The Nicoll Highway Collapse

D. W Hi ht D W. Hight T.O. Henderson + A.R. Pickles S. Marchand

Content
Background and construction sequence Observations up to the point of collapse, including monitoring The collapse Post-collapse investigations Design errors Jet grout layers Ground conditions and the buried valley The need for a trigger Back analyses of the collapse Relative vertical displacement and forced sway The bored piles Collapse mechanism and its trigger Lessons to be learnt

34m Diameter Temporary Staging Area (TSA) Shaft 35m deep 210m 2-cell cut & cover tunnels 200m cut & cover deep 30m - 33m stacked 4-cell scissors crossover tunnel 3-level cut & cover Station (NCH)- 30m deep 25m

3-level cut & cover Station (BLV)

550m 2-cell cut & cover tunnel and siding (BLS)

C824
Collapse site M3 area
800m twin bored tunnel

35m 2-cell cut and cover tunnel

370m stacked 2-cell cut & cover tunnel 20m - 25m deep

C825

C824 - 2km of cut and cover construction mostly in soft clay - up to 35m deep

Circle Li St Ci l Line Stage 1

Reclamation dates

Beach Road
The Concourse Plaza Hotel

NCH station

Cut and tunnels

Golden Mile Cover Tower

1930-1940 sMile 1930-1940sMil Golden G ld reclamation

Complex

Merdeka M d k Bridge Nicoll Highway

1970s reclamation

TSA shaft

Courtesy of Richard Davies

Kallang Basin

31700

54+900
ABH30 NOR RTHING (m)

ABH32

M3

ABH85 AC-2

ABH34

AC-4 ABH35

M2 31650 54+812

MC3007 ABH83

M3010 AC-3 ABH84 ABH31 MC3008 AC-1 ABH33

ABH82 ABH28 ABH81

ABH29

MC2026 ABH27 M2064 MC2025 M2065

31600 31550

31600

31650

31700

31750

31800

EASTING (m)

Pre- and post-tender site investigation

SOUTH
100
Fill Upper estuarine Upper U

ABH 31 M309 M304

Fill

ABH 32

NORTH

Upper estuarine Upper Marine Clay

90

Marine Clay

Upper F2

Upper F2

80

Lower Marine Clay

Lower Marine Clay

70

Base marine clay Top of OA

Lower o e estuarine Lower F2

Old Alluvium Old Alluvium

Typical section in M3

0 0

0.4 04

Cone resistance, qc (MPa) 0.8 08 1.2 12

1.6 16

0 0

20

Undrained shear strength (kPa) 40 60

80

100

AC1 AC2 AC3 AC4 10

AC1 AC2 AC3 AC4 10

Nkt=12
Depth (m)
Depth (m)

20

20

30

30

40

40

Piezocone data for M3 area

Construction details and sequence

SOUTH

NORTH
Fill

Formation level approx 33m bgl Temporary diaphragm walls,0.8m

Upper E

Upper Marine Clay

gp Driven kingpost Permanent bored piles supporting rail boxes 10 levels of steel struts at 3m +3.5m vertical centres

Upper F2

Lower Marine Clay

Sacrificial JGP

Permanent JGP
Lower E Lower F2 OA SW2 (N=35) OA SW1 (N 72) (N=72) OA (CZ)

General Excavation sequence for M3 up to level for removal of sacrificial JGP and installation of 10th level strut

Tidal sea

TSA Shaft

M3

Utility crossings

M2

Curved walls

Nicoll Highway

24/3/04

Support beam Shaft M3 No waler C channel stiffener Jacking point Walers Splayed strut Non-splayed strut Kingpost

M2

Two Struts Bearing Direct on Single panel Single Strut with Splays g y Bearing on Waler for Single panel Struts on Walers - No Splays

Gaps in Diaphragm Wall for 66kV Crossing

South side 13 March 04

Events and observations prior to collapse

Plate stiffener C channel stiffener

Replacement of plate stiffeners at strut-waler connection strut-

Strut bearing directly on Dwall C channel connection

Excavation for the 10th level of struts, including removal of the sacrificial JGP

Observations on the morning of the collapse

North wall
M301 M302 M303 M304 M305

333

334 H KP 181

335

336 H KP 182

337

338 H KP 183

339

340

M306

M307

M308

M309

M310

Order in which distortion to waler noted Excavation to 10th complete

Excavation in progress

Location of walers, splays and Dwall gaps

S335-9 S335 9

Strut 335-9 south wall

S338-9 S338 9

Strut 338-9 north wall 338-

Instrumentation and results of monitoring

M2/M3 plan at 9th level

M2 M3 TSA Shaft

GWV24 I 65

All struts at S335 instrumented for load measurements

I 104

Excavation Front Beyond S335 at the start of Day Shift on 18th April 2004
SG3358 Total Force SG3359 Total Force

5000

4500

Excavation Front Approaches S335 at the end of Day Shift on y 17th April 2004

4000

Excavation Front for 10th Level Advancing from S338

3000

2500

2000

1500

Excavation Front for 10th Level Between S336 and S335

1000

500

15th April
140 130 120 110

16th April
100 90

17th April
80 70 60

18th April
50 40

19th April
30 20 10

20th April
0 0

Hours before collapse

Measured strut load (kN) d

3500

Change in Measured Load at Strut 335 5000

336(N) & 337(N)

4500

Waler Buckling Observed 335(N) Support Bracket at 335(S) Drops Off Waler Buckling Observed

4000

Me easured strut load (kN)

3500

3000

2500

2000
Strut Load 335-9

1500

Strut Load 335-8

1000

338(N) & 335(S) Waler Buckling Observed

500

0
6.00 7.00 8.00 9.00 10.00 11.00 12.00 13.00 14.00 15.00 16.00

Time on 20April 2004

Observed trends in 8th and 9th strut loads

Load

In nstallation of 9th f

8th

Observed

9th

+ passes beyond S335

18 April 2004

20

3.30pm

Excavation approaches 9am m

The trends were consistent with there being g yielding of the 9th level strut-waler connection when the excavation passed beneath but with no f th significant changes i l d i either further i ifi t h in load in ith the 9th or 8th level struts until the collapse was initiated

Horizontal displacement (mm) 0 0 I 104 10 April 200 400

400

Horizontal displacement (mm) 200 I 65 10 April 0 0

10

10

Behind South Wall

North Wall

20 De epth (m)

20 De epth (m)

30

30

40

40

50

50

Horizontal displacement (mm) 0 0 I 104 10 April 15 April 200 400 400

Horizontal displacement (mm) 200 I 65 10 April 16 April 0 0

10

10

Behind South Wall

North Wall

20 De epth (m)

20 De epth (m)

30

30

40

40

50

50

Horizontal displacement (mm) 0 0 I 104 10 April 15 April 17 April 200 400 400

Horizontal displacement (mm) 200 I 65 10 April 16 April 17 April 0 0

10

10

Behind South Wall

North Wall

20 De epth (m)

20 De epth (m)

30

30

40

40

50

50

Horizontal displacement (mm) 0 0 I 104 10 April 15 April 17 April 20 April 10 200 400 400

Horizontal displacement (mm) 200 I 65 10 April 16 April 17 April 20 April 10 0 0

Behind South Wall

North Wall

20 De epth (m)

20 De epth (m)

30

30

40

40

50

50

500

Maximum horizontal displacemen mm nt,

400

I104 max I65 max

300

South wall

200

North N th wall ll
100

0 5-Jul

9-Aug

13-Sep

18-Oct 2003

22-Nov

27-Dec

31-Jan

6-Mar

10-Apr 2004

Comparison of inclinometer readings I65 and I104

 S338-9 stood for 8 days under load and was 20m from excavation front S335 9 stood for 2 5 days under load and was over S335-9 2.5 8m from excavation front B th S335-9S & S338 9N b kl d within 10 minutes Both S335 9S S338-9N buckled ithi i t Load in S335-8 and S335-9 was almost constant between 18 A il and i iti ti of collapse b t April d initiation f ll All C-channel connections failed downwards at both ends The south wall was pushing the north wall back

Key observations

The collapse

3.33pm

3.33pm

3.34pm

3.34pm

3.34pm

3.41pm

3.46pm

Post collapse Post-collapse investigations

Design errors

Errors Misinterpretation of BS5950 with regard to stiff bearing length b i l h Omission of splays Effects Design capacity of strut-waler connection was 50% of strut waler required design capacity where splays were omitted

Errors in structural design of strut-waler connection strut-

 Capacity based on BS5950:1990 = 2550 kN Average ultimate capacity based on physical load tests = 4100 kN Based on mill tests, 9 % of connections had 95% f capacity of 3800 kN- 4400kN Predicted 9th level strut load in 2D analyses which ignored bored piles was close to ultimate capacity therefore collapse was considered by the COI to be inevitable Inevitability of collapse

 Method A and Method B refer to two alternative ways of modelling undrained soil behaviour in Plaxis (Pickles, 2002) Method A is an effective stress analysis of an undrained problem bl Assumes isotropic elastic behaviour and a MohrCoulomb f C failure criterion As a result mean effective stress p is constant until y p yield Method A was being applied to marine clays which were of low over-consolidation or even under-consolidated over consolidation under consolidated because of recent reclamation Method B is a total stress analysis

Methods A and B

t
Cu from Method A

Cu from Method B

The shortcomings of Method A

Method A
105 100 95 90 85 RL (m) 80 75 70 65 60 55 -0.050

Method B
105 100 95 90 85 RL (m) 80 75 70 65 60 55 -0.050

0.000

0.050

0.100

0.150

0.200

0.250

0.300

0.000

0.050

0.100

0.150

0.200

0.250

0.300

Wall Disp. (m)

Exc to RL 100.9 for S1 Exc to RL 91.1 for S4 Exc to RL 81.6 for S7 Exc to RL 72.3 for S10 Exc to RL 98.1 for S2 Exc to RL 87.6 for S5 Exc to RL 78.3 for S8 Exc to RL 94.6 for S3 Exc to RL 84.6 for S6 Exc to RL 75.3 for S9

Wall Disp. (m)

Exc to RL 100.9 for S1 Exc to RL 91.1 for S4 Exc to RL 81.6 for S7 Exc to RL 72.3 for S10 Exc to RL 98.1 for S2 Exc to RL 87.6 for S5 Exc to RL 78.3 for S8 Exc to RL 94.6 for S3 Exc to RL 84.6 for S6 Exc to RL 75.3 for S9

M3 - S th Wall Di l South W ll Displacement t Method A versus Method B

Method A
105 100 95 90 85 RL (m) 80 75 70 65 60 55 -500 0 -3000 -2500 -2000 -1500 -1000 500 1000 1500 2000 2500 3000 3500 4000
RL (m) 105 100 95 90 85 80 75 70 65 60 55

Method B

-500

500

1000

1500

2000

2500

3000

3500

-3000

-2500

-2000

-1500

Bending Moment (kNm/m)

-1000

Bending Moment (kNm/m)

M3 - South Wall bending moments Method A versus Method B

4000

Method M th d A
g 3000 2800 2600 2400 2200 2000 1800 1600 1400 1200 1000 800 600 400 200 0 ( )
3000 2800 2600 2400 2200 2000 1800 1600 1400 1200 1000 800 600 400 200 0

Method B

2383

2206
Strut Load (kN/m)
S1 S2 S3 S4 S5 S6 S7 S8 S9

Strut Load (kN/m) _

S1 S2 S3 S4 S5 S6 S7 S8 S9

M3 strut forces Method A versus Method B

 Method A over-estimates the undrained shear strength of normally and lightly overconsolidated clays l Its use led to a 50% under-estimate of wall displacements and of bending moments and an under-estimate of the 9th level strut force of 10%

The larger than predicted displacements mobilised the capacity of the JGP layers at an earlier stage p y y g than predicted

Method A

 St t l design errors Structural d i Removal of splays at some strut locations p y Introduction of C-channel waler connection detail Use of Method A in soil-structure interaction analysis C ll Collapse was an i inevitable consequence of th it bl f the design errors which led to the applied loads on the struts increasing with time and equalling the capacity of the strut-waler connection

COI view on the principal causes of the collapse

 St t l design errors Structural d i Removal of splays at some strut locations p y Introduction of C-channel waler connection detail Use of Method A in soil-structure interaction analysis C ll Collapse was an i inevitable consequence of th it bl f the design errors which led to the applied loads on the struts increasing with time and equalling the capacity of the strut-waler connection

COI view on the principal causes of the collapse

Post collapse Post-collapse investigations

Jet grout

Excavation of sacrificial JGP in Type H

JGP quality in 100mm cores from borehole M1 in Type K

Shear wave velocity measurements in JGP at Type K

Shear wave section

4 Pressuremeter section JGP thic ckness (m) 3

Design thickness

0 0 2 4 6 8

Thicknesses of JGP in Type K

Post collapse Post-collapse investigations

Ground conditions and soil properties

BN3 P33 CN3

P36 P42/P48 P41/P47 P40 P54 P45 P52 AN2 TSAN2

31700

DN3 BN2E CN2 DN2E EN2 FN2 DN2W DN1E DN1W BN2W

54+900
CN1

M3

AN1 66KVNI

BN1W BN1E BN1Ea CL-1a

NOR RTHING (m)

M2 31650 54+812
FN1 EN1 CL 3g CL-3g CL-2

66KVS CFVN CTSN CPTNa CPTN FS1 DS2E DS2W DS3E CS3 ES3 DS3W EASTING (m) DS1E DS1W ES1 CS2 CS1

BS1Ea BS1Eb SPTSa BS1W AS1

TSAS2 TSAS2b

BS2E BS2W

AS2

31600
ES2

31550

31600

FS2

31650

BS3E BS3W

AS3

31700

31750

31800

BS4W

BS4E

CS4

BS4Ea

DS4W DS4E

PostPost-collapse ground investigation

0 0

0.4

Cone resistance, qt (MPa) 0.8 1.2

1.6

0 0

0.4

Cone resistance, qt (MPa) 0.8 1.2

1.6

BN1W qt u2 fs 10

BS2W qt u2 fs 10

North

South

D Depth (m)

20

De epth (m)
0 0.4 0.8 Pore pressure, u2 (MPa) 0.02 0.04 Friction, f s (MPa) 0.06 1.2

20

30

30

40

40 0 0.4 04 0.8 08 Pore pressure, u2 (MPa) 0.02 0.04 Friction, f s (MPa) 0.06 1.2 12

0.08

0.08

CPT profiles north and south of collapse area

100

=28.5o Ko=0.5 50 =34.2o

t (kP Pa)

0 0 50 100 150 200

s' (kPa)

-50 P-8 P-9 P-24 P-26 -100

=35o

Kisojibans CAU tests on Upper and Lower Marine Clay

P36 BN3

P42/P48 P33

50

55

60

65

70

75

80

85

90

95

100 105
CN3

P40

P41/P47 P54

P45

31700

DN3

P52 TSAN2 AN2 BN2E CN2 BN2W

Tender SI boreholes Tender SI CPTs Detailed design SI Detailed design SI CPTs Post collapse SI Post collapse SI CPTs Boreholes for bored piles Magnetic logging boreholes

DN2E DN2W EN2

54+900

ABH32

M3

AN1 ABH85

ABH34

31650

54+812
CFVN

CTSN CPTN

CPTNa

ABH27

BN1W 66KVNI CN1 WN2 BN1Ea BN1E WN6 WN4 WN8 WN3 WN1 ABH30 WN12WN10 WN7AWN5 DN1E FN2 WN14 WN11 WN9 WN16 DN1W WN13 WN18 WN15 WN20 WN17 WN22 WN19 WN24 WN21 MC3007 EN1 WN26 M3010 CL-1a WN23 ABH83 WN28 WN25 AC-3 AC 3 WN30A WN27 CL-2 FN1 WN29 CL-3g WN32 WN31 WN34 BS1Ea WN33 WS1 ABH84 SPTSa BS1Eb WS3 66KVS ABH31 WN37 WN35 WS9 WS7 WS5 WS11A WS12A WN36 WS13A WS11C WS12C WS15WS13C WS10 WS8 WS6 WS4A WS2 WN38 WS14 BS1W WS11B WS17 WS12B WS12 WS13B WS11 WS13 DS1E WS19 AS1 WN39 WS16 WS21 ABH29 WN40 WS18 ABH82 WS23 WS20 WS25 WS22 CS1 WS27 WS24 DS1W ABH28 WS29 WS26 WS34 WS31 WS28 WS30 ABH81 WS36 WS32 ES1 WS35A WS33 MC2026 WS37 AS2 BS2E WS38 FS1 CS2 BS2W M2064

AC-2 AC-4

NORTHI ING (m)

M2

ABH35 AC-1 AC 1 MC3008 ABH33

TSAS2 TSAS2b

M2065

31600 31550

MC2025 ES2

DS2E DS2W AS3 BS3E BS3W

31600

FS2

31650
DS3E DS3W ES3 CS3

31700

31750

31800

EASTING (m)
BS4W CS4 BS4Ea BS4E

DS4W DS4E

Evidence for a buried valley in the Old Alluvium

110

100

90 ELEVATION (mRL) (

80

70

60

50

Bottom of F2 Clay in Marine Clay Detailed design site investigation Bottom of F2 Clay in Marine Clay Post collapse site investigations Top of OLD ALLUVIUM Post collapse site investigations

Section along north wall

110

100

90 E ELEVATION (mR RL)

80

70

60

50

Section along south wall

P36 BN3 77.02 P42/P48 P33 86.39 CN3 82.27 P45 86.20 P52 85.98 BN2E 82.61 DN2E 81.94 P40 86.30 86.90 P41/P47 86 90 86.84 P54 86.58 86.07

31700

DN3 81.91

Tender SI boreholes Tender SI CPTs Detailed design SI boreholes Detailed design SI CPTs Post collapse SI boreholes Post collapse SI CPTs Boreholes for bored piles Magnetic logging boreholes

54+900
CN1 83.13

ABH32

M3
BN1Ea BN1E 81.57 80.68

83.41

AN1 82.91 ABH85 81.69 AC-2 81.73 AC-4 79.98 ABH35 AC-1 80.76 ABH33 80.51 81.69

NORTH HING (m)

M2 31650 54+812
WN38 WN29 WN30A WN31 78.34 77.51 WN32 WN33 77.71 WN34 78.02 WN35 80.58 WN36 80.53 80.66 WN37 81.43 78.35

WN6 WN7A 82.16 82.31 81.25 WN9WN8 81.38 WN10 WN11 81.42 81.38 81.43 WN12 DN1E 81.05 WN13 81.25 WN14 81.21 WN15 80.95 80.66 WN16 80.37 M304 WN17 81.86 80.72 WN18 M305 79.86 WN19 81.86 WN20 80.81 WN21 80.31 80.88 WN22 80.94 M303 WN23 79.93 WN24 79.91 M3010 WN25 80.15 MC3007 WN26 80.65 82.55 ABH83 80.95 WN27 80.36 M302 79.57 CL-3g KP182 KP183 79.72 AC-3 83.39 M209 M210 KP181 CL-2 CL 2 M211 80.22 81.69 BS1Ea M301

66KVNI 81.88 WN5WN4WN3WN2WN1

338

CFVN CTSN 81.97 82.17 ABH81 80.13

WN40 83.19

WN39 81.06

80.83

WS34 WS35A WS37 WS38 81.26 81 26 80.65 79.69

WS24 81.48 WS25 79.93 WS26 80.03 81.23 DS1E WS27 79.99 81.59 WS28 80.37 79.76 WS29 81.09 78.25 WS30 78.52 WS31 79.57 WS32 78.78 79.23 79.10 79.25

ABH82

74.60 66KVSSPTSa BS1Eb 74.72 ABH84 M308 M309 M310 WS3WS1 ABH31 WS2 71.09 71.52 M307 WS4A 78.20 WS9WS8 WS7 M212 & M306 WS10 76.59 82.07 75.20 73.84 72.61 missing82.57 72.70 76.77 WS15 72.28 WS16 AS1 WS17 81.39 M213 WS18 81.39 WS19 75.98 WS20 WS21 81.28 CS1 WS22 ABH29 81.45 81.38 WS23 80.84 82.01

78.04

TSAS2b 82.16

BS2E 82.28 M2064 79.00

ABH27 80.69

31600 31550

MC2025 81.70

DS2E 80.68 BS3E 81.29

31600

31650

EASTING (m)

31700

31750

31800

BS4Ea 80.96

BS4E 80.33

-14 -12 -10 -8

-6

-4

-2

10

12
DS4E 79.24

Distortion to upper F2 layer caused by the collapse

Buried valley in the Old Alluvium

Buried valley in the Old Alluvium

Coincidence between buried valley and distortion to upper F2 layer

 There was a buried valley crossing the site of the collapse diagonally from south-west to north-east The presence and setting of the buried valley explain the asymmetric conditions and the different collapse on the north and south sides The buried valley coincides with the major ground distortion on the south side and was clearly influential in the collapse Below the Lower Marine Clay the buried valley was infilled with estuarine organic clays on the south side and fluvial clays on the north side Gas exsolution almost certainly occurred in the deep y p organic clays as a result of stress relief, reducing their strength further

The buried valley

CN3

DN3 TSAN2

BN2E CN2 BN2W AN2

DN2E

DN2W EN2 CN1 ABH-32 AN1 ABH-85 WN1 I 67 I-67 AC-2

ABH-34

GA S MA
DN1E

BN1W BN1Ea I-65 WN10 BN1E

66KVN

ABH-30

FN2

AC-4

IN

DN1W EN1 WN26 94.9WN28 WN27 I-102 WN25 WN24

WN20

ABH-83 MC3007 CL-3G

M3010
KP181

CL-1A
KP183 KP182

FN1 WN34 WN35 WN37 WN38 CPTNa CFVN WN40 ABH-28 CPTN WN39

WN29 WN31WN30 WN32 WN33 WN36

AC-3

MC3008 AC-1

ABH-35

BS1Ea CL-2 66KVS I-104 ABH-84 SPTSa WS10 WS20 ABH-82 I-103 WS27 DS1E ABH-29 CS1 TSAS2 TSAS2B BS1Eb BS1W WS1 ABH-31 ABH 31 AS1

I-66

ABH-33

WS34 WS35 WS32

WS30 WS31

WS29 WS28

DS1W WS26

WS36 ABH-81 WS37 WS38 ABH 27 ABH-27 I-64

WS33

ES1 FS1

SE

AL W

L
MC2026 BS2E AS2 BS2W CS2

M2065

M2064

DS2E

DS2W ABH-80 ES2 FS2 DS3E ABH-26 ES3 DS3W CS3

BS3E

AS3

BS3W

BS4E

Post-collapse geometry and sequence of excavation superimposed on buried valley

CN1

A GA S S MA
DN1E

BN1W BN1Ea I-65 WN10 BN1E

66KVN

WN1

I AC

ABH-30

66kV Cable

IN

N1W 5 WN24

WN20

ABH-83 MC3007 CL-3G CL 3G

M3010
KP181

CL-1A
KP183 KP182

AC-3

BS1Ea CL-2 66KVS I-104 ABH-84 ABH 84 SPTSa WS10 WS20 ABH-82 DS1E ABH 29 ABH-29 CS1 BS1Eb BS1W WS1 ABH-31 AS1

DS1W WS26

LL

Sequence of excavation in relation to buried valley

 The buried valley was crossed without collapse developing Strut forces would have been a maximum in the buried valley and would have varied across the valley The collapse was not, therefore, inevitable An external influence (trigger) is required to explain the timing of the collapse and why it occurred after crossing the buried valley

Significance of crossing the buried valley without collapse

 S338-9 stood for 8 days under load and was 20m from excavation front S335-9 stood for 2.5 days under load and was over 8m from excavation front Both S335-9S & S338-9N buckled within 10 minutes Load in S335-8 and S335-9 was almost constant between 18 April and initiation of collapse All C-channel connections failed downwards at both ends The south wall was pushing the north wall back

Key observations

 3D effects cannot explain why the collapse was initiated at 9am S338 was 20-24m from the 20 24m excavation face and had stood for 8 days without distress, S335 was 8-12m from the excavation face Time effects cannot explain why the collapse was initiated at 9am there was no evidence of load increases in the monitoring

Potential triggers

In nstallation of 9th f

8th

Observed Minor additional loading <2mm yielding in 9th

9th

+ passes beyond S335

18 April 2004

20

3.30pm

Excavation approaches 9am m

Trigger

Load

Yielding of 9th

Post collapse Post-collapse investigations

Back analyses of the collapse

 Analyses by Dr Felix Schroeder and Dr Zeljko Cabarkapa using Imperial College Finite Element p g p g Program (ICFEP) 2D section through M307 (I104) and M302 (I65) Bored piles are not modelled ( p (enhanced JGP) ) Upper and Lower Marine Clays, F2 and lower Estuarine Clay modelled using Modified Cam Clay Coupled consolidation p Fill and OA sand modelled using Mohr Coulomb. OA-CZ clay/silt modelled as Tresca

Geotechnical analyses

 W ll EI reduced t allow f cracking, b Wall d d to ll for ki based on d reinforcement layout Bending moment capacities set according to reinforcement layout and ultimate strengths of steel a d concrete and co c ete as supp ed supplied JGP treated as brittle material 9th level strut capacity set and strut allowed to strain soften 72 hours after excavation to 10th level.

Geotechnical analyses

North
Fill Estuarine Upper Marine Clay F2 Lower Marine Clay F2 OA (Sand) - OA-CZ +71.00 +67.00 +63.00 +84.58 +81.58 +85.57 +82.07 +102.9 +98.58 +96.58 +102.9 +98.58 +97.07 Fill Estuarine

South
Upper Marine Clay F2 Lower Marine Clay +63.57 +61.57 +57.57 OA (Clay/Silt) - OA-SW-1 OA (Sand) - OA-CZ OA (Clay/Silt) - OA-CZ

OA (Clay/Silt) - OA-CZ

Stratigraphy assumed for ICFEP analyses

North wall
100

9th

South wall
100

95

95

90

90

measured

Elevation (m RL)

85

predicted predicted measured

85

80

80

75

75

70

70

65

65

60

60

55 50 0

Horizontal displacement (mm)

50 100 150 200 250 300 350

Horizontal displacement (mm)

55 450 -400 -350 -300 -250 -200 -150 -100 -50 0 50

North wall
100

10th

South wall
100

95

95

90

90

measured

Elevation (m RL)

85

predicted predicted measured

85

80

80

75

75

70

70

65

65

60

60

55 50

Horizontal displacement (mm)

0 50 100 150 200 250 300 350

Horizontal displacement (mm)

55 450 -400 -350 -300 -250 -200 -150 -100 -50 0 50

Predicted t e ds in 7th, 8th and 9th strut loads ed cted trends a d st ut oads
4500 4000

3500 3000

Prop for (kN/m rce m)

2500

Strut 7

9
2000

Strut 8 Strut 9

1500 1000

8 7

500 0 280.00

281.00

282.00

283.00 ( )

284.00

285.00

286.00

Time (days)

6 hours

Observed trends in 8th and 9th strut loads

Load

In nstallation of 9th f

8th

Observed

9th

+ passes beyond S335

18 April 2004

20

3.30pm

Excavation approaches 9am m

 The analyses matched reasonably well the build up in horizontal wall movements, and the trends in the forces in the 7th, 8th and 9th level struts To match movements and forces at all stages it was necessary to model the jet grout as a brittle material The upper JGP was predicted to pass its peak strength during excavation to the 6th level and the lower JGP to pass its peak strength during excavation to the 9th level Th 9th l The 9 h level strut reached i capacity d i excavation l h d its i during i to the 10th level

Key findings from geotechnical analyses

 The collapse had to be initiated by allowing the 9th level strut to strain soften a ductile failure of the connection was not associated with a collapse The bending moment capacity of the south wall was reached on the first stage of excavation below the 9th level, b t a hi l l but hinge did not f t form in th wall until th i the ll til the sacrificial JGP layer had been removed

Key findings from geotechnical analyses

Trigger required to initiate collapse

Relative vertical displacement between the kingposts and Dwall panels

V Vertical di isplacement of Dw (mm) wall )

100

Up
50

Top of southern wall-nolimit Top of southern wall-OA1 Top of southern wall-OA2 p

8th 9th 10th

0 0 5 10 15 20 25 30 35

-50

Down
-100

Excavation depth (m)

Predicted settlement of south wall during excavation

500

Ver rtical disp placement (mm) ts

Run 4apfnolimit 400 Run 4apfnolimit_OA1 Run 4apfnolimit_OA2 300

200

100

0 -5 -100 North Wall -200 South Wall 0 5 10 15 20 25 30

Distance (m)

Predicted vertical displacement of lower JGP layer after excavation to 10th level

20

Strut 262 no backfill 2 North wall

10 Diff ference in leve l between wall and kingpost (mm)

kin ngpost

Typical relative displacements of strut between walls and kingpost

1 0 7A 2 7B 3 -10 6A 4 6A -20 5 7A 7B 4 -30 5 1 3

Sout wall th

 Calibrated FE analyses predict downward displacement of Dwall and upward displacement of kingpost when sacrificial JGP layer excavated Survey data supports upward vertical displacement of centre of strut relative to ends

Relative vertical displacement

KP180

KP181

KP182

KP183

KP184

335

336

337

338

339

340

9th level
Sacrificial JGP

Excavation front Lower JGP

Kingposts in long section

Post collapse Post-collapse investigations Structural steel physical tests and numerical analyses. The effect of relative vertical displacement

12mm stiffener plates

5000

Waler

4000

Axial load (kN) (

3000

2000

1000

0 0 10 20 30 Axial displacement (mm) 40 50

Comparison of tests on connections with plate and C channel stiffeners

12mm stiffener plates

5000

Waler

4000

Axial load (kN) (

3000
C channel stiffener

Waler

2000

1000

C channel stiffener has similar capacity to double plate stiffener but becomes brittle after an initial ductile response Unforced sway

0 0 10 20 30 Axial displacement (mm) 40 50

Comparison of tests on connections with plate and C channel stiffeners

5000 Test result

4000

Axial loa (kN) ad

3000
C channel stiffener

Waler

2000

1000

0 0 10 20 30 Adjusted axial displacement (mm) 40 50

Calibration of FE model of strut-waler connection strut-

5000 Test result FE prediction 4000

Axial loa (kN) ad

3000
C channel stiffener

Waler

2000

1000

0 0 10 20 30 Adjusted axial displacement (mm) 40 50

Calibration of FE model of strut-waler connection strut-

5000

Ductile plateau 10-15mm Test result FE prediction

4000

Axial loa (kN) ad

3000
C channel stiffener

Yield 2000

Waler

1000

Ductile plateau allows failure to develop at both ends

0 0 10 20 30 Adjusted axial displacement (mm) 40 50

Calibration of FE model of strut-waler connection strut-

5000 8m strut

4000

Axial load (kN)

3000

2000

C channel stiffener

Waler

1000

0 0 10 20 30 Axial displacement (mm) p ( ) 40 50

Effect of strut length (bending restraint) on brittleness of connection

5000 8m strut 18m strut 4000

Axia load (kN) al

3000

2000

C channel stiffener

1000

Reduced restraint increases brittleness and reduces d til plateau d d ductile l t 8m strut effective kingpost 18m strut ineffective kingpost

Waler

0 0 10 20 30 Axial displacement (mm) 40 50

Effect of strut length (bending restraint) on brittleness of connection

Test on C channel stiffened connection by Nishimatsu

Test on C channel stiffened connection by Nishimatsu

5000 axial load only

4000

Axial load (kN)

3000

2000

C channel stiffener

Waler

1000

0 0 10 20 30 Axial displacement (mm) p ( ) 40 50

Effect of relative vertical displacement

5000 axial load only axial load+RVD 4000

Axial load (kN)

3000

2000

C channel stiffener

Waler

1000

RVD reduces ductile plateau, increases brittleness, makes stable situation unstable Forced sway

0 0 10 20 30 Axial displacement (mm) p ( ) 40 50

Effect of relative vertical displacement

Force, P

Ductile D til

strain

Effect of brittleness of strut to waler connection

2500

Ductile
2000 Run 4apf2100drop Run 4apf2100drop4 Run 4apf2100drop3 p p Run 4apf2100drop2 Run 4apf2100 ( p 1000 500 0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 )

Prop force (kN/m) e

1500

Total axial strain (%)

Effect of brittleness of strut to waler connection

2500 500

Run 4apf2100 R 4 f2100 Run 4apf2100drop2 Run 4apf2100drop3 Run 4apf2100drop4 Run 4apf2100drop

Ductile no collapse
2000

Prop force (kN/m) e

1500

1000

500

6 hours 9 days
0 -15 -10 -5 0 5 10 15

21 days
20 25

Time since reaching excavation level of 72.3m RL (days)

Force, P

Ductile D til No collapse

hours to collapse

Several days to collapse y

strain

 Th There was relative vertical di l l ti ti l displacement b t t between th the diaphragm wall, which settled when the sacrificial JGP was removed, and the kingpost, which rose, i.e. gp relative vertical displacement between the ends and centre of the strut (RVD) The strut-waler connection was ductile-brittle. The ductile plateau explains why both ends could fail p p y The brittleness of the connection determines the time taken for the collapse to develop

Trigger required to initiate collapse

 RVD reduces the length of the ductile plateau and increases the brittleness RVD can make a stable situation unstable k t bl it ti t bl RVD can shorten the time to collapse

Why downward failure at both ends? Why collapse after crossing the buried valley?

Trigger required to initiate collapse

 Free Fixed Forced sway

Downward failure at both ends

Trigger required to initiate collapse

Post-collapse positions of the Dwall panels

CN1

A GA S S MA
DN1E

BN1W BN1Ea I-65 WN10 BN1E

66KVN

WN1

I AC

ABH-30

66kV Cable

IN

N1W 5 WN24

WN20

ABH-83 MC3007 CL-3G CL 3G

M3010
KP181

CL-1A
KP183 KP182

AC-3

BS1Ea CL-2 66KVS I-104 ABH-84 ABH 84 SPTSa WS10 WS20 ABH-82 DS1E ABH 29 ABH-29 CS1 BS1Eb BS1W WS1 ABH-31 AS1

DS1W WS26

LL

Post-collapse geometry superimposed on buried valley

31650

CL-2

TUNNEL SOUTH WALL

BS1Ea SPTSa WS10 WS9 WS8

31640
NORT THING (m)

WS17

WS13C 12 2.5

WS1 12C 12.5

WS S11C 14.7

WS16

WS15

WS14

WS13B 13.7 WS13 16 WS13A 26 .2 WS12B 15. 12 WS12 9 WS12A 19 17.5 WS11B 1 WS11 19.6 WS11A 13.5

66KVS

WS12D

WS11D

31630

CS1 WS11E WS12E WS11F

Post collapse SI boreholes Magnetic logging boreholes Unsuccessful magnetic logging boreholes 2005 boreholes

WS13E WS13F

31620 31670 31680 31690

EASTING (m)

31700

Coincidence between inability to advance boreholes and missing Dwall panels

WS11E,11F, 12E,13F WS11-13A-D

Obstruction created by missing Dwall panels

1 M212

1 Gap

1 M306

1. Panels M306 and M212, each side of the gap, and panel 213 fail by toe kick-in and rotate back. back 2. Soil flows through resulting gap between M306 and M307, rotating panel 307. 3. Soil flows th 3 S il fl through resulting gap b t h lti between M307 and M308, rotating panel 308, etc.

Sequence of south wall panel movements

d drag

dr rag

304

305

drag

4
303

3
302

2
301

1
306

2a

3a

4a

Sequence of south and north wall panel movements

SOUTH

NORTH
Fill Upper E

Upper Marine Clay

Differing restraint imposed by bored piles at north and south walls

Upper F2

Lower Marine Clay

Sacrificial JGP

Permanent JGP
Lower E Lower F2 OA SW2 (N=35) OA SW1 (N 72) (N=72) OA (CZ)

 The bored piles had a major influence on the displacements of the Dwall panels during the collapse and on their post-collapse positions The bored piles restricted toe movements on the south side and prevented failure as the buried valley was crossed Loads carried by the bored piles contributed to the under-reading under reading of the strain gauges

Significance of the bored piles

North wall
M301 M302 M303 M304 M305

333

334 H KP 181

335

336 H KP 182

337

338 H KP 183

339

340

M306

M307

M308

M309

M310

Order in which distortion to waler noted Excavation to 10th complete

Excavation in progress

Location of walers, splays and Dwall gaps

CN1

A GA S S MA
DN1E

BN1W BN1Ea I-65 WN10 BN1E

66KVN

WN1

I AC

ABH-30

66kV Cable

IN

N1W 5 WN24

WN20

ABH-83 MC3007 CL-3G CL 3G

M3010
KP181

CL-1A
KP183 KP182

AC-3

BS1Ea CL-2 66KVS I-104 ABH-84 ABH 84 SPTSa WS10 WS20 ABH-82 DS1E ABH 29 ABH-29 CS1 BS1Eb BS1W WS1 ABH-31 AS1

DS1W WS26

LL

Repositioned bored piles at 66kV crossing

 Upward displacement of KP 180 and 181 accentuated by toe displacement of M306 and M212, where bored piles had been re-positioned and additional t movement was possible dditi l toe t ibl Resulting RVD fed back into buried valley

Relative vertical displacement

KP180

KP181

KP182

KP183

KP184

335 335

336 336 337 338 339 340

9th level
Upper JGP

Excavation front Lower JGP

Kingposts in long section

 C channel stiffened connection undergoes brittle failure Critical length of strut and of load in strut result in minimal lateral restraint to connection Restraint from rising kingpost results in downward force on connection RVD results in reduction in ductility of connection and increase in brittleness RVD makes a stable situation with overstress unstable bl RVD results in downward failure of connection

RVD was the trigger for the failure Failure mode of connection and RVD

Overall conclusions
The use of Method A in the numerical analyses to model near normally consolidated soils is fundamentally incorrect Its use led to under prediction of wall displacements and under-prediction bending moments and so to a reduction in the redundancy in the system. The JGP was strained beyond its peak and a plastic hinge formed in the wall as excavation of the sacrificial JGP was underway

Overall conclusions
There were errors in the design of the strut-waler connection resulting in a design capacity that was 50% of the required capacity where splays were omitted The collapse initiated some time after the excavation crossed the buried ll th b i d valley, where f h forces on th under-designed strutthe d d i d t t waler connections would have been a maximum An additional perturbation or trigger was necessary to explain the timing of the collapse, the downward failure of the walers at both ends and the trends in the monitoring data

Overall conclusions
The permanent bored piles in combination with the JGP p y played a significant role in p g preventing the collapse as the g p valley was crossed The collapse was triggered when working in the vicinity of the 66kV cable crossing At this location, the permanent bored piles had been repositioned and the JGP layout had been modified, allowing the wall toe to kick-in and cause additional uplift of g p the local kingposts

Overall conclusions
This additional upward displacement of the kingposts relative to the wall fed back into the system, introducing forced sway failure Downward movement of the walls has been predicted by analysis; the potential for relative upward movement of the kingposts has been confirmed by surveys Forced sway failure red ced the strain o er which the s a fail re reduced over hich connection remained ductile, increased the brittleness of the connection and allowed a stable situation to become unstable Forced sway failure can explain the timing of the collapse, the form of the observed distortions, the trends in the monitoring data, data and the speed at which the collapse developed

Overall conclusions
The collapse was not caused by hydraulic base heave and was not related to poor workmanship Wall rotation, which had been linked with inadequate , q penetration of the wall into the OA, was not the cause of the collapse Several factors had to act in combination to cause the collapse

Unforgiving site
Deepest excavation in marine clay in Singapore shortcomings in use of Method A not previously apparent because of depth dependence Ground conditions buried valley in OA infilled with soft fluvial and organic clay, rapid variation in depth of marine clay along and across th excavation resulting i an asymmetric section d the ti lti in ti ti p q g q Curvature of walls in plan requiring more frequent use of walings Presence of 66kVA crossing Need to adopt sacrificial JGP layer, removal of which caused p y , step increase in 9th level strut load and step increase in wall settlement

Lessons learnt
JGP is a brittle material The mass properties of JGP need to be more carefully evaluated Coring of JGP is not an adequate check The use of numerical modelling of soils in design should be carried out by specialists and its incompatibilities with current codes needs to be removed The potential for brittle failure of C channel connections must be recognised tb i d

Lessons learnt
Temporary and permanent works should be subject to independent checks The effects of relative displacement between kingposts g and walls should be considered in the design of strutted excavations Forced sway failure and its consequences should be y q recognised as a potential mechanism in design Monitoring did not warn of the impending collapse