Computers and Geotechnics 81 (2017) 307–321

Contents lists available at ScienceDirect

Computers and Geotechnics

journal homepage: www.elsevier.com/locate/compgeo

Research Paper

Pile driving formulas based on pile wave equation analyses

Rodrigo Salgado a,⇑, Yanbei Zhang a, Grace Abou-Jaoude b, Dimitrios Loukidis c, Vibhav Bisht a
a
Lyles School of Civil Engineering, Purdue University, 550 Stadium Mall Drive, West Lafayette, IN 47907, United States
b
Department of Civil Engineering, Lebanese American University, Lebanon
c
Department of Civil and Environmental Engineering, University of Cyprus, Cyprus

a r t i c l e i n f o a b s t r a c t

Article history: Pile driving formulas, which directly relate the pile set resulting from a hammer blow to the static load
Received 6 May 2016 capacity of the pile, are often used to decide whether a pile will have the required design capacity.
Received in revised form 3 September 2016 However, existing formulas do not consider soil or pile type, and do not, in general, reliably predict pile
Accepted 11 September 2016
capacity. In this paper, an advanced model for dynamic pile driving analysis was used to develop accurate
pile driving formulas. The proposed driving formulas are validated through well-documented case histo-
ries. Comparisons of predictions from the proposed formulas with results from static and dynamic load
Keywords:
tests show that they produce reasonably accurate predictions of pile capacity based on pile set
Pile dynamics
Pile load capacity
observations.
Numerical modeling Ó 2016 Elsevier Ltd. All rights reserved.
Pile driving formula
Case studies

1. Introduction distinction of the soil type surrounding the pile (e.g., clay or sand)
or the pile type (end-bearing versus floating or friction piles). As a
Proper modeling of pile driving is important for both planning result, the formulas used in practice often either largely over-
and inspection of pile driving operations. The dynamic response predict or under-predict capacity [2]. Safety factors as large as
of a pile during driving is very complex, involving the interactions six have been recommended when using these formulas [3].
of the hammer, cushion, pile and soil during application of an Existing pile driving formulas are based on a simple concept:
impact load. Because the pile driving process is variable and the energy of the hammer (ram) before impact is equal to the work
imposes significant changes to the state of the soil around the pile done by the total pile resistance for the observed pile head perma-
that are difficult to model, both the design and quality control of nent displacement (pile set) after a blow plus the energy dissipated
piling operations have been subject to considerable uncertainty inside the pile and within the soil during the blow, as well as the
and have been approached conservatively [1]. energy lost at impact in the various driving components between
Reliable estimation of the capacity of a driven pile based on the the hammer and the pile head. This can be written mathematically
ease or difficulty with which the pile is driven allows an inspector as:
to decide when pile driving can be discontinued. One of the tools
eh W H H ¼ Q ult ðs þ sc Þ ð1Þ
that may be used to decide whether an installed pile will have
the predicted capacity are pile driving formulas, which relate the where WH is the hammer (ram) weight; H is the hammer drop
pile set per blow to the capacity of the pile. Due to their simplicity, height; eh is the hammer efficiency; Qult is the ultimate pile capac-
these formulas have been widely and frequently used in practice. ity; s is the observed pile set; and sc is an empirical constant
Most follow the rational derivation of dynamic pile capacity using expressing the aforementioned energy losses and the energy stored
impulse-momentum principles to empirically relate the energy temporarily inside the pile due to elastic compression during the
generated by the driving system to the pile displacement. They blow process.
differ depending on simplifying assumptions and empirical Eq. (1) has been the basis for development of most empirical
adjustments. Moreover, existing pile driving formulas make no pile driving formulas. The equation is solved for Qult with the input
variable being the pile set s. The simplicity of these formulas,
⇑ Corresponding author. combined with budget limitations, has led to their significant use
E-mail addresses: [email protected] (R. Salgado), [email protected]
in practice. Pile driving formulas available in the literature include
(Y. Zhang), [email protected] (G. Abou-Jaoude), [email protected] the Gates formula [4], the modified ENR formula [5], the Eytelwein
(D. Loukidis), [email protected] (V. Bisht). formula [6], the Danish formula [7], the Janbu formula [7], the

http://dx.doi.org/10.1016/j.compgeo.2016.09.004
0266-352X/Ó 2016 Elsevier Ltd. All rights reserved.
308 R. Salgado et al. / Computers and Geotechnics 81 (2017) 307–321

Pacific Coast Uniform Building Code (PCUBC) formula [3], the Cana- 2. Static pile capacity calculations
dian National Building Code formula [3], and the Navy-McKay for-
mula [3]. Table 1 lists the five formulas that we use later for The limit resistance (Q L) of an axially loaded pile is the load at
comparison purposes. Salgado [8] suggests that consideration of which a pile plunges through soil. An ultimate limit state is gener-
existing empirical formulas in deep foundation quality control is ally expected to be reached at loads less than the limit load. For
typically not advantageous since the safety factors recommended piles in sand, the ultimate load Q ult is defined as the pile load
when they are used are large. McVay et al. [9] note that, although Q10% that causes a settlement equal to 10% of the pile diameter B
empirical dynamic formulas are very easy to use, their predictions [16]. For piles in clay, except heavily overconsolidated clay, Qult
are characterized by considerable scatter and, in some cases, bias. is practically equal to Q L, since the latter may be mobilized for a
A critique of these formulas can be found in Likins et al. [10]. settlement less than 0.1B [8].
Attempts to improve pile driving formulas [11,12] have been only The ultimate pile resistance is the summation of the ultimate
partly successful due to the complexities of the problem, a highly base resistance Qb,ult and limit shaft resistance Q sL:
nonlinear wave propagation problem that involves complex
Q ult ¼ Q b;ult þ Q sL ð2Þ
mechanics. Recalibration of dynamic formulae to local conditions
has been shown to substantially improve their predictive perfor- The base resistance Qb,ult is calculated using:
mance [13], but these formulae are constrained to regions where
Q b;ult ¼ qb;ult Ab ð3Þ
data is available. Improved pile driving formulas without these
drawbacks and exhibiting less scatter would require lower factors where qb,ult is the ultimate unit base resistance and Ab is the area of
of safety and would be useful in practice. the pile base.
In this paper, we simulate the pile driving process using the soil The shaft resistance Q sL is given by:
reaction models described in detail in [14] and, based on the X
results from a series of parametric simulations, propose pile driv- Q sL ¼ qsL;i As;i ð4Þ
i
ing formulas that explicitly account for the soil and pile type
and, as a consequence, exhibit reduced prediction scatter. Pile driv- where qsL,i is the limit unit shaft resistance along the segment of the
ing formulas are developed for five general cases: floating piles in shaft intersecting the ith sub-layer of the soil and As,i is the corre-
clay, piles in uniform sand deposits, end-bearing piles in sand, end- sponding shaft surface area.
bearing piles in clay and piles crossing soft clay and bearing on The ultimate unit base resistance (qb,ult) and limit unit shaft
sand. Here the term end-bearing pile refers to a pile for which resistance (qsL) of driven piles in sands and in clays are calculated
the base resistance is an appreciable fraction of the total pile resis- in this paper using the Purdue design equations [15] summarized
tance. The static pile capacities for these soil profiles were calcu- below.
lated using a set of recent static design methods [15]. For a pile base embedded in a sand layer, the limit unit base
Well-documented case histories of static load tests on driven resistance (qbL) is set equal to the cone penetration resistance
piles are used to validate the proposed formulas. Moreover, the (qc), which may be estimated using the cone resistance relation-
predictions of the proposed formulas are compared with those of ship of Salgado and Prezzi [17]:
existing formulas. In the next section, we present in detail the  0 0:8410:0047DR
methodology followed to generate the proposed formulas, along qc r
¼ 1:64 exp ½0:1041/c þ ð0:0264  0:0002/c ÞDR  h
with the necessary background information on static capacity cal- pA pA
culation and dynamic pile driving analysis. ð5Þ

Table 1
Traditional pile driving formulas.

Formula Equationsa Notes

pffiffiffiffiffiffiffiffiffiffi
Gates formula [4] Q u ¼ a eh Eh ðb  logðsÞÞ s in mm
a = 104.5
b = 2.4
  
Modified ENR [5] Qu ¼ 1:25eh Eh W H þn2 W P C = 0.0025 m
sþC W H þW P
n = 0.5 for steel-on-steel anvil on steel
or concrete piles
Danish formula [7] eh Eh
Q u ¼ sþC
qffiffiffiffiffiffiffiffiffi
1

C 1 ¼ e2AEh Eh L

Pacific Coast Uniform Building Code Q u ¼ esþC

h Eh C 1 k = 0.25 for steel piles and 0.1 for
2
(PCUBC) formula [3]b H þkW P
all other piles
C1 ¼ W
W H þW P

C 2 ¼ QAE
uL

Janbu [3] Q u ¼ eKhuEsh

WP
C d ¼ 0:75 þ 0:15 W
 ffi
qffiffiffiffiffiffiffiffiffiffiffiffi
H

K u ¼ C d 1 þ 1 þ Ckd
k ¼ eh Eh2L
AEs

a
Symbols in formula equations: Qu = predicted pile capacity (in kN), eh = hammer efficiency; Eh = maximum driving energy of the
hammer (in kJ); s = observed pile set (in m if not specified); WH = weight of the ram (in kN); WP = weight of the pile (in kN); L = length of
the pile (in m); A = cross-sectional area of the pile (in m2); E = Young’s modulus of the pile material (in kPa).
b
The calculation of predicted static capacity using PCUBC formula requires iterations.
 R. Salgado et al. / Computers and Geotechnics 81 (2017) 307–321 309

where r0h is the in situ horizontal effective stress, which is equal to

K 0 r0v , and pA is a reference stress equal to 100 kPa or its equivalent vo
in the desired unit system, DR is the sand relative density in %, and
/c is the critical state friction angle. K0 is the at-rest lateral earth
pressure coefficient, and r0v is the in situ vertical effective stress.
The ultimate unit base resistance qb,ult is correlated to qbL as fol-
lows (modified after Salgado [8]):
qb;ult ¼ qb;10% ¼ ð1  0:0058DR ÞqbL ð6Þ

shaft reaction models

The limit unit shaft capacity (qsL) at a point of a pile lateral surface
in contact with a sand layer is calculated using the equation:
qsL ¼ K r0v tan dc ð7Þ
where dc is the interface friction angle, which is taken as equal to
0.9/c for steel piles and 0.95/c for concrete piles, and K is the lateral
earth pressure coefficient given by Han et al. [18]:
 
h
K ¼ 0:2 þ ð0:02qc =r0v  0:2Þ exp 0:05 ð8Þ
B
where h is the distance from the pile base to the depth under
consideration. base reaction
For a pile base embedded in a clay layer, the gross limit unit model
base resistance qbL is evaluated using the following equation:
qbL ¼ Nc su þ q0 ð9Þ Fig. 1. Formulation of one-dimensional pile driving analysis.

where su is the undrained shear strength of the clay and q0 is the soil
surcharge at the pile base level. The bearing capacity factor Nc is set
equal to 12.3 according to Salgado et al. [15]. The energy losses in the assembly placed between the ram and
The limit unit shaft resistance qsL at a point of a pile interfacing the pile head (along with energy loses in the hammer) are indi-
with clay is calculated using the a method: rectly taken into account by using an overall efficiency factor of
the driving system eeff. The value of eeff is practically equal to the
qsL ¼ asu ð10Þ hammer energy transfer ratio (ETR) computed using measure-
The values of a are determined using the expression proposed by ments made in dynamic load tests using accelerometers (for
Basu et al. [19] for short-term pile capacity estimation, slightly recording velocity time history data) and strain gauges (for record-
modified as: ing force time history data) attached to the pile head. The eeff is less
  
 than the hammer efficiency values eh, the latter generally ranging
r0v0  A2
from 0.65 to 0.95 in the existing pile driving formulas for various
a ¼ A1 þ ð1  A1 Þ exp  /c  /r;min ð11Þ
pA hammer types [3]. According to the Rausche database [20], the
where /r;min is the minimum residual friction angle of the clay, and average ETR for diesel hammers, single acting air/steam hammers,
A1 and A2 are determined as follows: and hydraulic hammers on steel and concrete piles vary from 0.25
8 
to 0.8. The average ETR for drop hammers on steel and concrete
< 0:43 /c  /r; min ¼ 12
> piles is taken as 0.55 [21–24].

A1 ¼ 0:75 /c  /r; min ¼ 5 ð12Þ The shaft reaction model, shown in Fig. 2, is based on a contin-
>
:
1:00 /c  /r; min ¼ 0 uum approach [25] and has three components: (1) a soil disk rep-
resenting the near field soil surrounding the pile shaft; (2) a
  rheological model representing the thin shear band forming at
su
A2 ¼ 0:55 þ 0:43 ln 0
ð13Þ the soil-pile interface located at the inner boundary of the soil disk;
r v0
and (3) far-field-consistent boundaries placed at the outer bound-
Values of A1 can be obtained for a range of /c–/r,min through linear ary of the soil disk. The continuum approach accommodates a non-
interpolation. linear soil stress-strain relationship in the near-field disk and
provides solutions valid for the highly transient motion caused
3. Pile driving analysis by the impact of the hammer. The rheological model [26] consists
of a plastic slider and a viscous dashpot connected in parallel
Salgado et al. [14] developed a set of improved soil reaction (Fig. 2). In the absence of slippage between pile and soil, the soil
models for use in one-dimensional pile driving analysis that take reaction at the pile shaft wall is taken as the shear stress in the soil
explicitly into account soil non-linearity and hysteresis, are consis- segment between nodes 1 and 2. The strength of the plastic slider
tent with the mechanics of pile driving, and have input parameters is set equal to the static unit limit shaft resistance qsL. As long as
that are physically meaningful. the stress ss is less than the unit limit shaft resistance qsL, no slid-
In the one-dimensional pile driving analysis, the pile is dis- ing occurs and ss is transmitted to the pile segment. Sliding initi-
cretized into segments, each of which is modeled as a mass con- ates once ss = qsL, at which point the viscous dashpot is activated.
nected to the segments above and below through linear elastic The reaction of the viscous dashpot is a power function of the rel-
springs corresponding to the axial stiffness of the pile. The hammer ative velocity between the pile and the first node of the near field
impact is represented by the instantaneous application of an initial component [26,27].
velocity to the topmost mass (the ram mass) of the modeled The base reaction model, shown in Fig. 3, takes into account the
lumped mass system (Fig. 1). The pile helmet and cushion are nonlinear soil response under the pile base and the loading rate
not explicitly considered in the simulations of the present study. effect on base resistance; it also distinguishes between the differ-
310 R. Salgado et al. / Computers and Geotechnics 81 (2017) 307–321

Pile Shear band behavior

Far-field

CFF

1 2 i-1 i i+1 Ns KFF

Near field response

Fig. 2. Shaft reaction model consisting of three parts: continuous near field, far-field consistent spring and radiation dashpot, and rheological shear band model at the soil-
pile interface.

Nonlinear spring with rate pile driving simulations, the relative density of the sand layer takes
effect on strength various values from 10% to 90%, with 90% usually being of little
Pile practical significance and used only to cap the numerical results
Rb w
from above. The second case is that of a floating pile driven into
a normally consolidated clay layer (Fig. 4b). In the parametric
Radiation study, the value of the ratio of the normally consolidated
dashpot
undrained shear strength to vertical effective stress (su =r0v )NC
was varied from 0.20 to 0.30. In the third case, the pile crosses a
loose surficial sand layer and bears on a denser sand layer; the rel-
wb
ative density of the sand layer along the pile shaft is assumed to be
Fig. 3. Base reaction model consisting of a nonlinear spring and a radiation dashpot.
30%, and the relative density of the bearing layer ranges from 40%
to 90% (Fig. 4c). To simulate the case of an end-bearing pile in clay,
a pile crossing a normally consolidated clay layer and resting on an
ent types of damping. The total base reaction is the sum of the over-consolidated clay layer with OCR = 10 (Fig. 4d) was consid-
spring reaction and the radiation dashpot reaction. The stiffness ered. The ratio of undrained shear strength to vertical effective
of the non-linear spring at the pile base is formulated to follow a stress of the bearing layer was set to 6.3 (= OCR0.8, [8]) times that
hyperbolic load-settlement relationship so that the quasi-static of the overlying normally consolidated layer. The final case is a pile
response remains consistent with the load-settlement curve crossing a normally consolidated clay layer and resting on a rela-
observed in static pile load tests. The load-settlement curve pre- tively dense sand layer (Fig. 4e) with relative density varying from
dicted by the non-linear spring is set to reach an asymptote equal 40% to 90%. For cases 3–5, which contain two layers, only the tip of
to the limit base resistance QbL multiplied by a power function of the pile is assumed to rest on the dense bottom layer while the
the base velocity in order to account for the loading rate effects entire shaft is assumed to lie on the top layer. Table 2 lists the soil
on soil strength. The non-linear spring captures the energy loss parameters assumed for these five idealized profiles, which are
inside the failure mechanism due to material hysteresis. The radi- expected to approximate a wide range of soil profiles found in
ation dashpot coefficient, which accounts for the effect of embed- practice.
ment depth and hysteretic damping inside the far field soil, is set to The pile variables of the parametric analyses are the pile length
vanish as the spring reaction approaches the limit base resistance L, the pile diameter B, the pile wall thickness tw (in the case of
value. closed-ended steel pipe piles), and the pile material (concrete or
The predictions using these shaft and base reaction models steel). The hammer variables are the weight of the ram WH and
were shown to compare favorably with measured pile driving data the drop height H. In the series of parametric runs for the develop-
for two examples of fully-instrumented, full-scale pile load tests ment of the formulas, the initial potential hammer energy or rated
[14]. maximum energy Eh, calculated as the product of WH and H values
shown in Table 3, is multiplied by a fixed value eeff equal to 0.50 to
4. Proposed pile driving formulas simulate all hammer-pile material conditions. The exact value of
eeff is inconsequential to the parametric study results and the cali-
The goal of the present study is to derive pile driving formulas bration of the constants of the proposed equations because the eeff
for concrete piles and closed-ended steel pipe piles for five general enters the proposed equations solely through the product eeffEh,
pile-soil profile cases (Fig. 4). For cases involving sand, states rang- which is the energy transmitted to the pile head (referred to as
ing from loose to very dense are considered. For cases involving ENTHRU in PDAÒ dynamic testing). If the user plugs in the correct
clay, the ratio of undrained shear strength to vertical effective eeffEh values in the proposed formulas, the formulas would yield
stress for normally consolidated clay (NC) and the over- valid pile capacity predictions. The eeff, which is equal to the energy
consolidation ratio are used as variables. The static capacity equa- transfer ratio ETR, may be estimated from PDA measurements per-
tions and the 1D dynamic pile driving analysis methodology pre- formed under similar driving conditions or, if such information is
sented above were used in an extensive parametric study aiming not available, from databases present in the literature (e.g. [20]).
at producing pairs of values of total static pile capacity Qult and In routine onshore practice, the driven pile length is usually in
final pile set s for a wide range of values for each variable involved the range of 10–50 m, so pile lengths equal to 10 m, 20 m, 30 m
in the pile driving problem. and 40 m were considered. The values of the steel pipe wall thick-
The first general soil profile configuration we address is that of a ness were 9.5 mm (3/8 in.), 12.7 mm (1/2 in.), 15.9 mm (5/8 in.)
pile in a sand layer with uniform density (Fig. 4a). In the dynamic and 19.1 mm (3/4 in.). Table 3 lists the values (normalized with
 R. Salgado et al. / Computers and Geotechnics 81 (2017) 307–321 311

Normally
Relative density
consolidated clay
DR = 10 ~ 90%
OCR = 1

(a) (b)

Normally Normally
Relative density
consolidated clay consolidated clay
DR = 30%
OCR = 1 OCR = 1

Relative density Overconsolidated clay Sand

DR = 30 ~ 90% OCR = 10 DR = 40 ~ 90%

(c) (d) (e)

Fig. 4. Typical pile-soil profile systems found in pile design: (a) piles in sand of uniform density; (b) a floating pile in clay; (c) an end-bearing pile in sand; (d) an end-bearing
pile in clay; and (e) a pile crossing clay resting on sand.

Table 2
Soil parameters for typical soil profiles.

Case Shaft parameter Values of shaft parameter Base parameter Values of base parameter
Piles in sand of uniform density Relative density DR (%) 10,20,30,40,50, 60,70,80,90 Relative density DR (%) Same as for shaft
End-bearing in sand Relative density DR (%) 30 Relative density DR (%) 40, 50, 60, 70, 80, 90
Floating piles in clay Overconsolidation ratio OCR 1 Overconsolidation ratio OCR 1
su =r0v 0.2,0.23,0.25, 0.28,0.3 su =r0v Same as for shaft
End-bearing piles in clay Overconsolidation ratio OCR 1 Overconsolidation ratio OCR 10
su =r0v 0.2,0.23,0.25, 0.28,0.3 ðsu =r0v ÞNC Same as for shaft

Piles crossing clay resting on sand Overconsolidation ratio OCR 1 Relative density DR (%) 40, 50, 60, 70, 80, 90
su =r0v 0.25

Table 3
respect to reference force WR = 100 kN and reference length
Hammer and pile parameters.
LR = 1 m) considered for all variables. The range of pile, hammer
Controlling variable Values and soil parameters were such that the driving conditions of the
Normalized hammer weight: WH/WR 0.1, 0.2, 0.3, 0.5, 1.0, 1.5 parametric simulations ranged from hard (near refusal) to soft
Normalized drop height: H/LR 1.5, 2.5, 3.5, 4.5, 5.5 driving conditions.
Normalized pile length: L/LR 10, 20, 30, 40
The proposed pile driving formulas are expressed as products of
Normalized pile diameter: B/LR 0.178, 0.356, 0.534,
0.712, 0.89, 1.068 power functions of dimensionless variables. For each of the five
Normalized pile wall thickness: tw/LR (for closed- 0.0095, 0.0127, 0.0159, cases, they have the following forms:
ended steel pipe pile only) 0.0191  c 2   c4  c
Q 10% eeff Eh DR s WP 5
3
Note: reference force: WR = 100 kN = 2.25  10 lbf = 22.5 kips; reference length: ¼ c1 exp c3 ð14Þ
LR = 1 m = 3.28 ft = 39.3 in.
pa L2R W R LR 100% LR WR
312 R. Salgado et al. / Computers and Geotechnics 81 (2017) 307–321

for piles in a uniform sand layer, piles crossing a normally consoli- on the information in the field investigation report of a given pro-
dated clay layer and resting on a dense sand layer and end-bearing ject, which normally precedes pile installation.
piles in sand; and A series of 1D pile driving simulations using the Salgado et al.
 c2   c4  c [14] model were performed through the application of a single
QL eeff Eh su s WP 5 blow on a wished-in-place pile at the final installation depth for
¼ c1 exp c3 0 ð15Þ
pa L2R W R LR rv LR WR all values considered for the pile dimensions, hammer weight, drop
height and soil profile variables (Tables 2 and 3) to produce the pile
for floating piles in clay and end-bearing piles in clay. The relative set s. The corresponding ultimate pile capacity Qult were calculated
density in Eq. (14) is that of the bearing layer, while the ratio from Eqs. (5)–(13), which have been validated by comparison to
su =r0v in Eq. (15) is that of the normally consolidated (OCR = 1) clay the results of load tests usually performed between a few days
surrounding the shaft. The exponents and multipliers c1–c5 are and a few weeks after driving. Because of this and the fact that Q sL
obtained through non-linear regression analysis. is calculated from equations developed and validated for driven
The form of the proposed equations is related to that of the piles, correlation of the set s to the ultimate capacity Q ult has
Janbu formula listed in Table 1, as can be seen by rewriting them two implications: (1) the pile driving formulae implicitly corrects
as follows: for installation effects, so that, contrary to what might appear to
be the case, the numerical simulations of the pile set s due to a
eeff Eh
Q 10% ¼     single blow are not in effect analysis of a fresh blow on a
W R LR 1 eeff Eh 1c2  DR
 s c4 1 W c5
pa L3R c1 W R LR
exp c3 100 LR
P
WR
s wished-in-place pile and (2) the pile driving formulas account for
any setup that may occur during a time period ranging from
eeff Eh shortly after EOD to several days after installation. Values of c1,
¼ ð16Þ
As c2, c3, c4, and c5 in Eqs. (14) and (15) are parameters determined
by fitting the equations to the pairs of ultimate pile capacity Q ult
and
and corresponding pile set s produced by the series of parametric
eeff Eh analyses using non-linear least-squares regression. The values of
QL ¼      c4 1  c5  these coefficients for each typical soil profile are listed in Table 4
W R LR 1 eeff Eh 1c2 WP
pa L3R c1 W R LR
exp c3 rsu0 LsR WR
s for closed-ended steel pipe piles and Table 5 for concrete piles.
v
The coefficient of determination R2 quantifying the goodness of
eeff Eh
¼ ð17Þ fit ranges from 0.87 to 0.996. Any discrepancy between the Qult val-
Bs ues from the proposed formulas and those from Eqs. (5)–(13)
We see that both expressions can be written in the form of the (Figs. 5 and 6) is due to regression analysis residual.
Janbu pile driving formula. While the energy losses in many pile Figs. 5 and 6 compare the static ultimate capacity Qult values
driving formulas are accounted for by incrementing the final set s predicted by the proposed formulas to those predicted by the
(as in Eq. (1)), in the Janbu formula, s is multiplied by a factor Ku, PCUBC and Janbu pile driving formulas for set s values ranging
which is always greater than 1, with Ku  1 corresponding to the from about 30% of the pile diameter to nearly zero (refusal) for
percentage of the final set that is lost as energy. The factor Ku closed-ended steel pipe piles (Fig. 5) and concrete piles (Fig. 6)
depends on the pile weight, the hammer (ram) weight, the axial with W/WR = 0.2, H/LR = 2.5 and L/LR = 20. Each plot in Figs. 5 and
stiffness of the pile, the hammer energy ehEh and the final pile set 6 contains data for various values of pile diameter B and soil prop-
s. Similarly, in the proposed formulas, the set is multiplied by fac- erties su and DR. We see that, for soft driving conditions (s > 0.05B),
tors A or B, which depend on the energy transferred to the pile head there is agreement between the proposed and existing formulas.
eeffEh, the final pile set, the pile weight, and the soil characteristics. For hard driving conditions, the PCUBC and Janbu formulas appear
The novelty of the proposed formulas is that they contain, as vari- to significantly underpredict the pile capacity compared to the pro-
ables, soil properties, namely DR for sands and su for clays. Both posed formulas. The proposed equations were fitted to simulation
variables are expected to be available or be easily estimated based results that include near refusal cases. Under such conditions, the

Table 4
Coefficients of pile driving formulas for closed-ended steel pipe piles.

Case Variable
c1 c2 c3 c4 c5 R2
Piles in sand of uniform density 14.97 0.33 1.04 0.41 0.89 0.91
End-bearing piles in sand 8.11 0.41 0.74 0.53 0.69 0.87
Floating piles in clay 0.72 0.71 1.3 0.76 0.3 0.97
End-bearing piles in clay 1.57 0.59 1.51 0.67 0.46 0.96
Piles cross clay resting on sand 5.77 0.36 0.12 0.52 0.71 0.95

Table 5
Coefficients of pile driving formulas for precast concrete piles.

Case Variable
c1 c2 c3 c4 c5 R2
Piles in sand of uniform density 19.97 0.07 1.73 0.07 0.79 0.99
End-bearing piles in sand 22.05 0.08 1.13 0.08 0.76 0.99
Floating piles in clay 0.96 0.69 1.82 0.56 0.28 0.91
End-bearing piles in clay 3.01 0.24 3.1 0.19 0.75 0.96
Piles cross clay resting on sand 9.5 0.08 1.56 0.09 0.77 0.99
 R. Salgado et al. / Computers and Geotechnics 81 (2017) 307–321 313

500 25

400 20

300
Qult /(pALR 2)

15

Qult /(pALR 2)
W/WR = 0.2
H/LR =2.5
L/LR = 20
200 W/WR =0.2 10 tw/LR = 0.0127
H/LR = 2.5
L/LR = 20 Closed-ended steel pipe piles

100 tw/LR = 0.0127 5

Closed-ended steel pipe piles

0 0
0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.10 0.20 0.30 0.40
s/B s/B
(a) (c)
400 50

40
300

30
Qult/(pALR 2)

Qult /(pALR 2)

200
W/WR = 0.2
W/WR =0.2 20 H/LR = 2.5
H/LR = 2.5 L/LR = 20
L/LR = 20 tw /L R =0.0127
100
tw/LR = 0.0127 10 Closed-ended steel pipe piles

Closed-ended steel pipe piles

0 0
0.00 0.04 0.08 0.12 0.16 0.00 0.10 0.20 0.30
s/B s/B
(b) (d)
80

60
Qult /(pALR 2)

40
W/WR = 0.2
H/LR = 2.5
L/LR = 20
tw /LR =0.0127
20
Closed-ended steel pipe piles

0
0.00 0.04 0.08 0.12 0.16 0.20
s/B
(e)
Fig. 5. Comparison between capacity predicted by proposed formula, calculated static capacity, capacity predicted by Janbu formula and capacity predicted by PCUBC
formula for closed-ended steel pipe piles: (a) piles in sand of uniform density; (b) end-bearing piles in sand; (c) floating piles in clay; (d) end-bearing piles in clay; and (e) end-
bearing piles crossing clay and resting on sand.
314 R. Salgado et al. / Computers and Geotechnics 81 (2017) 307–321

500 30

400

20
300
Qult /(pALR 2)

Qult /(pALR2)
W/WR = 0.2
200 H/LR = 2.5
W/WR = 0.2 10 L/LR = 20
H/LR = 2.5
L/LR = 20 Precast concrete piles
100

Precast concrete piles

0 0
0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.10 0.20 0.30 0.40
s/B s/B
(a) (c)
400 60

300

40
Qult /(pALR2)

Qult /(pALR2)

200
W/WR = 0.2
H/LR = 2.5
W/WR = 0.2 20 L/LR = 20
H/LR = 2.5
100
L/LR = 20 Precast concrete piles

Precast concrete piles

0 0
0.00 0.04 0.08 0.12 0.16 0.00 0.10 0.20 0.30
s/B s/B
(b) (d)

250

200
Qult /(pALR 2)

150

100
W/WR = 0.2
H/LR = 2.5
L/LR = 20
50

Precast concrete piles

0
0.00 0.04 0.08 0.12 0.16 0.20
s/B
(e)
Fig. 6. Comparison between capacity predicted by proposed formula, calculated static capacity, capacity predicted by Janbu formula and capacity predicted by PCUBC
formula for precast concrete piles: (a) piles in sand of uniform density; (b) end-bearing piles in sand; (c) floating piles in clay; (d) end-bearing piles in clay; and (e) end-
bearing piles crossing clay and resting on sand.
 R. Salgado et al. / Computers and Geotechnics 81 (2017) 307–321 315

static pile base capacity is far from being fully mobilized. However, predicted using the pile driving formulas) for each case study are
the non-linear (hyperbolic) law used in the base resistance model listed in Table 7 (for closed-ended steel pipe piles) and Table 8
[14] of the 1-D pile driving simulations, with asymptote Q bL, which (for precast concrete piles).
is either equal to Q b,ult (in the case of clay bearing layer) or is well
correlated to Q b,10% through Eq. (6), effectively connects the near-
refusal set values to ultimate pile resistance. This is one the great- 5.1. Closed-ended steel pipe piles
est strengths of the proposed formulas. Existing pile driving formu-
las, on the other hand, have an inherent tendency to underpredict 5.1.1. Case 1: Paik et al. [28]
ultimate capacity based in values of set obtained near refusal The test site was located in Lagrange County, Indiana. The soil
because they are based on linear elasticity-perfect plasticity and profile consisted of loose gravelly sand with DR = 30% down to
do not link small sets to the resistance at values of displacement 3 m, followed by dense gravelly sand with DR = 80%. The ground-
associated with ultimate capacity. water table was 3 m below the soil surface. The test pile was an
instrumented closed-ended steel pipe pile, 8.24 m long, with an
outer diameter equal to 356 mm, and wall thickness of 12.7 mm.
5. Case studies It was driven by an ICE-42S single acting diesel hammer down to
6.87 m depth. The ram weight was 18.2 kN and the maximum
We validate the pile driving formulas through six cases studies hammer stroke was 3.12 m. The rated maximum driving energy
involving closed-ended steel pipe piles and five cases studies Eh was 56.8 kN m. According to the driving log, the observed final
involving precast concrete piles. The case studies contain the nec- pile set was 10 mm. The pile capacity at the end of the static load
essary pile driving information and at least one full-scale static test performed on this closed-ended pile 3 days after end of driving
load test. The ultimate loads predicted using the pile driving for- was 1765 kN. The ultimate load measured at a settlement of 10% of
mulas are compared to the ultimate capacities measured in the sta- the pile diameter in the load test was 1499 kN.
tic load tests. The pile base was located in the dense sand layer. The pile driv-
Table 6 lists the hammer, pile, and soil information for each of ing formula for end-bearing piles in sand was considered in this
the selected cases. Summaries of the pertinent quantities (includ- case. The ultimate pile capacity Qb,10% calculated using the pro-
ing the capacities measured in static load tests and the capacities posed pile driving formulas is 1671 kN. The CAPWAP [29]

Table 6
Summary of hammer, pile and soil information used in the pile driving formulas in the case studies.

Case Pile no. Hammer WH (kN) Eh (kN m) eh eeff a

tw (mm) WP (kN) DR (%) su/r0v s (mm)

Paik et al. [28] 1 ICE 42-S 18.2 56.8 0.85 0.38 12.7 8.8 80 N/A 10
Kim et al. [30] 2 ICE 42-S 18.2 56.8 0.85 0.38 12.7 18.2 90 N/A 9
Fellenius et al. [31] 3 APE D36-32 35.3 147.9 0.85 0.38 12.5 53.3 N/A 0.18 12.5
Yen et al. [32] 4 Delmag D62-22 60.9 224.3 0.85 0.38 12.0 59.0 37.5 N/A 20.3
Kulesza and Fellenius [34] 5 Delmag D30-13 29.4 89.7 0.85 0.38 9.5 9.5 85 N/A 18.8
Han et al. [18] 6 APE D30-32 29.4 94.1 0.85 0.38 9.5 12.9 85 N/A 6.4
Ismael [35] 7 Drop hammer 49.8 29.9 0.75 0.55 N/A 19.6 55 N/A 8
8 Drop hammer 49.8 37.4 0.75 0.55 N/A 18.4 100 N/A 2.6
Martin et al. [36] 9 Vulcan 010 44.5 44.1 0.85 0.40 N/A 57.7 45 N/A 22.8
10 Vulcan 010 44.5 44.1 0.85 0.40 N/A 55.8 30 N/A 7.8
11 Raymond 8/0 111.3 361.6 0.85 0.40 N/A 100.9 27 N/A 30.4
Meyerhof and Murdock [37] 12 Drop hammer 14.9 13.4 0.75 0.55 N/A 20.0 N/A 0.16 1.2
Altaee et al. [39] 13 Delmag 12 12.6 46.1 0.85 0.25 N/A 21.0 40 N/A 6.3
Fellenius and Samson [40] 14 Drop hammer 38 22.8 0.75 0.55 N/A 55.1 100 N/A 0.5
a
Assumed values: 0.38 and 0.25 for diesel hammers acting on steel and concrete piles respectively [20]; 0.54 and 0.40 for single acting air/steam hammers on steel and
concrete piles respectively [20]; 0.55 for drop hammers acting on either steel or concrete piles [21–24].

Table 7
Case studies for closed-ended steel pipe piles.

Case Pile-soil profile scenario Pile Static load Proposed CAPWAP Gates Modified Danish Pacific Coast Janbu
assumed no. test formula (kN) formula ENR formula formula formula
(kN) (kN) (kN) (kN) (kN) (kN) (kN)
Paik et al. [28] End bearing pile in sand 1 1499 1671 903 1016 3651 2744 2342 2302
Kim et al. [30] Pile crossing clay resting 2 1345 1280 1486 1050 3277 2289 1609 1840
on sand
Fellenius et al. [31] Floating pile in clay 3 1915 1399 980 1527 5752 2972 1825 2277
Yen et al. [32] Pile crossing clay resting 4 4150 2972 N/A 1576 6595 4093 2908 3301
on sand
Kulesza and Pile crossing clay resting 5 1010 644 N/A 1027 3654 2421 2212 2061
Fellenius [34] on sand
Han et al. [18] Pile in sand of uniform 6 3275 3300 3367 1490 8679 3445 2537 2752
density
316 R. Salgado et al. / Computers and Geotechnics 81 (2017) 307–321

prediction was 903 kN, based on a restrike test performed 11 days ied from 30 kPa at the top of the clay layer to 70 kPa at the bottom
after the end of pile driving [28]. The pile capacity calculated by the of the clay layer.
Gates formula is 1016 kN. Both of these estimates are conservative. The test pile was an instrumented, 45-m-long closed-ended
In contrast, the pile capacities calculated by the Modified ENR, steel pipe pile with an outer diameter equal to 406 mm and wall
Danish, PCUBC, and Janbu formulas are 3651 kN, 2744 kN, thickness equal to 12.5 mm. It was driven by an APE D36-32
2342 kN and 2302 kN, substantially overestimating the pile single-acting diesel hammer to a depth of 45 m. The ram weight
capacity. was 35.3 kN and the maximum hammer stroke was 4.2 m. The
rated maximum driving energy Eh was 147.9 kN m. The observed
5.1.2. Case 2: Kim et al. [30] pile set at EOD was 12.5 mm. Concrete was poured inside the pipe
The test site was located in Jasper County, Indiana. The soil pro- pile six days after EOD and the static load test was performed
file consisted of eleven different soils, mostly alternating between 48 days after EOD when pore pressures were fully dissipated. The
clayey silts and silty clays. The groundwater table was 1 m below test pile failed in plunging at 1915 kN.
the soil surface. Field exploration at the site included soil borings, For this case study, the driving formula for a floating pile in clay
CPT and SPT testing and laboratory testing. The test pile was an is assumed applicable. The vertical effective stress at the pile base
instrumented, 18.5-m-long closed-ended steel pipe pile with an was estimated to be equal about 382 kPa based on the unit weights
outer diameter equal to 356 mm and wall thickness of 12.7 mm. of the soil layers provided in Fellenius et al. [31]. For an su equal to
It was driven by an ICE 42S single-acting diesel hammer to a depth 70 kPa at the base of the pile, the ratio rsu0 is equal to 0.18, and the
v

of 17.4 m, slightly embedded in the 1.4-m-thick, very dense non- estimated pile capacity based on the proposed pile driving formula
plastic silt layer. The silt layer was underlain by a silty clay layer is 1399 kN. The CAPWAP measurement at restrike gave a pile
that extended to a depth of 25 m. The ram weight was 18.2 kN capacity of 980 kN based on restrike performed one day after end
and the maximum hammer stroke was 3.12 m. The rated maxi- of driving. It is likely that setup during the 47 day gap between
mum driving energy Eh was 56.8 kN m. The observed pile set at restrike and static load test may have resulted in the underpredic-
the end of pile driving was 7.8 mm. The static load test was per- tion. The pile capacities calculated from the Gates and PCUBC for-
formed 50 days after EOD. The ultimate load measured from the mulas are 1527 kN and 1825 kN, respectively, and are close to the
load test at a settlement of 10% of the pile diameter was 1345 kN. measured capacity. While, the pile capacities calculated using the
The relative density of the silt layer at the level of the pile base Modified ENR, Danish, and Janbu formulas are 5752 kN, 2972 kN,
was calculated to be 90% based on the CPT log. The driving formula and 2277 kN, respectively, all overestimating the pile capacity by
for piles penetrating through clay and bearing on sand is the most a large margin.
applicable in this case. The ultimate pile capacity Qb,10% based on
the proposed driving formula is equal to 1280 kN, in close agree- 5.1.4. Case 4: Yen et al. [32]
ment with the measured ultimate load. The mobilized pile resis- The test site was located on the southern coast of Taiwan. The
tance from the CAPWAP [29] prediction based on the restrike soil profile at the test site was composed of a 6-m-thick layer of
test 35 days after EOD on an identical neighboring pile was hydraulic sand fill and natural sand underlain by a 13-m-thick clay
1486 kN. Close estimates are also obtained using the Gates and layer and a 21-m-thick sand layer. The groundwater table was
PCUBC formulas of 1050 kN and 1609 kN, respectively. The pile found approximately 1 m below the ground level. The sand layer
capacity calculated using the Modified ENR, Danish, and Janbu for- is interbedded with 1- to 2-m-thick clay sublayers that are located
mulas are 3277 kN, 2289 kN, and 1840 kN, respectively, again at approximate depths of 28 m and 36 m. The test pile was an
severely overestimating the pile capacity. instrumented closed-ended steel pipe pile, 36-m-long, with an
outer diameter of 609 mm, and a wall thickness of 12 mm. The
5.1.3. Case 3: Fellenius et al. [31] embedment depth of the test pile was 34.25 m. Field exploration
The test site was located in Sandpoint in Idaho. The soil profile at the site included soil borings and CPT and SPT testing. The rela-
consisted of a 9-m-thick sandy layer followed by a layer of nor- tive density of the sand layer was estimated from the SPT log using
mally consolidated postglacial alluvial deposit that extended to the Skempton [33] equation:
48 m from the ground surface. The water table was 4 m below vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
DR u N60
the ground level. Field exploration at the site included soil borings, ¼t r0 ð18Þ
CPTu soundings, field vane shear tests, and laboratory tests. The 100% A þ BC v pA
undrained shear strength determined from the field vane test var-

Table 8
Case studies for precast concrete piles: measured and estimated loads in kN.

Case Pile-soil profile scenario Pile Static load Proposed Gates Modified Danish Pacific Coast Janbu
assumed no. test formula formula ENR formula formula formula
Ismael [35] (two End bearing pile in sand 7 1135 1515 741 2103 1579 1331 1344
piles) End bearing pile in sand 8 2391 2678 1098 5475 3006 2193 2576
Martin et al. [36] Pile in sand of uniform 9 2225 3251 667 1067 1158 703 801
(three piles) density
Pile in sand of uniform 10 1958 2633 965 2649 2180 1330 1694
density
Pile in sand of uniform 11 4183 4228 1680 7511 5840 4041 4587
density
Meyerhof and Floating pile in clay 12 956 585 769 1938 1912 1081 1508
Murdock [37]
Altaee et al. [39] End bearing pile in sand 13 1100 1338 1049 2968 2499 1374 1869
Fellenius and Samson Pile crossing clay resting 14 3740 4795 1167 3963 1786 836 1232
[40] on sand
 R. Salgado et al. / Computers and Geotechnics 81 (2017) 307–321 317

where A is between 27 and 46; B is approximately 27, and C is 1 for imately 25 m, with a 1.4–m-thick silty layer located at a depth of
normally consolidated sand. For the SPT N value of 22 at the bearing 9.1 m. The water table was 4.3 m below the ground surface.
layer of the test pile, a relative density of 41% was calculated using The test pile was driven by a single-acting impact hammer (APE
Eq. (18) with estimated vertical effective stress equal to 342.5 kPa. D30-32). The hammer had a weight of 29.4 kN, stroke of 3.2 m and
It should be noted, however, that the N60 term in Eq. (18) represents a maximum rated energy of 94.1 kN m. The pile embedment depth
the corrected SPT N value. This requires knowledge of the SPT ham- was 15.4 m and the total length of the pile was 16 m. The observed
mer type, rod length, sampler type and borehole diameter. These pile set at the end of driving was 6.4 mm. The ultimate pile capac-
details were not provided by Yen et al. [32]. Therefore, the relative ity measured in the static load test, performed 9 days after end of
density of 41% was obtained using the uncorrected SPT N value. driving, was 3275 kN.
The relative density of the sand layer was also determined from The case of a pile in sand of uniform relative density was
CPT data using Salgado and Prezzi [17] equation: assumed appropriate for this case. The relative density was calcu-
  r0  lated at each sand layer using Eq. (19) by using a critical state fric-
qc
ln pA
 0:4947  0:1041/c  0:841 ln p h tion angle of 33° and a horizontal effective stress equal to 0.4 times
DR ¼ r0  A ð19Þ the vertical effective stress. The relative density was estimated to
0:0264  0:0002/c  0:0047 ln pAh
be 85% when averaged across the pile shaft. The estimated capac-
ities by the proposed formula, CAPWAP (restrike 22 days after
The qc value at the bearing layer of the test pile was about 10 MPa,
EOD) and the Danish formula are 3300 kN, 3367 kN and 3445 kN
which yields a relative density of 34% using Eq. (19) with an esti-
respectively, all of which are quite close to the measured pile
mated horizontal effective stress of 154 kPa and critical state fric-
capacity. The Modified ENR formula significantly overestimates
tion angle equal to 30°.
the capacity to be 8679 kN. The capacities predicted by the PCUBC
The test pile was driven with a Delmag D62-22 diesel hammer.
and Janbu formula are 2537 kN and 2752 kN respectively. The
The observed pile set at the end of pile driving was 20.3 mm. The
Gates formula underestimates the pile capacity to be 1490 kN.
load settlement curve obtained from the static load test, performed
32 days after EOD, yielded an ultimate pile capacity of 4150 kN,
which occurred at a settlement equal to about 0.1B.
5.2. Precast concrete piles
The pile capacity predicted using the proposed pile driving for-
mula for a pile crossing clay and resting on sand is 2972 kN using a
5.2.1. Case 7: Ismael [35]
relative density of 37.5% (average of SPT and CPT based estimates)
Ismael [35] reported tests on two 300-mm-square precast con-
for the sand layer at the pile base. The pile capacity predictions
crete piles that were driven at two locations in Salmiya and Shu-
based on the conventional pile driving formulas varied from
waikh in Kuwait. Both sites are located close to the Gulf, at a
1576 kN based on the Gates formula to 6595 kN using the Modified
distance of 25 km apart. At both sites, auger borings, sampling,
ENR formula. The Danish, PCUBC, and Janbu formulas produced
and SPT testing were carried out to characterize the soil. The fol-
estimates of 4093 kN, 2908 kN, and 3301 kN, respectively. The
lowing paragraphs present the soil profile, test pile details, and
Gates formula significantly underestimates the pile capacity, while
predictions for each site.
the Janbu formula yields a good estimate.
The soil profile in Salmiya consisted of an 8-m-thick fine-to-
medium silty sand followed with a grey dense sand layer up to
5.1.5. Case 5: Kulesza and Fellenius [34] 16 m. The water table was 2 m below the ground surface. The test
The test site was located near Briech in Morroco. It lies 3.6 km pile was driven by a 49.8 kN drop hammer to an embedment depth
east of the coastline in a river flood plain at an elevation of 1.2– of 9.25 m. The drop height was 0.6 m throughout the test. The
2.2 m above sea level. At the test pile location, the soil profile con- observed pile set at the end of pile driving was 8 mm. The SPT N
sisted of a 4-m-thick fill layer underlain by a 5-m-thick NC clay value at the pile base was reported as 30, which produces an esti-
layer and an 8-m-thick dense sand layer. The groundwater table mated relative density of 55% using Eq. (18). The pile capacity mea-
was approximately 2.7 m below the soil surface. The SPT N value sured in the static load test was 1135 kN. The estimated pile
for the sand layer at the base of the test pile was 42, which yielded capacity using the proposed pile driving formula for an end-
a relative density of 85% estimated using Eq. (18). bearing pile in sand is 1515 kN. The pile capacity calculated by
The test pile was a closed-ended steel pipe pile, driven to a the Gates formula is 741 kN. The pile capacities calculated using
depth of 10.5 m, with an outer diameter of 406 mm, and a wall the Danish, PCUBC, and Janbu formulas are 1579 kN, 1331 kN,
thickness was 9.5 mm. It was driven with a Delmag D30-13 and 1344 kN, respectively. The pile capacity calculated using the
single-acting diesel hammer. The observed pile set at the end of Modified ENR is 2103 kN, which significantly overestimates the
pile driving was 18.8 mm. The pile reached plunging failure during pile capacity measured in the static load test.
static load testing at a pile head movement of 16 mm (less than The soil profile in Shuwaikh consisted of a 5-m-thick fine-to-
0.1B) and an applied load of 1088 kN. coarse sand fill followed with a fine to medium silty sand layer
The predicted pile capacity with the proposed pile driving for- up to 10.5 m. The water table was 1.5 m below the ground surface.
mula for a pile crossing clay and resting on sand is 644 kN using The test pile was driven with the same 49.8 kN drop hammer to an
a relative density of 85% for the sand layer at the pile base. The pile embedment depth of 8.7 m. The hammer drop height was 0.75 m
capacity calculated by the Gates formula is 1027 kN. The pile throughout pile driving. The observed pile set at the end of the pile
capacities calculated using the Modified ENR, Danish, PCUBC, and driving was 2.6 mm. The SPT N value at the pile base was reported
Janbu formulas are 3654 kN, 2421 kN, 2212 kN, and 2061 kN which as 100, corresponding to an estimated relative density of 100%
overestimate significantly the pile capacity at EOD determined according to Eq. (18). The pile capacity measured in the pile load
from the static load test. test was 2391 kN. The estimated pile capacity by the proposed pile
driving formula for the scenario of an end-bearing pile in sand was
5.1.6. Case 6: Han et al. [18] 2678 kN. The pile capacity calculated by the Gates formula is
The test was performed at a bridge construction site located in 1098 kN, which significantly underestimates the pile capacity
Marshall County, Indiana, USA. Two SPTs and two CPTs were per- measured in the static load test. The pile capacities calculated
formed close to the site location. The soil profile predominantly using the Danish, PCUBC, and Janbu formulas are 3006 kN,
consisted of dense sand intermixed with silt to a depth of approx- 2193 kN, and 2576 kN, respectively. The pile capacities calculated
318 R. Salgado et al. / Computers and Geotechnics 81 (2017) 307–321

using the Modified ENR formula is 5475 kN, severely overestimat- The test pile was driven with a drop hammer weighing 14.9 kN.
ing the pile capacity. The drop height of the hammer was 0.90 m throughout pile driv-
ing. The observed pile set at the end of pile driving was 1.2 mm.
The static load test was performed one month after EOD and the
5.2.2. Case 8: Martin et al. [36]
measured pile capacity was 956 kN. Although the proposed pile
The test site was located in the Tidewater region of Virginia.
driving formula for floating piles in clay was derived based on sim-
This paper reported static load tests on three precast concrete piles
ulations that assume the clay is normally consolidated throughout,
with different cross sections and length. The three test piles were
and not overconsolidated as in the case of this London clay, we pre-
driven through a relatively weak silty, clayey sand layer to rest
sent this case to examine the performance of the proposed formu-
on a loose-to-medium dense sand layer. The ground water table
las beyond its applicability limits. The pile capacity predicted by
is approximately 2.5 m below the ground surface.
the pile driving formula for floating piles in clay is 585 kN. The
The first and second test piles were 360-mm-square, 21-m-long
Gates formula predicts a close capacity in this case of 769 kN.
precast concrete pile driven by a Vulcan 010 single-acting, air-
The PCUBC and Janbu formulas predict capacities of 1081 kN and
driven hammer. The third pile was a 460-mm-square, 26-m-long
1508 kN, respectively. The largest pile capacities of 1938 kN and
precast concrete pile driven by a Raymond 8/0 single acting, air-
1912 kN are calculated by Modified ENR and Danish formulas,
driven hammer.
The embedment depth of the first pile was 18.9 m. The observed respectively.
pile set at the end of pile driving was 22.8 mm. The SPT N value at
the pile base was reported as 15, which gave an estimated relative 5.2.4. Case 10: Altaee et al. [39]
density of 45% using Eq. (18). The pile capacity measured in the The test site was located at the Baghdad University Complex,
pile load test was 2225 kN. The prediction using the proposed pile Iraq, near the bank of the River Tigris. The soil profile consisted
driving formula for piles in a sand of uniform density was 3251 kN. of two main soil layers: an upper 3-m-thick clayey silty sand
The Gates formula, Modified ENR, Danish, PCUBC, and Janbu for- layer and a lower, thick uniform sand layer with some silt.
mulas predicted pile capacities of 667 kN, 1067 kN, 1158 kN, The groundwater table was approximately 6.4 m below the
703 kN, and 801 kN which all underestimate the measured capac- ground surface. The instrumented 285-mm-square precast con-
ity of the static load test. crete test pile was driven to a depth of 11 m into a sand deposit
The embedded length of the second pile was 18.3 m. The using the Delmag D12 diesel hammer. The SPT N value at the
observed pile set at the end of pile driving was 7.8 mm. The SPT pile base was about 15, which yields a relative density of 44%
N value at the pile base was reported as 8, which leads to an esti- using Eq. (18). The qc from the cone penetration test at the pile
mated relative density of 30% using Eq. (18). The pile capacity mea- base was about 6.5 MPa, yielding a relative density of 36% using
sured in the pile load test was 1958 kN. The prediction by the pile Eq. (19).
driving formula for piles in uniform sand layer was 2633 kN. The The observed pile set at the end of pile driving was 6.25 mm.
Janbu and Danish formulas predict the closest capacities of The pile capacity measured from the static load test was
1694 kN and 2180 kN, respectively. The Gates and PCUBC formulas 1100 kN. The pile capacity predicted using the pile driving for-
underestimated the pile capacity, giving values of 965 kN and mula proposed for end-bearing piles in sand is 1338 kN for
1330 kN, respectively. The Modified ENR overestimated the capac- the average of the SPT and CPT-based estimates of relative den-
ity with an estimated resistance of 2649 kN. sity (equal to 40%) for the sand layer at the pile base. The pile
The embedment depth of the third pile was 20.4 m. The capacities estimated using the Gates and PCUBC formulas are
observed pile set at the end of the pile driving was 30.4 mm. The 1049 kN and 1374 kN. The pile capacities calculated using the
SPT N value at the pile base was 7, which corresponds to an esti- Modified ENR, Danish, and Janbu formulas are 2968 kN,
mated relative density of 27% using Eq. (18). The pile capacity mea- 2499 kN, and 1869 kN, respectively, largely overestimating the
sured in the pile load test was 4183 kN. The prediction using the pile capacity.
pile driving formula for piles in uniform sand was 4228 kN. The
pile capacity calculated by the Gates formula is 1680 kN signifi-
5.2.5. Case 11: Fellenius and Samson [40]
cantly underestimating the pile resistance. The PCUBC and Janbu
The test site was located on the Sidbec property at Contre-
formulas predict capacities of 4041 kN and 4587 kN, respectively,
coeur, Quebec. The test pile was a 300-mm Herkules H800 con-
which are close to the measured resistance. The Modified ENR
crete pile driven to a depth of 26 m. The soil profile consisted
and Danish formulas overpredict the measured pile capacity with
of an 18-m-thick, sensitive marine clay layer followed by a very
estimates of 7511 kN and 5840 kN, respectively.
dense silty sand layer down to a depth of 32 m. The groundwater
table is 1 m below the soil surface. The SPT N value of this dense
5.2.3. Case 9: Meyerhof and Murdock [37] silty sand layer at the pile base was 158, which suggests an
Meyerhof and Murdock [37] reported testing done on piles at a extremely dense sand with perhaps some gravel content; the rel-
site located in Barnet, near London. The test pile was a 305-mm- ative density using Eq. (18) would be higher than 100%; a value
square, 9.14-m-long, precast concrete pile. The embedded length of 100% was assumed.
of the test pile was 8.53 m. The soil profile consisted of a 1-m- The test pile was driven by a drop hammer weighing 38 kN. At
thick soft Brown London Clay layer followed by a stiff to hard the end of pile driving, the drop height of the hammer and the
Brown London Clay layer throughout. The plastic limit and liquid observed set of the test pile were reported as 0.6 m and 0.5 mm
limit for the clay layer were about 27% and 80%, respectively, lead- (50 blows per inch), respectively. The static load test was per-
ing to a plasticity index of 53. The clay layers were reported as formed 43 days after EOD and the measured pile capacity was
highly overconsolidated, but the OCR values were not provided. 3740 kN. The pile capacity estimated using the pile driving formula
No groundwater was found during the field investigation. The tri- proposed for piles crossing a clay layer and resting on a sand layer
axial compression shear strength for the fully disturbed clay sam- is 4795 kN. The pile capacities calculated using the Gates, Danish,
ple was reported as 24 kPa at the location of the test pile base. The PCUBC, and Janbu formulas are 1167 kN, 1786 kN, 836 kN, and
unit weight of the Brown London Clay was assumed to be 18.8 kN/ 1232 kN, respectively, significantly underestimating measured pile
m3 according to Skempton [38]. The estimated rsu0 of the clay layer capacity. The pile capacities calculated using the Modified ENR for-
v

at the pile base was 0.16. mula is 3963 kN.

 R. Salgado et al. / Computers and Geotechnics 81 (2017) 307–321 319

Table 9
rQ values for predicted pile capacities with different pile driving formulas.

Case Pile no. Proposed formula CAPWAP Gates formula Modified ENR Danish formula PCUBC formula Janbu formula
Paik et al. [28] 1 1.11 0.60 0.68 2.44 1.83 1.56 1.54
Kim et al. [30] 2 0.95 1.10 0.78 2.44 1.70 1.20 1.37
Fellenius et al. [31] 3 0.73 0.51 0.80 3.00 1.55 0.95 1.19
Yen et al. [32] 4 0.72 0.38 1.59 0.99 0.70 0.80
Kulesza and Fellenius [34] 5 0.64 1.02 3.62 2.40 2.19 2.04
Han et al. [18] 6 1.01 1.03 0.46 2.65 1.05 0.77 0.84
Ismael [35] (two piles) 7 1.33 0.65 1.85 1.39 1.17 1.18
8 1.12 0.46 2.29 1.26 0.92 1.08
Martin et al. [36] (three piles) 9 1.46 0.30 0.48 0.52 0.32 0.36
10 1.34 0.50 1.41 1.14 0.69 0.89
11 1.01 0.40 1.80 1.40 0.97 1.10
Meyerhof and Murdock [37] 12 0.61 0.80 2.03 2.00 1.13 1.58
Altaee et al. [39] 13 1.22 0.95 2.70 2.27 1.25 1.70
Fellenius and Samson [40] 14 1.28 0.31 1.06 0.48 0.22 0.33

2
Average r Q

0
Proposed CAPWAP Gates Danish Janbu

Closed-ended steel pipe piles Precast concrete piles Combined results

(a)
200

160
MAPE %

120

80

40

0
Proposed CAPWAP Gates Danish Janbu

Closed-ended steel pipe piles Precast concrete piles Combined results

(b)
Fig. 7. Bias and variability – as expressed by (a) average rQ and (b) MAPE – for proposed formula, CAPWAP prediction, Gates formula, Modified ENR, Danish formula, PCUBC
formula, and Janbu formula.
320 R. Salgado et al. / Computers and Geotechnics 81 (2017) 307–321

5.3. Summary of the pile driving formula performance las have the capability to reliably predict pile capacity in a variety
of conditions, so long as the design soil profile can be adequately
To assess the performance of the proposed and existing formu- fitted into one of the five general soil profile configurations for
las in predicting the capacity data from the previously presented which the formulas were derived.
case histories, we consider two performance measures. The first
one is the average rQ ratio of predicted to observed pile capacity Acknowledgements
(rQ = Qult,predicted/Qult,observed), which gives an indication of predic-
tion bias. The relative performance of the proposed and existing The research presented in this paper was funded by the Indiana
pile driving formulas is shown in Table 9. We see that the proposed Department of Transportation and the Federal Highway Adminis-
formulas perform very well. The second measure is the Mean Abso- tration – United States through the Joint Transportation Research
lute Percentage Error (MAPE): Program.

N Q ðiÞ ðiÞ
1X ult;predicted  Q ult;observ ed 1 X ðiÞ
N
MAPE ¼ ¼ jr  1j ð20Þ References
N i¼1 Q
ðiÞ N i¼1 Q
ult;observ ed
[1] Paikowsky SG, Caniff M, Robertson S, Budge AS. Load and resistance factor
where N is the total number of case history piles. The MAPE helps design (LRFD) pile driving project: phase II study. St. Paul (MN): Minnesota
Department of Transportation Research Services; 2014.
quantify the dispersion of the predictions. Two pile driving formulas
[2] Hannigan PJ, Goble GG, Likins GE, Rausche F. Design and construction of driven
may both exhibit minimal bias (average rQ close to 1.0), but one pile foundations – volume II. Federal Highway Administration, Pub. No. FHWA
may have much larger MAPE than the other, thus yielding predic- NHI-05-043, Washington (DC); 2006.
tions with greater uncertainty. [3] Bowles EJ. Foundation analysis and design. 5th ed. New York: McGraw-Hill
Inc.; 1996.
The ratio rQ of the proposed formulas has an average value [4] Gates M. Empirical formula for predicting pile bearing capacity. Civ Eng
equal to 1.04 for all case histories examined (i.e., the formulas on 1957;27:65–6.
average overestimate the actual pile capacity by 4%). In addition, [5] ENR. Michigan pile test program test results are released. Eng News Rec 1965;
May 20:26–8, 33–4.
the overall MAPE for the proposed formulas is equal to 23%. Tradi- [6] Chellis RD. Pile foundations. 2nd ed. New York: McGraw-Hill; 1961.
tional formulas produce estimates with greater dispersion; some [7] Olson RE, Flaate KS. Pile-driving formulas for friction piles in sand. Soil Mech
formulas are also strongly biased. The Gates formula is overly con- Found Div 1967;93:279–96.
[8] Salgado R. The engineering of foundations. 1st ed. New York: McGraw-Hill;
servative (Fig. 7a), with average rQ = 0.61 (almost 40% underesti- 2008.
mation of pile capacity) and a MAPE of 39% (Fig. 7b). On the [9] McVay M, Birgisson B, Zhang L, Perez A, Putcha S. Load and resistance factor
other hand, the modified ENR formula is overly unconservative, design (LRFD) for driven piles using dynamic methods—a Florida perspective.
Geotech Test J 2000;23:55–66. http://dx.doi.org/10.1520/GTJ11123J.
with average rQ equal to 2.1 and MAPE of 117%. The Danish formula
[10] Likins GE, Fellenius BH, Holtz RD. Pile driving formulas—past and present. Full-
is also unconservative, exhibiting an average rQ = 1.42 and MAPE of scale test. Found. Des. 2012:737–53.
57%. The Janbu formula has an average rQ = 1.14 and MAPE of 40%. [11] Paikowsky SG, Chernauskas LR. Energy approach for capacity evaluation of
Among the existing formulas considered, the best performance is driven piles. In: Barends FBJ, editor. Fourth international conference on applied
stress. Theory to piles. Rotterdam (The Netherlands): Balkema; 1992. p.
observed for the PCUBC formula, which has an average rQ = 1.0, less 595–601.
than the proposed formulas, but has a substantially higher disper- [12] Paikowsky SG, Regan JE, McDonnell JJ. A simplified field method for capacity
sion and thus, uncertainty, with a MAPE of 36%. This also becomes evaluation of driven piles, Report FHWA-RD-94-042; 1994.
[13] Reddy SC, Stuedlein AW. Accuracy and reliability-based region-specific
evident by comparing the minimum and maximum rQ values of the recalibration of dynamic pile formulas. Georisk Assess Manag Risk Eng Syst
PCUBC formula and the proposed formulas. The minimum and Geohazards 2013;7:163–83. http://dx.doi.org/10.1080/
maximum rQ values for PCUBC are 0.22 and 2.19, respectively. In 17499518.2013.779833.
[14] Salgado R, Loukidis D, Abou-Jaoude G, Zhang Y. The role of soil stiffness non-
contrast, the rQ range for the proposed formulas is much narrower linearity in 1D pile driving simulations. Géotechnique 2015;65:169–87. http://
(minimum rQ = 0.61 and maximum rQ = 1.46). dx.doi.org/10.1680/geot.13.P.124.
[15] Salgado R, Woo SI, Kim D. Development of load and resistance factor design for
ultimate and serviceability limit states of transportation structure foundations
6. Summary and conclusions Joint Transportation Research Program. West Lafayette (IN): Indiana
Department of Transportation and Purdue University; 2011. http://dx.doi.
org/10.5703/1288284314618.
Pile quality assurance and control rely on pile driving formulas,
[16] Terzaghi K. Discussions on the progress report of the committee on the bearing
dynamic load tests during or after driving, and static load tests. value of pile foundations. Proc Am Soc Civ Eng 1942;68:311–23.
Each tool has a different application domain. Pile driving formulas, [17] Salgado R, Prezzi M. Computation of cavity expansion pressure and
so long as they are reliable, can be useful in routine projects, par- penetration resistance in sands. Int J Geomech 2007;7:251–65. 10.1061/
(ASCE)1532-3641(2007) 7:4(251).
ticularly low- to medium-budget projects. This paper uses predic- [18] Han F, Prezzi M, Salgado R, Zaheer M. Axial resistance of closed-ended steel
tions of 1D pile driving simulations with improved soil reaction pipe piles driven in multilayered soil. J Geotech Geoenviron Eng 2016. http://
models to develop reliable pile driving formulas. There are two dx.doi.org/10.1061/(ASCE)GT.1943-5606.0001589 [in press].
[19] Basu P, Prezzi M, Salgado R, Chakraborty T. Shaft resistance and setup factors
general forms for the pile driving formulas: one applies to piles for piles jacked in clay. J Geotech Geoenviron Eng 2014;140. 10.1061/(ASCE)
with their base resting in sand (piles in a sand of uniform density, GT.1943-5606.0001018. 04013026-1-004013026-16.
end-bearing piles in sand, piles crossing normally consolidated [20] Rausche F. Pile driving equipment: capabilities and properties. In: Keynote
lectures and proceedings of the 6th international conference on applied stress,
clay with their base penetrating a sand layer) and another for piles Theory to piles. Rotterdam.
with their base resting in clay (either floating piles or end-bearing [21] Lam J. Termination criteria for high-capacity jacked and driven steel H-piles in
piles with their bases in a stiff clay layer). Each of the two general Hong Kong. University of Hong Kong; 2007.
[22] Lim PC, Broms BB. Influence of pile driving hammer performance on driving
pile driving formulas include parameters that are a function of the criteria. Geotech Eng 1990;21:63–9.
type of pile (closed-ended steel pipe piles or precast concrete [23] Mostafa YE. Onshore and offshore pile installation in dense soils. J Am Sci
piles). 2011;7:549–63. http://dx.doi.org/10.7537/marsjas070711.80.
[24] Allen T. Development of the WSDOT pile-driving formula and its calibration
Eleven well-documented case histories that include load test
for load and resistance factor design (LRFD) Report WA-RD 610.1. Washington
and soil profile information for five closed-ended steel pile piles State Department of Transportation; 2005.
and eight precast concrete piles, were used to validate the pro- [25] Holeyman AE. Dynamic non-linear skin friction of piles. In: Proceedings of the
posed pile driving formulas. Comparison of pile driving formula international symposium on penetrability driveability piles. p. 173–6.
[26] Randolph MF, Simons HA. An improved soil model for one-dimensional pile
predictions with results from static pile load tests and traditional driving analysis. In: Proceedings of the 3rd international conference on
pile driving formulas shows that the proposed pile driving formu- numerical methods in offshore piling. p. 3–17.
 R. Salgado et al. / Computers and Geotechnics 81 (2017) 307–321 321

[27] Lee SL, Chow YK, Karunaratne GP, Wong KY. Rational wave equation model for editors. Full-scale testing in foundation design. Oakland (CA): American
pile-driving analysis. J Geotech Eng 1988;114:306–25. http://dx.doi.org/ Society of Civil Engineers; 2012. p. 452–70. http://dx.doi.org/10.1061/
10.1061/(ASCE)0733-9410(1988) 114:3(306). 9780784412084.0033.
[28] Paik K, Salgado R, Lee J, Kim B. Behavior of open- and closed-ended piles driven [35] Ismael NF. Analysis of load tests on piles driven through calcareous desert
into sands. J Geotech Geoenviron Eng 2003;129:296–306. http://dx.doi.org/ sands. J Geotech Geoenviron Eng 1999;125:905–8. http://dx.doi.org/10.1061/
10.1061/(ASCE)1090-0241(2003) 129:4(296). (ASCE)1090-0241(1999) 125:10(905).
[29] GRL and Associates I. GRL software: case pile wave analysis program [36] Martin RE, Seli JJ, Powell GW, Bertoulin M. Concrete pile design in tidewater
(CAPWAP); 1997. Virginia. J Geotech Eng 1987;113:568–85. http://dx.doi.org/10.1061/(ASCE)
[30] Kim D, Bica AVD, Salgado R, Prezzi M, Lee W. Load testing of a closed-ended 0733-9410(1987) 113:6(568).
pipe pile driven in multilayered soil. J Geotech Geoenviron Eng [37] Meyerhof GG, Murdock LJ. An investigation of the bearing capacity of some
2009;135:463–73. bored and driven piles in London clay. Géotechnique 1953;3:267–82. http://
[31] Fellenius BH, Harris DE, Anderson DG. Static loading test on a 45 m long pipe dx.doi.org/10.1680/geot.1953.3.7.267.
pile in Sandpoint, Idaho. Can Geotech J 2004;41:613–28. http://dx.doi.org/ [38] Skempton AW. Slope stability of cuttings in Brown London Clay. Sel Pap Soil
10.1139/t04-012. Mech 1984:241–50. http://dx.doi.org/10.1680/sposm.02050.002 [Thomas
[32] Yen T-L, Lin H, Chin C-T, Wang RF. Interpretation of instrumented driven steel Telford Publishing].
pipe piles. In: Foundation engineering: current principles and practices. ASCE; [39] Altaee A, Fellenius BH, Evgin E. Axial load transfer for piles in sand. I. Tests on
1989. p. 25–9. an instrumented precast pile. Can Geotech J 1992;29:11–20. http://dx.doi.org/
[33] Skempton AW. Standard penetration test procedures and the effects in sands 10.1139/t92-002.
of overburden pressure, relative density, particle size, ageing and [40] Fellenius BH, Samson L. Testing of drivability of concrete piles and disturbance
overconsolidation. Géotechnique 1986;36:425–47. http://dx.doi.org/ to sensitive clay. Can Geotech J 1976;13:139–60. http://dx.doi.org/10.1139/
10.1680/geot.1986.36.3.425. t76-015.
[34] Kulesza RL, Fellenius BH. Design and testing of piles on a telecommunications
project in Morocco.. In: Hussein MH, Massarsch KR, Likins GE, Holtz RD,