Professional Documents
Culture Documents
Aser Ibrahim2017
Aser Ibrahim2017
Abstract: This paper presents a methodology to predict the pile-head load versus displacement curve, and the associated mobilized stiffness
of a pile embedded into sandy soils, on the basis of the soil-pile interaction. The mobilized tensile resistance and displacement of the pile shaft
are determined up to failure on the basis of the soil and pile properties, including the soil stress-strain curve. The proposed technique also pro-
Downloaded from ascelibrary.org by University of Saskatchewan on 09/11/17. Copyright ASCE. For personal use only; all rights reserved.
vides the t-z curves for piles in sandy soils that can be separately used in a combined model for a complete superstructure-substructure solution.
A detailed flowchart is presented to describe the proposed methodology that has been compiled through a computer program. A number of
comparisons with full-scale load tests at different sites were employed to validate the suggested technique. In addition, a finite-element model
has also been developed to compare the results of the presented method with those determined from the finite-element analysis and full-scale
load tests results. DOI: 10.1061/(ASCE)BE.1943-5592.0001142. © 2017 American Society of Civil Engineers.
did not account for the elastic extension of the pile in calculating the lus, Gavg , suggested by De Nicola and Randolph (1993).
mobilized side resistance. Reddy et al. (1998) attempted to improve
the model proposed by Sulaiman and Coyle (1976) by modifying
several aspects, which included that the elastic extension of the pile Variations of Shear Displacement and Strain in the Soil
was considered in computing the mobilized side resistance. Despite Surrounding the Pile
the improvement, the modified model was still dependent on the
work suggested by Sulaiman and Coyle (1976). The pile in question is divided into a number of segments as
The assessment of the mobilized pile response is essential to in the method developed by Coyle and Reese (1966). The soil
evaluate the interaction between the foundation and superstructure. around the pile is also modeled in horizontal rings and sublayers
The literature shows that when it comes to the pile response under with a thickness of Hs each, which is equal to the length of each pile
axial tension forces in sandy soils, most available methods are either segment, as shown in Fig. 1(a). Ashour and Helal (2017) suggested
empirical or only concerned with the ultimate pile capacity. Other a nonlinear, parabolic, degrading pattern for the soil vertical dis-
methods used equal side-resistance values for tension and compres- placement, Zs , versus radial distance, r, as shown in Figs. 1(b and c)
sion loading. The work of O’Neil and Raines (1991), Lehane et al. and given by Eq. (4). The shear strain, g i, in sand slice i between ri
(1993), De Nicola and Randolph (1993), and Alawneh et al. and riþ1 along with the associated soil vertical displacement can be
(2007) showed that there is a strong basis for considering lower determined using Eq. (5).
side-resistance values for tension loading than for compression
2
loading. Therefore, the objective of this paper is to develop a r0
method that predicts the mobilized pile response under axial ten- Zs ¼ Z0 (4)
r
sion forces in sandy soils, including the development of the t-z
curves along the length of the pile, on the basis of soil and pile
properties. It is noteworthy that the proposed method accounts for zi ziþ1 Dzi
gi ¼ ¼ (5)
variations of the shear stress (resistance) and confining pressure riþ1 ri Dri
of sands around the pile with the progress of pile displacement.
Such a methodology reflects a representative scenario for pile where r = radial distance from the pile center ðr ¼ r0 Þ; and Zs ¼ Z0
behavior in sand under tension forces. at the pile-soil interface (pile surface).
Fig. 1. Variations of shear displacement and strain in soil around the pile: (a) soil rings around a pile segment; (b) cross section in soil rings; (c) radial
variation of z
Fig. 3. Relationships between stress level, friction angle, and horizontal stress changes in sandy soils
g ¼ ɛð1 þ y s Þ (12)
t
Gmob ¼ (13)
g
The presented approach allows the assessment of the t-z curves Fig. 4. Stress-strain soil model (adapted from Ashour et al. 1998,
along the pile length in sandy soils. For a given displacement of a © ASCE; adapted from Norris 1986)
particular pile segment, Zmid , the accompanying shear strain, g 0 , at
the pile-soil interface using Eqs. (4) and (5) is determined. The
mobilized shear stress, t 0 , caused by Zmid (i.e., g 0 Þ is estimated by
assuming a very small value of t 0 and then calculating the corre- caused by Zmid and the one determined based on t 0 is achieved.
sponding values of s 3 , s df , SL; and y s , as shown in Eqs. (8)–(11). This process provides a point (Zmid versus t 0 ) on the t-z curve for
It is noteworthy that the initial shear stress starts with 0.001 t max at that particular pile segment.
each depth and increases in very small increments to capture the A larger displacement value, Zmid , for that pile segment is
appropriate converging values with the desired tolerance. The effect assumed, and the corresponding t 0 can be estimated through the
of the pile tensile iterative process described to construct the entire t-z curve for the
force on the vertical effective stress of the sur- pile segment in question at a particular depth until t 0 equals t max .
rounding soil i:e:; ðs v0 Þr is determined by using Eq. (6) and
assuming a unit value for X in the first trial. The flowchart presented in Fig. 5 demonstrates the iterative process
As described in Fig. 4 and Eq. (10), SL is used to compute the of constructing the t-z curve.
associated « and thus a new value of g 0 and Gmob , as given in Eqs.
(12) and (13), respectively. X is adjusted on the basis of the new
value of Gmob [Eqs. (6) and (7)]. New values of s 3 , s df , SL, y s , g 0 , Mobilized Pile-Head Load-Displacement Relationship
and Gmob at t 0 are calculated on the basis of the adjusted values of
X until a sufficient convergence (i.e., less than 1%) of X is obtained. Pile segments are numbered in ascending order from tip to top
Thereafter, the value of t 0 is increased, and the described proce- assuming the thickness of the pile segment, Hs, equal to the pile
dures are repeated until an acceptable convergence between g 0 diameter as shown in Fig. 6. Linear springs are used to model the
pile segments to determine the stretching in the pile materials subjected to axial forces at the top and bottom, Qt and Qb , respec-
under tensile loads. One of the advantages of the proposed model tively, and a vertical-side shear force, T, as seen in Fig. 6.
is the consideration of the horizontal stress relaxation at the wall Because of the permeability of cohesionless soils, any negative
interface in association with the pile extension. The pile side re- pore-water pressure that may develop as a result of the voids cre-
sistance with surrounding soils (i.e., pile-skin friction) is modeled ated between the pile base and underlying soil (i.e., suction) will
through a set of nonlinear springs along the pile length (i.e., t-z dissipate quickly. Therefore, for the first segment (i = 1) at the
curves). In addition to its own weight, wo , each pile segment is pile tip, Qb is assumed to be 0.
Qb þ Qt Hs
Z 00 ¼ (18)
2 Ap Ep
Zt ¼ Zb þ Z 00 (19)
Fig. 7. t-z curves for a bored concrete pile at a depth of 10 m in sand with different relative densities: (a) proposed model; (b) O’Neil and Reese
(1999) method
Sand state Dr (%) SPT-N60 g unsat (kN/m3 Þ w (degrees) d (degrees) e Uniformity coefficient (cu ) ɛ50 Gavg (kN/m2)
Loose 30 8 17 30 30 0.764 1.6 0.0052 79,692.35
Medium 60 30 19 35 35 0.671 1.6 0.0037 130,589.8
Dense 80 50 21 40 40 0.61 1.6 0.0028 161,705.2
Note: e = the soil void ratio.
diameters develop higher values of shear stress and lower values curves can be noticed. Fig. 10 shows the predicted cumulative-load-
of displacement at failure. In other words, the shear stress and distribution curves along the pile versus the curves reported by
strain zone of influence around the pile grows with an increase in Carvalho and Albuquerque (2013) for pile-head loads of 160, 240,
pile diameter. and 360 kN. Good agreement was seen between the predicted and
The results of a full-scale load test performed by Carvalho and reported curves. The larger differences observed in Fig. 9 compared
Albuquerque (2013) on a 10-m-long and 0.35-m-diameter concrete to those shown in Fig. 10 may be referred to the type of data. While
pile were used to exhibit the capability of the presented technique Fig. 10 displays only a comparison of shear stresses, Fig. 9 presents
(Figs. 9–11) (test details are presented in a subsequent section both pile displacements and shear stresses. For a pile-head load of
herein). Fig. 9 presents the t-z curves at 4.2, 6.4, and 8.6 m below 372.76 kN, Fig. 11 shows the progress of the vertical displacement
the ground surface as predicted by the proposed model using the of soil and the associated degradation of shear stress and strain at ra-
input data presented in Table 2 versus those back calculated from dial distances away from a pile wall at 9.75 m below the ground sur-
data reported by Carvalho and Albuquerque (2013). A reasonable face. As shown in Fig. 11, rapid degradations of soil displacement
agreement between the predicted and field back-calculated t-z and shear strain and stress can be observed in the soil around the
pile followed by a decrease at a very slow rate; however, the signifi- Case Studies for Comparison and Validation
cant values of shear stresses develop in the radial range approxi-
mately 10 times the pile diameter. The proposed method was validated through a number of compari-
sons between the evaluated load-displacement curves and those
obtained from full-scale load tests in sandy soils under axial-tensile
Numerical Analysis loads. The results of the numerical analysis are presented for some
of the cases.
A numerical model was developed to evaluate pile-tensile loading The following case studies were based on field test data that
using the finite- element program Plaxis 2D. The cases presented in relied on devices, installation efficiency, and methods of data acqui-
the current study were modeled as axially loaded cylindrical piles. sition and analysis. On the other hand, the proposed model assumes
Therefore, a two-dimensional (2D) axisymmetric model was used isotropic soil conditions. Although the measured data in the case
for modeling the pile and surrounding soil. The piles in the case studies came from actual soil properties, used and reported soil
studies were either steel or RC piles that were modeled as linear, properties (i.e., input data), such as the soil unit weight and friction
elastic, nonporous material. Unreported unit weights of RC and angle, represent an average value for the entire soil layer (not the
steel were assumed to be 25 kN/m3 and 78.5 kN/m3 , respectively. A actual values at each depth). This data use may explain the varying
value of 20 107 kN/m2 was used for the elastic modulus of steel. levels of agreement between the measured and predicted pile-head
The elastic modulus of concrete in tension was assumed the same as loads and displacements.
in the compression up to the tensile strength (Park and Paulay
1974). Therefore, the value of the elastic modulus is equal to 29:2 Full-Scale Pile-Load Tests in South-Central Brazil
106 kN/m2 . The Poisson’s ratio was assumed to be 0.15 for concrete
and 0.33 for steel (Bowles 1996). Carvalho and Albuquerque (2013) performed full-scale axial-
Soil was modeled as a drained elastoplastic material using the tension load tests on three bored concrete piles of 0.35, 0.40, and
Mohr-Coulomb failure criteria. Soil properties used in the current nu- 0.50 m in diameter and 10 m in length. Strain gauges were installed
merical analysis, such as unsaturated unit weight, g unsat , and friction at five levels along the pile length (0.6, 3.1, 5.3, 7.5, and 9.7 m). The
angle, w , are reported in all the case studies. The pile-soil interface water table was located below the base of the piles. The top two
friction angle, d , with concrete piles was assumed to be the same as layers of soil, listed in Table 2, were silty and clayey sands [SM and
w . For steel piles, d was reported, except in the U.S. tests for which SC according to the Unified Soil Classification System (USCS)]
the pile-soil interface friction angle was not reported; in these cases, with 62 and 61% of sands. The remaining percentage was fines (silt
the full soil-friction angle was used instead. The saturated unit weight, and clay). The geotechnical parameters reported by Carvalho and
g sat , of soil was calculated or estimated. The ranges of values sug- Albuquerque (2013) and interpreted for the use in the numerical
analysis and proposed model are presented in Table 2. Fig. 12
gested by Bowles (1996) were used to estimate the soil elastic modu-
shows good agreement between the pile-head load versus displace-
lus, E, and the Poisson’s ratio, y s . The dilatancy angle, W, was taken
ment curves predicted by the proposed model and those obtained
as equal to w −30 for cases in which it was not reported. The interac-
from the field tests of the three different piles.
tion between the pile and adjacent soil was simulated via interface ele-
ments placed around the pile. Eq. (21), as follows, was used to calcu-
late the friction angle d at the pile-soil interface, which is a function Full-Scale Field Tests in Christchurch, New Zealand
of the interface strength reduction factor, Rint : McManus (1997) reported a study on nine bored piles. The
tand ¼ Rint tan w (21) objective of the study was to investigate the effect of cyclic load-
ing on the uplift capacity of bored piles in granular soil. Three
Rint is a strength reduction factor that is applied to the pile-soil piles were successfully loaded monotonically up to failure in
interface element in the finite-element analysis. It is assumed by the tension before being subjected to cyclic loading (EQCO2, 3, and
user and depends on pile-surface roughness. 4). Some problems were encountered in testing the EQCO3 pile,
Fig. 9. t-z curves at different depths for a bored concrete pile that is10 m long and 0.35m in diameter
Table 2. Reported and Interpreted Input Soil Parameters Used for Tests in the South Central Brazil
Reporteda Interpreted
Thickness (m) 3
g unsat (kN/m ) w (degrees) SPT −N72 Gs e ɛ50 d (degrees)
0–6 16.3 30 4 2.73 0.94 0.007 30
6–2 18.9 23 7 2.76 0.71 0.005 23
Note: e = the soil void ratio.
a
Data were from Carvalho and Albuquerque (2013).
Fig. 10. Predicted load distribution versus measured data along a 10-m-long bored concrete pile under different pile-head tension loads
so the test was repeated as EQCO3A. The piles were 0.75 m in axial-tension load of 0.00638 m and 1,481.6 kN, respectively. A
diameter at 5.5 m embedded length and with 1 m extension reasonable agreement can be noticed between the predicted and
above ground level. The water table was well below the pile measured data over the entire pile-head load versus displacement
bases. The instrumentation used consisted of a displacement curve up to failure.
transducer, a load cell, and three strain gauges connected to one
of the longitudinal steel bars. The geotechnical parameters Full-Scale Load Tests in Alton, Illinois
reported by McManus (1997) were used in both the numerical
analysis and proposed model as described in Table 3. A compari- The U.S. Army Corps (USACE) of engineers performed a pile-load
son between predicted and measured pile-head responses under test program for the new Lock and Dam 26 replacement project
axial tension loads is presented in Fig. 13. (Tucker and Briaud 1988). Sixteen tension-load tests were reported,
Although the piles had the same dimensions and were tested 13 of which were on H-piles and 3 on pipe piles. The three closed-
under the same conditions, there was a considerable difference in ended driven pipe piles were 0.305, 0.356, and 0.406 m in diameter
their responses. At the initial loading stages of the test, it can be and 10.97 m long. The piles were filled with sand during the load
observed that the measured responses of piles lay between the tests. The river bottom was lowered before any pile driving took
results of the numerical analysis and the proposed model and are place. Soil properties reported by Tucker and Briaud (1988) and
closer to the latter. other interpreted parameters used in the numerical analysis and pro-
posed model are presented in Table 5. The predicted pile response
Field Tests in Dunkirk, France was compared to numerical and field test results (Fig. 15).
A good agreement can be seen between the predicted and meas-
Jardine and Standing (2000) performed full-scale load tests on ured results for the 0.305 and 0.356-m-diameter piles. The 0.406-m-
eight open-ended steel pipe piles. The objective was to study the diameter pile maintained a stiffer measured pile-head response
effect of cyclic loading on those piles in sand through multiple compared to those in the proposed model and numerical analysis.
full-scale static and cyclic load tests. However, the only pile Such a soft predicted response of the 0.406-m-diameter pile can be
tested in tension before being subjected to any other type of load- attributed to the soil properties used, which were the result of aver-
ing was an open-ended driven steel pipe pile that was 19.32 m aging the coefficient of variation (COV) of the SPT values for the
long with a 0.457 m diameter. The 2.5-m section at the pile head 13 borings from Tucker and Briaud (1988). The predicted failure
was composed of steel Grade 52 with 20-mm wall thickness, and the load (724 kN) was 29% less than the measured failure load (1,020
remainder of the pile was made of steel Grade 44 with 13.5-mm wall kN), which is less than the 38% average COV reported by Tucker
thickness. The water table was located 4 m below ground level. and Briaud (1988) for the 13 SPT borings.
The piles were modeled, as reported, as equivalent solid elastic
cylinders with y p of 0.3 and Ep of 35:2 106 kPa over the first
1.5 m and y p of 0.3 and Ep of 24:9 106 kPa over the remaining Conclusions
length of the pile. Four displacement transducers were attached to
a reference beam, and a high-quality load cell was used. Soil The proposed approach allows the assessment of the pile behavior
properties and parameters reported by Jardine and Standing under axial-tensile forces on the basis of pile and soil properties and
(2000) were used in the numerical analysis and proposed model, pile-soil interactions. The presented procedures clearly illustrate the
as shown in Table 4. The predicted pile-head response was com- mobilized responses of axially loaded piles and the creation of the
pared to numerical and field results, and the results are presented associated t-z curves along the pile length through the use of nonlin-
in Fig. 14. According to the proposed method, slippage at the ear stress-strain curves for sandy soils. The developing stress and
pile-soil interface would occur at a pile-head displacement and an strain values for the soil around the pile and soil-pile skin resistance
Fig. 11. Degradation of vertical displacement and shear strain and stress at radial distances in surrounding sand
Fig. 12. Predicted pile-head responses versus field results of Brazil tests on piles of different diameter
Table 3. Reported and Interpreted Input Soil Parameters Used for the Christchurch, New Zealand, Test
Reporteda Interpreted
Thickness (m) 3
g unsat (kN/m ) w (degrees) D50 (mm) W (%) g sat (kN/m3) E (kN/m2) y W (degrees) ɛ50 d (degrees)
0–6 17.2 43 15 4.6 19 25,000 0.3 13 0.002 43
a
Data were from McManus (1997).
Fig. 13. Predicted pile-head responses versus field and numerical results of New Zealand tests
Table 4. Reported and Interpreted Input Soil Parameters Used for the Dunkirk, France, Test
Reporteda Interpreted
Thickness (m) 3
g unsat (kN/m ) w (degrees) d (degrees) W (degrees) 3
g sat (kN/m ) E (kN/m2) y Rint ɛ50
0–4.4 17.1 35 28 17.5 19.5 80,000 0.3 0.76 0.004
4.4–7 20 35 28 17.5 22 80,000 0.3 0.76 0.004
7–8 19.5 32 28 15 21.5 77,000 0.3 0.85 0.006
8–14.5 20 35 28 17.5 22 80,000 0.3 0.76 0.004
14.5–15.5 19.5 32 28 15 21.5 77,000 0.3 0.85 0.005
15.5–17 20 35 28 17.5 22 80,000 0.3 0.76 0.004
17–18 19.5 32 28 15 21.5 77,000 0.3 0.85 0.005
18–30 20 35 28 17.5 22 80,000 0.3 0.76 0.004
Note: E = the soil Young’s modulus.
a
Data were from Jardine and Standing (2000).
1600
1400
Pile-head load, Q, kN
1200
1000
800
Fig. 14. Predicted pile-head responses versus field and numerical results of Dunkirk test
Reporteda Interpreted
Average SPT
Thickness (m) g unsat (kN/m3) w (degrees) W (%) (BL/0.3 m) g sat (kN/m3) E (kN/m2) y W (degrees) Rint ɛ50 d (degrees)
12.19 18.23 34 19 65 21 7,430 0.3 5 1 0.007 34
a
Data were from Tucker and Briaud (1988).
Downloaded from ascelibrary.org by University of Saskatchewan on 09/11/17. Copyright ASCE. For personal use only; all rights reserved.
Fig. 15. Predicted pile-head responses versus field and numerical results of Alton, Illinois, tests on piles of different diameter
can be determined in a progressive and mobilized fashion up to fail- in the surrounding soil while considering confining-pressure relaxa-
ure. The current method determines the progressive pile-head load tion. The good agreement observed between predicted and meas-
versus displacement according to developing shear stress and strain ured pile-head responses, as obtained from field tests, highlights the