earth-retaining systems adopted in this excavation made its ground settlements distinctively different from those of other multipropped excavations. Within $1.5H_e$ from the pit, the ground along C1-1 to C1-6 behind the east pit middle span had greater ${\delta }_{fv}$ than the corresponding C2-1 to C2-6 and C3-1 to C3-4 behind the west and north pit middle span. In spite of this discrepancy, they had similar ground settlement influence zones, which can be bounded by envelope (5). Along the same side, C1-1 to C1-6 behind the middle span had apparently greater settlements than both C5-1 to C5-5 and C6-1 to C6-5 distance away, which exhibited clear evidence of spatial corner stiffening behavior like those cases in Tan et al. (2014).

For this excavation, zone I of Peck (1969) for sand and soft to hard clay highly underestimated its ${\delta }_{fv}/H_e$ and ground settlement zone; ${\delta }_{fv}/H_e$ in this case was up to 0.8% within zone II of Peck (1969) for very soft to soft clay. Envelope (2) of O’Rourke (1976) for soft to medium stiff Chicago glacial clay would highly overestimate ${\delta }_{fv}/H_e$ within $1.5H_e$ for this case, but underestimate its ${\delta }_{fv}/H_e$ beyond $2H_e$. Regarding envelope (3) of Clough and O’Rourke (1990) for stiff to hard clay, it could make a reliable estimation on the ground settlement zone, but would highly underestimate ${\delta }_{fv}/H_e$. Envelope (4) of Hashash et al. (2008) for the CA/T project in medium stiff BBC would highly underestimate ${\delta }_{fv}/H_e$, but overestimate the ground settlement zone. As for the other four Suzhou excavations, envelope (3) of Clough and O’Rourke (1990) could make a good prediction in terms of both ${\delta }_{fv}/H_e$ and the ground settlement zone. With the exception of this case, most ${\delta }_{fv}/H_e$ in Suzhou and Moormann (2004) were within zone I of Peck (1969). Apparently, the measured ground settlements in this case were much greater than those of Cases 11, 12, 16, and 17 in Suzhou.

It has been recognized in the literature (e.g., Galloway et al. 1999; Xu et al. 2016; Wu et al. 2017) that discharging of ground water (phreatic and artesian water) could cause pronounced land subsidence. Considering the significant drawdown in artesian water levels behind this excavation (Fig. 7), it was suspected that the exceptionally larger ground settlements of this excavation compared to the other excavation cases in Suzhou might be associated with dewatering. As illustrated in Table 2, Case 11 had $H_e=26m$ and $L_w=46\mathrm{m}$; i.e., its DW had penetrated through the confined aquifer at 38–44 m BGS and extended into the underlying impermeable clay and silty clay stratum (aquitard). Because the DW had relatively good watertightness, pumping of aquifer water inside the pit for releasing artesian pressure in Case 11 would not cause substantial drawdown in the water level outside the pit; i.e., dewatering-induced ground settlement would not be significant for Case 11. As for Cases 12, 16, and 17, their $H_e$ were no more than 20 m, and dewatering for releasing underlying confined aquifer pressure was not required; i.e., as mentioned previously, for excavations with $H_e\le 20\mathrm{m}$ in Suzhou, the weight of overburden soils was able to resist upward confined aquifer pressure. Like this case, the DWs of Cases 12, 16, and 17 were toed in the upper clayey strata and did not penetrate into the underlying confined aquifer. In light of the good watertightness of the DW and the low permeability of the upper clayey strata (aquitard), pumping of phreatic water inside the pit would not cause a substantial drawdown in the water level outside the pit for Cases 12, 16, and 17. Different from Cases 12, 16, and 17, pumping of confined aquifer water inside the pit was conducted in this case because of its great excavation depth (21.5–30.7 m). However, its waterproof curtain was toed at 27.3 m BGS and did not penetrate deeply through the confined aquifer stratum [see Figs. 3 and 8(a)]. Consequently, significant drawdown in the water level outside the pit was monitored during dewatering (see Fig. 7). On the basis of these analyses, it was reasonable to infer that the dramatically larger ground settlements of this excavation compared to those of Cases 11, 12, 16, and 17 should, to some extent, result from the significant drawdown in the aquifer level.

The preceding analyses disclose that the drawdown in water level imposed a significant adverse influence on ground settlement. As plotted in Fig. 7(a), the measured drawdown in phreatic level by the 20-m-deep observation wells at 2 m behind CBP walls was up to 2 m along the east pit side and up to 11 m along the other three sides. The drawdown in the underlying confined aquifer level measured by the 40-m-deep observation wells at 25 m behind the CBP walls was up to 14 m along the pit perimeter; see Fig. 7(b). As schematically illustrated in Fig. 8(b), drawdown of the water (phreatic or artesian) level in the ground behind the pit would incur consolidation-induced settlement $S_t$ for the upper 38 m silty clay and clayey silt strata (aquitard) and compression $S_c$ of the underlying silty sand layer (confined aquifer). Therefore, dewatering-induced total ground settlement $S$ can be given by

As for the upper aquitard [Fig. 8(b)], dewatering of its phreatic water levels from $h_1$ to $h_2$ caused increment of vertical effective soil stress $\mathrm{\Delta }{\sigma }_1^{\prime }$ with a magnitude of $\mathrm{\Delta }{\sigma }_1^{\prime }=u_1={\gamma }_{\omega }(h_1-h_2)$ in the aquitard, in which ${\gamma }_{\omega }$ = density of water, $h_1$ = phreatic level before dewatering, and $h_2$ = phreatic level after dewatering. Meanwhile, releasing (dewatering) of artesian water in its underlying confined aquifer led to reduction of $u_2=({\gamma }_{\omega }{h}_3-{\gamma }_{\omega }{h}_4)$ in the upward artesian pressure against the aquitard bottom, which can be treated as a downward pressure $\mathrm{\Delta }{\sigma }_2^{\prime }=u_2$ applied to the bottom of the upper aquitard after dewatering, where $h_3$ = confined aquifer level before dewatering and $h_4$ = confined aquifer level after dewatering. According to the one-dimensional consolidation theory of Terzaghi (1925), the change in pressures acting on the boundaries of soil strata would result in consolidation (compression). Although the water level in the upper perched aquifer at 6–12 m BGS was not monitored during excavation, it should rarely affect ground settlement. As illustrated in Fig. 8, the waterproof curtains (CBP wall and MIP) extended to 27.3–34.05 m BGS, and this perched aquifer was intermediate between the two impermeable aquitards; therefore, dewatering conducted inside the excavation hardly influenced the water level of the perched aquifer outside the excavation. Hence, it can be postulated that the upper perched aquifer should rarely affect ground settlement behind the excavation. Based upon the preceding analyses, it can be concluded that for the upper 38-m-thick soil strata overlying the confined aquifer at 38–44 m BGS, its $S_t$ primarily consisted of $S_{t1}$ due to drawdown $u_1$ in its phreatic level and $S_{t2}$ due to drawdown $u_2$ in its underlying confined aquifer level. In accordance with the one-dimensional consolidation theory of Terzaghi (1925), $S_t$ can be approximately estimated by the following equations:where, $S_t$ = consolidation-induced ground settlement at time $t$ for the upper aquitard; $S_{t1}=S_t$ associated with drawdown in phreatic level; $S_{t2}=S_t$ associated with drawdown in the underlying confined aquifer level; $U$ = consolidation ratio; $S_{\infty }$ = final consolidation-induced ground settlement; $S_{\infty 1}=S_{\infty }$ associated with drawdown in the phreatic level; $S_{\infty 2}=S_{\infty }$ associated with drawdown in the underlying confined aquifer level; $H_1$ = thickness of the upper aquitard, equal to 38 m; $a_v$ = compressibility coefficient of the upper aquitard; $e_0$ = initial void ratio of the upper aquitard; $M$ = constrained modulus of the upper aquitard; and $k$ = permeability coefficient of the upper aquitard.

As shown in Fig. 7, the reductions in the phreatic level and underlying confined aquifer level were up to 2 and 14 m behind the east pit side ($\mathrm{\Delta }{h}_1=2\mathrm{m}$; $\mathrm{\Delta }{h}_2=14\mathrm{m}$) and 11 and 14 m behind the other three sides ($\mathrm{\Delta }{h}_1=11\mathrm{m}$; $\mathrm{\Delta }{h}_2=14\mathrm{m}$). Thereafter, the phreatic levels showed limited variation throughout Stages S3 to S7. According to Eqs. (2)–(10), the roughly estimated $S_t$ during excavation was approximately 68 and 55 mm for the upper aquitard behind the east and the other three sides, respectively.

Regarding the underlying confined aquifer layer (silty sand), drawdown in the aquifer level would increase its effective stress with a magnitude of ${\gamma }_{\omega }·(h_3-h_4)$, which would lead to compression settlement $S_c$ of the silty sand layer. It can be approximately estimated bywhere $H_2$ = thickness of the confined aquifer layer, equal to 6 m; and $M$ = constrained modulus of the silty sand (confined aquifer layer). The estimated $S_c$ could be up to 60 mm. Based on the previous analysis, the total ground settlement ($S=S_t+S_c$) related to dewatering could be up to 128 mm behind the east pit side and 115 mm behind the other three sides.

As illustrated in Fig. 8(a), the 44-m-deep relief wells for releasing confined aquifer pressure were completely penetrating wells; i.e., they penetrated through the confined aquifer and reached the underlying aquitard. The influence zone due to pumping of aquifer water can be roughly delineated with the empirical equation proposed in Liu and Wang (2009), i.e.,where $R$ = radius of influence from relief well; $\mathrm{\Delta }h$ = drawdown in aquifer level; and $k$ = permeability of confined aquifer layer measured by in-situ injectivity test. The estimated $R$ could be up to 45 m distant from this pit, more than $2H_e$. This in part explains why the ground settlement of this case measured at a distance of $2H_e$ behind the pit was still significantly greater than the other cases in Suzhou; see Fig. 14.

The previous theoretical analyses disclose that the majority of the pronounced ground settlement measured behind this excavation should derive from the significant drawdown in the aquifer level outside the pit, whose influence could reach as far as $2H_e$ behind the pit. The significant drawdown in the water level largely arose from the previously mentioned deficiency in the waterproof design. This deficiency led to the water flow path in the confined aquifer layer not being cut off between the pit and the ground outside [Fig. 8(a)] during discharging underlying confined aquifer water inside the pit; consequently, a significant drawdown in water level was monitored behind the pit. If such a project were constructed in a densely populated urban environment, a safer design would have to be adopted to minimize ground subsidence to protect adjacent structures or facilities, e.g., (1) extending the waterproof curtain deeply enough to penetrate through the underlying confined aquifer despite the risk of increasing project cost dramatically, or (2) recharging the underlying confined aquifer with the water discharged by relief wells inside the pit in the proximity of the structures or facilities to be protected (see the illustrations in Fig. 15).