CASE STUDIES OF FILL STABILITY IN LAND RECLAMATION

Laifa Cao1, Cee Ing Teh1, Ming-Fang Chang1, Victor Choa1, Yung Mook Na2 and Bo Myint Win3

ABSTRACT Two case studies of embankment failure at a reclaimed site in Singapore were carried out. Both the limit equilibrium slip surface method and the finite element method were employed in the studies. The analysis indicated that the critical fill height predicted by the limit equilibrium method was affected by the shear strength of the embankment fill, whereas that assessed from the finite element analysis was controlled by the input parameters of the foundation soil. Both the limit equilibrium method and the finite element method give reasonable prediction of the critical fill height when proper soil parameters and slope are chosen. The finite element analysis provides a better prediction of the actual slip surface than the limit equilibrium analysis. The finite element analysis also gives a reasonable estimate of the surface settlement. INTRODUCTION Limited land space in many countries in the East and Southeast Asian region has led to extensive foreshore land reclamation for diverse uses. In the planning and design of land reclamation, the stability of the fill resting on foundation soils is important. Slope failure may seriously disturb the original subsoil, damage the field monitoring instruments, and delay the working schedule. Conventionally, slope stability is analyzed using the limit equilibrium slip surface method (SSM). The slip surface method is simple and fast. The calculated factor of safety provides a direct indication of the safe height and slope for the constructed fill. However, the analysis method provides no information on the change of stress and strain in the soil. For a detail understanding of the problem, the finite element method (FEM) may be useful. This paper describes two case studies involving embankment failure at a land reclamation site at Changi East in Singapore, one involving the placement of hydraulic sandfill and surcharge over a slurry-like soil mass in a silt pond and the other over a natural soft marine clay beneath the seabed. Both SSM and FEM are used in the analysis. A limit equilibrium based computer program PCSTABL developed at Purdue University is employed for the slip surface analysis. SAGE CRISP, a critical state finite element program developed by Sage Engineering and Cambridge University, is adopted for stress and deformation analyses. The critical fill heights obtained from the SSM and FEM analyses are compared with the field observation. The vertical movement and lateral displacement estimated from the FEM analysis are compared with the field measurements in the silt pond. The shear bands estimated from both the SSM and the FEM are compared with that explored by field investigation at the location of fill failure in natural seabed clay. The difference between the SSM and the FEM in the analysis of fill stability is discussed. RECLAMATION OVER SILT POND The silt pond covers a rectangle area of 180 hectares at Changi East. It was formed in 1986 from hydraulic dumping of fine material washings originated from sand quarrying activities into a man-made containment bund. The top surface of the silt deposit in the slurry-filled pond was extremely soft, making direct hydraulic filling or dumping of sand over the deposit impossible. Spreading of sand with high water content was used in the reclamation up to an elevation of -1 to 0 mCD (Chart Datum). Although care was taken in the slow spreading of sand, failure of foundation was detected at one location. To strengthen the foundation at the
1

Geotechnical Research Centre, School of Civil & Structural Engineering, Nanyang Technological University, Singapore 639798 2 Hyundai Engineering & Construction Co. Ltd, Changi Airport P. O. Box 0310, Singapore 918151 3 SPECS Consultants Pte Ltd, Singapore

failure location for further sand spreading, high strength polyester woven textile was placed in two layers over the location and the surrounding areas after the heave-up mud was removed. Following the placement of geotextile, sand was spread evenly on the entire area in the silt pond until the fill reached the elevation of 5 mCD. Then vertical drains were installed in the entire area. A maximum settlement of about two meters had taken place two months after the installation of the vertical drains. However, site investigation indicated that the soil was still very soft with the water content as high as 130%. The undrained shear strength (su) determined by the field vane test (FVT) was as low as 1 to 3 kPa above elevation -10 mCD. Considering the extremely low values of su, a trial embankment was built using the hydraulic filling method to investigate the possible highest level of surcharge that could be added. A slope inclination of 1:17.5 was adopted, which was the gentlest slope that could be achieved with hydraulic filling. Prior to the construction of the trial embankment, settlement plates and inclinometers were installed to monitor the surface settlement and horizontal displacement during the surcharge. Site investigation at the location indicated that the soil profile comprised a 5.5 metre thick sandfill which was underlain by a layer of very soft silty clay with a thickness of 22 m. A dense silty sand layer underlay the very soft silty clay. Figure 1 shows the cross section of the trial embankment, the foundation soils and the locations of settlement plates (S1, S2, S3, S4) and inclinometer (I1). Soil properties such as the bulk density (), su and the angle of internal friction () are also shown in the figure. The values of and were obtained from laboratory tests and su from the FVTs. A stability analysis was carried out prior to the construction of the embankment.
Sand = 32o = 1.9 Mg/m3 S1 1:17.5 S2 S3 I1 S4 = 1.3 Mg/m3 = 1.45 Mg/m3 = 1.45 Mg/m3 = 1.45 Mg/m3 70 m 2.5 mCD -3 mCD -7 mCD -14 mCD -19 mCD -25 mCD

su = 1.5 kPa Very soft silty clay su = 3.5 kPa su = 5 kPa su = 7 kPa 20 m

70 m

20 m

30 m

Figure 1 : Cross section of trial embankment Limit Equilibrium Slip Surface Analysis The cross-section with the centre line in the extreme left and soil parameters shown in Figure 1 were used in the slip surface calculation. The bottom of the very soft silty clay was considered as a stable boundary due to the existence of dense silty sand beneath the soft silty clay. The water table was assumed to be at the ground surface. Since the thickness of the very soft soil was greater than 20 m, failure should be confined within this layer. Bishop's and Janbu's circular failure analyses were carried out. The factors of safety calculated by both methods are shown in Table 1. Both methods gave similar results. The factors of safety estimated by the two methods for embankment height of less than 3.5 m were found to be great than 1.0. When the height was greater than 3.5 m, the factor of safety approached one and the embankment became unstable. The critical fill height is significantly affected by the input shear strength of the sandfill. The critical height increases when a larger value of is adopted. During the construction of the embankment, the settlement plates located at the toe of slope of the embankment showed significant heave when the surcharge height was increased from 2.5 to 3.7 m. The hydraulic Table 1 : Factor of safety for embankment filling operation was stopped immediately and monitoring Embankment Factor of Safety was carried out on all the instruments installed in the trial Height Bishops Janbus embankment area. Measurements of the inclinometer (m) Method Method indicated continuous lateral movement with a rate as high as 1.0 3.014 3.113 100 cm/day. A slip failure was judged to have occurred and 1.5 2.182 2.145 the lateral movement would continue if the surcharge 2.0 1.661 1.70 loading was not reduced. This observation is consistent 2.5 1.393 1.429 with the stability analysis carried out before the 3.0 1.232 1.264 construction of the embankment. After reducing the 3.5 1.172 1.146 surcharge and adding fill at the toe of the slope, the rate of 4.0 1.043 1.066 lateral movement was significantly reduced.

Finite Element Analysis The trial embankment was analysed as a plane strain problem. The soil cross section used in the limit equilibrium analysis was adopted in the finite element analysis. The silty clay was divided into four horizontal layers to account for the variation of soil properties with depth. The existing sandfill with a thickness of 5.5 m was taken as another horizontal layer. The surcharge was treated as a separate layer. A total of 455 linear-strain triangle elements were used to simulate the soil mass and the surcharge. An undrained analysis was carried out because the time of surcharge of three days was relatively short for the low permeability soil. A drained analysis would required consideration of the consolidation of soil under both the self-weight of the soft silty clay and the surcharge, which was very complex and was not discussed in this paper. The sand and surcharge fill were modelled as a linear elastic material. The bulk density of sand was taken as 1.9 Mg/m3 based on the field measurement. It is well known that the effective Young's modulus (Es) depends on the level of strain. Laboratory tests indicated that the value of Es ranged from 5 to 20 MPa for an axial strain of about 1% under in-situ stress. In the finite element analysis, taking Es as 10 MPa would not cause significant computing error since the modulus of sand is normally much larger than that of clay and the main source of deformation should come from the strain in the clay. The very soft silty clay was simulated as an elastic-perfectly plastic material obeying the Mohr-Coulomb failure criterion. The value of su obtained from FVTs as shown in Figure 1 was used in the analysis. As discussed by Britto and Gunn (1984), an undrained analysis can be carried out by the use of effective stress parameters and a very larger bulk modulus of water (Kw). The value of Es for the clay was estimated from oedometer test, which was believed to be similar to the field condition since the very soft soil was still consolidating under its self-weight. The soil modulus was calculated from the compression index, initial void ratio, effective in-situ pressure and effective loading with an assumed Poisson's ratio of 0.3. The calculated Es increased with depth and ranged from 245 to 440 kPa. The value of Kw was taken as 106 kPa. Figure 2 shows the variation of lateral displacement with increasing fill thickness at -4 mCD elevation at the location of inclinometer I1. The lateral displacement measured by the inclinometer increased rapidly as the fill height approached 3.7 m, which signified the inception of failure. The undrained analysis also showed a significant increase in the lateral displacement when the embankment height increased from 3 to 3.5 m. As observed from Figure 2, the undrained analysis seems to be able to provide an indication of the inception of instability, although the lateral displacement magnitude was overestimated. The overestimation of lateral displacement may be attributed to several factors, including the accuracy of input soil parameters and the plane strain condition assumed in the analysis.
0.8 1

Lateral displacement (m)

Vertical movement (m)

Predicted 0.6 Measured

0.5 0 -0.5 -1 -1.5 Predicted at Predicted at Measured at Measured at 0 70 140 H = 2.5 H = 3.5 H = 2.5 H = 3.7 m m m m 210

0.4

0.2

0 0 1 2 3 4 5

Height of fill (m)

Distance from centre line (m)

Figure 2 : Lateral displacement response during the construction of embankment

Figure 3 : Comparison between measured and predicted surface settlements

A comparison between the measured and predicted surface settlements for fill heights of 2.5 m and 3.5 m is shown in Figure 3. There was a significant increase in the settlement under the embankment when fill height increased from 2.5 m to 3.5 m. The undrained analysis gave reasonable estimate of the settlement within the embankment. The observed heave in the vicinity of the toe was significantly smaller than prediction. As discussed earlier, this discrepancy may be partly attributed to the approximate nature of the soil parameters used in the analysis. In order to predict heave accurately, loading-extension tests following

the field stress-path at the toe may be required. Also, the actual soil heave near the toe was reduced because a surcharge was added at the toe when the fill approached 3.7 m in the actual construction. As the embankment height increased, the extent of subsoil yielding became greater. The plots of the yielding zone where the ratio of shear stress to the undrained shear strength reached a value of 1, as illustrated in Figure 4, indicated that the yielding was initiated close to the bottom of sandfill and subsequently propagated towards the bottom of the very soft silty clay as the fill height increased. When the height of embankment reached 3.5 m, practically the whole area under the embankment yielded, indicating incipient slope failure.
15

Emabnkment height = 2.5 m Elevation (mCD)

Embankment height = 3.5 m

Yielded zone

Yielded zone
70 140 210

Distance from centre line (m)

Distance from centre line (m)

Figure 4 : Distribution of yielding zone RECLAMATION OVER NATURAL CLAY Prior to land reclamation, the mean sea level was at 1.6 mCD. The seabed elevation, varied between -3 to -5 mCD, was underlain by the Kallang Formation comprising of soft to firm marine clay. The marine clay consisted of two distinct layers separated by a sand or stiff silty clay layer that was formed from exposure of the seabed to the atmosphere during the rise and fall of the sea level in the geological past. The upper layer is known locally as the upper marine clay (UMC) and the lower layer as the lower marine clay (LMC). At one location, sand filling up to 5.5 mCD was completed in August 1994. Vertical drains were installed in April 1998. Subsequently, surcharge was placed by hydraulic pumping. When the surcharge height reached 8.5 mCD, a slope failure occurred. Cone penetration tests (CPTs), FVTs and borings were carried out to check the soil stratification and to determine the slip surface after the failure. Back analyses using SSM and FEM were conducted. Based on the results of CPTs carried out prior to the installation of vertical drains and the actual slope of surcharge, the cross section used for the back analysis was simplified as shown Figure 5. The deposit of dense sand at the bottom of LMC was considered as a stable boundary. Limit Equilibrium Slip Surface Analysis The main input soil parameters for the analysis is for the sand layers and su for the clay layer. Laboratory tests showed that for the loose sand was about 32o. The values of su obtained prior to the installation of vertical drain and the placement of surcharge were used for the analysis. FVTs conducted
8.5 mCD 3 mCD Sand -4.5 mCD -16 mCD -20 mCD -23 mCD Upper marine clay Silty clay Lower marine clay - 43 mCD 70 m 9m 61 m 70 m su = 15 to 27 kPa su = 35 kPa su = 100 kPa su = 55 to 85 kPa = 1.57 Mg/m3 = 1.6 Mg/m3 = 1.9 Mg/m3 = 1.65 Mg/m3 Water table 1 mCD = 32o = 1.9 Mg/m3

Figure 5 : Cross Section for slope stability analysis

under marine condition indicated that the su value for the upper marine clay ranged from 10 to 30 kPa. For the lower marine clay, su ranged from 40 to 80 kPa and a su value greater than 90 kPa was observed for the intermediate stiff clay layer. The values of s u were also estimated from CPTs by using the method suggested by Cao et al. (2000) and shown in Figure 5. For the upper part of UMC, su increased linearly with depth from 15 to 27 kPa. An average value of 60 kPa was chosen for LMC. The factor of safety for different embankment heights was analyzed using Bishop's and Janbu's methods. The calculated factors of safety are presented in Table 2. Bishop's method gave a slightly higher factor of safety than Janbu's method. Janbu's method indicated that the slope would become unstable when Table 2 : Factor of safety during surcharge surcharge elevation reaches 7 mCD. When surcharge Fill Factor of Safity elevation reached 8.5 mCD, both Bishop's and Janbu's Elevation Height Bishop's Janbu's methods indicated that the slope was unstable. Field (mCD) (m) method method observation showed that the slope actually failed when 6.0 10.5 1.174 1.058 the embankment reached 8.5 mCD high and the after 6.5 11.0 1.122 1.011 failure slope corresponded approximately to the design 7.0 11.5 1.076 0.970 slope. This indicated that a steeper than design slope 7.5 12.0 1.036 0.934 might have contributed to the failure. 8.0 12.5 1.002 0.902 8.5 13.0 0.971 0.874 Finite Element Analysis Similar to the case study in the silt pond, a plane strain analysis was also carried out to evaluate the stress and deformation in the natural foundation soil under the surcharge. The cross section shown in Figure 5 was used in the finite element analysis. The natural soil layers were divided into 315 quadrilateral elements and the sandfill and surcharge into 15 triangle elements and 76 quadrilateral elements. The sea-water was represented as an elastic material with a shear modulus of 0.5 kPa. The actutral construction sequence was simulated in the analysis. The elements of sea-water inside the sandfill were replaced by the elements of sandfill during the reclamation using imposed elements. The fully coupled (Biot) consolidation analysis was carried out because an undrained analysis was not suitable for this case for which the marine clay was consolidated under the load of sandfill for 44 months prior to the installation of vertical drains as mentioned earlier. The natural soil was modelled as modified Cam clay. The sandfill was modelled as a linear elastic material with Young's modulus of 10 MPa. The soil parameters determined from laboratory tests were summarized in Table 3. These parameters included the necessary critical state parameters (, , ecs and M) (Britto and Gunn, 1984), Poisson's ratio (), earth pressure coefficient at rest (Ko), and horizontal and vertical permeabilities (kh and kv). The bulk densities were as shown in Figure 5. Table 3 : Soil parameters used in the finite element analysis M Ko kh (m/day) ecs 0.065 0.44 3.6 0.94 0.3 0.79 6.0 x 10-5 0.065 0.33 3.4 0.94 0.3 0.64 3.0 x 10-5 0.030 0.11 2.2 0.98 0.3 0.93 3.6 x 10-5 0.065 0.37 3.5 0.94 0.3 0.70 7.0 x 10-5

Elevation (mCD) - 4.5 ~ -16 -16 ~ -20 -20 ~ -23 -23 ~ -43

kv (m/day) 3.0 x 10-5 1.5 x 10-5 1.8 x 10-5 3.5 x 10-5

The coupled consolidation analyses indicated that both the settlement and the lateral displacement increased significantly during surcharging from the elevation of 5.5 to 8.5 mCD. Figure 6 shows the maximum lateral displacement at the distance of 70 m from the centre line. Figure 7 shows the surface settlement. In order to evaluate the critical embankment height, the zone approaching the critical state was drawn in Figure 8 based on the drained analysis. The zone in which the soil has reached the critical state expanded quickly when the fill height increased from 7.5 to 8.5 mCD, indicating incipient slope failure, which is consistent with the field observation. Based on the maximum displacement vectors and the boundary of zone approaching the critical state, a shear band could be roughly estimated. The shear band estimated from FEM was compared with that from SSM and field sliding depths determined by CPTs and boring as shown in Figure 9. It was observed that the finite element analysis provided a better agreement with the actual slip surface in comparison with the slip-circle predictions.

Maximum lateral displacement (m)

0.8

0.4

Vertical movement (m)

0.6

0.4

-0.4

Ele va tio n 3 m C D Ele va tio n 5.5 m C D

0.2

-0.8
Ele va tio n 5.5 m C D (c o ns o lida te d 44 m o nths ) Ele va tio n 8.5 m C D

0 -4.5

-1.2 -2.5 -0.5 1.5 3.5 5.5 7.5 9.5 0 50 100 150 200 250

Fill elevation (mCD)

Distance from centre line (m)

Figure 6: Lateral displacement during filling

Figure 7 : Settlement at the surface of UMC

Elevation (mCD)

-3 -13 -23 -33 -43 0 70 140 210 7.5 8.5 mCD Silty Clay LMC UMC

Elevation (mCD)

-3 -13 FEM -23 -33 SSM -43 0 70 140 210 CPT Borehole

Distance from centre line (m)

Distance from centre line (m)

Figure 8 : Zones approaching critical state during surcharging CONCLUSIONS

Figure 9 : Comparison of shear band predicted from finite element and slip surface methods

The behaviour of a slurry foundation and a natural soil foundation subjected to embankment loading can be predicted reliably by the finite element analysis. The actual soil condition and the drainage condition need to be assessed accurately prior to the adoption of drained or undrained analysis. The surface settlement, the critical fill height and the shear band may be reasonably estimated by the finite element analysis. However, the accurate prediction of lateral displacements and heave requires careful assessment of soil parameters corresponding to the stress-path prevailing in the field during embankment loading. The limit equilibrium slip surface method can provide directly the factor of safety against sliding. However, the accurate estimation of the critical fill height and slip surface depends on a reasonable consideration of the shear strengths of embankment fill and foundation soil and a representative cross section. In the assessment of fill stability at reclaimed sites, the shear strength of the foundation soil should be based on that of the unimproved natural soil. The calculated factor of safety is sensitive to the assumed strength of embankment fill when the slip circle method is used for the analysis of embankment over a slurry-like soil. REFERENCES Cao, L.F., Chang M.F., Teh, C.I. and Choa, V. (2000). "Evaluating soil improvement by cone penetration test". Proc. of Int. Conference on Geotechnical & Geological Engineering, Melbourne, Australia. Britto, A.M. and Gunn, M.J. (1984). CRISP user's and programmer's guide. Cambridge University, Cambridge, England.