Sei sulla pagina 1di 9

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/282375524

Findings of the ISSMGE jack-up leg penetration prediction


event

Article · May 2015


DOI: 10.1201/b18442-192

CITATIONS READS

2 66

2 authors, including:

Gülin Yetginer
Equinor
15 PUBLICATIONS   57 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

NPD Seabed Project View project

Pile Reliability JIP View project

All content following this page was uploaded by Gülin Yetginer on 14 December 2018.

The user has requested enhancement of the downloaded file.


Frontiers in Offshore Geotechnics III – Meyer (Ed.)
© 2015 Taylor & Francis Group, London, ISBN: 978-1-138-02848-7

Findings of the ISSMGE jack-up leg penetration prediction event

B.F.J. van Dijk


Fugro Engineers B.V., Nootdorp, The Netherlands

A.G. Yetginer
Statoil ASA, Stavanger, Norway

ABSTRACT: This paper presents the findings of an international jack-up leg penetration prediction event,
organized by the International Society for Soil Mechanics and Geotechnical Engineering (ISSMGE) Technical
Committee for Offshore Geotechnics (TC209). Professionals within the offshore geotechnical industry were
invited to predict leg penetration behaviour for four different soil conditions based on geotechnical information
available, spudcan/footing dimensions and preloads. Evaluation of the predictions and comparison with actual
measured leg penetrations provided insight in industry practice with regard to selection of calculation methods,
soil parameter values and the range of predictions that can be expected for various soil conditions.

1 INTRODUCTION 2 CASE 1 WEAK CLAY

The ISSMGE Technical Committee for Offshore Case 1 concerns a LeTourneau Design, Super Gorilla
Geotechnics (TC 209) organized an international XL Class jack-up rig installed in the Gulf of Mexico.
jack-up leg penetration prediction event. Predictors
were invited to present their assessment of spudcan leg
penetration behaviour for four cases. Three located in 2.1 Jack-up rig details
the Gulf of Mexico, and one located in the North Sea. The Super Gorilla jack-up rig is among the largest rigs
For each case soil data and jack-up details, includ- in the world. The preload was 155.9 MN per spudcan.
ing preload and spudcan geometry were provided. The lightship-plus-variable (LPV) load was 84.5 MN.
Predictors were asked to provide information on the The slightly irregular spudcan shape is presented in
calculation method(s), selected (soil) parameter val- Figure 1. The spudcan has an effective area of 307 m2
ues, spudcan penetration curves, expected penetration and a volume of 1,327 m3 .
depth and foundation risks.
In total there were 25 Predictors, representing 7 con-
sultants and 6 institutes, including 2 universities. 2.2 Soil conditions
Some Predictors used more than one prediction model, The soil conditions consist of normally consolidated
resulting in over 30 predictions per case. The Predic- (NC) soft clay over slightly over-consolidated clay.
tors provided best estimate (BE), lower bound (LB) From 45 m below seafloor (bsf) medium dense silt
and/or upper bound (UB) predictions. The predic- to very dense fine sand was found. Measured sub-
tions were compared with actual measured spudcan merged unit weight (γ  ) and undrained shear strength
penetrations for each case.
Limit equilibrium methods, such as given in ISO
19905-1 (2012) and SNAME (2008) were mostly
used, either by utilizing proprietary and commercially
available software or Excel sheets. Three Predic-
tors used the Coupled Eulerian-Lagrangian model in
Abaqus/Explicit, a Large Deformation Finite Element
(LDFE) method. One participant used an in-house
CPT based method. Since all predictions were made
after the event, they may be classified as Type C
(Lambe 1973). Figure 1. Case 1 spudcan geometry.

1267
Figure 2. Soil conditions cases 1 and 2.

(su ) are presented on Figures 2a and 2b, together with


the design lines used by Predictors.
High quality 3 inch diameter push samples were
collected in the clay. A borehole log to 60 m bsf was
provided, presenting soil stratigraphy, soil descrip-
tions, submerged unit weight and undrained shear
strength. The borehole log indicated gas blisters
between 19.2 m and 30.5 m bsf. Additionally, lab-
oratory results included Atterberg Limits (AL) and
(remoulded) Unconfined Undrained triaxial (UU)
tests and Miniature Vane (MV) tests.

2.3 Predicted and measured penetrations


Best estimate predictions and measured spudcan pen-
etrations are presented in Figure 3. Most predictions
are within a few metres of measured penetrations at
LPV load. However at preload, there is a large spread
in predicted penetration results, which are all consis-
tently shallower than observed. The LDFE predictions
lie within the same range as the predictions based on Figure 3. Case 1 BE predictions and measured penetration.
limit equilibrium models.
Figure 4 presents predicted LB, BE and UB, as
provided by the Predictors, measured spudcan leg
penetrations at preload and corresponding probability
density graphs, assuming normal distributions. Most
predictions, including the LB predictions are shallower
than the final measured penetrations. There is little
overlap between the predicted and measured proba-
bility density graphs (Fig. 4.). The mean measured
penetration (35 m bsf) is approximately 8 m deeper
than the mean of the BE predictions (27 m bsf). For the
bow leg (spudcan), which penetrated deepest (37 m),
this difference increases to 10 m. At deep water sites
this difference between predicted and actual penetra- Figure 4. Case 1 predicted and measured spudcan
tions could lead to insufficient leg length, once the penetration.
jack-up rig has been mobilized to location.

methods of Skempton (1951), Brinch Hansen (1970),


2.4 Back analysis by Menzies & Roper (2008)
Houlsby & Martin (2003) and Hossain et al. (2006)
This case has already been analysed by Menzies & with measured spudcan penetrations. Hence some
Roper (2008). They compared predictions using predictions may have been of type C1 (Lambe

1268
2.5 Comparison of methods used by
predictors
The failure mechanism for this case is general shear in
clay. SNAME (2008), ISO 19905-1 (2012), Skempton
(1951), Houlsby & Martin (2003) and Hossain et al.
(2014) methods were mostly used by Predictors.
BE penetration predictions are shown (as sym-
bols) in Figure 6 against the average undrained shear
strength below the spudcan base (su,avg ) as used by the
Predictors. su was averaged to a depth (davg ) of B/2 for
most methods, with the exception of Skempton (1951)
(davg = B) and Brinch Hansen (1970) (davg = B/4).
For Hossain et al. (2014) and Houlsby & Martin
(2003) BE predictions are shown against su at spud-
can base (su0 ). Figure 6 also includes theoretical lines
for Vesic (1973), Skempton (1951), Hossain et al.
(2014), Houlsby & Martin (2003), Young et al. (1984)
and Brinch Hansen (1970) methods, assuming 155.9
Figure 5. Case 1 BE predictions and measured penetration.
MN preload, full backflow and parameters used by
Menzies & Roper (2008). It is noted that lines are not
related to a su profile, except for Hossain et al. (2014)
and Houlsby & Martin (2003) methods. In case of par-
tial backflow, the theoretical lines would shift down by
1973); i.e. measured results were known to Predictors, approximately 6 to 7 kPa (i.e. an decrease of su,avg ).
although Figure 3 does not suggest this. Some predictions are above the theoretical lines for
Menzies & Roper (2008) adjusted the vertical bear- full backflow. This could have been caused by neglect-
ing capacity of all four prediction methods with an ing the spudcan volume, differences in averaging su
empirical factor of 0.85 to better match measured and/or modification of methods used.
spudcan penetrations. Figure 5 shows their adjusted Both Young et al. (1984) and Houlsby & Martin
vertical bearing capacity for Houlsby & Martin (2003) (2003) methods are included in ISO 19905-1 (2012).
method only. Menzies & Roper (2008) indicated that Young et al. (1984) is actually Skempton (1951)
the adjustment factor is probably related to the spud- method where su is averaged over a depth of 0.5 ∗ B
can geometry, which is not circular, but features short below spudcan base level instead of B. SNAME (2008)
(9 m) and long (10.1 m) radii and has a relatively flat includes a Main Text method (Vesic 1973) and men-
bottom. They mentioned that this empirical factor has tions other methods such as Houlsby & Martin (2003)
been used in practice at other sites where this type of in the Commentary section. Hence for Predictors who
jack-up was employed. used ISO 19905-1 (2012) and SNAME (2008), it
The Houlsby & Martin (2003) method provided the is unclear which methods were actually used. One
best match with measured penetrations. The method SNAME (2008) prediction is close to the theoretical
gives a theoretical lower bound to soil resistance and Vesic (1973) line, while the others are closer to the
should therefore predict the deepest expected spudcan theoretical Young et al. (1984) and Houlsby & Martin
penetration (ISO 19905-1 2012). Houlsby & Martin (2003) lines. ISO 199051 (2012) predictions are close
(2003) consider spudcan tip cone angle (β) and rough- to the Houlsby & Martin (2003) theoretical line, prob-
ness factor (α). In their assessment Menzies & Roper ably used by Predictors. Differences found for these
(2008) used α = 1.0. ISO 19905-1 (2011) recommends predictions are probably caused by differences in soil
α = 0.5 for stiff clay found at final penetration depth. parameter value selection.
For α = 0.5 instead of α = 1.0, the ratio between mea- One Predictor obtained a spudcan penetration of
sured and calculated spudcan capacity increases from 33.6 m bsf, close to the measured penetration. Predic-
0.86 to 0.90. tor reported using Brinch Hansen (1970), combined
Menzies & Roper (2008) did not consider a stable with a modified version of Hossain et al. (2006)
cavity depth for this case, but considered full back- for partial backflow and in-house experience. Using
flow. Using Houlsby & Martin (2003) with Hossain the Predictors soil parameters with Brinch Hansen
et al. (2006) partial backflow recommendations, the (1970) and Hossain et al. (2006) recommendations, the
vertical bearing capacity increases by approximately Authors found a soil resistance of 211 MN at 33.6 m
9% (Fig. 5), leading to a shallower penetration. bsf. The difference of 35% is believed to be a result
According to Menzies & Roper (2008), the Hossain of modifications made to the approach, based on the
(2006) method over-predicted soil resistance by 20%, Predictor’s in-house experience.
in addition to the above mentioned empirical reduc- One LDFE Predictor calculated a penetration of
tion factor. Hossain et al. (2014) is a modification of 33.1 m bsf. Using approximately the same soil parame-
Hossain (2006) which includes an adjustment factor ters, another LDFE prediction resulted in a penetration
of 20% on the deep bearing capacity factor. of 24.3 m bsf. Not enough detail was provided by the

1269
Figure 7. Case 2 spudcan geometry.

For the Houlsby & Martin (2003) method, the pre-


dicted penetrations are between 26.2 and 30.2 m bsf.
Predictions are close to the theoretical line. Differ-
ences may have been caused by assumptions made
(e.g. spudcan roughness). su0 varied between 56.5 and
63.9 kPa, whereas the increase of su with depth (ρ)
varied between 1.2 and 2.0 kPa/m. It is noted that
for these penetration depths, higher ρ value decreases
resistance. Two predictions with higher su0 values are
related to partial backflow, which is not as expected.
Predictors using Hossain et al. (2014) method pre-
dicted penetrations of 25.9 and 31.4 m bsf. Latter
Predictor used Menzies & Roper (2008) reduction fac-
tor of 0.85 on the bearing capacity, which resulted in
a relatively high su0 .
One Predictor used the Young et al. (1984) method.
Figure 6. Case 1 su,avg for BE penetration predictions The prediction is well below the theoretical line,
(symbols) and theoretical lines, for 155.9 MN preloading caused by a partial backflow assumption.
condition. The discrepancy between predictions and observa-
tions indicates that the models and the parameter selec-
tion may not always represent the actual behaviour of
predictors to explain the difference. The third LDFE deep penetrating spudcans correctly. Possible causes
Predictor used a probabilistic approach. The prediction are spudcan size, spudcan shape and/or a large dis-
showed a 5% probability that the spudcan penetrates turbed zone around the spudcan, degrading the su along
more than 31.2 m, which was selected as final pene- an important part of the failure zones.
tration depth, and a 50% probability that the spudcan
penetrates more than 26.7 m.
Using a simple hand calculation of 9 ∗ su to 10 ∗ su , 3 CASE 2 WEAK CLAY OVER SAND
one Predictor assessed a penetration depth of 32 m bsf.
Interestingly, the bearing capacity factor Nc for the Case 2 concerns a Marathon LeTourneau Design,
methods used by other Predictors are lower than 9. Class 82-SD-C jack-up rig, installed in the Gulf of
The relatively large predicted penetration depth is due Mexico.
a low estimate of the su value (Fig. 6).
CPT-based prediction resulted in a penetration 3.1 Jack-up rig details
depth of 31.5 m bsf. Predictors used a cone fac-
tor Nk = 15 to derive cone resistance from selected The spudcans were preloaded to approximately
undrained shear strength design line. 32 MN. The spudcan shape is presented in Figure 7.
Penetration predictions using the Skempton (1951) The spudcan has an effective area of 115 m2 and a
method vary between 20.0 m and 30.6 m bsf. It can volume of 288 m3 .
be seen from Figure 6 that the selection of soil con-
ditions is not the only factor for the wide spread in
the predictions. If selection of soil conditions were the 3.2 Soil conditions
main factor for the differences, the trend in predictions The soil conditions consist of alternating clay and sand
would have been expected to be parallel to the the- layers. The stratigraphy consists of normally consoli-
oretical lines. Backflow conditions or modifications dated soft clay at seafloor over medium dense silty fine
of the method may attribute to the main cause of the sand. Measured γ  , su , internal friction angle (φ ), and
difference. Most Skempton (1951) predictions have a CPT data are presented on Figures 2c and 8, together
higher su,avg than the theoretical line. Possibly these with the design lines used by Predictors. High qual-
Predictors used a lower value for davg , e.g. davg = B/2, ity 3 inch diameter push samples were collected in
as recommended by Young et al. (1984). This would the clay. Borehole logs to 30 and 60 m bsf were pro-
reduce su,avg by 7 to 14 kPa. vided, presenting soil stratigraphy, soil descriptions,

1270
Figure 8. Soil conditions case 2.

Figure 10. Case 2 predicted and measured spudcan


penetration.

4 CASE 3 FISSURED CLAY

Case 3 concerns a jack-up rig installed in the North


Sea.
Figure 9. Case 2 BE predictions and measured penetration.

γ  and su . Additionally, laboratory results included AL 4.1 Jack-up rig details


and (remoulded) UU and MV tests. The spudcans were preloaded to approximately
32 MN, which was less than the design preload
of 3,800 tonnes originally submitted to Predictors
3.3 Predicted and measured penetrations (Fig. 12). The spudcan was rectangular (7.7 m by
Predictions and measured spudcan penetrations are 8.3 m). No details were provided regarding the volume
presented in Figures 9 and 10. The predicted penetra- of the spudcan.
tion is in most cases very close to the actual measured
penetration. All BE predictions lie within 2.4 m from
the measured depths at final preload. 4.2 Soil conditions
The predictions for the vertical bearing capacity just The soil conditions consist of over-consolidated heav-
above and into the first sand layer show a large scatter, ily fissured clay. Only CPT data was available at this
ranging between about 60 MN and 600 MN, and one location. No laboratory data, such as γ  and su were
extreme of around 1400 MN. This scatter may have available. The predictors were informed that samples
been caused by differences assessing the su profile for in a nearby borehole indicated heavily fissured clay.
the second clay layer and/or the failure model for sand Cone resistance, and undrained shear strength derived
over clay. from CPT data using cone factors Nk = 20 and Nk = 30
In case the preload would have been higher than are presented on Figures 11b and 11c. Figure 11
32 MN this might have led to significant differences also includes design lines for γ  and su used by the
in predicted probability of punch-through. Predictors.

1271
Figure 11. Soil conditions case 3.

Figure 14. Case 4 spudcan geometry.

The deep measured penetration at this site is caused


by the heavily fissured clay. Depending on the spac-
ing size, fissures can significantly reduce the su for
large foundations when compared to the measured su in
undrained triaxial tests and CPT derived values using
an Nk = 20 (Terzaghi et al. 1996). For this case, selec-
tion of su values for the assessments using Nk = 30 to
35 may have been more appropriate. The above men-
tioned two Predictors used Nk factors between 30 and
40 to account for fissured clay. One other Predictor
Figure 12. Case 3 BE predictions and measured penetration. mentioned possibility of reduced su due to fissures,
and used an Nk = 30 for the LB prediction.

5 CASE 4 SAND OVER WEAK CLAY

Case 4 concerns a Marathon LeTourneau Design,


Class 116-C jack-up rig, installed in the Gulf of
Mexico.

5.1 Jack-up rig details


The spudcans were preloaded to approximately
53.4 MN. The spudcan shape is presented in Figure 14.
Figure 13. Case 3 predicted and measured spudcan The spudcan has an effective area of 144 m2 and a
penetration.
volume of 383 m3 .

4.3 Predicted and measured penetrations


5.2 Soil conditions
Figures 12 and 13 present predicted and measured
jack-up leg penetrations. All Predictors but two under The soil conditions consists of a 7 m thick medium
predicted the measured penetration. dense to dense fine sand layer overlaying soft to stiff

1272
Figure 15. Soil conditions case 4.

Figure 18. Case 4 predicted and measured punch-through


load.

The soil was expansive due to the presence of bio-


genic gas. To account for the small diameter sample
size and sample expansion, the measured low su val-
ues were adjusted by applying several correlations
commonly used in the Gulf of Mexico (Quiros et al.
2000). Both measured and adjusted data was provided
Figure 16. Case 4 BE predictions and measured penetration.
to Predictors.

5.3 Predicted and measured penetrations


Predictions and measured spudcan penetrations are
presented in Figures 16 and 17. Measured spudcan
penetrations show punch-through at approximately
4 m bsf near the preload. It is noted that, between 4 m
bsf and final penetration depth, the measured verti-
cal bearing capacity has been estimated, considering
percentages of applied load and hull buoyancy.
One Predictor indicated low punch-through risk.
The LDFE prediction resulted in a shallow penetration
to 3.5 m bsf and no clear risk of punch though.All other
Figure 17. Case 4 predicted and measured spudcan Predictors indicated a (high) risk of punch-through
penetration. and/or rapid leg penetration. Shallow penetration (less
than 4 m bsf) was predicted 3 times for the BE.
Most Predictors anticipated punch-through for a
clay to 56.1 m bsf. Figure 15 presents measured γ  and spudcan load between 25 MN and 50 MN (Fig. 18), i.e.
su , as well as the design lines for γ  , su and φ used by much less than the measured leg load between 53 MN
the Predictors. The clay samples were obtained with a and 56 MN at which punch-through occurred. Predic-
2.25 inch diameter (small) sampler. tors used either the load spread model (ISO 19905-1

1273
2012) or the Hanna & Meyerhof (1980) model to assess selection, the sophistication of the prediction method
the punch-through load. and engineering judgement.
Three Predictors assessed a punch through load
close to the measured one (less than 10% difference).
The first Predictor performed a parametric study, ACKNOWLEDGMENTS
using ISO 19905-1 (2012) and SNAME (2008) recom-
mendations with load spread model with factors (ns ) The authors gratefully acknowledge Fugro, Statoil,
varying between 1 and 5. ISO 19905-1 (2012) with an David Menzies and ISSMGE’s commitment and
ns of 2 and SNAME (2008) with an ns of 3 resulted in support to improving geotechnical practice.
punch through loads close to the measured one. Vari-
ation of ϕ between 25 and 35 degrees for the ISO
199051 (2012) recommendations did not influence the REFERENCES
penetration curves, as expected. The second Predictor
combined SNAME (2008) and Osborne et al. (2011) Brinch Hansen, J. 1970. A Revised and Extended Formula
for Bearing Capacity, The Danish Geotechnical Institute
recommendations with the punching shear method by
Bulletin, No. 28, pp. 5–11.
Hanna & Meyerhof (1980). The last Predictor used a Hanna A.M. & Meyerhof G.G. 1980. Design Charts for Ulti-
newly developed discrete analysis approach, where the mate Bearing Capacity of Foundations on Sand Overlying
soil is discretised into thin layers. The punch through Soft Clay, Canadian Geotechnical Journal Vol. 17, 1980
load was assessed by peripheral shear strength along Houlsby, G.T. & Martin, C.M. 2003. Undrained Bearing
the sand plug, bearing resistance at the plug base, plug Capacity Factors for Conical Footings on Clay, Géotech-
weight, spudcan buoyancy and soil backfill. nique, Vol. 53, No. 5, pp. 513–520.
As can be seen in Figure 16, there is a large spread Hossain, M.S., Randolph, M.F., Hu, Y. & White, D.J.
in predicted vertical bearing capacity, both in the top 2006. Cavity Stability and Bearing Capacity of Spudcan
Foundations on Clay, Proceedings Offshore Technology
sand layer and the underlying clay. This is caused by
Conference, Houston, 1–4 May 2006. OTC paper 17770.
the spread in soil parameter values, load spread factors Hossain, M.S., Zheng, J., Menzies, D., Meyer, L. & Randolph,
and calculation models used. M.F. 2014. Spudcan Penetration Analysis for Case Histo-
ries in Clay, Journal of Geotechnical and Geoenvironmen-
tal Engineering, Vol. 140, No. 7, 13 p.
ISO International Organization for Standardization 2012.
6 CONCLUSIONS Petroleum and Natural Gas Industries – Site Specific
Assessment of Mobile Offshore Units – Part 1: Jack-ups,
For all cases, there is high scatter in predicted verti- International Standard ISO 19905-1:2012.
cal bearing capacity to spudcan penetration. It can be Lambe, T.W. 1973, Predictions in Soil Engineering, Géotech-
concluded that accurate prediction of the penetration nique, Vol. 23, No. 2, pp. 151–201.
depth for cases 1, 3 and 4 is difficult and for most pre- Menzies, D. & Roper, R. 2008. Comparison of Jack-up
dictions was inaccurate. Case 1 indicated that current Rig Spudcan Penetration Methods, Proceedings Offshore
practice for selection of parameter values and calcu- Technology Conference, Houston, 5–8 May 2008. OTC
paper 19545.
lation models for general shear in clay may lead to
Osborne, J.J., Teh, K.L., Houlsby, G.T., Cassidy, M.J., Bienen,
overestimation of soil resistance. The predictions for B. & Leung, C.F. 2011. InSafeJIP: Improved Guide-
case 2 compared well to the measured penetration. lines for the Prediction of Geotechnical Performance of
However, as for case 4, there was a large scatter in Spudcan Foundations during Installation and Removal
prediction of the punch-through load for the first sand of Jack-up Units: Joint Industry-funded Project, Woking,
layer. Report No. EOG0574-Rev1c.
For case 1 the rig might have been installed but Quiros, G.W., Little, R.L. & Garmon, S. 2000. A Normalized
would not have had sufficient leg length to work the Soil Parameter Procedure of Evaluating In-Situ Undrained
location. The predictions for case 3 show that the large Shear Strength, Offshore Technology Conference, 1–4
May 2000, Houston. OTC paper 12090.
scale strength of the fissured clay is much lower than
Skempton, A.W. 1951. The Bearing Capacity of Clays,
generally expected. Some predictions for case 4 might Building Research: 1st International Congress, London,
have led to an unforeseen punch-through. Sept. 1951 Papers Division 1 and 2, pp. 180–189.
The more advanced LDFE assessments did not SNAME Society of Naval Architects and Marine Engineers
provide better predictions than those using limit equi- 2008. Recommended Practice for Site Specific Assess-
librium methods showing that use of more advanced ment of Mobile Jack-up Units, First Edition, May 1994
models does not necessarily increase prediction accu- (Revision 3,August 2008),Technical & Research Bulletin,
racy. Comprehensive soil data was not available for 5-5A.
all of the cases, possibly causing some of the differ- Terzaghi, K., Peck, R.B. & Mesri, G. 1996. Soil Mechanics
in Engineering Practice, 3rd ed., Wiley, New York.
ences between predictions and measurements. How-
Vesic,A.S. 1973Analysis of Ultimate Loads of Shallow Foun-
ever, since all Predictors were provided with the same dations, Journal of the Soil Mechanics and Foundations
information, the observed wide range in the predic- Division, Vol. 99, No. SM1, pp. 45–73
tions is mostly due to variances in data interpretation, Young, A.G., Remmes, B.D. & Meyer, B.J. 1984. Foundation
selection of parameters and calculation models used. Performance of Offshore Jack-up Drilling Rigs, Journal of
This paper illustrates that the accuracy of any pre- Geotechnical Engineering, Vol. 110, No. 7, pp. 841–859.
diction depends on the quality of the data, parameter

1274

View publication stats

Potrebbero piacerti anche