Innovative Geotechnologies for Energy Transition | The Society for Underwater Technology

doi:10.3723/GPCS7406

Evaluation of pile driveability predictions in sand

N.B. Yenigul, Y. Yan, L.C.H. Braakenburg & V.M. Thumann
Seaway7, Zoetermeer, The Netherlands

ABSTRACT: Pile driveability is a critical component of pile design process to ensure that the selected impact
hammer has sufficient energy to drive piles to final depth, without refusal, and acceptable fatigue damage during
driving. Pile driving records often show considerable scatter because of variations in soil conditions and
behaviour, pile dimensions, and set up during interruptions in driving. However, representative pile drivability
prediction remains a challenge. Over the past decade, Seaway7 have installed numerous foundations for
offshore wind turbine generators. Recent Seaway7 experience in North Sea revealed that pile driveability
predictions still have room for improvement installation feasibility. This paper considers pile driving records
from different offshore campaigns in North Sea to cover a range of subsea applications including varying pile
properties, impact hammers, geographical locations, and sandy soil conditions. Driving records have been back-
analysed with as-installed blow count and hammer energies to assess the offset of existing in-house prediction
model.

1 Introduction Bryn et al (2018), Cardoso et al ( 2018), Maynard

et al. (2019), Perikleous et al (2020) are some exam-
Pile installation is a high-risk activity in offshore con- ples of various recent research that focus on drivea-
struction projects. Pile driveability predictions and bility predictions and improvement.
their accuracy are critical to assess installation feasi- This paper presents the back-analysed pile driving
bility by selecting an appropriate hammer to success- records from several offshore campaigns in the North
fully advance the piled foundation to its required Sea that cover a wide range of varying pile properties,
depth. Furthermore, inefficient pile driving can cause impact hammers, geographical locations, and sandy
project delays and/or material damage to pile as well soil conditions.
as driving system. Particularly during offshore foun- The main purpose is to assess the offset of the ex-
dation installations such delays and damages can have isting prediction model, and to identify whether there
significant financial costs and associated risks due to is ground for correction factors to be applied with re-
the extreme environmental working conditions. spect to soil or pile/hammer configuration. Moreover,
Performing back analysis regularly is of high im- this paper aims to serve as contribution of Seaway7’s
portance to ensure that the theoretical and empirical vast experience and observations to the understanding
methods used for predictions are appropriate and al- of pile driveability in scarcity of public information
low realistic predictions. regarding measured pile driving data and observa-
Over the past decades, Seaway7 have installed nu- tions for large diameter piles with relatively shallow
merous foundations including those for offshore wind embedment depth.
turbine generators (WTGs). Pile driving records often
show considerable scatter because of variations in soil
conditions and behaviour, pile dimensions, and set up 2 Pile driveability assessment
during interruptions in driving. Recent Seaway7 ex-
perience in North Sea revealed that performing repre- The dynamic soil resistance to driving, Rdyn can be de-
sentative pile driveability predictions remains a chal- scribed as the product of a static soil resistance to
lenge and that there is still room for improvement to driving, SRD and a damping constant as in the
enhance installation feasibility.

758
 Session 8 - Installation of Piled Foundations

following formula (Alm et.al, 1989, Pile Dynmics 𝑓𝑓𝑓𝑓𝑠𝑠𝑠𝑠,𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑠𝑠𝑠𝑠 = 0.2 × 𝑓𝑓𝑓𝑓𝑠𝑠𝑠𝑠,𝑖𝑖𝑖𝑖 (4)
Inc, 2010).
where:
𝑅𝑅𝑅𝑅𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 = 𝑆𝑆𝑆𝑆𝑅𝑅𝑅𝑅𝑆𝑆𝑆𝑆(1 + 𝐽𝐽𝐽𝐽 × 𝑣𝑣𝑣𝑣 𝑑𝑑𝑑𝑑 ) (1) d – Depth to actual layer (m)
p – Pile penetration (m)
where,
σ′vo – Effective overburden pressure (kN/m2)
𝑣𝑣𝑣𝑣 – Pile segment velocity (m/s)
δf – Soil-pile interface friction angle at failure (°)
𝐽𝐽𝐽𝐽 – Damping coefficient (s/m)
k – Shape factor of degradation (-)
n – Damping exponent
SRD is mobilized as a function of the pile displace- The lateral stress coefficient, K is directly linked to
ment and is one of the most important variables in a the cone resistance using the following formula:
driveability analysis. SRD is generally evaluated on ′ )𝑞𝑞𝑞𝑞 (𝜎𝜎𝜎𝜎 ′ ⁄ )0.2
basis of pile bearing capacity principles, where the ca- 𝐾𝐾𝐾𝐾 = (0.132⁄𝜎𝜎𝜎𝜎𝑣𝑣𝑣𝑣0 𝑡𝑡𝑡𝑡 𝑣𝑣𝑣𝑣0 𝑝𝑝𝑝𝑝𝑎𝑎𝑎𝑎 (5)
pacity is contributed by pile tip resistance and side where:
friction along the pile surface. σ′vo – Effective overburden pressure (kN/m2)
In this study, pile driving is modelled in two separate pa – Reference pressure =100 kN/m2
steps: qt – Cone tip resictance (kN/m2)
1. Derivation of the Soil Resistance to Driving The shape factor for the “rate” of degradation, 𝑘𝑘𝑘𝑘
(SRD) of the encountered ground conditions was also related to the cone tip resistance:
for a particular pile geometry, using best esti-
mate and high estimate soil profiles for pile 𝑘𝑘𝑘𝑘 = (𝑞𝑞𝑞𝑞𝑡𝑡𝑡𝑡 ⁄𝑝𝑝𝑝𝑝0 ′)0.5 /80 (6)
driving prediction analysis. In sands unit base resistance offered by the pile tip
2. Modelling of the dynamic behaviour of the during driving was formulated as follows:
hammer-pile-soil system using a wave equa- ′ )0.2
tion analysis to estimate the blow count vari- 𝑞𝑞𝑞𝑞𝑡𝑡𝑡𝑡𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = 0.15 × 𝑞𝑞𝑞𝑞𝑡𝑡𝑡𝑡 (𝑞𝑞𝑞𝑞𝑡𝑡𝑡𝑡 ⁄𝜎𝜎𝜎𝜎𝑣𝑣𝑣𝑣0 (7)
ation with penetration depth. The driveability A database of 178 piles from 18 different jacket
analysis calculations have been performed in- structures at the North Sea were used to develop the
house by Seaway7 using commercially avail- model. The database pile diameters ranged from 72 to
able software GRLWEAP. 108 inches and pile penetrations extended up to 70m.
Various models such as Stevens (1982), Toolan & The back-analysis was calibrated with a single set of
Fox (1977), Jardine et al. (2005), Lehane (2005), Alm dynamic soil parameters. Quake values were taken as
Hamre (2001) exist to assess SRD. Seaway7’s drive- 2.5mm for both shaft and tip. The tip damping was
ability calculations are performed based on independ- taken as 0.5s/m and the shaft damping was 0.25s/m
ent interpretation of soil information from installation for all case studies.
contractor’s perspective, using the GRLWEAP pro- The model was shown to predict resistances lying
gram and theoretical models for soil driving re- close to or slightly above the average of the back-cal-
sistance by Alm & Hamre (2001). culated values and hence provided a reasonable basis
The pile driving data from pile driving records are for prediction of best estimate resistance profiles.
back analysed with as-installed blow count and ham- Alm and Hamre (1998) recommend considering the
mer energies to assess the accuracy of the prediction effect of soil variability in an upper bound resistance
and the parameters used within these assessments. profile, using a factor of 1.25.
2.1 Soil Resistance to Driving (SRD)
2.2 Dynamic Behaviour of The Hammer- Pile- Soil
Alm and Hamre (2001) have presented SRD predic- System
tion model using Cone Penetration Test (CPT) data
directly and including a friction degradation concept GRLWEAP has been used to model the dynamic be-
for the calculation of the pile shaft resistance. haviour of the hammer-pile-soil system and perform
The model defines a start and a residual static fric- driveability analyses.
tion, along with a shape function describing the rela- GRLWEAP is a one-dimensional Wave Equation
tive fatigue friction reduction. The shaft friction is es- (Smith, 1960) program that simulates the pile re-
timated to exponentially decay from a maximum to a sponse to pile driving equipment. The program re-
residual value. Further, the model for maximum skin quires the input from soil resistance during driving,
friction is based on the Mohr Coulomb failure criteria. damping and quake parameters, hammer and cushion
The formulation that this method uses for sand is de- properties and pile properties.
scribed below: The hammer and pile are modelled as a series of
discrete masses interconnected by springs and damp-
𝑓𝑓𝑓𝑓𝑠𝑠𝑠𝑠 = 𝑓𝑓𝑓𝑓𝑠𝑠𝑠𝑠,𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑠𝑠𝑠𝑠 + �𝑓𝑓𝑓𝑓𝑠𝑠𝑠𝑠,𝑖𝑖𝑖𝑖− 𝑓𝑓𝑓𝑓𝑠𝑠𝑠𝑠,𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑠𝑠𝑠𝑠 �𝑒𝑒𝑒𝑒 −𝑘𝑘𝑘𝑘(𝑑𝑑𝑑𝑑−𝑝𝑝𝑝𝑝) (2) ers. The resistance of the soil at the pile toe is mod-
′ elled by a point force on the last pile node and the
𝑓𝑓𝑓𝑓𝑠𝑠𝑠𝑠,𝑖𝑖𝑖𝑖 = 𝐾𝐾𝐾𝐾 × 𝜎𝜎𝜎𝜎𝑣𝑣𝑣𝑣0 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡�𝛿𝛿𝛿𝛿𝑓𝑓𝑓𝑓 � (3)
resistance of the soil along the pile shaft is

759
 Innovative Geotechnologies for Energy Transition | The Society for Underwater Technology

represented by a set of point resistances at each pile hammers, geographical locations, and sandy soil con-
node. The magnitudes of these resistances are mod- ditions.
elled as a bi-linear function of force versus displace- Site 1 and Site 2 include monopiles with outer di-
ment and of the damping constant, J. ameter of 8.0m and 6.5m, respectively. Site 3 and 4
The velocity of the hammer ram and hence impact includes jacket piles with a diameter of 2.1m. Table 1
of this mass produces a displacement in the discrete presents the pile and hammer data for four sites.
masses. The displacement of the subsequent sections
leads to a compression or to an extension in the Table 1. Pile and hammer data for four projects in North Sea.
springs and dampers between them, and in the springs Site 1 Site 2 Site 3-4
and dampers representing the pile-soil interaction. Pile outer diameter [m] 8.0 6.50 2.134
The force generated by the springs and dampers pro- Penetration depth [m] 24.5-30 27-29 50-52
duces a resultant force on the discrete mass, which in Wall thickness [mm] 58-80 54-75 50-80
turn results in a change of the element’s velocity. The Pile length [m] 64.6-72 59-60 71-73.4
transferred force and velocity then become an input Pile weight [kg] 760-860 613-647 229-237
for the next step. This procedure is ongoing until the Hammer type MHU S3000 S1800
3500S
velocity reduces to zero. Rated Energy [kJ] 3500 3000 1440
The selection of the damping factors, J for toe and Ram weight [ton] 175 150 75
shaft resistance are soil type dependent. In the analy- Hammer weight [ton] 312 290 200
sis presented in this study, for shaft damping and toe Anvil weight [ton] 228 120 16.6
damping values of 0.16 s/m and 0.50 s/m have been Hammer efficiency [%] 3-64 16-68 13-86
chosen respectively. Quake (shaft and toe) values in
the force-displacement relationship were assumed to The soil types encountered at all four sites primar-
be 2.54 mm. ily comprise quaternary marine sands. Relative den-
sity, RD ranges from between from 40-90% for Site 1,
between 60-90% for Site 2, between 40-100% for
2.3 Back Analysis of Pile Driving Records Sites 3 and 4.
At all four project sites, cone penetration tests
Seaway7 perform regularly back analysis as to evalu- (CPT) were carried out at most of the pile locations.
ate whether the pile driveability predictions are ap- Interpreted CPT profiles for some selected example
propriate and how to incorporate the field experience pile locations are illustrated in Figures 1, 2 & 3.
for improved/realistic predictions in remainder of on-
going projects or for future applications.
Due to the non-linear relationship between Rdyn
and blowcount, a back-analysis based on blowcounts
only is insufficient.
The back-calculations presented in this paper,
therefore include calculating of the actual encoun-
tered Rdyn using the recorded blowcount and energy,
and the specific Rdyn versus blowcount graphs for dif-
ferent energy levels.
To account local variation of soil conditions High
Estimate is considered as factor of 133% and Low Es-
timate is considered as factor 75% of BE soil condi-
tions, respectively. The ratio of the recorded dynamic Figure 1. CPT profiles for example pile locations for Site 1.
soil resistance to driving, Rdyn;rec, against the predicted
dynamic soil resistance to driving, Rdyn;pred, is plotted
as logarithms. If this ratio is within a range of -0.12
(75% → LE) and 0.12 (133% → HE) then the predic-
tion is considered as acceptable.

3 Pile driveability assessment

During pile installation the energy delivered by the

hammer and the blowcount per 0.25m are recorded in
the pile driving records. The pile driving records from
four recently installed offshore wind farm projects in
North Sea have been considered in this paper to cover
a wide range of varying pile properties, impact Figure 2. CPT profiles for example pile locations for Site 2.

760
 Session 8 - Installation of Piled Foundations

For monopiles at Site 1 the model overestimates

blowcounts in the shallow depth up to approximately
5-7m whereas below 5-7m penetration it produces
generally lower blow-counts, namely underpredicting
the measured response by approximately a factor of
2-2.5. The deviation between the model predictions
and the driving records became more significant with
increasing depth.
For Site 2 model produces the lower blow-counts
from the depth impact driving (i.e., ~ 16-17m depth)
starts. The deviation between the model predictions
of blowcount and the driving records ranges mostly
Figure 3. CPT profiles for example pile locations for Sites 3&4. between factor of 2 and 3 times and became more sig-
nificant with increasing depth. Furthermore, at some
Driving records for all four sites have been back- locations an offset factor of 4.5-6 has been observed
analysed with as-installed blow count and hammer at depths where CPT logs indicate cone resistance, qc
energies from driving records. From driving reports, values exceeding 80MPa (Figure 2 and Figure 4).
it has been observed that pile driving with impact The model yields better matching results for jacket
hammer started around 3-4 m depth at monopile piles with relatively smaller diameter and deeper em-
locations for Site 1, around 16 m depth for Site 2, and bedment depth (Site 3 and 4) compared to monopiles
around 4-5 m at jacket pile locations for Sites 3 and with larger diameter and shallow embedment depth
4. (Site 1 and 2). The model produces higher blowcounts
Figures 4 and 5 present the variation in hammer in the shallow depth of approximately 10-15 m depth
efficiency during driving for some selected monopile (Figure 5). Except from the depths where high qc val-
locations in sites 1 and 2, for jacket piles at Sites 4 & ues (i.e., higher than 80 MPa) observed the offset be-
5, respectively. For reference, the actual blowcounts tween predicted and measured response is factor of
together with the predicted blowcounts for the actual 1.1-1.4. The model predicts 2.5 to 4.5 times higher
measured energy were plotted. blow counts where qc values is higher than 80MPa
(Figure 3) and deviation become more significant
with increasing depth.
A back analysis based on blowcounts only is not
adequate to evaluate the soil resistance calculation
model because of the non-linear behaviour of dy-
namic resistance to driving and blow count. There-
fore, the ratio of recorded Rdyn to predicted Rdyn has
been analysed and the results for some selected loca-
tions per site have been plotted in Figure 6 and 7. If
the value of ratio Rdyn,rec/Rdyn,pred is within a range of
0.75 to 1.33, then the prediction is considered ac-
ceptable.
Figure 4. Blowcount and hammer energy versus penetration for
selected example location for Sites 1 & 2.

Figure 6. Dynamic resistance versus penetration for selected

example locations for Sites 1 & 2.
Figure 5. Blowcount and hammer energy versus penetration for
selected example location for Sites 3 & 4.

761
 Innovative Geotechnologies for Energy Transition | The Society for Underwater Technology

A modified friction degradation mechanism has

been examined for some of the monopile locations at
Site 1 and 2. In the modified Alm & Hamre model the
residual shaft friction was raised to 50% of the initial
value and therewith additional back analyses were
performed.
As can be derived from Figures 8 and 9, the mod-
ified Alm and Hamre model yields higher blowcount
values compared to the original model. The back
analyses clearly show that the drivability predictions
for monopiles are noticeably improved using the
modified model.
However, further analyses have to be done in com-
Figure 7. Dynamic resistance versus penetration for selected bination with different values of empirical coeffi-
example locations for Sites 3 & 4. cients included in the original model to identify more
representative friction degradation concept for mono-
For Site 1, in shallow penetration depths approxi- piles with shallower embedmenth depth compared to
mately less than 5m, the model tends to overestimate jacket piles.
the soil resistance. For sand layers between 5m to
15m the predictions are better correlated to qc values Blow count [blows/0.25 m]

between 25-40MPa. The soil resistance is underesti- 0

0 50 100 150 200 250 300

mated for qc values less than 50MPa below 15m depth

(Figure 3). The model produces good prediction for
5

monopiles at penetration depths below 15m for qc val- 10

Penetration depth [m]

ues exceeding 50MPa. This is contrary to the results 15

observed for jacket piles at Site 3 and 4.

For Site 3 and 4 in medium dense to dense sand for
20

shallow penetration depths up to 13m model shows a 25

tendency to overestimate resistance for soils with qc 30

values between 20MPa to 40MPa. Layers with the

same qc values in greater penetration depths show a
35

Site 1 [ Actual ] Site 1 [ Alm & Hamre ] Site 1 [ Modified Alm & Hamre ]

good fit. For deeper sand (~ 20m to 25m depth), the Site 2 [ Actual ] Site 2 [ Alm & Hamre ] Site 2 [ Modified Alm & Hamre ]

tendency of overestimating soil resistance appears for Figure 8. Comparison between Alm & Hamre and Modified
qc>50MPa and increases with increasing qc. Alm & Hamre for blowcount and hammer energy versus
For jacket piles the model produce good predic- penetration for selected example locations for Sites 1 & 2.
tions as expected because the pile dimension, pile
penetration and the soil properties fall in the same
range of those in data set used by Alm & Hamre
(2001) and many others later. Furthermore, the obser-
vation of the overestimation of the model for higher
qc values at greater penetration depths is also coin-
cides with findings of other studies (Maynard et al.
2019).
However, for monopiles (i.e., with shorter embed-
ment depth compared to jacket piles) the model tends
to underestimate the soil resistance as well as blow-
counts. This is considered to be due to the effect of
the friction degradation concept, embedded in the
Alm and Hamre model, whereby the friction re-
Figure 9. Comparison between Alm & Hamre and Modified
sistance applied by the overlying layers degrades as Alm & Hamre dynamic resistance versus penetration for
the pile penetrates deeper into the ground, might be selected example locations for Sites 1 & 2.
overestimated. The embedment depth of monopiles is
significantly shorter than jacket piles. Since the em-
bedment depth has direct effect on relative fatigue 4 Conclusions
friction reduction, a raise of lower limit of residual
resistance, which is defined in the model as 20% According to analysis presented in this paper, the cur-
based on back-analyses of data set of much longer rent prediction model based on Alm & Hamre (2001)
jacket piles in the North Sea, is expected to yield an does not provide good estimation for monopile with
improved prediction. relatively short embedment in contrast to jacket piles

762
 Session 8 - Installation of Piled Foundations

with much deeper embedment depth. The original Alm, T. and Hamre, L. (1998). Soil Model for Driveability Pre-
model appears to systematically underestimate blow- dictions. OTC 8835, Offshore Technology Conference, No.
OTC 8835, 13.
counts for monopiles by approximately a factor of 2 Alm, T. and Hamre, L. (2001). Soil Model for Pile Driveability
to 2.5. Predictions Based on CPT Interpretations. Proceedings of
For jacket piles the observed offset between pre- the 15th International Conf. on Soil Mechanics and Founda-
dicted and measured blowcounts is factor in range of tion Engineering, Istanbul, Vol. 2, pp. 1297–1302.
1.1 to 1.4. However, for qc values greater than 80MPa Byrne T., Gavin K., Prendergast L.J., Cachim, P., Doherty, P,
at greater depths the model predicts 2.5 to 4.5 times and Pulukul S.C. (2018). Performance of CPT based meth-
ods to assess monopile driveability in North Sea sands.
higher blowcounts and this deviation becomes more Ocean Engineering, 166, 76-91.
significant with increasing penetration depth. Cardoso, A., Raymackers, S., Davidson, J. and Meissl, S (2018).
The results showed the blowcount as well as soil Interpreting properties of glacial till from CPT and its accu-
resistance was underpredicted for monopiles particu- racy in determining soil behaviour type when applying it to
larly at greater depths. There appears to be evidence pile driveability assessments. Proceedings of the 4th Interna-
tional Symposium on Cone Penetration Testing, Delft, , pp.
that the reasoning behind the underestimations is the 199–204.
method’s friction degradation concept not adequately Jardine, R. J., Chow, F. C., Overy, R. & Standing, J. R. (2005).
reflecting all pile dimensions. ICP design methods for driven piles in sands and clays.
Therefore, a modified version of the original Alm Thomas Telford.
and Hamre model, with a reduced friction degrada- Lehane B.M., Schneider J.A. and Xu X. (2005). The UWA-05
tion effect was examined and found to yield more ac- method for prediction of axial capacity of driven piles in
sand. Proc. Int. Symposium on Frontiers in Offshore Ge-
curate results in cases analysed. otechnics, Perth, 683-690.
The result of this study highlights that the model Maynard, A.W., Hamre, L., Butterworth, D., and Davidson, F.
requires modification for driveability predictions of (2019). Improved Pile Installation Predictions for Mono-
monopiles which have significantly shallower em- piles. Proc. Stress wave theory and testing method for deep
bedment depth as compared to the jacket piles that Foundations, West Conshohocken, 426-449.
Perikleous, G.,Stergiou, T., and Meissl, S. (2020). An assess-
were considered in development of the original ment of the accuracy of SRD methodologies for OWF mono-
model. Further investigations and modifications will pile installation against a North Europe driving records data-
have to be performed to improve the friction degrada- base. 4th International Symposium on Frontiers in Offshore
tion concept for monopiles, focusing on increasing of Geotechnics (ISFOG).
the residual shaft friction as a next step, and by ad- Pile Dynamics Inc. (2010); GRLWEAP: Wave Equation Analy-
justing of the empirical coefficients considered for the sis of Pile Driving. Procedures and Models Manual, Cleve-
land.
friction degradation factor. Smith, E.A.L. (1960). Pile Driving Analyses by the Wave Equa-
tion. Journal of the Soil Mechanics and Foundation Division,
ASCE, Vol. 86, pp. 35-61.
References Stevens, R.S., Wiltsie, E.A. and Turton, T.H. (1982). Evaluating
Pile Driveability for Hard Clay, Very Dense Sand and Rock.
Alm, T., Bye A., and Kvalstad T.J. (1989). A new interpretation OTC 4205. Proc Offshore Tech Conf, Houston, USA.
of soil resistance for pile driveability analysis. 12th Interna- Toolan, F.E. and Fox, D.A. (1977). Geotechnical Planning of
tional Conference on Soil Mechanics and Foundation Engi- Piled Foundations for Offshore Platforms. Proc ICE, Lon-
neering , Rio de Jenerio., pp.1085-1088. don, Part 1, 62, 221.

763