J. Xu & G. Barton
CH2M Hill/Halcrow Group, Glasgow, UK
ABSTRACT: This paper presents FLAC modeling and analysis of an abutment pile foundation for a road overbridge on soft ground. The lateral force acting on a pile due to the existence of a soft compressible layer (sandy silt) is analyzed. This lateral load arises from lateral soil move ment under the unsymmetric loading generated by a new approach embankment. The Contrac tors construction sequence is simulated and its effect on the calculation is considered. The effect of strength/stiffness parameters and pile coupling springs on the calculation results is also exam ined. The result from FLAC modeling is compared with an analytical solution (elastic continu um approach).
1 INTRODUCTION
During construction of pile foundations for bridge abutments, the abutment is often located over a weak deformable layer, typically soft alluvial claysilt. Horizontal and vertical movements are likely to occur in the soft layer in response to loading from the approach embankment. As the embankment loading increases the soft claysilt will be under plastic deformation and tend to flow away from the abutment. When a row of vertical piles are installed below the abutment the claysilt layer will ‘squeeze through’ between the piles attracting horizontal pressure to the face of piles and an upward pressure on the underside of the pile cap. This loading, in combination with additional lateral loading from earth pressure on abutment structures retaining approach embankments, will induce deflection and bending moments in the piles. Ultimately this may lead to structural distress or failure of the pile foundation or bridge structure. The horizontal pressure is expected to be low at pile cap level where the pile and soil are moving together. It is also low at the interface with deep stiff layers because the difference be tween the pile and soil movement gradually decreases to this level. This feature has been con sidered by various semiempirical or theoretical analyses to determine the lateral pressure distri bution on piles within a soft claysilt layer. In this paper FLAC (Itasca 2000) software is used to analyze an abutment pile foundation for a new bridge at Glendoick Interchange on the A90 in Perthshire, Scotland. The A90 is a dual carriageway connecting Perth and Dundee and there was a requirement to maintain the flow of traffic throughout construction. A pile foundation had been selected due to a 7 m thick soft layer of sandy silt at relatively shallow depth beneath the existing ground surface. Various modeling issues arose in relation to the analysis but in particular how best to simulate construction sequences in order to predict the lateral loading on piles more accurately. One of the analytical methods (Stewart et al. 1994) is presented for comparison with the findings of the analysis using FLAC software.
2
PROJECT DATA AND MODEL ASSUMPTIONS
2.1 Construction sequence
FLAC (version 4.0) software was used and the contractors proposed construction sequence sim ulated, as follows:
Ground preparation for embankment construction： Remove any unsuitable material. Band drain was installed to accelerate the consolidation process in the subsoil but this is not includ ed in the model as consolidation is not directly modeled in the FLAC analysis. Construction of embankment divided into three stages. Stage 1 lift of 3.5 m. Stage 2 lift of 4 m. Stage 3 lift of 2.85 m. (Note that construction of embankment also includes a 9.74m wide temporary surcharge which is constrained by proximity to an existing live carriageway  see Fig. 1). Excavate 6m depth of embankment/surcharge material to level of piling platform. Installation of piles. Construct concrete abutment wall on the pile head, including wingwall. Backfill behind the wall and complete the final slope. Place deck on top of the abutment wall. Remove remaining temporary surcharge.
Figure 1. Illustration of the arrangement of abutment/embankment construction
2.2 Site geotechnical conditions and parameters
The parameters adopted for soils are summarized in Table 1.
Table 1. Soil parameters.
__________________________________________________________________________________________________________
Strata sequence
Thickness
(m)
Bulk
Density
3
(kN/m )
Young’s modulus, E
(MN/m ^{2} )
Undrained
strength, Cu
2
(kN/m )
Friction
Angle
(degrees)
__________________________________________________________________________________________________________
Firm CLAY Loose SAND Soft sandy SILT Firm to stiff Glacial Till Engineered Fill
1.3
0.6
7.0
16

20 
30 
66 
 
19 
11.2 
 
30 
18 
10 
30 
 
21 
58 
58 
 
19 
30 
 
29 
__________________________________________________________________________________________________________
The upper layers represent late glacial raised marine deposits which is normally consolidated and this is underlain by an overconsolidated deposit, glacial till, of low compressibility. Relatively few insitu test data are available for the layers of firm CLAY and loose SAND. A CPT records cone resistances of about q _{c} = 3 MPa, and the triaxial test recorded C _{u} = 66 kPa.
Below these is a soft laminated sandy SILT. The SPT N value is generally very low from 0 to 6 but could increase with depth. A typical value of N = 5 is used to derive the undrained strength of C _{u} = 30 kPa. Underlying this alluvium is stiff to very stiff sandy gravelly CLAY (Glacial Till). SPT N val ues range from 25 to >50. A triaxial test in a sample near to the top of the layer records C _{u} = 92 kPa. A more conservative value of 58 kPa was assumed for design and in view of the paucity of testing only a simple constitutive model has been used (See table 1).
2.3 
Piles 

A single row of 5 piles is installed for each abutment. The pile type is 0.9 m diameter concrete bored pile. The spacing between piles is 2.35 m centre to centre. The model is 2 dimensional and planestrain in nature. The spacing effect is considered in FLAC with the results presented inside FLAC being multiplied by 2.35 to obtain the actual values per pile. The following materi al parameters are adopted: 


E = 30 GPa 


s Shear couplingspring cohesive strength (cs_scoh) = 5.7×10 ^{4} N/m s ear li ri ri i al resis a e s ri 


Couplingspring shear stiffness (cs_sstiff) = 1.3×10 ^{9} N/m/m 


Normal couplingspring frictional resistance (cs_nfric) = 0 


Normal couplingspring cohesive strength (cs_ncoh) = 8.5×10 ^{4} N/m 


Couplingspring normal stiffness (cs_nstiff) = 1.3×10 ^{9} N/m/m 

2.4 
Structural loadings 
Nominal loading per pile is:

Min. axial = 895 kN 

Max. axial = 1895 kN 

Max. shear = 412 kN 

Max. moment = 1590 kN.m 
These are unfactored service loads derived from structural software (Superstress) and act at the top of pile. In fact a row of pile tops are connected with a concrete abutment wall about 6m high. The pile head boundary condition should therefore be something between fixed and free rotation. For the simplicity, this abutment is not included in the FLAC model as it has been modeled in the structural program and load derived from this is considered to be a reasonable approximation.
3 FLAC ANALYSIS AND THE CONSTRUCTION SEQUENCE SIMULATION
The Contractors construction sequence as listed in section 2.1 was simulated by the FLAC mod el: 25 m long piles being installed after preload by the embankment, and partial removal to the level of piling platform. Backfill to the abutment follows installation of piles and construction of abutment. This sequence is defined as Model 1. The static nature of the solving process implies that any residual consolidation process within the soil layer is not considered in this model. Potential lateral forces on the pile due to residual movement of soil are effectively discounted. To simulate lateral forces acting on the pile due to the existence of a soft layer in the ground required the sequence of construction to be changed in the FLAC analysis. That is, the pile is installed first and then the embankment constructed in 3 lifts. This approach is called Model 2. The Model 2 simulation, however, produces an overestimation of lateral force on the pile as it does not consider the benefit from the consolidation process of the soft layer being stiffened un der the preloading of each lift. As an approximation, the C _{u} in the soft layer is modified based on the increase in effective stresses using a linear equation (Skempton & Henkel, 1953):
Then the stressstrain modulus is estimated by assuming E=400C _{u} . So after the 1 ^{s}^{t} lift Young’s modulus E for the soft layer increases from 10 MPa to 15 MPa, to account for the stiff ness gained in the consolidation process during the 1 ^{s}^{t} lift; increasing to 20 MPa after the 2 ^{n}^{d} lift. It is assumed that the consolidation process completes at each stage.
3.1 Model 1
Figure 2 shows the history of Ydirection displacement (vertical settlement) at a point in the soil where the proposed pile meets the existing ground level. The three leap stages from 8000 to 18000 solving steps corresponded to the settlement after each lift of embankment, which is 1 ^{s}^{t} lift: 68 mm 2 ^{n}^{d} lift: 140 mm 3 ^{r}^{d} lift: 175 mm In Figure 2, after the embankment is complete, an excavation takes place and a heave of 30 mm is observed. This number is obtained by ‘zeroing out’ the settlement accumulated so far. It should be noted that the heave is overestimated since the Young’s modulus is higher on unload ing than that on loading. The actual heave can be estimated at about half of the calculated value i.e. 15 mm. After installing the pile and imposing a structural load, the selected point settles about 13 mm, shown in Figure 2 after zeroing out the heave. After backfilling behind the abutment and removing any remaining surcharge, the settlement increases up to 33 mm. Subtracting the original heave produces a net settlement of about 18 mm at the selected point. Figure 3 shows the normal pressure distribution along the pile. It indicates active pressure (at the left side of pile) existing only in the lower part of the soft sandy SILT layer. The value for this pressure is quite small; the maximum value being 16 kPa. This arises from backfill behind the abutment and removal of surcharge and does not include the effect of the three lifts of em bankment fill.
Figure 2. History of Ydirection displacement at a point in soil where the pile meets existing ground level (Model 1).
Figure 3. Normal pressure distribution along the pile (Model 1).
During FLAC analysis it is found that the parameter of normal coupling spring (especially cohesion strength) has a great impact on the pile resistance behavior under lateral loading. To make the point clearer, Table 2 lists the results of maximum pile xdisplacement for different parameters of normal coupling spring. Note that the results correspond to a construction stage in which the pile is installed and subject to structural loading but is not yet backfilled. From theory, the cohesive strength of normal coupling spring can be taken from the un drained shear strength of soil surrounding the pile multiplying the pile perimeter. If we take C _{u} = 30 kPa, then cs_ncoh = 30×10 ^{3} ×π×0.9 = 8.5×10 ^{4} . The maximum xdisp is 132 mm. This large value is due to relatively high value of shear force and moment acting on the pile head. In reali ty, the pile head movement will not be so large as it is restricted by the abutment above it which in turn connects to the bridge deck. When C _{u} = 35 kPa is assumed, the displacement is immedi ately reduced by half to 63.9 mm. Case 3 and 4 represent perhaps more realistic strength of normal coupling spring which increase with the depth of soil. This results in smaller displace ment of 15 to 17 mm. The difference caused by the different normal stiffness of coupling spring is not significant in comparison to those caused by assumptions on cohesive strength.
Table 2. Sensitivity of maximum lateral displacement of pile to parameters of coupling spring.
__________________________________________________________________________________________________________
Case Cohesive strength 
Friction Angle, 
Normal stiffness, 
Maximum x 2 

cs_ncoh 
cs_nfric 

(N/m) (degree) cs_nstiff (N/m ^{2} ) displacement of pile (mm ) __________________________________________________________________________________________________________ 


^{4} 
0 
1.3×10 
^{9} 
132 

^{4} 
0 
1.3×10 
^{9} 
63.9 

^{4} 
70 
1.3×10 
^{9} 
14.9 

^{3} 
70 
1.3×10 
^{8} 
16.9 
__________________________________________________________________________________________________________
3.2
Model 2
In Model 2, the construction sequence is modified to capture ground consolidation effects. The most construction stages keep the same as model 1 except that the installation of piles is moved before the three embankment lift stages. It is appreciated that this will result in an overestima tion of lateral pressure on piles if a correction is not made for ground consolidation due to pre load prior to pile installation. To simulate this process, the Young’s modulus of soft sandy SILT is increased after each lift of embankment fill using the same values as in Model 1. The magni tude of lateral loading is depending on the progress of the consolidation process. In other words, it will depend on the time period which is allowed for excess pore water pressure to dissipate and the settlement of the ground to progress. Changing stiffness of the soft layer is only an ap proximate simulation of the consolidation process and potentially an overestimation of lateral force may result. Figure 4 shows the history of vertical displacement at the same point where the pile meets ex isting ground level. Heave during the excavation stage to prepare a pile platform is about 15mm, and correcting for a likely overestimation reduces this to about 7.5 mm. Thereafter, settlement is about 5mm after imposing structural loadings on the pile; 17 mm aggregated after backfilling and removing surcharge. Overall, after subtracting the original heave, the predicted settlement is about 10mm. It is noted that both heave and settlement are considerably lower than those in Model 1. This is attributed to the change of construction sequence for Model 2. Since the pile has already been installed when the excavation takes place, the pile and soil heave together, which leads to lower vertical displacement at the selected point compared to soil heave without the pile. Settlement after structural loading is imposed on the pile (4 mm) is also considerably lower than that in Model 1 (13 mm) consistent with the overall settlement of the soil at the se lected point being less than that in Model 1. Figure 5 shows normal pressure on the pile after the 3 ^{r}^{d} lift of the embankment fill. It can be seen that the average pressure in the range of sandy SILT increases to about 21 kPa. The actual normal force acting on a single pile is 50 kPa in this case. In reality, the pile is under the com bined effect of lateral loading due to the existence of the soft layer and the loadings from the bridge structure. Figure 5 didn’t consider the effect of structural load so it can be compared with the analytical solution presented below.
Figure 4. History of Ydirection displacement at a point in soil where the pile meets existing ground level (Model 2).
Figure 5. Normal force distribution along the pile (Model 2).
4 ANALYTICAL SOLUTION
An analytical approach (Stewart et al. 1994) is adopted to derive lateral pressure acting on the current pile foundation and compared with the FLAC analysis result. This method is based on a simple deformation mechanism and accounts for the main features of the problem through an approximate representation of the embankmentsoilpile interaction. As an approximation of a pile group connected by a rigid cap, the piles are represented as single beams with a fixed support at the base, as well as fixed at the top to prevent rotation, as shown in Figure 6.
Moment to prevent rotation
Figure 6. Simplified model for normal pressure distribution along the pile.
Pile behavior within the underlying stiffer stratum can be simulated by assuming this section of the pile as an equivalent cantilever with length l _{e} . With a suitable l _{e} this cantilever has similar deflection and rotation at the soft/stiff interfaces as the pile in the elastic soil. The length can be obtained as below (Springman & Bolton 1989)
e
(1)
where l _{c} is the critical pile le r la eral l adi a d ρ _{c} is a parameter describing the relative homogeneity of the soil (Randolph 1981). A uniformly distributed load of magnitude p _{m} is applied on a section of the pile (over a length of h _{1} ) where the soil displacement is greater than the pile deflection. The pile is considered to be unloaded over a length h _{u} below this loaded section, since relative soilpile displacements are assumed to be much smaller in this zone. l _{e} is included in the length of h _{u} , as shown in Figure 6. The mean pressure acting on the idealized pile is obtained as:
m
(2)
where G _{m} = average shear modulus of soft layer; G _{r} = reduced shear modulus around pile due to installation effects and strain; n _{r} = number of rows of piles; s = pile spacing across the face of the embankment; d = pile diameter, and
1
4
3
2
2
3
u
4
u
f
3
3
u
f
2
f
u
1
Before calculating p _{m} by equation 2, the length of the loaded section h _{1} has to be determined firstly by the following expression
where
2
2
u
u
u
2
u
f
3
5
u
f
2
f
f
(3)
Equation (3) must be solved iteratively. The resulting value of h _{1} is substituted into equation (2) to obtain p _{m} . The critical slenderness ratio of the pile is determined from (Springman, 1990):
(4)
where E _{p} Y ’s m d l s of concrete pile; G _{c} = soil characteristic shear modulus; r = radi us of pile section. Here Gc is taken from the stiff layer (Glacial Till) under the soft layer. To represent the ef fect of variations in Poisson’s ratio on the deformation of a laterally loaded pile, a modified shear modulus parameter is given by (Randolph, 1981)
So Gc can be calculated from
From equation (4)
m
e
e
s s
il s
ρ _{c} = 1.
and from equation (1)
Here
la
ial
ill is ass
med as a
With E _{p} = 35×10 ^{6} kN/m ^{2} and G _{c} = 27.3×10 ^{3} kN/m ^{2} , l _{c} = 6.95 m and l _{e} = 2.36 m.
Therefore, L
With h u = 5m,
.
G _{m} for soft layer can be calculated from
It is assumed
. So with G _{m} = 4710 kPa, G _{r} = 3140 kPa.
Substituting these values into equation (2), produces a mean pressure (p _{m} ) = 47 kPa. This value is slightly less than that derived from FLAC analysis but is considered to be reasonably close.
5 CONCLUSIONS
This paper uses FLAC software to study piles subjected to loading from lateral soil movements in a real project. It is found that in order to assess the average active pressure acting on the piles (p _{m} ) correctly, careful consideration should be given to the construction sequence simulated in the model, as well as improvements in the stiffness of soft layers after each lift of embankment adjacent to the bridge abutment. Otherwise too low or too high active pressure will be derived. It is also found that the horizontal displacement of piles is highly sensitive to the cohesive strength of normal coupling spring but not so to the normal stiffness. Finally, the FLAC results are com pared with an analytical solution. The mean pressure (p _{m} values) from both approaches is rea sonably close with the analytical solution producing slightly lower values.
REFERENCES
Itasca Consulting Group, Inc. 2000. FLAC – Fast Lagrangian Analysis of Continua, Ver. 4.0. Minneap olis: Itasca. Randolph, M.F. 1981. The response of flexible piles to lateral loading. Geotechnique 31, No. 2, 247259. Seaman, J.W. 1994. A guide to accommodating or avoiding soilinduced lateral loading of piled founda tions for highway bridges. Project Report 71, Transport Research Laboratory, UK. Springman, S.M. & Bolton, M.D. 1990. The effect of surcharge loading adjacent to piles. Contractor Re
port 196, Transport Research Laboratory, UK. Stewart, D.P., Jewell, R.J. & Randolph, M. F. 1994. Design of piled bridge abutments on soft clay for loading from lateral soil movements. Geotechnique 44, No. 2, 277296. Symons, I.F. 1976, Assessment and control of stability for road embankments constructed on soft sub soils. TRRL laboratory report 711, Transport and Road Research Laboratory, UK.
Tomlinson, M.J. (7
th
ed.) 2001. Foundation design and construction. Prince Hall.
Tomlinson, M.J. &Woodward, J. (5 ^{t}^{h} ed.) 2008. Pile design and construction practise. London & New York: Taylor & Francis.