Lateral pile analysis based on subgrade reaction and comparison with

other methods
Y. Gong & D. Wang
AECOM Australia Pty Ltd

ABSTRACT: Subgrade reaction method with linear soil springs is widely used in structural analysis for piles
because of its simplicity. Geotechnical engineers are often required to provide subgrade modulus for evalua-
tion of the soil spring stiffness. However, subgrade modulus is not an inherent soil property and highly de-
pendent on structure geometry and magnitude of deflection. Due to the limitation of the linear spring model, it
is not unusual for the validity of the analysis result to be questioned. This paper provides a review of the sub-
grade reaction method and the various approaches in determining subgrade modulus. Effect of the variation of
subgrade modulus on pile analysis is studied using structural software Space Gass. Lateral pile analyses based
on the p-y curve and finite element methods are also carried out using software packages LPile/Group and
Plaxis 3D. Analysis results are compared and the reliability of the subgrade reaction method is discussed.

1 INTRODUCTION soils modelled as uncoupled linear springs. Ge-

otechnical engineers are requested to provide sub-
Pile foundations are frequently designed to support grade modulus for evaluation of the soil spring stiff-
both vertical and lateral loads. Typical examples in- ness. Although soil spring model is convenient to
clude pile-supported bridges, buildings and offshore use, there is difficulty in deriving appropriate spring
structures. Lateral pile analysis is one of the key de- constant. Subgrade modulus is not an inherent soil
sign issues. Pile deflections and design actions property but load and deflection dependent. A lot of
(bending moments, shear forces) are assessed from research has been undertaken, and quite a number of
lateral pile analysis and used in pile design. approaches have been proposed for assessing K.
Various methods are available for lateral pile Those approaches generally correlate K with other
analysis, like Broms method, Brinch Hanson meth- soil parameters such as elastic modulus or shear
od, subgrade reaction method, p-y curve method, strength, and give highly variable K values.
elastic method, finite difference and finite element This paper provides a review of the various ap-
methods, and so on. Among the various methods the proaches in determining subgrade modulus. Effect of
subgrade reaction method is probably the most con- the variation of K on analysis is studied using the
venient method for structural modelling, and thus software package Space Gass. Lateral pile analyses
widely used by structural engineers. The theory of based on more sophisticated methods, i.e. p-y curve
subgrade reaction assumes that the unit soil reaction and finite element methods, are also carried out us-
(p) for a laterally loaded pile increases linearly with ing software packages LPile/Group and Plaxis 3D.
the lateral deflection (y), as expressed in the follow- The Space Gass analysis results are compared with
ing equation (Terzaghi 1955) . the LPile/Group and Plaxis to study the reliability of
the subgrade reaction method in lateral pile analysis.
𝑝 = 𝑘𝑠 𝑦 (1)
In various publications ks is called by many dif-
ferent names such as coefficient of subgrade reac- 2 SUBGRADE MODULUS CORRELATIONS
tion, modulus of subgrade reaction, subgrade reac-
tion, subgrade modulus, etc. To avoid confusion ks is As discussed above, subgrade modulus is a con-
called coefficient of subgrade reaction in this paper, ceptual parameter for lateral pile analysis, and vari-
while ks multiplied by structure size (pile diameter ous approaches have been proposed to determine K
for instance) is called subgrade modulus (K). by correlation with other soil parameters. Some cor-
In structural analysis, piles are commonly incor- relations are shown Table 1 based on Sadrekarimi &
porated into the analysis model with surrounding Akbarzad (2009) and CIRIA Report 103 (2004).
Table 1. Common correlations proposed for ks (K=ksB) sand, K=nhz according to Terzaghi (1955), where nh
Investigator Suggested Expression is the rate of increase with depth, and z is depth. For
Biot 0.95 𝐸𝑠 𝐵4 𝐸𝑠
ks = [ ]0.108 medium dense dry sand, nh varies from 3.5 to 10.9
𝐵(1−𝑣𝑠2 ) (1−𝑣𝑠2 )𝐸𝐼
𝐸𝑠 MN/m3 as suggested by Terzaghi. In this study, nh is
Bowles
𝑘𝑠 =
𝐵′ (1
− 𝑣𝑠2 )𝑚𝐼 𝑠 𝐼𝑓
adopted to be 7 MN/m3, while the lower and upper
bounds are adopted to be 3.5 MN/m3 and 14 MN/m3.
Jamiolkowski 𝑘𝑠 = 500 𝑡𝑜 700𝑐𝑢 /𝐵

Klopple and Glock 2𝐸𝑠

𝑘𝑠 = 3.2 Analysis methodology
𝐵(1 + 𝑣𝑠 )
Meyerhof and Baike 𝐸 Single pile analysis is performed using three soft-
𝑘𝑠 =
𝐵(1 + 𝑣𝑠 ) ware packages, Space Gass, LPile and Plaxis 3D.
Selvadurai 0.65 𝐸𝑠 Space Gass is commonly used in structural analysis,
𝑘𝑠 = .
𝐵 1 − 𝑣𝑠2 while the other two in geotechnical analysis.
Skempton 𝑘𝑠 = 80 𝑡𝑜 320𝑐𝑢 /𝐵 In Space Gass model, horizontal ground reaction
Terzaghi 𝐵+1 2 is simulated by linear springs at 1 m interval. Spring
For sand 𝑘𝑠 = 𝑘𝑠1 ( )
1
2𝐵 stiffness is obtained by multiplying K by contribu-
For clay 𝑘𝑠 = 𝑘𝑠1 tive pile length. Pile is restrained vertically at base.
𝐵
Terzaghi For sand 𝑘𝑠 = 𝑛ℎ 𝑧/𝐵 LPile (Ensoft 2017) analysis is based on the p-y
For clay 𝑘𝑠 = 67𝑐𝑢 /𝐵 curve method (Matlock & Reese 1960, Reese & Van
Vesic 0.65𝐸𝑠 12 𝐸𝑠 𝐵4 Impe 2011), which adopts non-linear relationship
𝑘𝑠 = √ between ground reaction and lateral deflection. In
𝐵(1 − 𝑣𝑠2 ) 𝐸𝑝 𝐼
this study, built-in p-y curves and default settings in
𝐸𝑠 (1 − 𝑣𝑠 ) μ
Vlassov
𝑘𝑠 = ( ) LPile are adopted. Clay is modelled by Reese’s Stiff
(1 + 𝑣𝑠 )(1 − 2𝑣𝑠 ) 2𝐵 Clay without Free Water, with the Strain Factor
The table shows that K (=ksB) is generally corre- (E50) set to be determined by the program. For sand,
lated with the following parameters: a) measurement Reese’s Sand model is used, and the rate of modulus
from field testing such as plate-load test; b) elastic increase (k) is set to Default.
property of soil; c) undrained shear strength; d) In Plaxis 3D (2017), Mohr-Coulomb model is
ground depth; and e) bearing capacity. Structure size used for soil, while pile is modelled as “Embedded
is included in all the correlations, while flexural Beam”. The embedded beam element models pile as
stiffness is included in some correlations. The table a beam interacting with soil by means of special in-
also indicates a high variation of K value. Due to the terface elements. Skin resistance of the embedded
fact that K is load and strain dependent, it is thought beam is set to be layer dependent, while ultimate end
that each correlation has its rationality and limita- bearing pressure is set to be 1000 kPa.
tion, and is applicable in a certain scenario. Ac-
knowledging the limitation of the correlations, ge-
otechnical engineers normally provide a range, not a 3.3 Analysis results
single value of K to account for the variation. Single pile is analysed for two lateral load cases ap-
plied at pile top, which are 300kN and 600kN. The
analysis is undertaken using three software packag-
3 STUDY CASE 1: SINGLE PILE es, Space Gass, LPile and Plaxis 3D. The analysis
3.1 Analysis cases results are discussed below.
A 20 m long, 0.9 m diameter concrete pile fully em- 3.3.1 Spring stiffness
bedded in ground is assumed in the study. Two soil In Space Gass analysis the subgrade modulus (K) is
types are considered, i.e. stiff clay and medium sand. assumed to be constant and evaluated from estab-
Properties of the two soils are shown in Table 2. lished correlations. LPile analysis is based on p-y
curve where the spring stiffness (p/y) varies with de-
Table 2. Soil parameters adopted in the analysis
flection. The spring stiffness adopted in the Space
Soil Unit weight Cohesion Friction Angle Modulus
Gass model and those from LPile are plotted in Fig-
kN/m3 kPa Degrees MPa
Clay 18 75 0 15
ure 1 for clay and sand under the two load cases.
Sand 18 0 34 30 The figure shows that spring stiffness values from
the correlations and p-y curves are comparable with-
For stiff clay, Vesic’s correlation (1961) is used
in shallow depth. With depth increase the difference
to evaluate subgrade modulus. Assuming modulus of
becomes larger and is up to about 300 times for clay
pile Ep = 34500 MPa, K is calculated to be about 10
and 45 times for sand. As known p-y curve charac-
MPa. To study the effect of variation of K, K is di-
terises non-linear soil pressure-deflection relation-
vided by 2, and multiplied by 2, to give likely lower
ship. With increasing depth and reducing loads, de-
bound of 5 MPa, and upper bound of 20 MPa. For
flection is smaller and soil is stiffer. Terzaghi’s K
for sand also increases with depth, but the rate of in- strain relationship, which would give soil stiffness
crease is apparently lower than that from p-y curve. comparable to the correlations. Accordingly the pile
internal forces from Plaxis are similar to Space Gass.
3.3.2 Deflection and internal forces Note deflections from Plaxis are also similar to
Soil spring stiffness from correlation is significantly Space Gass results with base case spring stiffness
different from that assessed from LPile. Impact of being adopted. Sensitivity analysis by Space Gass
the difference on pile analysis is investigated in this shows that with reducing spring stiffness, pile de-
paper. The Space Gass analysis results based on flection and maximum moment increase, maximum
spring stiffness from correlations are plotted in Fig- shear is not affected but its depth increases.
ures 2-3 for clay, Figures 4-5 for sand, together with
the LPIle and Plaxis 3D analysis results.
It can be seen that varying soil spring stiffness in 4 STUDY CASE 2: PILE GROUP
Space Gass affects lateral deflection apparently, but
has less impact on bending moment, and the least ef- 4.1 Analysis cases
fect on shear force. For pile in sand, the moment and A four-pile pile group is analysed for lateral loads.
shear force curves from Space Gass are comparable The square pile group comprises four concrete piles
to LPile and Plaxis. For pile in clay, however, the with rigid connection to a pile cap on ground sur-
moment and shear force curves from LPile are dis- face. The piles are 20 m long, 1 m diameter and at 3
tinctive and the maximum values could be signifi- m center to center spacing. The pile cap is 5 m
cantly larger. That is due to the significant difference square and 1.5 m thick. Ground conditions are the
between soil spring stiffness used in LPile and Space same as those adopted in single pile analysis, i.e.
Gass for clay, whereas the difference is relatively stiff clay and medium sand. To simplify the analysis,
small for sand, as indicated in Figure 1. Mohr- pile bases are assumed to be on competent rock.
Coulomb model used in Plaxis adopts linear stress

(a) Clay: full pile depth (b) Clay: 4m depth (c) Sand: full pile depth (d) Sand: 4m depth
Figure 1. Soil spring stiffness in Space Gass & LPile.

Figure 2. Analysis results of single pile in clay (lateral load = 300 kN).
Figure 3. Analysis results of single pile in clay (lateral load = 600 kN).

Figure 4. Analysis results of single pile in sand (lateral load = 300 kN).

Figure 5. Analysis results of single pile in sand (lateral load = 600 kN).

GROUP analysis is undertaken using the same p-

y curves as used in the single pile analysis. Piles are
4.2 Analysis methodology assumed to have 0.5 m socket in rock. Rock is simu-
Pile group is analysed using Space Gass, GROUP lated by the Strong Rock model in GROUP (Ensoft,
and Plaxis 3D. In Space Gass analysis, only the base 2015) with 15 MPa UCS. Group effect is set to be
case soil springs (K = 10 MPa for clay and nh = 7 automatically determined by the software. Fixed
MN/m3 for sand) are considered. Piles are modelled connection between pile and pile cap is adopted.
in the same way as that for single pile. Group effect Plaxis analysis is also undertaken with the same
is not considered in the analysis. Pile cap is mod- soil models as for single pile analysis except that a
elled as a frame with rigid connection with piles. 0.5 m rock layer at the base is included. Rock is
modelled as a Mohr-Coulomb material with c = 200 Space Gass analysis, the four piles behave the same
kPa,  = 40° and E = 800 MPa. Pile cap is modelled as group effect is not considered. For GROUP and
as “Plate” with rigid connection with piles modelled Plaxis, group effect is automatically considered in
as “Embedded Beam”. analysis, and the front (F) pile and rear (R) pile be-
have differently as indicated in the plots.
The figures show that Space Gass gives smaller
4.3 Analysis results deflection than Plaxis, which may be due to group
Pile group is analysed for two lateral loads, 2000kN effect being ignored in Space Gass analysis. Deflec-
and 4000kN acting at pile top (bottom of pile cap). tion from GROUP is the smallest under a small load,
The analysis results are presented in Figures 6-9. In but increases quickly with load due to the non-linear

Figure 6. Analysis results of pile group in clay (lateral load = 2000 kN).

Figure 7. Analysis results of pile group in clay (lateral load = 4000 kN).

Figure 8. Analysis results of pile group in sand (lateral load = 2000 kN).
Figure 9. Analysis results of pile group in sand (lateral load = 4000 kN).

nature of p-y curve. Similar to the findings of single 4) Mohr-Coulomb model is a linear elastic-
pile, p-y curve method (GROUP) gives significantly perfectly plastic model. Plaxis analysis results based
different internal forces for piles in clay compared on Mohr-Coulomb model is similar to Space Gass
with the other two methods, which is thought to be analysis based on linear soil springs. If a non-linear
due to the large difference of soil stiffness assumed soil model is adopted, Plaxis analysis results may be
for clay. The difference is apparently smaller for more comparable to LPile/GROUP analysis.
piles in sand. Pile internal forces from Space Gass 5) Front pile in a pile group takes more loading
and Plaxis are generally comparable. Analysis shows than rear pile. Ignoring group effect in analysis tends
that front pile takes more loading than rear pile. Ig- to underestimate pile deflection and internal forces.
noring group effect in analysis tends to underesti- 6) Lateral pile analysis based on the subgrade
mate pile deflection and internal forces although the reaction method does not give irrational results.
difference may not be significant in some cases. With K properly assessed, the simple method could
be reasonably used. However, due to the high varia-
tion of K, the method is liable to under- or over-
5 CONCLUSIONS AND DISCUSSIONS design of pile. Verification with more sophisticated
approaches like the p-y curve or finite element
This paper discussed subgrade reaction method in method is recommended for detailed design, espe-
lateral pile analysis, and compared that method with cially for piles subject to large lateral loads and
p-y curve and finite element methods based on the structures sensitive to lateral deflection.
analysis of a single pile and a four-pile pile group
using Space Gass, LPile/GROUP and Plaxis 3D.
Main conclusions drawn are discussed below. 6 REFERENCES
1) Subgrade reaction method based on linear
soil springs (constant subgrade modulus) is widely CIRIA Report 103. 2004. Design of laterally-loaded piles.
used in structural pile analysis. However, subgrade Ensoft Inc. 2017. Technical Manual for LPile 2018.
modulus is not an inherent soil property but load and Ensoft Inc. 2015. Technical Manual for GROUP 2016.
deflection dependent. Various correlations in deter- Matlock, H. & Reese, L. C. 1960. Generalized solutions for
mining K give highly variable K values. laterally loaded piles. Journal of the Soil Mechanics and
2) Variation of K has apparent influence on lat- Foundations Division, ASCE 86: 63-91.
eral deflection in structural analysis, less effect on Plaxis 3D. 2017. Material Models Manual.
bending moment, and the least effect on shear force. Reese, L.C. & Van Impe, W.F. 2011. Single Piles and Pile
3) p-y curve characterizes soil’s non-linear be- Groups Under Lateral Loading. 2nd edition. Boca Raton,
havior under lateral load. Soil spring stiffness values CRC Press.
assessed from p-y curve are significantly different Sadrekarimi, J. & Akbarzad, M. 2009. Comparative study of
from subgrade reaction method. Comparatively, the methods of determination of coefficient of subgrade reac-
difference is larger for clay than for sand. Conse- tion. Electronic Journal of Geotechnical Engineering 14E.
quently, difference of pile analysis between LPile/ Terzaghi, K. 1955. Evaluation of coefficient of subgrade reac-
GROUP (based on p-y curves) and Space Gass tion. Geotechnique 5(4), 297- 326.
(based on constant subgrade modulus) is larger for Vesic, A. B. 1961. Beams on Elastic Subgrade and Winkler’s
piles in clay than for piles in sand. Hypothesis. Proc. 5th Int. Conf. on Soil Mechanic and
Foundation Engineering, Paris: 845-850.