Damage analysis of the new Sanyi railway tunnel in the 1999 Chi-Chi earthquake: Necessity of second lining reinforcement

Tunnelling and Underground Space Technology 73 (2018) 48–59

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology journal homepage: www.elsevier.com/locate/tust

Damage analysis of the new Sanyi railway tunnel in the 1999 Chi-Chi earthquake: Necessity of second lining reinforcement

T

⁎

Chih-Chieh Lua, , Jin-Hung Hwangb a b

National Center for Research on Earthquake Engineering, No. 200, Sec. 3, Xinhai Rd., Taipei 10668, Taiwan, ROC Department of Civil Engineering, National Central University, No. 300, Jhongda Rd., Jhongli City, Taoyuan County 32001, Taiwan, ROC

A R T I C L E I N F O

A B S T R A C T

Keywords: MCSRD Seismic Mountainous tunnel NATM Reinforcement

This study documented the case history of the new Sanyi railway tunnel. The New Austrian tunneling method (NATM)-built mountainous tunnel was seriously damaged during the 1999 Chi-Chi earthquake. In order to better understand the vulnerability and the deformation of the underground structure subject to strong ground motion, the modified cross-section racking deformation (MCSRD) method was used to evaluate the seismic performance of the tunnel. The analyses carefully took the nonlinear soil-structure interaction into account in order to derive damage pattern of the tunnel. Based on the analysis results, in addition to the extremely strong ground motion, numerical testing identified the causes of the disaster to include rectangle-like geometry at the refuge section, non-reinforcement of the second lining, and imperfect backfilling. The results also showed that the second lining of the NATM-built tunnel sustained substantial seismic loading and should be suitably reinforced in seismically active areas. The effect of the reinforcement of the second lining was demonstrated with a reinforcement example.

1. Introduction The seismic performance of underground structures has become an important issue due to an increase in incidents of damaged tunnels caused by catastrophic earthquake events such as the 1923 Kanto earthquake, 1995 Kobe earthquake, 1999 Chi-Chi earthquake, and 2008 Wenchuan earthquake (Okamoto, 1973; Asakura and Sato, 1996; Wang et al., 2000; Hwang and Lu, 2007; Lu and Hwang, 2008; Wang and Zhang, 2013; Shen et al., 2014). As demonstrated by historical evidence, underground structures are still at a high risk of damage from compression by the surrounding ground, triggered by strong shaking. The documented case histories of damaged tunnels are valuable and worth further study for the practical engineer to take the lesson learned for future design work. In this study, the object of interest was the new Sanyi railway tunnel in central Taiwan. The NATM-built mountainous tunnel was seriously damaged by the 1999 Chi-Chi earthquake. Because of this event, the NATM concept that the second lining will not take any loading should be modified, and it is strongly recommended that the second lining in a seismically active area be reinforced. The purpose of this study was to determine the nonlinear seismic behavior of the studied mountainous tunnel subject to strong ground motion, and provide a guide to the reinforcement of a second lining according to the results of the parameter sensitive analysis.

⁎

To assess the seismic performance of underground structures, several references provide an overview of design work regarding seismic issues in underground structures; these include FHWA (2009), ISO 23469 (2005), and Hashash et al. (2001). According to these references, the seismic evaluation approach can be roughly divided into three categories, which are the CSRD (cross-section racking deformation) method, the MCSRD method, and full dynamic analysis. Of these, the pseudo-static CSRD and MCSRD methods prescribe a seismic ground deformation and consider the interaction between the underground structure and the surrounding ground using certain assumptions (Wang, 1993; Penzien, 2000; Nishioka and Unjoh, 2002; Gil et al., 2001; Huo et al, 2006; Kontoe et al., 2008, 2014; Park et al, 2009; Lu and Hwang, 2017). Note that the contribution of inertial force to the underground structure during an earthquake is ignored in these pseudo-static approaches. The CSRD and MCSRD methods are thought to be more applicable to the cases when the contribution of inertial force to the seismic performance of the concerned underground structure is small, i.e. mountainous tunnels. In contrast, the full dynamic analysis can consider the dynamic characteristics of an underground structure and the surrounding ground during dynamic loading without ignoring the inertial force, and it is believed that the comprehensive approach can better capture the seismic performance of an underground structure. To evaluate the performance of these methods, Tsinidis et al. (2016a,

Corresponding author. E-mail addresses: [email protected] (C.-C. Lu), [email protected] (J.-H. Hwang).

https://doi.org/10.1016/j.tust.2017.12.009 Received 29 June 2017; Received in revised form 8 November 2017; Accepted 6 December 2017 0886-7798/ © 2017 Elsevier Ltd. All rights reserved.

Tunnelling and Underground Space Technology 73 (2018) 48–59

C.-C. Lu, J.-H. Hwang

overburden depths range from 20 to 150 m. The ground formations that the tunnel passes through are the Miocene Kuantaoshan Sandstone, the Shihliufeng Shale, the Toukeshan Gravel, and the river gravel terrace. The tunnel crosses over two fault zones, Sanyi fault and Shihliufen fault. Based on engineering characteristics, the rock mass along the tunnel can be classified into 6 grades, as shown in Table 1, and the geological profile accompanying with construction conditions during tunneling are shown in Fig. 2. In addition, the geological investigation, conducted by United Geotech (1989), indicated that the ground formations can be roughly divided into eight kinds of formations. The physical and mechanical properties of these formations are summarized in Table 2. Since the tunnel cross section was designed for an electric doubletrack railway system, waterproofing was installed to prevent leakage of ground water. Furthermore, to satisfy the demands of operation and maintenance, refuge spaces were excavated on the side wall at 20 m intervals for small refuges and 300 m intervals for large ones. The three types of cross-sections, including the standard one, of the tunnel are shown in Fig. 3. In general, the tunneling adopted the Drill and Blasting (D&B) method with bench excavation. When difficult ground conditions were encountered, ground treatment or special excavation methods were adopted. From the monitored records during construction, the horizontal convergence deformation ranged from 15 to 555 mm, and the crown settlement ranged from 2 to 155 mm in most typical sections. The problems encountered during excavation included (1) roof spalling due to fractured rock, (2) rock mass sliding along the planes of cleavage and join, and (3) rock softening owing to ground water leakage. The problematic locations are indicated in Fig. 2.

2016b), and Tsinidis (2017) conducted a series of centrifuge tests that demonstrate better performance of the full dynamic analysis when compared with the test results. It is recommended to conduct a full dynamic analysis for the seismic performance of an underground structure if the target of concern is near the ground surface and confined by a soft medium, i.e. an urban shallow tunnel. Since the rigorous dynamic analysis has a more complicated theoretical basis and is time consuming to use in comparison with the pseudo-static approaches, the CSRD and MCSRD methods are favored in practical design work if the inertial force does not obviously affect the seismic behavior of the studied case, such as a mountainous tunnel, like the old Sanyi railway tunnel and new Sanyi railway tunnel. Hwang and Lu (2007) documented the information for the old Sanyi railway tunnel and employed the MCSRD method to assess the seismic capacity of the mountainous railway tunnel in terms of peak ground velocity (PGV), seismic shear strain, and JMA intensity scale. The evaluated seismic capacity of the tunnel by the MCSRD method agreed well with the field observation during 1935 Hsinchu-Taichung and 1999 Chi-Chi earthquakes. Also, the MCSRD method can capture the seismic mechanism of the tunnel in comparison with the results by dynamic analysis. Contrary to the old Sanyi railway tunnel safely coming through the 1999 Chi-Chi earthquake, it is surprising that the “new” Sanyi railway tunnel, which is near the old ones, was seriously damaged by the shaking of the earthquake. To figure out the failure mechanism of the tunnel, a preliminary study of the seismic performance for the new Sanyi railway tunnel during the 1999 Chi-Chi earthquake was conducted by Lu and Hwang (2008) using the MCSRD method. In order to simulate the damage pattern of the tunnel, the strength of the second lining was reduced to 10% of its initial strength when the loading condition is over the corresponding P-M strength curve. The simplified approach, which gave a reduction in the strength factor of the yielding structural member to model the nonlinear mechanism of structural elements, was improved in 2017. Lu and Hwang (2017) compiled the Axial forceMoment-Curvature (A-M-C) surface of the structural elements in FLAC2D using the program’s built-in FISH language to model the nonlinear behavior of the structure. When this model is combined with the FLAC built-in nonlinear constitutive model for the geotechnical material, the nonlinear interaction between the underground structure and the surrounding ground can be well understood. The framework of the MCSRD method was also modified accordingly. To better reflect the seismic performance of the new Sanyi railway tunnel, the suggested implementation of the MCSRD method by Lu and Hwang (2017) was adopted to re-evaluate the failure mechanism of the most seriously damaged section at Sta. 161K+300 and review the inferred damaging factors including excessive earthquake shaking, rectangle-like geometry at the refuge section, imperfect backfilling, nonreinforced second lining, and geological weak zone (Hsu and Weng, 2000). From the results of numerical analyses, the simulation results agreed well with the field damage pattern observed after the 1999 ChiChi earthquake. Besides the bad geological conditions, the results indicated that rectangle-like geometry at the refuge section and the unreinforced second lining were the main causes of damage to the tunnel during the earthquake. The results also showed that the second lining sustained substantial seismic loading and the seismic capacity of the lining could increase significantly when the amount of reinforcement of the second lining is over a threshold value. Thus, the second lining of the NATM tunnel should be suitably reinforced in a seismically active area.

2.2. Seismic ground motions On 21 September 1999, a strong quake with a magnitude of 7.3 on the Richter scale occurred near the town of Chi-Chi. A large earthquake like this had not been experienced in Taiwan for over 100 years. The maximum ground accelerations measured by the strong motion seismographic stations in the Nantou-Wufeng area were as high as 0.7–1.0 g. It caused significant damage in nearby areas in central Taiwan. Because the seismographic stations in that area are dense, the acceleration history records of the earthquake near the concerned area were available, which can be obtained from the nearby seismographic stations. The closest seismographic station, 5 km from the tunnel, is Jian-Jhong elementary school, and its recording acceleration history is used in the following analysis. Since the tunnel axis runs in a roughly South-North direction, the critical motion is supposed to be in the eastwest direction which mainly provided the racking action on the concerned tunnel section. Fig. 4 shows the acceleration, velocity, and displacement histories in the E-W direction at that station. Note that although the maximum ground acceleration is only 0.14 g, the maximum velocity is 26.5 cm/s, which is a more important motion index than acceleration when using MCSRD method to assess the seismic capacity of the tunnel. 2.3. Field investigation after the 1999 Chi-Chi earthquake The new Sanyi railway tunnel was one of the most seriously damaged tunnels after experiencing the strong shaking of the 1999 ChiChi earthquake. It is located in the western foothills of central Taiwan, which passes through the Sanyi fault zone, Shihliufen fault zone, and other weak ground. The new Sanyi railway tunnel was designed and excavated using NATM, the total length of the tunnel is 7261 m, and its overburden ranges from 20 to 150 m. The tunnel was completed and opened to vehicle traffic in 1998 to replace the old Sanyi railway tunnels built in 1908. Unfortunately, after operating for just one year, the tunnel was shaken by the 1999 Chi-Chi earthquake and severely damaged. Railway traffic was interrupted for several days by this catastrophic earthquake. After field investigation, eight main damaged

2. New Sanyi railway tunnel 2.1. Basic information The new Sanyi railway tunnel is in central Taiwan and passes through a series of small mountain ridges and terraces neighboring a small valley, as shown in the Digital Terrain Model (DTM) of Fig. 1. Its 49

Tunnelling and Underground Space Technology 73 (2018) 48–59

C.-C. Lu, J.-H. Hwang

Fig. 1. The geographical location of the tunnel and the surrounding topography (after Hwang and Lu, 2007).

Liyutan Reservoir SUNSEN SANYI

1

2

Shihkang Dam

3 4 5 6 7 8

Seismographic station

New

Old railway 9

railw

ay

FONGYUAN

New Sanyi railway tunnel

Sun Ya

t-sen Fr

Note:

N

eeway

= The number of old Sanyi railway tunnel

Table 1 Rock mass classification of the new Sanyi railway tunnel (after Hsu and Weng, 2000). Classification

Rating

Description

Stable

I

Slight fracture

II

Moderate fracture or swelling

III

High fracture or swelling

IVa

Sandstone or alternations of sandstone and silty sand. Some local zones have the conditions of leakage or seepage but would not affect the strength of the rock mass Alternations of sandstone and shale, which intercalate thin shale. Degree of fracture is slight with some local severe fracture. If there is not suitable support, wedge failure would occur Alternations of sandstone and silty sand, which intercalate thin shale. Degree of fracture increases and usually coincides with fault zones. Unfavorable orientations of discontinuity exist Thin alternations of sandstone and shale. Strength of rock mass decreases due to high fractured zones and faults. Quantity of water leakage is not serious The geological characteristic is similar to IVa, but the excavated surface would be unstable due to a great quantity of water leakage and a wide fault zone Lateritic gravel. It is a different type overburden, a great part of which contains large pebbles and gravel intercalated with sandy and silty soil The geological characteristic is similar to Va, but the excavated surface would be unstable due to a great quantity of water leakage and a wide fault zone

IVb Weak and non-cohesive

Va Vb

intersection would happen at these locations when subjected to ground shaking. From the results of the field investigation, it was also found that the cross sections with refuges were most easily damaged due to their easily causing stress concentration. (3) Imperfect backfilling: Since the new Sanyi railway tunnel was constructed for an electric double-track railway system, the leakage needed to be avoided. Therefore, a complete circumferential waterproof membrane was installed after construction of the primary support to prevent the leakage of ground water. This causes an interface between the first and second linings and might generate some voids during the construction of the second lining, and the second lining not to bond tightly with the surrounding ground, which will cause spalling of the crown and sidewall concrete during strong shaking. (4) Non-reinforced second lining: Since the tunnel was designed using the NATM concept, the second lining was not reinforced with steel rebar. Therefore, it is weak in resisting shear force and bending moment. When a strong earthquake happens, a large shear force and bending moment will be induced in the second lining due to dynamic squeezing of the surrounding ground. Thus, the lining

sections with various types of damage conditions were observed along the tunnel. Some damaged conditions are shown in Fig. 5. In order to understand the damage conditions of the second linings caused by the earthquake, a non-destructive inspection technique, ground penetration radar (GPR), was used in combination with visual inspection to map the 3-D failure conditions; the mapping results are shown in Fig. 6. It was observed that there are eight main damaged sections, and the relevant information fro these sections is summarized in Table 3. Based on the field investigation and construction record, the following five damaging factors were deduced (Hsu and Weng, 2000): (1) Excessive earthquake loading: The 1999 Chi-Chi earthquake was the largest earthquake in the past 100 years in Taiwan. The violent ground motion of the earthquake seriously squeezed the underground structure, which increased the internal force of the tunnel close to its limit. (2) Rectangle-like geometry at the refuge section: The small and large refuges were symmetrically installed on side walls of the tunnel every 20 and 300 m, respectively. The irregular geometry of the tunnel caused stress concentrations at the corners, and serious 50

Tunnelling and Underground Space Technology 73 (2018) 48–59

C.-C. Lu, J.-H. Hwang

500m

0

Elevation (m)

S:1/2000

400 300 200 161K +000

Mileage

162K +000

500

163K +000

500

164K +000

500

165K +000

500

166K +000

500

167K +000

500

Ground formation Rock mass classification

Va

Construction accident

IVa

1

1

III IVa

1 21 1

Seismic damage section

III

1,2

III IVa III IVa IVIII a III IVa

3 2 1

1

IVa

III IVa

3

III

1 2

2

IVa III

2

2

Gravel

Sandstone intercalated fine shale

Sandstone and sandstone intercalated fine shale

Sanyi fault

Sandstone and alternations of sandstone and shale

Alternations of sandstone and shale

Sandstone

Thick sandstone intercalated fine shale

Shale

3

III III IVa III IVa III IVa IVa

1

5

8

Classification Stable

I II

The softening of rock mass owing to encountering water leakage

III

High fracture or swelling

IVa

Weak and non-cohesive

Va

Roof spalling after excavation due to fractured rock

3

Rating

Slight fracture

Construction accident

Rock mass sliding along cleavage and joint

1 11 1 2 3

7

Moderate fracture or swelling

2

III IVa Va IVa

Rock mass classification of the new Sanyi railway tunnel

Loessial gravel

1

III

2

1 2

6 4

IVa III IVa

III

3 1 2

Legend of ground formation

III IVa IVa

IVb

Vb

Fig. 2. Geological profile and construction conditions along the new Sanyi railway tunnel (after United Geotech, 1989).

Table 2 The geological parameters of the formations. Formation type

γ (kN/m3)

G (GPa)

B (GPa)

Vp (m/sec)

VS (m/sec)

ϕ (°)

c (kPa)

Description

A B C D E F TK FK

25.0 24.0 24.5 23.7 22.0 19.7 –a –

9.3 4.5 5.4 3.1 4.4 0.8 – –

30 6 10 8 7 3 – –

3968 2190 2803 2300 2387 1423 2000 1900

1933 1364 1484 1150 1419 638 1000 950

45 44 32.5 32 31 30 – –

85 80 60 50 4.7 50 – –

Sandstone Sandstone intercalated siltstone or shale Siltstone or Shale intercalated fine sandstone Siltstone or Shale Fault gouge and mudstone Weak bonding sandstone River gravel terrace or lateritic gravel Toukeshan gravel

a

Lack of experimental data.

section is located at Sta. 161K+300. It was therefore selected to be analyzed in detail using the modified cross section racking deformation method. The MCSRD method imposes a prescribed seismic shear deformation on the boundaries of a ground domain wherein an underground structure has been constructed to simulate the seismic ground deformation caused by an earthquake. By gradually applying seismic deformation on the boundary in the numerical model and letting the numerical program automatically calculate the complex soil-structure interaction, the MCSRD method can accurately quantify the seismic deformation of an underground structure and identify the vulnerability of the structure when subject to different magnitudes of earthquakes. Moreover, the nonlinear deformation of the underground structure system can also be identified if the nonlinear properties of the structure and soil materials are defined. The MCSRD method can better account for the complicated soilstructure interaction between the underground structure and surrounding ground compared with the CSRD method. Furthermore, the MCSRD method can capture the seismic performance of a mountainous tunnel with a relatively easy theoretical basis, and the calculation time is much shorter in comparison with the dynamic time history analysis. The MCSRD method is the preferred method for seismic design. The

will be easily damaged, and can even collapse. (5) Geological weak zones: Most of the eight damaged sections are at geological weak zones based on the field investigation and the record of construction conditions. Sta. 161K+300-400 is located at the Sanyi fault, and its overburden is about 23–65 m. There was a cave in and collapse of the support system during construction, and the crown deformation was 8–44 cm. For Sta. 164K+740-880, the ground formation is sandstone intercalated with fine-grained shale. Its mass classification is III. The crown deformation was 3.5–25 cm, and there was a nearby collapse of the support system. For Sta. 165K+620-800, the ground formation has an alternation of sandstone and shale, and its rock mass classification is III-IV. The crown deformation was about 40 cm. There were many emergent ground treatments at 165K+757-785 and 165K+805-818 while tunneling across these two sections. 3. The failure mechanism of the section at the Sta. 161K+300 3.1. The MCSRD method Along the new Sanyi railway tunnel, the most seriously damaged 51

Tunnelling and Underground Space Technology 73 (2018) 48–59

C.-C. Lu, J.-H. Hwang

Y

CL

Shotcrete

M1+2.170

M3

CL

+6.720

Shotcrete Inner lining Waterproof

Inner lining

R11 R12

+6.720

C L

-2.080

C L

eastbound

Standard cross section (non-invert)

westbound ±0.000

C L

C L

eastbound

+1.850

westbound ±0.000

X

-1.710

X

-1.710

Invert section

Non-invert section

CL

Waterproof layer

350

+2.170 Y

Shotcrete Inner lining Waterproof

+3.350

X

±0.000

-0.650 -0.800

Y

Y

200

A1

1

M2

CL

Taichung

Sanyi

A1

R1 R3

Non-invert section

Invert section

Shotcrete +0.850 ±0.000

-0.150

-0.650

A1

±0.000

1

-0.650

M3

200

450

±0.00 0

-0.650 -0.800

200

+1.100

Inner lining

350

R= 22

5

+1.850 0

Taichung

10 R=

M5 R Sanyi R3 1 A1 M2

X Cross section of large refuge hole

Cross section of standard refuge hole

Y

Standard cross section (invert) Fig. 3. The cross sections of the main tunnels and the tunnels with refuge spaces (after Hsu and Weng, 2000).

Displacement (cm) Velocity (cm/sec)

Acceleration (g)

0.2

4. Repeat steps 2 to 3 until the accumulated seismic shear-strain applied on boundaries reaches the design value, which can be evaluated by the following equation.

Max. acceleration=0.141g

0.1 0.0

γdes =

-0.1

v Vs

(1)

40 Max. velocity=26.5cm/sec

where γdes is the design seismic shear-strain, which is the maximum shear-strain of the site induced by the earthquake; v is the peak ground velocity (PGV) of the site during the earthquake; and Vs is the average shear wave velocity of the ground.

20 0 -20 15

3.2. Numerical modeling

Max. displacement=11.3cm

10 5

Before conducting the MCSRD method, to decide an appropriate boundary distance is important in order to consider the calculation time and the accuracy of the analysis results. In this study, the dimensions of the numerical mesh is about three times of the width of the tunnel after testing. The geological parameters and the grid mesh of the analyzed model are shown in Fig. 8. The soil properties herein are uniform in the whole numerical model due to the geological data, and the improvement of the surrounding ground due to the installation of support during tunneling was not considered. Of course, the improvement of the surrounding ground will reduce the external load applied to the second lining. However, it is hard to quantify the effect of rock reinforcement on the primary support during excavation. To be conservative, the authors ignored the improvement effect of primary support in this study. In addition, the grid mesh herein is uniform, while typically the grid mesh should be much denser at the location of interest, which is the zone closest to the tunnel. However, a feature of the MCSRD method is to apply seismic shear strain on the boundaries of the numerical mesh; the applied seismic shear strain should propagate from the boundary to the center of the mesh, so the near- and far- field grid mesh would be considered equally important – the density of the grid mesh in the whole model would eventually be uniform. The boundaries were

0 -5 -10 -15

0

10

20

30

40

50

60

70

80

90

time (sec) Fig. 4. Seismic motion histories in East-West direction near the tunnel (after the Central Weather Bureau).

computational steps of the MCSRD method are summarized in Fig. 7 and described below (Lu and Hwang, 2017): 1. Set up a numerical model with a suitable mesh domain; initialize insitu geostatic stresses; and construct a tunnel in the domain. 2. Apply a small seismic shear-strain with a slow shear-strain rate on the boundaries of the analyzed domain. 3. Monitor the internal forces and deformation, such as the axial force and bending moment, as well as the curvature of the lining, and force them to follow the nominal lining member’s capacity curve once they depart from the elastic state. 52

Tunnelling and Underground Space Technology 73 (2018) 48–59

C.-C. Lu, J.-H. Hwang

Fig. 5. Typical damage conditions of the tunnel (Hsu and Weng, 2000).

(a) Spalling of lining extended to the opening

(c) Longitudinal cracks of crown

(b) Spalling of crown

(d) Inclined cracks of side wall trace the nonlinear behavior of the structure element in FLAC2D. The feasibility of the above concept was demonstrated with an example of a cantilever beam subject to a prescribed rotate at the free end as shown in Fig. 10. In this study, the mentioned simplified approach was used to simulate the nonlinear behavior of the second lining of the new Sanyi railway tunnel while conducting the MCSRD method.

located 25 m distant from the center line of the tunnel. The shear strain was imposed on the boundaries with a sufficiently slow rate to make the shear strain distribution uniform throughout the analyzed domain. In the numerical model, the ground formations were simulated by solid elements obeying Mohr-Coulomb failure criteria. Based on the in-situ drilling data, the geological formation near the concerned section is type C (please refer to Table 2 for the geological parameters of the type C formation). The second lining was simulated as piece-wise beam elements with three different interface conditions between the lining and surrounding ground. The parameters of the second lining and three different considered interfaces, which represent three different qualities of backfilling, are shown in Table 4. Based on the measured PGV of 26.5 cm/s in the E-W direction, the total imposed shear strain on the boundaries was 0.0179% according to Eq. (1), and the applied shear strain rate was as slow as 5.12 × 10−5 1/s.

3.4. Results and discussion While conducting MCSRD analysis, the ground motion in the numerical domain would be stimulated by the application of seismic shear strain on the boundaries. The soil-structure interaction between earthquake-caused ground motion and the underground structure would be very significant due to the differences in stiffness and geometry of structure. Fig. 11 shows the local displacement field after applying the rightward seismic shear strain (0.0046%). The displacement field shows that the ground displacement near the tunnel is restrained by the tunnel. Fig. 12 shows the enlarged deformation behavior of a tunnel lining before and after being squeezed by the surrounding ground. By comparison between the deformed mesh (solid part) and undeformed mesh (dotted part), the tunnel was squeezed to the right by the surrounding ground due to the application of a clockwise seismic shear deformation. The internal forces on the tunnel lining in different cases are shown in Fig. 13. The result of Case 2 is not shown owing to the final collapse of the tunnel lining. It can be seen that the internal forces of the lining are quite small at the beginning (before earthquake shaking). That is compatible with the concept of NATM, which assumes the second lining does not take any forces from geostatic stresses. However, when seismic shear strain was applied to the model boundaries, stimulating the ground motion to squeeze the tunnel, the internal forces of the second lining increased significantly, especially at the corners of refuge spaces, which demonstrate that the geometry at the refuge section is unfavorable to the safety of the tunnel. In addition, it should be noted that the

3.3. Simulation of nonlinear behavior of the second lining For the simulation of nonlinear behavior of the seconding lining, because the quantity of axial force is not uniform throughout the underground structure during ground shaking, the relationship between moment and curvature subjected to different axial loading conditions needed to be derived, and the “Response2000” software was used to derive the Axial Force-Moment-Curvature (A-M-C) surface of a concrete section with different degree of reinforcement. Fig. 9 shows the A-M-C surface of the non-reinforced concrete lining of the new Sanyi railway tunnel. It is hard to have an appropriate constitutive model to well define the fully nonlinear behavior shown in Fig. 9. In order to approximately and reasonably consider the nonlinear of the structural element, a simplified approach proposed by Lu and Hwang (2017) was adopted, which first compiles the database of the Axial Force-MomentCurvature (A-M-C) surface in the FLAC2D built-in “FISH” language and adjusts the setting of the “plastic moment” according to the structural member’s corresponding A-M-C surface in each few numerical steps to 53

Tunnelling and Underground Space Technology 73 (2018) 48–59

C.-C. Lu, J.-H. Hwang

Fig. 6. Mapping results of GPR at the vault of the crown (after Hsu and Weng, 2000).

members would appear when the external loading surpasses the strength of the members. Once some of the lining segments yield, the load-carrying capacity of the lining system would be damaged and the other non-yielding members must share the extra loading. If the squeezing force of the surrounding ground keeps increasing, the other non-yielding members might also yield and finally make the whole lining system completely fail. Therefore, it is worthwhile to recognize the key members of the lining system, which could be a good reference for engineering policy regarding reinforcement. Fig. 14 shows the development of yielding mechanism along the tunnel lining in three different cases. For Case 1 and Case 3, both yielding states occur at the

axial forces at some parts of the lining were in a tensile condition after applied seismic shear strain, which is very unfavorable to non-reinforced lining and demonstrates again the weakness of the tunnel built using NATM in a seismic area. The computed results of Case 1 (with interface) and Case 3 (without interface) seem to be similar. This is because in Case 3, the void in the disturbed zone surrounding the tunnel is assumed to be filled with backfill, and a better parameter of tensional bond was therefore assigned to the interface. Since the second lining can bond well with the surrounding ground, the seismic behavior of this case tended to be similar to the case without an interface. During the application of seismic shear deformation, the yielding

Table 3 Basic information of the eight damaged sections along the new Sanyi railway tunnel after the 1999 Chi-Chi earthquake (after Wang et al., 2001). Sec.

Location

Damage typesa

Overburden (m)

Geological condition

Rock mass classification

Construction hazard

Opening

Concrete condition

①

Approx. 161k + 300

2, 4

55–65

Sanyi Fault

IVa

–

Large refuge

② ③ ④

Approx. 161k + 360 161k + 375–410 Approx. 164k + 740

2, 4 1, 3, 4 1, 4

35–45 25–35 120

Sanyi Fault Sanyi Fault

IVa IVa III

Cave-in and collapse – –

Small refuge Small refuge Small refuge

⑤ ⑥ ⑦

164k + 758–810 164k + 842–880 165k + 600–660

1, 3 1 1, 2, 3

125 130–150 105–110

III IVa III

– – –

⑧

Approx. 165k + 800

1, 4

125

– – Squeezing and support damage –

Local void existing Good Void existing Local void existing Good Void existing Void existing

Small refuge

Void existing

a

Fractured zone

IVa

Damage types: (1) longitudinal cracks; (2) transverse cracks; (3) inclined cracks; (4) cracks nearby the opening.

54

Tunnelling and Underground Space Technology 73 (2018) 48–59

C.-C. Lu, J.-H. Hwang

Fig. 7. Flow chart of the MCSRD analysis procedure (modified by Lu and Hwang, 2017).

50m

50m

Tunnel

Applied shear deformation

γ = 24.5 kN/m3 c = 60kPa

φ = 32.5 Vs = 1484 m/sec

55

Fig. 8. Geological parameters and grid mesh of the model.

Tunnelling and Underground Space Technology 73 (2018) 48–59

C.-C. Lu, J.-H. Hwang

Table 4 The parameters of lining and interfaces. Case

Thickness (m)

fc′ (MPa)

E (GPa)

Moment inertia (10−4 m4)

Tensional bond of interface (kPa)

1 2 3

0.3

23.5

23

22.5

–a 42 3070

a

No interface between lining and surrounding ground (No slip).

Moment (ton-m)

20 15 10 5

Fig. 11. The local displacement fields (Case 1).

0 700

600

500 400

300

Axial force (ton)

200

100

0

400

350

350

250

200

150

100

0

50

Curvature (rad/km)

Fig. 9. Axial force-moment-curvature surface of the concrete section.

20

θ

Input M-C curve

Moment (10kN-m)

16

Simulation by FLAC2D

12 8 Fig. 12. An enlargement of deformation of the mesh near the tunnel (Case 1).

4

4. Guide to the reinforcement of the second lining

0 0

50

100

150

200

250

300

350

4.1. The support system in NATM concept

Curvature (rad/km)

NATM is one of the most popular tunneling methods in the world. In NATM, the structural components of a tunnel consist of a primary support and secondary lining. The role of the primary support is directed at enabling the surrounding ground to support itself rather than taking the entire weight of the surrounding ground. By this way, the primary support and inherent strength of the surrounding ground could cooperatively bear the weight of loosening rock mass. Since the inherent strength of surrounding ground would be rationally considered to share part of the extra loading, the design of the tunnel’s primary support using NATM can be more economic in comparison with the traditional American Steel Support Method (ASSM), which assumes that the rock pressure is fully taken by supports. During the construction of NATM-built tunnels, the cross section would deform until the interaction between the primary support and surrounding ground reaches a stabilized state of equilibrium; theoretically, the secondary lining installed afterward will not bear any rock load in the NATM approach, and therefore, the purpose of constructing a second lining is only for aesthetic demand. Because of this, the second lining is rarely reinforced in practice. However, when a large earthquake occurs, the surrounding ground will shake and squeeze the tunnel, and an extra load will be

Fig. 10. Comparison between the theoretical and simulated results of the of the demonstrated example (after Lu and Hwang, 2017).

upper-left and lower-right corners of the refuge owing to the rightward seismic shear strain. Nonetheless, the development of yielding states for Case 1 was faster than Case 3, which demonstrates that the allowance of relative slip between the tunnel lining and surrounding ground slows the yielding state development under the same loading conditions. Differently from Case 1 and Case 3, Case 2 simulated a weak bonding between the lining and surrounding ground. This was hard to hold during the squeeze of the surrounding ground, and the whole weight of the crown would eventually be applied on the two upper corners of the refuges. Since the strength of the two non-reinforced members is too weak to sustain the weight of the crown, the tunnel would finally collapse (as shown in Fig. 15). This indicates that imperfect backfilling between the lining and the surrounding ground is harmful.

56

Tunnelling and Underground Space Technology 73 (2018) 48–59

C.-C. Lu, J.-H. Hwang

Applied shear strain = 0.0179%

Shear force

Bending moment

Initial condition

Max. value = 0.27 kN-m

Max. value = 0.5 kN

Case 1 (without interface)

Max. value = 37.5 kN-m

Max. value = 48.1 kN

Case 3 (with interface)

Max. value = 38.4 kN-m

Max. value = 49.0 kN

Axial force

Tensile loading

Max. value = 4.9 kN Min. value = -1.9 kN

Max. value = 654 kN Min. value = -362 kN

Max. value = 61.5 kN Min. value = -347 kN

Fig. 13. The internal forces of the lining before and after ground squeezing due to an earthquake.

4.2. The effect of reinforcement

unavoidably applied on the second lining. If this extra load is not considered at the beginning of the second lining design, the non-reinforced lining will be easily damaged and even collapse during a strong earthquake. For instance, in the 1999 Chi-Chi earthquake, a total of fifty tunnels were reported to have been damaged, and some of them were built using NATM (Wang et al., 2000). After the earthquake, seismic capacity of second linings became an important topic to tunnel engineers in Taiwan, a seismically active area.

From the results of the MCSRD analysis, the causes of the collapse of the crown could be traced back to irregular geometry at the refuge section, imperfect backfilling, and/or a non-reinforced second lining. Unfortunately, due to the demands of operation and maintenance, the irregular geometry at the refuge section and imperfect backfilling might be unavoidable, which leads to tunnels in a condition similar to Case 2. The better strategy to avoid the damage condition in the second lining Fig. 14. The yielding sequence of the lining.

Case

1

2

Development of yielding condition (solid circle is the yielding location)

= 0.0069%

= 0.0117%

= 0.0131%

= 0.0035%

= 0.00660.0068%

= 0.007-0.008%

= 0.007%

= 0.012%

= 0.0134%

3

57

Tunnelling and Underground Space Technology 73 (2018) 48–59

C.-C. Lu, J.-H. Hwang

that the tunnel would be seriously damaged and collapse in a case with the area of steel (AS) = 4.8 cm2. However, in the case with AS = 5.5 cm2, the tunnel remained stable and did not collapse. The only yielding member of the tunnel in this case appeared at the lower-right corner of a right refuge, and the corresponding shear strain is 0.0207%. Fig. 18 displays the improvement in the tunnel’s seismic capacity due to different degrees of reinforcement. It can be seen that the tunnel’s seismic capacity increased significantly when the amount of reinforcement exceeds 5.5 cm2. The results can be attributed to the finite strength of ground material, which makes an upper limit on the squeezing force acting on the tunnel. Fig. 19 shows the moment and horizontal shear stress at monitors during the MCSRD analysis. It can be seen that the moment at lower-left refuge and the horizontal shear stress at the monitored soil element did not noticeably increase when the free-field shear strain was greater than 0.04%. Therefore, when the strength of the tunnel was greater than the maximum possible external force, the tunnel could resist any level of earthquake. Note that the required tension steel reinforcement ratio for temperature and shrinkage is just about 0.0018–0.002, which means that the studied case can prevent the damage from the disastrous earthquakes as long as the reinforcement ratio of the second lining is larger than the demand for temperature and shrinkage. The results could demonstrate the necessity and profitability of the reinforcement for the second lining.

Fig. 15. The collapse of the tunnel lining in Case 2.

30 cm Reinforcement position

5. Conclusion Size of cross section

100 cm

Protective layer

100cm x 30 cm

fc’

23.5 MPa

fy

420 MPa

As (Area of steel)

1. An approach for considering nonlinear behavior of structure element proposed by Lu and Hwang (2017) was adopted and combined with the MCSRD method. Since the improved MCSRD method could simultaneously consider the nonlinear behavior of the ground medium and structural elements, the nonlinear interaction between an underground structure and the surrounding ground could be well simulated. 2. The damaged condition of the tunnel simulated by the improved MCSRD method agreed well with field observation. Moreover, the damage-causing factors to the tunnel, including irregular geometry at the refuge section, imperfect backfilling, and non-reinforced second lining, were also identified based on the results of numerical analyses. 3. The second lining in a NATM-built tunnel will sustain a substantial extra load after being squeezed by seismic ground motion even

4 cm

0, 4.8, 5.5, 6.0 cm2

Reinforcement ratio 0, 0.0016, 0.0018, 0.0020 Fig. 16. The parameters of the reinforcement per unit meter.

during an earthquake is to reinforce the second lining. In order to study the effect of reinforcement, the basic information of reinforcement that is summarized in Fig. 16 was implemented in Case 2. Fig. 17 displays the development of yielding mechanism along the tunnel lining when it is strengthened by different degrees of reinforcement. The figure shows

As (cm2)

Development of yielding condition (solid circle is the yielding location)

0

= 0.0035%

= 0.00660.0068%

= 0.007-0.008%

4.8

= 0.00650.0067%

= 0.00680.0069%

= 0.00710.0079%

5.5

= 0.0207%

58

Fig. 17. The yielding sequence of different reinforced linings.

Tunnelling and Underground Space Technology 73 (2018) 48–59

C.-C. Lu, J.-H. Hwang

6

Fig. 18. The corresponding relationship among the critical shear strain, PGV, and the amount of reinforcement.

7

70 60

4

50 3

40 30

2

PGV (cm/s)

The seismic shear strain making first yielding member happen (10-4)

80 5

6

5

20 1 0 0

1

2

3

4

5

6

7

10

4

0

1-3 JMA intensity scale

12

0.7

10

0.6

8

0.5

6

0.4

4

0.3 Monitor item Horizontal shear stress Moment

2 0 -2 0.00

0.2 0.1

buried structures. Soil Dyn. Earthquake Eng. 21, 735–740. Hashash, Y.M.A., Hook, J.J., Schmidt, B., Yao, J.I.C., 2001. Seismic design and analysis of underground structures. Tunn. Undergr. Space Technol. 16 (4), 247–293. Hsu, L.P., Weng, S.L., The geological treatment for railway tunnel after seismic damage – a case study of Sanyi no. 1 railway tunnel. Treatment Technology of Engineering Geology on Tunnel, pp. 125–153, 2000 (in Chinese). Huo, H., Bobet, A., Fernández, G., Ramírez, J., 2006. Analytical solution for deep rectangular structures subjected to far-field shear stresses. Tunn. Undergr. Space Technol. 21 (6), 613–625. Hwang, J.H., Lu, C.C., 2007. Seismic capacity assessment of old Sanyi railway tunnels. Tunn. Undergr. Space Technol. 22 (4), 433–449. ISO 23469. Bases for design of structures—Seismic actions for designing geotechnical works. ISO International Standard. ISO TC 98/SC3/WG10; 2005. Kontoe, S., Zdravkovic, L., Potts, D.M., Menkiti, C.O., 2008. Case study on seismic tunnel response. Can. Geotech. J. 45 (12), 1743–1764. Kontoe, S., Avgerinos, V., Potts, D.M., 2014. Numerical validation of analytical solutions and their use for equivalent-linear seismic analysis of circular tunnels. Soil Dyn. Earthquake Eng. 66, 206–219. Lu, C.C., Hwang, J.H., 2008. Damage of new Sanyi railway tunnel during the 1999 ChiChi Earthquake. Geotechnical Special Publication, No. 181, ASCE. Lu, C.C., Hwang, J.H., 2017. Implementation of the modified cross-section racking deformation method using explicit FDM program: a critical assessment. Tunn. Undergr. Space Technol. 68, 58–73. Nishioka, T, Unjoh, S., 2002. A simplified seismic design method for underground structures based on the shear strain transmitting characteristics. SEWC2002 Structural Engineers World Congress. Okamoto, S., 1973. Introduction to Earthquake Engineering. University of Tokyo Press, Tokyo, pp. 29–40. Park, K.H., Tantayopin, K., Tontavanich, B., Owatsiriwong, A., 2009. Analytical solution for seismic-induced ovaling of circular tunnel lining under no-slip interface conditions: a revisit. Tunn. Undergr. Space Technol. 24 (2), 231–235. Penzien, J., 2000. Seismically induced racking of tunnel linings. J. Earthq. Eng. Struct. Dyn. 29, 683–691. Tsinidis, G., 2017. Response characteristics of rectangular tunnels in soft soil subjected to transversal ground shaking. Tunn. Undergr. Space Technol. 62, 1–22. Tsinidis, G., Rovithis, E., Pitilakis, K., Chazelas, J.L., 2016a. Seismic response of box-type tunnels in soft soil: experimental and numerical investigation. Tunn. Undergr. Space Technol. 59, 199–214. Tsinidis, G., Pitilakis, K., Madabhashi, G., 2016b. On the dynamic response of square tunnels in sand. Eng. Struct. 125, 419–437. Shen, Y., Gao, B., Yang, X., Tao, S., 2014. Seismic damage mechanism and dynamic deformation characteristic analysis of mountain tunnel after Wehchuan earthquake. Eng. Geol. 180, 85–98. United Geotech, 1989. Drilling and testing report of route changing section of Sanyi no. 1 runnel – route changing and double track project. Technical Report. United Geotech, Taipei (in Chinese). Wang, J.N., 1993. Seismic Design of Tunnels: A State-of-the-art approach. Monograph, monograph 7. Parsons, Brinckerhoff, Quade and Douglas Inc, New York. Wang, W.L., Wang, T.T., Su, Z.J., Lin, J.H., Huang, T.H., 2000. The seismic hazard and the rehabilitation of the tunnels in central Taiwan after Chi-Chi. Sino-Geotechnics 81, 85–96 (in Chinese). Wang, W.L., Wang, T.T., Su, J.J., Lin, C.H., Huang, T.H., 2001. Assessment of damage in mountain tunnels due to the Taiwan Chi-Chi earthquake. Tunnel. Undergr. Space Technol. 16, 133–150. Wang, Z.Z., Zhang, Z., 2013. Seismic damage classification and risk assessment of mountain tunnels with a validation for the 2008 Wenchuan earthquake. Soils Found. 45, 45–55.

Horizontal shear stress (MPa)

Moment (104 N-m)

Area of reinforcement (cm2)

0.0 0.02

0.04

0.06

0.08

0.10

0.12

0.14

Shear strain (%) Fig. 19. The variations of the moment and horizontal shear tress at monitors during the increase of free-field shear strain.

though it does not bear extra loading at initial conditions. It is suggested they be suitably reinforced with steel bars to resist earthquake loading. 4. According to the study of the effect of reinforcement, the seismic resistance of a tunnel could efficiently improve and become more reliable when the amount of reinforcement is greater than a threshold value, which could be a useful index for designers to develop a profitable reinforcement strategy. Note that the threshold amount of reinforcement may be not easy to clearly decide and vary in each case due to the site specific characteristics. Also, the uncertainty of the simulation results may exist because of the limit of the numerical simulation and the analyst’s personal assumption while conducting the seismic performance evaluation for the underground structure. To be on safe side, it is recommended that the yielded threshold amount of the reinforcement based on the proposed procedure should be considered with a certain degree of conservative bias for the design demand. References Asakura, T., Sato, Y., 1996. Damage to mountain tunnels in hazard area. Soils and Foundations, Japanese Geotechnical Society, Special Issue, pp. 301–310. FHWA, 2009. Technical Manual for Design and Construction of Road Tunnels—Civil Elements, U.S. Department of Transportation, Federal Highway Administration, Publication No. FHWA-NHI-10-034. Gil, L.M., Hernandez, E., Fuente, P.D.l., 2001. Simplified transverse seismic analysis of

59