Professional Documents
Culture Documents
To cite this article: T. Bhatkar, D. Barman, A. Mandal & A. Usmani (2017) Prediction of behaviour
of a deep excavation in soft soil: a case study, International Journal of Geotechnical Engineering,
11:1, 10-19, DOI: 10.1080/19386362.2016.1177309
In this paper, the nature of wall deformation and surface settlement for deep excavation in soft soil was
studied. Excavation was carried out in eastern metro city of India in the bank of Hooghly River. This
excavation was the part of underground metro station construction. Numerical analysis was performed
to predict the soil behaviour during excavation. Accordingly, the most suitable excavation method was
finalised. Vertical and horizontal deformations were measured during construction on adjacent area to
ensure the safety of nearby existing structures. Predicted deformation was relatively high when compared
with measured one. Back calculation from measured deformation shows the conservative estimation
of stiffness during investigation and anisotropic soil behaviour. This study indicates that the numerical
analyses can be very effective tools to predict the behaviour of soil during excavation with installation of
support systems. Thus, it helps for necessary planning to ensure the safety of nearby structures. However,
accuracy of such analysis depends on the judicial estimation of soil parameters.
Keywords: Deep excavation, Deformation, Soft soil, Surface settlement
stress conservatively. However, this conservative prediction adjacent structures which further could save both time and cost.
by numerical simulation for deep excavation depends on the However, in the present study only an attempt had been made
efficiency of constitutive laws to represent few critical features to verify the influence of soil stiffness estimated from observed
of soil model e.g. estimation of soil parameters, location and deformation up to second stage (S-2) of excavation.
nature of fictitious boundary conditions, effect of surcharge
from existing structures, sequence and schedule of excavations,
etc. Nonetheless, many researchers have studied these aspects
Description of site location
(Holt and Griffith 1992; Yoo and Lee 2007; Hsiung 2008). The project consisted of approximately 3.3-km long twin-bored
In clay soil, excavation-induced stress and deformation in 5.4-m internal diameter tunnels. A total of three similar under-
surrounding ground are also influenced by consolidation process ground stations were proposed along the stretch of the tun-
and creep. Thus, numerical analysis with effective constitutive nel; so out of three, one of the underground stations known as
law should consider this phenomenon along with stress–strain Howrah Maidan Station located near Howrah Railway Station
relationship to predict the surrounding ground behaviour more in Kolkata, India was selected for study which was located on
realistically during excavation. Few researchers also developed the junction of national and state highway on the western side
coupled simulation techniques (Yong et al. 1989; Whittle et al. of Hooghly river. This urban area is occupied by a mixture
1993; de Lyra Nogueira et al. 2009, 2011) to incorporate this of commercial and residential buildings. Therefore, it was of
aspect. Although the available numerical techniques are suffi- utmost importance to ensure the stability of existing structures.
ciently effective to predict the behaviour of deep excavation, The location of Station is shown in Fig. 1. The B2–B2 cross sec-
the accuracy very much depends on the estimation of different tion of Howrah Maidan Station was considered for analysis as
parameters of surrounding ground, which are normally non-ho- shown in Fig. 1. The maximum elevation of groundwater level
mogeneous in nature. Therefore, more research needs to be con- was observed at approximately 1.0 m below the ground surface.
ducted to verify the model response with field study to develop
confidence among researchers and practitioners. Moreover, such
study must address the need to calculate the non-homogeneous
Investigation
soil parameters more realistically to minimise the gap between Design methodology of this project was mainly based on
prediction and observed ground behaviour. parameters evaluated from field and laboratory testing of soil
Dinakar and Prasad (2014) studied the performance of dia- through vertical core drilling prior to excavation. Different in
phragm wall on the stress distribution and deformation character- situ and laboratory testing was carried out to get a realistic
istics of ground below adjacent structure at vulnerable locations. value of the geotechnical properties of the soil encountered.
A typical building load is idealised and its effects on excavation In situ standard penetration testing (SPT) was undertaken in
and supporting system is analysed in terms of bending moment, boreholes drilled near station location and corrected to N60. The
shear force and displacement of diaphragm wall. Results of the N60 is the normalised value for SPT, corrected for overburden
study revealed that diaphragm wall method of excavation was and field procedures. More details of soil profile in this location
able to limit ground movements under buildings considerably are presented in Table 1.
and the excavation could be carried out safely. In absence of Cambridge self-boring pressure meter (SBP) was used in
wall, the soil fails at very shallow depth. It was observed that four boreholes. The purpose of the pressure meter testing was
for 20.0 m depth of excavation with diaphragm wall, the wall to obtain field values for the engineering properties of the ground,
movements should be restricted within 0.1–0.3% of depth of specifically strength, stiffness and in situ lateral stress. Some
excavation. Horizontal and vertical displacements as well as tests were also carried out to measure the consolidation char-
shear stress in soil decrease linearly with increase in distance acteristics of the less permeable material. The instrument was
of the structure from the excavation. Therefore, the minimum fully calibrated prior to bringing to the site, and the properties
distance of excavation from existing structure could be estimated of the membrane were checked at intervals during the course of
based on allowable stresses and displacement of ground. the fieldwork. The fieldwork took place between 28 August and
In present investigation, an effort has been made to predict 29 October 2010. At this station location, a total of 11 additional
the deformation pattern of surrounding ground due to deep tests were conducted at varying depths in two positions (SBP1 and
surface excavation. This excavation was part of an underground SBP2). SBP1 and SBP2 were progressed at the main station and
metro rail station in proposed East–West metro line of Kolkata, crossover, respectively. The tests were undertaken by a specialist
India. Excavation work and necessary support structures were sub-contractor between 28 August and 12 September 2010.
already completed. This paper mainly focused on the prediction Necessary soil parameters were derived from field and lab-
of horizontal and vertical deformations during each stage of oratory tests to get a broad view of the geotechnical properties.
excavation. The predicted deformation was also compared with Laboratory tests were performed in different soil samples col-
the available measured values during the course of construc- lected during drilling. It includes, drained and un-drained shear
tion. Finally, back calculation method was proposed to estimate strength, Atterberg’s Limit test, Permeability test, Oedometer
the important soil parameters to understand the accuracy of esti- test and some Chemical tests. The relevant parameters from
mated non-homogeneous soil parameters. Even under similar these tests are reported in Tables 3 and 4. Cohesion values were
ground condition, some soil parameters could vary significantly calculated by curve fitting method from laboratory tests using
due to the size and method of excavation. This approach can an adaption of Carter et al. (1986). The relevant soil parameters
prove to be the best practice in any geotechnical engineering from HSSmall model were obtained from the Oedometer test,
project along with initial field and laboratory investigations. Traiaxial test and the Self-boring Pressure meter tests which
Thus, it could minimise risks of construction and damage to were carried out near the station (Brinkgreve et. al. 2008;
Table 1 Geological profile for Howrah Maidan station the city (this may be a result of defective/inadequate drain-
age). Hence, for modelling purposes the maximum elevation
Depth m bgl
N60 value of groundwater is assumed to be located at 1.0 m above the
Stratum Strata Top Bottom (SPT) ground level. Thus, excavation process was simulated con-
Howrah Maidan station sidering ground water flow condition. The adopted method
crossover of construction was bottom-up excavation, using diaphragm
Made ground Unit 1 0 1.2 3 walls to limit the degree of ground movement. According
Organic clay/silt Unit 2 1.2 15.65 6 to assessment report for the station, unit 3b and unit 3a are
Clay/silt Unit3a 15.65 29.65 33
Sandy silt/silty sand Unit 3b 29.65 65.65 53
likely to behave in a cohesive manner based on the classifi-
cation tests. Therefore in terms of its engineering behaviour,
these two could be treated as one unit. Table 1 illustrates the
Hardin and Black 1969 and Truty 2009). For the estimation of geological profile for Howrah Maidan Station and the soil
HSSmall model stiffness m was varied between 0.5 and 1.0 to classification summary.
obtain the most conservative values.
Excavation schedule and stages
Soil conditions Underground basements were built by bottom-up method,
The likely ground condition was expected to be encountered whereas substructure and superstructure floors are constructed
include alluvial deposits consisting of silts, clays, sands and sequentially from the bottom of the substructure or lowest level
organic silt. In general, the top 1.2 m of soil consists of made of basement to the top of the superstructure. In the bottom-up
ground overlying 15.65 m of soft clayey silt that will be method, a trench is excavated from the surface within the tunnel
required some form of retention during excavation. The site and then the trench is backfilled and the surface restored after-
topography is generally flat. The surveyed levels of the bore- wards. The trench can be formed using open cut (sides sloped
holes suggest that the ground level is approximately +5.65 m back and unsupported), or with vertical faces using an exca-
above mean sea level (MSL). It is known that groundwater vation support system. In this method, the tunnel is completed
abstraction has caused groundwater levels to reduce over time before it is covered up and the surface reinstated.
in this region (Sengupta 2009). Seasonal rains during the mon- Conventional bottom-up sequence of construction generally
soon period are known to cause flooding in many parts of consists of the following steps:
• Installation of temporary excavation support walls, such • Monitoring lateral movement in diaphragm wall using
as soldier pile and lagging, sheet piling, slurry walls, tan- inclinometer.
gent or secant pile walls. Different monitoring instruments were installed at various loca-
• Dewatering within the trench if required. tions to ensure the safety of all nearby structures. Tilt meter was
• Excavation and installation of temporary wall support planned for thirty numbers in total quantity as 27 were installed
elements such as struts or tiebacks. due to unavailability of space. To monitor surface settlement, two
• Complete construction of the walls and then the roof. instruments were installed as planned. Pavement settlement point
• Backfilling to final grade and restoring the ground surface. was planned for 43 in total quantity but only 34 were installed as
As such, bottom-up construction sequence was used in the two of instruments were not installed due to unavailability of space
analyses for present station construction. However, complete and seven instruments were not installed as located on D-wall. Soil
construction of wall and roof along with backfilling was not settlement points with five instruments were planned and all were
considered in analysis. Only the excavation of various layers installed. However, most of these instruments didn’t show any
was simulated which was carried out between January and significant reading during excavation and some instruments were
June 2012. The various excavation stages used in modelling damaged during construction. Therefore, only the lateral move-
i.e. Stage 1 (S1) – excavation level at 2.0 m, Stage 2 (S2) ment in diaphragm wall and surface settlement were considered
– excavation level at 8.0 m, Stage 3 (S3) – excavation level in the present study. To monitor lateral movement in diaphragm
at13.0 m, Stage 4 (S4) – excavation level at 21.0 m and vari- wall, inclinometer was planned for four locations and all were
ous strut levels were at 1.65, 7.85 11.85, and 15.85 m, closely installed as shown in Fig. 1.
simulating actual excavation sequences. Monitoring frequencies prior to excavation of all the instru-
ments were nearly weekly, during excavation it was daily, during
backfilling it was weekly and after backfilling it was monthly.
Monitoring plan and status The alarming value for horizontal deformation was 63 mm and
Geotechnical monitoring has now become an important part of for surface settlement near the diaphgram wall was 60 mm which
the construction all over the world and especially for excavation gradually reduces to 10 mm up to the last boundary of the soil
projects in urban area where occupancy level is quite high and model. These values are decided considering the present site
proper monitoring could result in preventing failures and conse- condition (Long 2001), and required action needs to be taken
quent damages. The support philosophy in this project is based once it is exceeded. Necessary facilities must be mobilised to
on staged excavation followed by incremental installations of ensure the safety of adjacent buildings e.g. structural support
support measures and verifications by monitoring. In addition, system will be enhanced, construction activity will be stopped
progressive support enhancements based on actual performance and design will be reviewed if this alarming value was exceeded.
are an essential part for any geotechnical project, which lead to
cost-effective and practical application of support. Therefore,
an adequate geotechnical monitoring plays an important role in
Numerical analysis
such excavation. However, in this project the scope to minimise Modelling
the support system based on monitoring report was limited, A finite element modelling was used to analyse the staged
keeping in mind the stability of existing operating railways, excavation (cross section B2–B2) and dewatering for one of
road ways and old high-rise buildings. The sole purpose of the underground stations. The ground was discretised using
geotechnical monitoring was to ensure the safety and service- six-nodded triangular plane strain elements. While the slabs
ability of nearby structures during excavation. were modelled as a plate-element having flexural rigidity sim-
The field monitoring plan and interpretation covers the ilar to concrete structures. The model boundary was extended
period during construction, and contains monitoring results of approximately three times the width of excavation to mini-
all the instrumentation installed and in operation. In the present mise the effect of artificial boundaries on the ground move-
project, the objectives were to monitor wall deflections, move- ment. This was decided based on few trials considering the
ments of adjacent buildings and ground settlement outside the effect on deformation of excavated area and optimising the
excavation during construction. time required to run the model. Thus, complete excavation was
The specific purposes and instruments of monitoring are carried out in 11 phases as shown in Table 2. The soil model
as follows: consisting of 435 m wide and 65.65 m deep with diaphragm
• Monitoring building movement using building settlement wall of 29.65 m and barrette of 38.65 m depth as shown in Fig.
pin and prism target. 2. The properties for various layers of soil are given in Tables
• Monitoring surface (ground/pavement/roads) settlement 3 and 4. Sequential staged construction and dewatering were
using ground settlement point. simulated using consolidation type analysis for 7 and 14 days.
50 kPa
75 kPa
20kPa 20kPa 20kPa
(0,0) 20kPa (100,0) (120.5,0) Unit 1 : Made Ground (0.0 m) (435,0)
Interface
38.65 m Unit 3a : Clay/Silt (15.65 m)
65.65 m
Slabs
20.5 m
(65.65,0) (435,65.65)
435 m
The existing maximum building load of 75 kPa was about 10 m construction phases. The properties of structural elements are
away from the diaphragm wall on the right side. Construction presented in Table 5. Stiffness properties were derived based on
live load of 20 kPa at ground level was considered symmetri- detailed design of reinforced concrete structures and subsequent
cally. Additional 10 kPa load were taken on all internal slab for analysis using SAP 2000.
Table 5 Properties of structural members for different model soil structure interaction and assigning Rinter < 1.0 may
stages in numerical model lead to unrealistic soil behaviour. Therefore the extended inter-
Structure elements EA (kN/m) EI (kNm2/m) face, the Rinter is set to 1.0. However, it was only activated during
plastic analysis stages.
29.58 × 106 2.465 × 106
Comparison
In the present investigation, the predicted deformation values
both in horizontal and vertical directions were compared with
observed values. Figure 7 shows the comparison of horizontal
deformation from model analysis versus measured one. The
variation in maximum horizontal deformation was within the
range of 20–50%, whereas variation in maximum vertical
deformation was much higher with a range of 100–110% as
shown in Fig. 8. In all these cases, model study predicted
the higher deformation. In case of horizontal deformation,
the variation was comparatively less and expected to be the
combined effect of stiffness of wall and soil. In contrast, the 8 Variation of maximum vertical diformation for different
variation of vertical deformation was comparatively higher stages of excavation
and expected to be due to conservative estimation of soil
stiffness. The maximum vertical deformation was much
higher than the maximum horizontal deformation after com- Back calculation approach
plete excavation. Installed supports had restricted horizontal
movement. In addition, the cumulative effect of previously Monitoring deformation during the excavation could not only be
installed strut and wall from both sides of excavation may used for the safety of the construction site but also provide impor-
have reduced the deformation. The horizontal movement as tant information for back analysis at initial stages of excavation.
predicted was around 0.25%H (where H is the depth of exca- Method of back calculation certainly helps to re-assess the already
vation). In actual ground condition, the quantity of maximum determined soil properties from laboratory and field investigation.
horizontal movement was also less than the maximum verti- Ideally, the selection of soil parameters is an iterative approach
cal deformation. As stated earlier, it is difficult to determine to minimise the uncertainties involved by comparing predicted
the accurate value of soil stiffness beforehand from initial and measured soil behaviour during excavation (Rechea et al.
investigation. Because of its sensitiveness to different ground 2008; Wang et al. 2014). Particularly those parameters, which are
conditions i.e. water condition, compactness, particle compo- case-sensitive and could get affected by excavation method, exca-
sition, size and method of excavation, etc., it is recommended vation size and any other project-specific parameters. Thus on most
to measure stiffness value from actual site conditions. The occasions, this approach could result in considerable cost-effective-
prediction could have been more accurate if the relevant soil ness by optimising the design (Mansour et al. 2015). However, in
parameters for present study were obtained from back cal- the present study, there was limited scope for minimisation of the
culation using the observed deformation value at the initial support system based on observed deformation and subsequent
stage of excavation. back analysis due to proximity of sensitive structures. From the
9 Comparison of maximum (a) vertical and (b) horizontal defromation for left side wall from back calaculation
earlier comparison, it was observed that though the horizontal were studied and compared. Present investigation shows that
deformation was in reasonable agreement the vertical deforma- Two-dimensional Finite Element analysis is an effective way
tion was overestimated in the present model. Back analysis also to investigate the performance of deep excavation, in which
supported the fact that the soil parameters, more importantly the detailed geotechnical and structural aspects such as size of
stiffness, were estimated conservatively. The measured deforma- excavation, structural support system and actual excavation
tions were normally used to back analyse soil parameters including sequence can be well accounted. From the above study, the
Young’s Modulus (E) and shear strength parameters i.e. cohesion following conclusions can be drawn:
and friction angle. It was acknowledged (Pakbaz et.al. 2013) that • Two-dimensional plain strain model can predict excava-
for low-level deformations at shallow depth, the Young Modulus tion-induced ground movement for long station reason-
(E) is effective; however, for considerably higher depths, where ably well.
deformations are due to plastic behaviour of the soil, cohesion and • It was observed in model study as well as in measured data
friction angle parameters are more sensitive. The back analysis that the maximum ground surface settlement is more com-
method was based on iterative variation of involved parameters to pared to the maximum lateral wall deflection in soft soil.
match the obtained results with the real measured data. Therefore, • The maximum settlement on the ground as predicted is
an attempt had been made to back-calculate the soil modulus by located 15–25 m away from the wall.
matching model prediction with observed response from initial • Predicted amount of horizontal and vertical movements
stage of excavation. In this approach, the Young Modulus value with numerical model were more than observed deforma-
(E) was changed from 12 × 103 to 24 × 103 kN/m2 (after few trials) tions, which could be due to lack of reliance on laboratory
for first and second layers to match the model deformation with parameters of representative soil samples. However, this
observed one and finally both the deformations keeping the third little conservative approach should be well accepted to
layer properties unchanged were compared. After complete exca- ensure the safety of nearby sensitive structures.
vation, the predicted vertical deformation was in good agreement The accurate prediction of the performance of deep exca-
with observed one as shown in Fig. 9 (a, b). However, the predicted vation, especially the ground movements requires a realistic
horizontal deformation was almost 50% of measured deformations. soil model that is able to consider the small-strain stiffness
This indicates anisotropic nature of soil behaviour, which was not non-linearity of the soil. In addition, the soil model adopted
considered in the present study. The anisotropic nature of soil needs to be calibrated with soil properties corresponding to
behaviour influences the wall movement and settlement during geotechnical condition at construction site.
deep excavation, particularly in soft soil condition (Clough and Further, the numerical simulation can be incorporated as an
Hansen 1981; Finno et al. 1991). From the above study, it was integral part of the observational method to permit maximum
understood that the value of the Young Modulus calculated from economy and assurance of safety during construction. The
back analysis is higher than the designed Young Modulus based results of computation are not more than working hypothesis
on initial investigation. Thus back calculated stiffness can well pre- that have to be subjected to confirmation or modification during
dict the deformation during excavation; otherwise it could be an construction. In the past, the numerical simulations could not
important factor in the performance of deep excavation (Schweiger benefit from these types of field observations but with this pro-
and Freiseder 1994). posed approach, there can be a direct link between precedence
and numerical modelling. As construction advances and new
field data are obtained, the material model can be improved
Conclusions as new behaviour observed. Thus, the back calculation should
In this paper, the surface settlements and lateral deformation of be the compulsory part of geotechnical site investigation and
the diaphragm wall and their influence on the adjacent structure geotechnical design should continue till the end of such project.