Professional Documents
Culture Documents
CHEON DISSERTATION Analysis of Spatial Variability in Geotechnical Data For Offshore Foundations
CHEON DISSERTATION Analysis of Spatial Variability in Geotechnical Data For Offshore Foundations
by
2010
The Dissertation Committee for Jeong Yeon Cheon Certifies that this is the
approved version of the following dissertation:
Committee:
Jorge G Zornberg
Lance Manuel
David Mohrig
Chadi S. El Mohtar
Analysis of Spatial Variability in Geotechnical Data
for Offshore Foundations
by
Dissertation
Presented to the Faculty of the Graduate School of
in Partial Fulfillment
of the Requirements
Doctor of Philosophy
I would like to express my deepest and sincere appreciation to Dr. Robert Bruce
Gilbert for the motivation, guidance, encouragement and patience throughout my doctoral
study. I was fortunate to have him as a supervising professor. He is one of the most
creative and talented individuals I have ever known. Whenever I made a simple
Whenever I got stuck with complex equations and analyses, he helped me to solve the
problems using the basics. I have learned there is no exact right or wrong answer in the
real world and the all things are from the basics. He also has a warm personality. He has
always encouraged me to push forward with challenges. “Go get 'em, Tiger, Doogie!”
Manuel, and Chadi S. El Mohtar for reviewing the dissertation and giving me valuable
comments. I would like to express special thanks to Dr.Mohrig for providing essential
comments and references which were very helpful for me to understand the soil from
of Texas at Austin: Drs. Stephen G. Wright, Robert B. Glbert, Charles M. Woodruff Jr.,
Roy E. Olson, Kenneth H. Stokoe II, Jorge G Zornberg, Fulvio Tonon, Ellen M. Rathje,
and Dr.Chadi S. El Mohtar for creating this unique and one of the best geotech programs.
I am also grateful to Dr. Ghislaine De Regge for encouraging me whenever I meet across
v
her at ECJ. Her oral and written presentation classes are the best English class I have ever
taken. It has been a great pleasure, excitement, and also challenge to study in this
program.
I would like to thank administrative staffs, Teresa Tice-Boggs, Chris Trevino and
John Hall, and a technical staff, Federico Castro from the Geotechnical Engineering
Center.
Special thanks go to two friends of mine. Mary Nodine was a research partner for
my first research project at UT. She helped me to get used to the research project and UT.
She has full of positive energy that makes people happy. Sang Yeon Seo started and
finished his graduate study in the same year as mine. He always listened to me, helped
good neighbor is better than the near relatives who live far away. Sincere gratitude is
expressed to the Korean student fellows and their family for their continuous
encouragement and emotional help during the last four years. They have been more than
I would like to thank Young Jin Park. He encouraged me to start the graduate
study at UT, taught me how to drive, helped me settle down at Austin, and give me
advice and encouragement when I had a hard time with any selecting issues.
I cannot express my deep gratitude and love to my family in a few words: Geum
Nyeon Goh, Jinho Cheon, Jeong Eun Cheon, Taekyu Park, Soohee Park, Soohyung Park,
Danu Park and Taerin Cheon. Their continuous love and support have kept me surviving
vi
Finally, I would like to confess a secret before I leave UT: I rode a bike in the
hallway of ECJ 9th floor almost every Sunday during the fall semester 2006. One short
excuse goes with this statement: “it was my first semester at UT.” In other longer words,
(1) I did not know that riding a bike inside a building is prohibited by the university; (2) I
had no place to go during the weekend except ECJ; and (3) I needed to spend significant
amount of time at school to finish reading assignments for CE 387L.1. I preferred riding
clockwise so that I could see the photos of all the geotech people, by the way.
Somebody said that life goes by at the speed of age. Even when my life speed
reaches almost its speed limit, I will remember my four years at UT Austin as one of the
August, 2010
vii
viii
Analysis of Spatial Variability in Geotechnical Data
for Offshore Foundations
Publication No._____________
Deep foundations, such as piles and suction caissons, are used throughout an
offshore oil and gas production facility in deepwater. Ideally, the values of geotechnical
investigation programs performed at the site of the foundation. However, the locations for
facilities are not known exactly when soil borings are drilled and the footprint of a
facility in deepwater can be very large with numerous foundation elements spread out
The objective of this research is to assess, analyze and model spatial variability in
from 14 offshore project sites covering the past twenty years of deepwater development
ix
in the Gulf of Mexico are compiled into a database. The geologic setting is primarily a
normally to slightly overconsolidated marine clay, and the property of interest for the
strengths are analyzed quantitatively and graphically. Geostatistical models that describe
spatial variability in the design shear strength properties to the distance away from the
available information are developed and calibrated with available information from the
capacity. Design examples are presented to demonstrate the use of the reliability
methodology.
Based on the design undrained shear strength profiles for the past 20 years in this
Gulf of Mexico deepwater area, the design undrained shear strength varies spatially but
does not depend on the time or method for site investigations. There are nonlinear spatial
relationships in the point shear strength laterally and vertically due to stratigraphy such
that depth-averaged shear strengths are correlated over further distances than point shear
strengths. The depositional forces are an important factor causing spatial variations in the
undrained shear strength, with greater variation and less spatial correlation in the more
recent hemipelagic deposits (about upper 60 feet) than the deeper turbidite deposits and
along the shelf versus off the shelf. The increased conservatism required in deep
foundation design due to spatial variability when site specific strength data are not
available is generally small with less than a five percent increase required in design
x
Table of Contents
xi
3.5 Summary ........................................................................................................... 42
4.3.2 Sample statistics for Sur (z) and Sur,avg (L) ........................................... 59
4.4 Spatial Variability in Design Shear Strength .................................................... 65
4.4.1 Introduction ............................................................................................ 65
4.4.2 Sources of variability in design shear strengths ..................................... 65
4.4.3 Analysis of spatial variability in design shear strength.......................... 70
4.4.3.1 Spatial variability in design undrained undisturbed shear strength70
3.4.3.2 Spatial variability in design undrained remolded shear strength .. 71
4.4.3.3 Relationship between a point value and a depth- averaged value . 73
4.4.3.4 Relationship between design undisturbed and design remolded
strengths......................................................................................... 73
4.4.4 Analysis of spatial correlation.............................................................. 103
4.5 Summary ......................................................................................................... 112
xii
Chapter 6: Model for Design Shear Strength of Soil..................................................... 135
6.1 Introduction..................................................................................................... 135
6.2 Basic Model Framework................................................................................. 135
6.2.1Basic form of model .............................................................................. 135
6.2.1.1 Unconditioned (generic) model................................................... 135
6.2.1.2 Spatially conditioned model ........................................................ 136
6.2.2 Model calibration ................................................................................. 138
6.3 Models for Design Undrained Shear Strength ................................................ 139
6.3.1 INTRODUCTION ..................................................................................... 139
6.3.2 Models for the design undrained shear strength at a point ( Su ) .................. 139
6.3.3 Models for the depth-averaged design undrained shear strength ( Su,avg )
.............................................................................................................. 143
6.3.4 Models for the equivalent linear profile of design shear strength ( Su1 )145
6.3.5 Models for the design remolded shear strength at a point ( Sur )........... 147
6.3.6 Models for the depth-averaged design remolded shear strength ( Sur,avg )
.............................................................................................................. 148
6.3.7 Models for the depth-averaged design shear strength ( Su,avg ) with
anisotropic coordinates......................................................................... 150
6.4 Examples: Predicted Design Shear Strength Profile....................................... 153
6.4.1 Example: predicted generic shear strength profile ............................... 153
6.4.2 Example: Spatially conditioning the profile on existing data points.... 154
6.4.3 Example: Ways of depicting extrapolated design profiles................... 157
6.5 Summary ......................................................................................................... 159
References...................................................................................................................... 203
Vita................................................................................................................................. 212
xiv
List of Tables
Table 4.1: Sample Statistics for the Design Undrained Shear Strength, Su (z) ......... 55
Table 4.2: Sample Statistics for the Depth-Averaged Design Shear Strength,
Su,avg (L) . .................................................................................................... 55
Table 4.3: Sample Statistics for the Slope of the Equivalent Linear Undrained
Shear Strength Profile, Su1 (L) . ................................................................ 57
Table 4.4: Sample Statistics for the Design Undrained Remolded Shear Strength,
Sur (z) ......................................................................................................... 62
Table 4.5: Sample Statistics for the Depth-Averaged Design Undrained Remolded
Shear Strength, Sur,avg (z) . ......................................................................... 62
Table 5.1: Statistics Relating to Design Ultimate Axial Capacity. .......................... 116
Table 6.2: Model parameters for SU,AVG (L) model ................................................ 144
Table 6.3: Correlation distance for SU (L) and SU,AVG (L) model.......................... 145
Table 6.4: Model parameters for SU1 (L) model ..................................................... 146
Table 6.5: Correlation distance for SU1 (L) model .................................................. 146
Table 6.6: Model parameters for SUR (L) model .................................................... 148
xv
Table 6.8: Correlation distance for SUR (L) and SUR,AVG (L) model ...................... 149
Table 6.9: Model Parameters for the Correlation Coefficient for SU,AVG (L) Based
on ( a shelf , d shelf ) Coordinates. ................................................................... 151
Table 8.2: Example-Extrapolated Axial Capacity and Partial Factor of Safety for a
15-ft Caisson ( ΩS =0.2, ΩR =0.4, β t arg et =4.0) ....................................... 191
xvi
List of Figures
Figure 1.2: Floating Production Systems (taken from “Hoover Deep Draft Caisson
Vessel in Perspective” by Sandstrom, Slocum and Heideman with
ExxonMobil). .............................................................................................. 2
Figure 2.2: Major physiogeographic features of the Gulf of Mexico outer continental
shelf (MMS, 2000). ..................................................................................... 9
Figure 2.3: Cone penetrometer and Remote vane (images from Fugro-
McCLELLAND, 2006). ............................................................................ 12
Figure 2.4: Torvane Tester Set (image taken from www.humboldtmfg.com) ............ 13
Figure 2.12: Example-Design Profiles for a Data Set with Test Results....................... 22
Figure 2.13: Design Undrained Undisturbed Shear Strength versus Depth .................. 23
xvii
Figure 2.17: Database Screen Showing the Folders. ....................................................... 27
Figure 2.22: Example of Charts for Design Shear Strength Collected in One Place..... 31
Figure 3.1: Load and Resistance Component of a Suction Caisson for Axial
Loading...................................................................................................... 35
Figure 3.2: Plasticity model (a) failure mechanism (Murff and Hamilton, 1993, and
(b) simplified analysis ( Aubeny et al., 2001 and 2003) ........................... 37
Figure 3.4: Suction Caisson under Inclined Loading Condition (Aubeny et al.,2003).40
Figure 3.6: Effect of the α Factor on Combined Capacity for a Suction Caisson
with L/D=6, N C =9 and D=15 feet........................................................... 43
Figure 3.7: Effect of the End Bearing Factor on Combined Capacity for a Suction
Caisson with L/D=6, α =0.8 and D=15 feet............................................. 44
Figure 3.8: Effect of Su1 on Combined Capacity for a Suction Caisson with L/D=6,
α =0.8 , N C =9 and D=15 feet.................................................................. 44
Figure 3.9: Combined Capacity for Suction Caissons with Various L/D.................... 45
Figure 3.10: Normalized Combined Capacity for Suction caissons with Various L/D. 45
Figure 4.2: Equivalent Linear Profile of Design Undrained Shear Strength. .............. 48
Figure 4.3: Profiles of Su (z) and Su,avg (L) in One Tenth-Foot Intervals. ............... 52
xviii
Figure 4.4: Sample Mean and Standard Deviation for Su (z) ..................................... 53
Figure 4.6: Sample Mean and Standard Deviation for Su,avg (z) ................................. 54
Figure 4.8: Sample Mean with 5-, 10-, 90-, and 95-Percentile Values for Su (z) ....... 56
Figure 4.9: Sample Mean with 5-, 10-, 90-, and 95-Percentile Values for Su,avg (L) . . 56
Figure 4.11: Sample Mean and Standard Deviation for Sur (z) .................................... 60
Figure 4.13: Sample Mean and Standard Deviation for Sur,avg (z) ................................ 61
Figure 4.15: Sample Mean with 5-, 10-, 90-, and 95-Percentile Values for Sur (z) ...... 63
Figure 4.16: Sample Mean with 5-, 10-, 90-, and 95-Percentile Values for Sur,avg (L) . 63
Figure 4.18: Comparison between Su,avg (L) and Sur,avg (L) ......................................... 64
Figure 4.34: Spatial Variability in Design Remolded Shear Strength at z=60 feet (Site
Average). ................................................................................................... 87
Figure 4.35: Spatial Variability in Design Remolded Shear Strength at z=60 feet
(Each Data Set).......................................................................................... 88
Figure 4.36: Spatial Variability in Design Remolded Shear Strength at z=60 feet
(Each Data Set).......................................................................................... 89
xx
Figure 4.38: Spatial Variability in Depth-Average Design Remolded Shear Strength
at z=60 feet (Each Data Set)...................................................................... 91
Figure 4.40: Spatial Variability in Design Remolded Shear Strength at z=120 feet
(Site Average)............................................................................................ 93
Figure 4.41: Spatial Variability in Design Remolded Shear Strength at z=120 feet
(Each Data Set).......................................................................................... 94
Figure 4.42: Spatial Variability in Design Remolded Shear Strength at z=120 feet
(Each Data Set).......................................................................................... 95
Figure 4.48: Design Undrained Undisturbed Shear Strength versus Design Undrained
Remolded Shear Strength........................................................................ 101
Figure 4.51: Rivers along the Coast and Contour of Distance off the Coast............... 105
Figure 4.53: Distance along The Shelf Line and Distance off the Shelf Line. ............ 106
Figure 4.54 Spatial Correlation based on Distance from the Mouth of Mississippi
River. ....................................................................................................... 107
xxi
Figure 4.55: Spatial Correlation based on Distance from Rivers and Distance off the
Coast........................................................................................................ 108
Figure 4.57: Spatial Correlation Based on Distance along the Continental Shelf Line
and Distance off the Continental Shelf Line. .......................................... 110
Figure 5.1: Axial Capacities for 60- and 120-ft Caissons.......................................... 117
Figure 5.2: Spatial Variability in Axial Capacity for 60-ft Caisson (Site Average
Value). ..................................................................................................... 118
Figure 5.3: Spatial Variability in Axial Capacity for 120-ft Caisson (Site Average
Value). ..................................................................................................... 119
Figure 5.4: Lower-Bound versus Design Axial Capacities for 60- and 120-ft
Caissons................................................................................................... 122
Figure 5.5: Ratio of Lower-Bound to Design Axial Capacity for 60- and 120-ft
Caissons................................................................................................... 123
Figure 5.6: Lateral Capacity for Loading at the 2/3 Point. ........................................ 126
Figure 5.7: Spatial Variability in Lateral Capacity for a 60-ft Long Caisson (Site
Average). ................................................................................................. 127
Figure 5.8: Spatial Variability in Lateral Capacity for 120-ft Caisson (Site Average).128
Figure 5.9: Design Axial Capacity Versus Design Lateral Capacity for 60- and 120-
ft Caissons. .............................................................................................. 129
Figure 5.10: Taut Leg System (Vryhof Anchors Manual 2000).................................. 130
Figure 5.11: Spatial Variability in Total Capacity under Combined Loading for a 60-
ft Long Caisson (Site Average). .............................................................. 132
Figure 5.12: Spatial Variability in Total Capacity under Combined Loading for 120-ft
Caisson (Site Average)............................................................................ 133
xxii
Figure 6.1: Correlation Coefficient and Correlation Distance................................... 141
Figure 6.3: Plot for SU (L) and SU,AVG (L) model ................................................... 144
Figure 6.5: Plot for FSur (L) and FSur,avg (L) model ................................................... 150
Figure 6.8: Example-Existing Design Profiles for the P-J1 Data Set. ....................... 155
Figure 7.1: Uncertainty in Axial Capacity versus Correlation between FS (L) and
FB (L) ....................................................................................................... 162
Figure 7.3: Linear Relationship between Su1 and Lateral Capacity. ............................ 165
Figure 7.4: Example: Generic Side Capacity for a 90-ft Long, 15-ft Diameter
Caisson. ................................................................................................... 167
Figure 7.5: Example: Predicted Axial and Lower-Bound Axial Capacity (Generic)
for a 15-ft Diameter Caisson. .................................................................. 169
Figure 7.6: PDF and CDF for Generic Design Axial Capacity (L=90ft, D=15ft). .... 169
Figure 7.7: Plan View of the Site A(Numbers are Fs values in ksf for 90-ft
penetration).............................................................................................. 171
Figure 7.8: Reduction in Uncertainty versus Distance from the Existing Data Point.171
Figure 7.9: Example: Spatially Conditioned Side Capacity for a 90-ft Long, 15-ft
Diameter Caisson .................................................................................... 174
xxiii
Figure 7.10: Example: Extrapolated Axial Capacity at P-J1. ...................................... 175
Figure 7.11: PDF and CDF for Extrapolated Design Axial Capacity (L=90ft, D=15ft).176
Figure 8.2: Lower Bound and Median Axial Capacity (Generic). ............................ 182
Figure 8.3: Partial Safety Factor versus Spatial Variability in Capacity (R and S are
modeled with a lognormal distribution). ................................................. 184
Figure 8.4: Spatial Coefficient of Variation for Generic Axial Capacity. ................. 185
Figure 8.7: Design Axial Capacity versus Caisson Length for Generic Site (D=15ft)189
xxiv
Chapter 1: Introduction
As oil production has moved into deepwater (i.e., water depth greater than 3,000
feet), many challenges have arisen in designing foundations for offshore structures
(Figure 1.1). Geotechnical properties, particularly the shear strength of the soils, are the
primary quantities used for designing offshore foundations. Ideally, the design values of
geotechnical properties are determined by results from geotechnical investigation
programs, including soil borings, drop piston cores and cone penetration probes,
performed at the site of the foundation. However, the specific location of a foundation
may not be known until after the site investigation. Also, the footprint of a facility may be
large with numerous foundation elements spread out over miles (Figure 1.2). These
elements can include anchors for mooring systems, foundations for wells and well
manifolds, and foundation for pipelines and flow lines. Lastly, the cost and time required
to perform site investigations is significant in deepwater. Therefore, it is not generally
possible to perform a site-specific investigation for every foundation element in
deepwater.
The motivation of this research is to assess, analyze and model spatial variability
in geotechnical properties in a deepwater region. The potential contributions of this
research are: (1) to understand the magnitude of and sources of spatial variability in
geotechnical properties for foundations in a deepwater offshore region; (2) to be able to
use this understanding to design a foundation in this region to achieve a target reliability
without having a site-specific investigation; and (3) to assess the value of information to
decide whether obtaining additional geotechnical information at a site would be
worthwhile.
1
Figure 1.1: Deep-water Development Systems (Minerals Management Services,
www.mms.gov).
Figure 1.2: Floating Production Systems (taken from “Hoover Deep Draft Caisson
Vessel in Perspective” by Sandstrom, Slocum and Heideman with
ExxonMobil).
2
1.2 BACKGROUND
Oil exploration and production have moved into deeper water off the continental
shelf. It is difficult and comparatively expensive to perform geotechnical investigations
for a foundation design. Therefore, much work has been conducted previously to
understand spatial variability in geotechnical properties in order to achieve a more
accurate characterization of soils for offshore foundation design.
Quiros et al. (1983 and 2003), Young et al (2000) and Randolph (2004)
interpreted and compared geotechnical properties of soil in the Gulf of Mexico from
various geotechnical investigations and different sampling methods. They concluded that
the sampling technique and interpretation of test results have a significant influence on
having reliable strength data.
Much effort has been devoted to assess and model the magnitude and
characteristics of uncertainty regarding spatial variability in soil properties and capacities
for offshore foundations. Vladez-Liamas (2003) analyzed the spatial variability of the
thickness and water content of the soft clay layer and developed a model of spatial
variability. He emphasized the usefulness of geostatistics as a method for modeling the
spatial variability of soil properties. Gambino (1998), Al-Awar (2002) and Altuntas
(2002) developed geostatistical models for the spatial variability of axial and lateral pile
capacity for different offshore regions in relatively shallow water. They presented a
reliability-based design methodology for designing offshore pile foundations without a
site specific boring and provided design charts that relate the magnitude of uncertainty in
pile capacity to the design factors in the current design code. Wang (2002) studied
methods to calibrate geostatistical models with non-normally distributed and correlated
data points.
3
From a reliability perspective, a realistic model for the left-hand tail of the
capacity is important (Fjeld 1977). Najjar (2005) provided evidence for the existence of
the lower-bound capacity by analyzing the databases of load tests conducted on deep
foundations. He concluded a lower-bound capacity can cause a significant increase in the
calculated reliability for a geotechnical design. He suggested incorporating the lower-
bound capacity into the current design code.
1.3 OBJECTIVE
6
Chapter 2: Database
2.1 INTRODUCTION
Geotechnical and geological data for the study area (Figure 2.1) are compiled
from site investigations conducted for a variety of different projects over the past twenty
years. In this chapter, the source of the data, location and geologic setting of the region
are briefly described. Soil properties of the study area and test methods are then
summarized. Lastly, the design and contents of the database are described. The database
constructed in the chapter is used for the analyses of spatial variability in Chapter 4 and
for the model calibration process in Chapter 6.
7
2.2 SOURCE OF DATA
The data used in this dissertation came from a total of 63 engineering reports for
15 different offshore project sites. These reports were obtained in a digital document
format from a proprietary source. For some project sites, electric files containing the raw
data of cone penetration tests were also obtained. Information about the name and exact
location of project sites are confidential and are not presented in this dissertation.
A plan view of the study area in the Gulf of Mexico is shown in Figure 2.1.
Across the site, the seafloor slopes generally downward from the northwest to the south.
The average and maximum water depths of the Gulf of Mexico are about 5,000 feet and
13,000 feet, respectively (R. E. Turner, 1999). The water depths in the study area range
from 3,000 feet to 9,000 feet. Approximately one-half of the Gulf of Mexico is comprised
of shallow and inter continental shelf areas where the water depth is less than
approximately 650 feet (R.H. Gore, 1992) as shown in Figure 2.2. The continental shelf
is characterized by a gentle slope of less than one degree. The continental shelf width
ranges from 10 miles off the Mississippi River to 220 miles offshore west Florida (MMS,
2000) The continental slope and canyon area, where the most site investigations for this
study were performed, extends from the shelf edge at approximately 650 feet water depth
to an abrupt change in the bathymetric gradient at a water depth of approximately 10,000
ft The overall gradient of the continental slope is 3 to 6 degrees.
The shallow sub surface stratigraphy within the research area is inferred to be
primarily clays based on the Unified Soil Classification System with minor zones of sand
seams. The soil condition is normally to slightly over-consolidated, marine clay.
Typically, less than 20 feet of Holocene sediments (less than 12,000 years old) cover
8
several hundred feet (considering only layers investigated) of Pleistocene sediments
(greater than 12,000 years old) in the investigation locations. Holocene deposits are
composed of very soft, highly plastic clay having high water content. Pleistocene deposits
are comprised of very soft clay at the top of the Pleistocene deposit to stiff or very stiff at
depth. The upper part of soil layer in the study area (from the mudline to a depth of about
50 to 60 feet) consists of hemipelagic sediments which were deposited by relatively high
sea level and shows high variation of its soil properties (Flemings et al. 2006). Thereafter,
the soil layer is composed of turbitites whose properties may vary with distance from the
shoreline at the low sea level (Figure 2.1, Anderson et al. 2004). Numerous shallow faults
and some landslide scars are present in the study area. All landslide deposits seen in the
study area were buried by Holocene deposits and are believed to have occurred in the late
Pleistocene (greater than 12,000 years ago). Shallow faults and buried landslides are
detected by geophysical surveys and are thus generally to be avoided.
Figure 2.2: Major physiogeographic features of the Gulf of Mexico outer continental
shelf (MMS, 2000).
9
2.4 DESCRIPTION OF DATA
The stratigraphy and properties of soil are obtained from the field investigations
and the laboratory tests. In this section, the testing methods that were used in the
geotechnical reports are described
For soil borings, the penetration depth ranges from 50 to 1,300 feet below the
mudline and the average depth is on the order of 200 feet below the mudline. Two types
of soil sampling techniques were used for soil borings: driven and pushed (ASTM
D1587-08, D 6282-98). Typical driven samples were obtained in 2.250-inch outside and
2.125-inch inside diameter thin wall tubes that were driven by a 175-pound sliding
weight operated by a wire-line. The weight was dropped approximately 5 feet a sufficient
number of times to drive the sampler the desired 24 inches. Typical pushed samples were
retrieved by advancing 3.00 inch outside and 2.83 inch inside diameter about 24 inches
into the ground. For the very soft to soft clay near mudline, either 2.00-inch outside and
1.875-inch inside diameter or 2.50-inch outside and 2.125 inch liner sampler were used.
Immediately after the soil sampling operations, the remote vane tests were performed in
the soil borings to measure the in situ shear strengths of the soil. The vane shear test
involves placing a four-bladed vane (Figure 2.3 (a)) into the soil at the desired depth,
typically 5 feet deeper than the previous sampling depth, and rotating it to determine the
torque required to shear a cylindrical surface with the vane. This torque is then converted
to the unit shearing resistance of the failure surface by limit equilibrium analysis (ASTM
D6282-98).
For jumbo piston cores (JPC), the penetration depth ranges from 45 to 68 feet
below the mudline. Jumbo piston core samples were 4 inches in diameter with about 45
to 65-ft soil samples recovered using a jumbo piston core barrel which typically consisted
10
of 5.25 inch outside and 4.625 inch inside diameter well casing and the PVC pipe as a
sample liner. This large diameter of the core barrel allowed us to obtain a high quality
sample by minimizing edge effects and sample disturbance (ASTM D 4823 – 95).
Thirty-four cone penetration tests (CPT) were performed in the study area with
the penetration depths ranging from 105 to 135 feet below the mudline. From the cone
penetration tests, continuous data were retrieved with depth by pushing the cone into the
soil at a steady rate (2 cm/second or 0.8 in./second, ASTM D 5778-07). A cone tip has an
apex angle of 60 degrees with a cross-sectional area of either 10 or 15 cm² (Figure 2.3
(b)). The undrained shear strength of soil was estimated by dividing the net cone tip
resistance, q net , by the cone resistance factor, N kt , (Schmertmann, 1978). The cone
factor, N kt , is normally determined by correlating q net with the specific type of
measured field and laboratory strength tests (e.g., miniature vane, undrained
unconsolidated triaxial and remote vane tests).
The undrained shear strength of the clay samples was determined from the
laboratory tests: Torvane, pocket penetrometer, miniature vane, undrained unconsolidated
(UU) triaxial tests, and the direct simple shear (DSS) tests.
The Torvane is a hand-held vane shear device for rapid determination of shear
strength in cohesive soils either in the laboratory or the field (Figure 2.4). The Torvane
consists of a vane with blades on the lower surface, which are pressed into the soil to be
tested. When the upper knob is rotated with finger pressure, the knob is slowly turned at a
constant rate to provide torque on the vane until the soil is sheared. A scale on the knob
reads the approximate ultimate shear strength of the sample.
The pocket penetrometer is a small, hand-held device that is spring loaded (Figure
2.5). The pocket penetrometer consists of a loading piston and a spring in a cylindrical
body. The flat-faced piston is pushed against of a soil surface until the soil experiences a
11
punching type bearing failure. The penetration resistance read on the side of the device
provides the approximate ultimate compressive strength.
(a) (b)
Figure 2.3: Cone penetrometer and Remote vane (images from Fugro-McCLELLAND,
2006).
12
Figure 2.4: Torvane Tester Set (image taken from www.humboldtmfg.com)
13
The miniature vane is a small, four-bladed device used to determine rapidly the
shear strength on undisturbed or remolded soils. The test procedure based on ASTM
D4648–05 is as follows: “The miniature vane shear test consists of inserting a four-
bladed vane in the end of an undisturbed tube sample or remolded sample and rotating it
at a constant rate to determine the torque required to cause a cylindrical surface to be
sheared by the vane. This torque is then converted to a unit shearing resistance of the
cylindrical surface area. The torque is measured by a calibrated torque spring or torque
transducer that is attached directly to the vane.” The miniature vane tests were performed
with the combination of two blade sizes and four different calibrated torque springs.
The unconsolidated-undrained (UU) triaxial test is used to determine the strength
and stress-strain relationships of a cylindrical specimen of either undisturbed from
pushed sampling or remolded clay. Specimens are enclosed in a thin rubber membrane
and are subjected to a confining fluid pressure in a triaxial chamber. The specimen is not
allowed to consolidate under the confining pressure. The specimen is sheared in
compression without drainage at a constant rate of axial deformation. The undrained
shear strength of clay is determined in terms of the total stress as one-half the maximum
observed stress. The UU triaxial tests were performed according to ASTM D2850.
The direct simple shear (DSS) test is used to measure soil strengths under the
constant volume condition which is equivalent to the undrained condition. The soil
specimens are loaded axially and allowed to consolidate one-dimensionally. The
specimens are then sheared at a constant rate of deformation and the resulting shear
forces are measured. The DSS tests were performed according to ASTM D 6258.
Submerged unit weights are measured for each extruded soil sample, including
each UU triaxial compression test specimen, by weight a soil sample of known volume.
14
2.4.2 Measured geotechnical data
This section discusses summary information for the measured data. The primary
data includes water content, plasticity, unit weight, and undrained shear strength.
The water content is in excess of 100 percent in the upper about 8 to 20 feet of
sediments. Below this high water content layer, the water content gradually decreases to
about 50 percent at about 60 to 70 feet of penetration. Thereafter, the water contents
decrease and fluctuate between about 35 and 70 percent (Figure 2.6).
The plastic index is a measure of the plasticity of a soil. The plastic index is the
difference between the liquid limit and the plastic limit as follows:
PI = LL - PL (2.1)
where PI is the plastic index in percent, LL is the liquid limit in percent, and PL is the
plastic limit in percent. The soil in the study area is classified as a highly plastic clay
(CH) according to the Unified Soil Classification System and the plasticity chart is shown
in Figure 2.7. The liquidity index (LI) is used for scaling the natural water content of a
soil sample to its Atterberg limits and is expressed as:
ω − PL ω − PL
LI = = (2.2)
LL − PL PI
15
Figure 2.6: Water Content.
The submerged unit weight was measured for samples in the laboratory tests by
subtracting the density of seawater from the measured total unit weight. In many cases,
sample expansion was found, which reduces the apparent unit weight of the samples. In
order to account for the effects of sample expansion, submerged unit weights are
calculated from natural water contents and specific gravity using the following equation:
ω
G s (1 + )
γ ' = γw 100 − γ (2.3)
ω w,sea
1 + Gs
100
where γ ' is the computed unit weight in pcf, γ w is the unit weight of fresh water (62.4
pcf), γ w,sea is the unit weight of sea water (64.0 pcf), ω is the natural water content of
soil sample in percent, and G s is the average measured specific gravity of soil solids.
The submerged unit weight for Holocene deposits is relatively constant with depth,
ranging from about 15 to 25 pcf. The average value of the submerged unit weight is about
40 pcf for the depth ranging 0 to 200 feet from the mudline.
The undrained shear strengths of the clays were evaluated by in situ tests such as
remote vane and cone penetration tests, and conventional laboratory tests such as torvane,
miniature vane and undrained unconsolidated (UU) triaxial tests. The measured
undrained shear strengths versus depth from all test results are shown in Figure 2.8. The
lines of 5 psf/ft and 14 psf/ft are inserted as the lower and upper bound for the increasing
rate of undrained remolded shear strength. Typically, undrained shear strengths from the
miniature vane and the undrained unconsolidated (UU) triaxial tests are used as the basis
17
when the undrained shear strength is interpreted and the design profile is selected.
Generally the miniature vane tests yield slightly higher shear strength values than the UU
triaxial tests because the miniature vane tests are performed on the samples before
extrusion from the sample tube and the UU tests are performed on the samples after
extrusion. There might be a sample disturbance due to stress relief.
In this study, the cone factors ( N kt ) in each design report, which is constant with
depth, were used to determine undrained shear strength from the net cone tip resistance.
In the case that there was no information on the constant cone factor, the N kt value of
16 was used, which is the typical value for normally consolidated clay in the Gulf of
Mexico area. The average value of N kt is 17.5 and the number of use for N kt is shown
in Figure 2.9.
The undrained remolded shear strengths of the clays were evaluated by both
miniature vane and undrained unconsolidated (UU) triaxial tests. The undrained
remolded shear strengths versus depth from all test results are shown in Figure 2.10. The
lines of 1.5 psf/ft and 5 psf/ft are inserted as the lower and upper bound for the increasing
rate of undrained remolded shear strength.
Soil sensitivity is defined as the ratio of peak undrained shear strength to
remolded undrained shear strength with no change of water content on the samples. The
remolded and undisturbed shear strengths from both miniature vane and UU triaxial test
were used to evaluate soil sensitivity. The evaluated soil sensitivity values in this study
area range between about 2 to 8. For the highly-plasticity clays with the PI great than 80
percent, the values of soil sensitivity range between 2 to 4 in the study area, which is
typical for the clays in the Gulf of Mexico.
18
Figure 2.8: Undrained Shear Strength versus Depth
16
Total number of CPT performed =34
14 13
12
12 Average
Nkt =17.4
10
No. of use
8
8
2 1
0
16 16.5 17 17.5 18 18.5 19
Nkt
19
Figure 2.10: Undrained Remolded Shear Strengths versus Depth
20
2.4.3 Design profiles for undrained shear strength
Offshore foundations are designed using profiles of unit weight and shear strength
that are developed by a geotechnical engineer based on the measured data and the
stratigraphy. Figure 2.12 shows examples of design profiles for undrained shear strength
for one data set. The solid and dotted lines represent the design profiles for undrained
strength and undrained remolded strength, respectively. The in situ and laboratory test
results are also shown in the same figure for comparison. The design properties are
selected based on several factors such as the measured test results, the empirical
relationships, and the engineers’ judgment.
A total of 97 design undrained shear strength profiles, 61 design undrained
remolded shear strength profiles, 97 submerged unit weight profiles were selected in this
manner. For 36 data set locations, the design remolded strength profile was not selected
or recommended, where no test for remolded shear strength was performed. Figure 2.13
and Figure 2.14 show the design undrained undisturbed shear strength and the design
undrained remolded shear strength versus depth, respectively. The design undisturbed
profiles range from a linear profile of about 7 psf/ft to 10 psf/ft and the design remolded
profiles range from from a linear profile of about 2 psf/ft to 4 psf/ft. (The design
undrained shear strength at the mudline is small and assumed to be zero for simplicity for
linear profiles.) The variation in design values of undrained shear strength and undrained
remolded shear strength are analyzed in detail in Chapter 3. The design submerged unit
weight of soil with depth is shown in Figure 2.15 with a profile of average value. The
average value of the design submerged unit weight is about 40 pcf for depths of 0 to 200
ft.
21
Shear Strength (ksf)
Torvane
Penetrometer
Minivane
100
UU
DSS
Design-undisturved
UU remolded
Depth (ft)
Design-Remolded
300
400
500
Figure 2.12: Example-Design Profiles for a Data Set with Test Results
22
Design Undrained Shear Strength of Soil (ksf)
50
Depth(ft)
100
Average
150
10 psf/ft
7 psf/ft
200
50
Depth(ft)
100
Average
4 psf/ft
150
2 psf/ft
200
23
Design Submerged Unit Weight of Soil (kcf)
0.00 0.02 0.04 0.06 0.08 0.10
0
50
Depth (ft)
100
Average
150
200
24
2.5 DATABASE
25
Figure 2.16: Overview of Location of the Data Sets.
26
examples of charts that show laboratory and field shear strength test results with the
design profiles for one of the data sets.
The second function of a data workbook is to present the compiled data. For
example, in the “B-02 Design shear strength plots” folder, shown in Figure 2.17, the
charts of the shear strength data for all 97 data sets are collected in one worksheet, as
shown in Figure 2.21. In addition, the charts of the design shear strength profiles are
collected by group, with linear shear strength profiles of 3 and 10 psf per foot shown for
reference in Figure 2.22. These collections facilitate ready comparisons of shear strength
profiles among data sets.
27
Figure 2.18: Example of Data Workbook-Test Results.
DSS
40
Design-undisturved
Depth (ft)
UU remolded
Depth (ft)
Design-Remolded 60
300
80
400
100
500 120
(a) (b)
29
Figure 2.21: Example of Shear Strength Charts Collected in One Place.
30
Figure 2.22: Example of Charts for Design Shear Strength Collected in One Place.
31
Table 2.2: Summary of Database.
32
required for calculations are stored in an input worksheet. Calculation and data analysis
are performed in a calculation worksheet that contains formulas, functions and macros.
Results of calculations and analyses are collected and presented in an output worksheet.
An output worksheet can be queried or exported for further analysis and visualization.
Figure 2.23 shows an example of a calculation workbook that is for calculation of generic
and spatially conditioned values of undrained shear strength. There are two input
worksheets and one calculation worksheet in this example workbook. The input
worksheets contain the data required for calculations such as Su values with the three-
33
Chapter 3: Undrained Shear Strength and Deep Foundations Design
3.1 INTRODUCTION
This chapter describes how the undrained shear strength is used for design of a
deep foundation. Deep foundations support facilities, including float production systems,
well systems, and pipeline systems. The most common foundations are driven steel pipe
piles and suctions caissons, which are jacked into the sea floor with under pressure.
Suction caissons are the most prevalent and are therefore the focus of this study.
However the design methodology is similar for all deep foundations in offshore
application.
Typical diameters of current caissons are 12 to 25 ft and length to diameter ratios
are about six. The loading condition of a suction caisson varies from horizontal to vertical
depending on the type of mooring system.
Axial capacity of suction caissons in clay is estimated using the API (2003)
design guide. Axial capacity of a suction caisson due to soil is composed of two
components: side shear capacity, QS and reverse end bearing, Q P . Axial capacity of
This formulation is based on a limit equilibrium analysis of the caisson (Figure 3.1). In
this study, spatial variability in the design axial capacity only due to soil is analyzed;
34
however. the submerged weight of the caisson is added to the right side of Equation (3.1)
to calculate the total axial capacity.
The axial side shear capacity is calculated as a function of average unit side shear
over the caisson length:
L L
1 1
QS = AS FS = AS (
L0 ∫
fs(z) dz) = AS (
L0∫α(z)s u (z) dz) = ASα avgSu,avg (L) (3.2)
where FS is the average unit side shear, AS (= π DL) is side surface area of caisson, z is
the depth of penetration and fS (z) is unit side shear as a function of depth, z. L is the
length of a caisson and D is the diameter of a caisson. The design axial side shear is
proportional to the depth-averaged design shear strength, Su,avg (L) , which is described
and analyzed in Chapter 4. The factor α(z) is a dimensionless friction coefficient that
depends on Su /σ 'v and is less than or equal to 1.0. A depth-averaged α value, α avg , is
assumed here to be 0.8 throughout the depth of penetration, which is the typical value
used in design of suction caissons for normally consolidated clay in the Gulf of Mexico.
Figure 3.1: Load and Resistance Component of a Suction Caisson for Axial Loading.
35
The reverse end bearing is developed at the tip of caisson under undrained loading
conditions. It is assumed that a tensile failure occurs at the bottom of the soil plug and
that the plug will come out with the caisson. The reverse end bearing is directly
proportional to undrained strength and formulated as:
Q p = A p q b = A p (N cSu ) (3.3)
where q b is the unit reverse end bearing of caisson and A p (= π D 2 / 4) is the cross-
sectional area of the caisson tip. A bearing capacity factor of 9 is adopted here, as
recommended in API (2003). The design end bearing capacity is proportional to the
design undrained shear strength at a point, Su (z) , which is described and analyzed in
Chapter 4.
The design lower-bound axial capacity for a suction caisson in normally and
slightly over-consolidated clay is estimated in a similar manner to the design axial
capacity by using the design remolded shear strength of the clay with an α value of 1.0
and the bearing capacity factor, N C , of 9.0 (Najjar, 2005). The design remolded shear
strength at a point, Sur (z) and the depth-averaged design remolded strength, Sur,avg (L)
The plasticity model initially proposed by Murff and Hamilton (1993) was
adopted to estimate the lateral capacity of a suction caisson under undrained conditions.
This model is based on a three-dimensional collapse mechanism. This mechanism
36
combines a deforming conical soil wedge in the near surface with plane strain
deformation below the conical wedge, and a hemispherical failure surface around the tip
(Figure 3.2(a)). This model was further simplified by Aubeny et al (2001 and 2003) as
shown in Figures 3.2(b) and 3.3.
Figure 3.2: Plasticity model (a) failure mechanism (Murff and Hamilton, 1993, and (b)
simplified analysis ( Aubeny et al., 2001 and 2003)
37
For the laterally loaded suction caisson (Figure 3.3), the internal rate of energy
dissipation, E , at any point along the caisson is simply the product of the mobilized
pressure ( P ) times the velocity at the point in question times the elemental area. The total
rate of energy dissipation due to side resistance is as follows:
⋅ L z
Ds = ν o D ∫0
1−
L0
(P + γ 'z)dz (3.4)
P = N pSu (3.5)
where P is the ultimate average unit lateral pressure (Matlock, 1970; Reese et al., 1975),
N p is a dimensionless bearing factor, ν o is the velocity at the mudline, D is the caisson
diameter, L is the depth of penetration, and Su is the undrained strength of soil which is
where R is the radius of suction caisson. R1 is the distance from center of rotation to
caisson tip, R 2 is the radius of spherical surface, φ is the angular coordinate about
caisson centerline in horizontal plane, ω is the angular coordinate from the caisson
centerline, and Su is the linear profile of undrained strength of soil.
where Li is the depth of load attachment. In the simplified model, the lateral capacity
for a given load attachment point is obtained by minimizing F with respect to the center
of loading. The simplified plasticity model has been validated by comparing it with
numerical solutions and experimental tests (Aubeny et al., 2001, and Clukey and
Plhillips, 2003).
One of the limitations of this model is that only a linearly increasing shear
strength profile and an average value of unit weight can be used. A spreadsheet program
that uses nonlinear profiles of design shear strength and unit weight was developed for
this research. This program is presented in detail in Appendix. An advantage of using this
spreadsheet program is that input values such as intercepts and slopes of shear strength
profiles and average unit weights do not required reevaluation whenever a new design
depth of suction caisson is applied. Using this spreadsheet program the equivalent linear
profile of design undrained shear strength can be evaluated, which gives the same lateral
capacity as the design undrained shear strength.
39
3.4 TOTAL DESIGN CAPACITY OF A SUCTION CAISSON UNDER COMBINED LOADING
Depending on the types of mooring system attaching the suction caissons to the
structure (Figure 1.1), suction caissons carry various loading conditions from horizontal
to vertical. The capacity of suction caissons subjected to inclined load is discussed in this
section. Aubeny et al. (2003b) extended their work on lateral capacity to general
conditions of combined loading (Figure 3.4). They used a simplified upper bound
analyses to determine the estimate of capacity by equating the rate of internal energy
dissipation due to side and bottom resistance to the rate of work due to an externally
applied load. The capacity of caisson is obtained by minimizing F with respect to the
optimization parameters, ξ and L0 , where ξ is a parameter relating the magnitude of
vertical to horizontal motion ,and L0 is the center of the rotation. The most recent
Figure 3.4: Suction Caisson under Inclined Loading Condition (Aubeny et al.,2003).
40
The optimum depth of load attachment for the ultimate capacity is a depth at
about two-thirds the length of the caisson, called the 2/3 point (Aubeny et al., 2003 and
Choi, 2007). In this study, a 2/3 point was used for the load attachment depth. One
example of interaction curves for the combined capacities for suction caissons is shown
in Figure 3.5. There are three zones of interaction between the axial and the lateral
capacities that cause failure. In the “axial failure” and “lateral failure” zones, an axial and
a lateral load, respectively, controls the failure. In the “combined failure” zone, the
failure mode is more complicated. The interaction between the axial and lateral capacities
of suction caisson results in the failure of the suction caisson.
The parameter study associated with the friction coefficient, α the bearing
factor, N C , and the rate of strength increase per unit depth, Su,linear , was performed and
is presented in Figures 3.6 to 3.8. The aspect ratio of 6 (i.e., L/D=6) was used for this
parameter study. As expected, the combined capacity increases with an increase in
parameters, α , N C and Su1 . The friction coefficient, α has a similar effect on both
axial and lateral capacities (Figure 3.6), while the bearing capacity factor, N C , barely
affects the lateral capacity (Figure 3.7). The rate of strength increase per unit depth, Su1 ,
has more influence on the lateral capacity than on the axial capacity (Figure 3.8).
The combined capacity with various aspect ratios ranging from 4 to 9 was
evaluated and is presented in Figure 3.9. Each point in Figure 3.9 represents the ultimate
capacity at the padeye corresponding to angles of load inclination of 0, 5, 10, 15, 20, 25,
30, 35, 40, 50, 60, 70, 80, and 90 degrees from the horizontal. These interaction diagrams
are normalized by the axial ultimate capacity ( Vmax ) and the lateral ultimate capacity
( H max ) and shown in Figure 3.10.
41
In order to account for nonlinear variation in design undrained shear strength
versus depth, the following approach is proposed to estimate total capacity of a suction
caisson under combined loading:
1. Determine the profiles of the design undrained shear strength at a point, Su (z) , the
depth-averaged design shear strength, , Su,avg (L) and the slope of the equivalent
and the ultimate design lateral capacity ( H max ) using either Su (z) or Su1 (L) as
3.5 SUMMARY
The description of methods for estimating caisson capacity including design axial
capacity, design lateral capacity and total capacity of a caisson is included in this chapter.
The design ultimate axial capacity can be evaluated from the design undrained shear
strength at a point and the depth-average design undrained shear strength. The design
lower-bound axial capacity can be evaluated using the design undrained remolded
strength at a point and depth-averaged design remolded strength. The design lateral
ultimate capacity can be evaluated using the equivalent linear profile of design shear
strength. An approach to estimate total capacity is proposed.
42
5000
D=15 ft
4000 L/D=6
Nc=9
Axiyal capacity, V (kips)
α=0.8
3000 Axial Failure Su1=8 psf/ft
Su0=0 psf
Combined Failure
ψ=90° ψ=70° ψ=50° ψ=40° ψ=30°
2000
ψ=20°
1000
ψ=10°
Lateral Failure
0 ψ=0°
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Figure 3.5: Example of Interaction Curve for Combined Capacity of Suction Caissons.
5000
alpha=0.6
ψ=70°
ψ=50° alpha=0.7
4000 ψ=40°
alpha=0.8
ψ=30°
Axiyal capacity, V (kips)
alpha=0.9
3000 alpha=1
ψ=20°
2000
ψ=10°
1000
0
0 1000 2000 3000 4000 5000 6000 7000 8000
Latral capacity, H (kips)
Figure 3.6: Effect of the α Factor on Combined Capacity for a Suction Caisson with
L/D=6, N C =9 and D=15 feet.
43
5000
Nc=6
ψ=70° ψ=50° ψ=40°
Nc=9
Nc=12
4000
Nc=15
ψ=30°
Axiyal capacity, V (kips)
Nc=18
3000
ψ=20°
2000
ψ=10°
1000
0
0 1000 2000 3000 4000 5000 6000 7000 8000
Latral capacity, H (kips)
Figure 3.7: Effect of the End Bearing Factor on Combined Capacity for a Suction
Caisson with L/D=6, α =0.8 and D=15 feet.
8000
Su1=6psf/ft
Su1=7psf/ft
Su1=8psf/ft
6000 ψ=70° ψ=50° Su1=9psf/ft
Axiyal capacity, V (kips)
ψ=40° Su1=10psf/ft
ψ=30° Su1=11psf/ft
4000
ψ=20°
2000 ψ=10°
0
0 2000 4000 6000 8000 10000 12000
Figure 3.8: Effect of Su1 on Combined Capacity for a Suction Caisson with L/D=6,
α =0.8 , N C =9 and D=15 feet
44
10000
L/D=4
L/D=5
8000 L/D=6
ψ=70°
ψ=50° L/D=7
ψ=40°
Axiyal capacity(kips)
L/D=8
6000 ψ=30°
L/D=9
ψ=20°
4000
ψ=10°
2000
0
0 2000 4000 6000 8000 10000 12000 14000 16000
Latral capacity(kips)
Figure 3.9: Combined Capacity for Suction Caissons with Various L/D.
1.2
L/D=4
L/D=5
L/D=6
1 L/D=7
L/D=8
L/D=9
0.8
V / Vmax 0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1 1.2
H / Hmax
Figure 3.10: Normalized Combined Capacity for Suction caissons with Various L/D.
45
Chapter 4: Analysis of Spatial Variability in Design Undrained Shear
Strength
4.1 INTRODUCTION
In this study, three metrics are chosen as a way of representing the undrained
undisturbed shear strength of the soil layer: (1) the design undrained shear strength at a
point, Su (z) where Su is design undrained shear strength and z is depth below
mudline; (2) the depth-averaged design undrained shear strength at the depth of
penetration, Su,avg (L) where L is the penetration depth; and (3) the equivalent linear
design undrained shear strength at the depth of penetration, Su,linear (L) . These three
metrics are selected because, in designing foundations, each component of the total
capacity of a foundation is related to these metrics. The end bearing capacity and the side
shear capacity under the axial loading condition are approximately proportional to Su (z)
46
and Su,avg (L) , respectively. The lateral capacity is approximately proportional to
Su,linear (L) .
Based on Su (z) , the depth-averaged design undrained shear strength at the depth
of penetration, Su,avg (L) , is calculated as follows:
L
1
L ∫0
Su,avg (L)= Su (z)dz (4.1)
where L is the depth of penetration. For example, if the values of the design undrained
shear strength in 10-foot intervals down to a depth of 100 feet is given as shown in Figure
4.1(a), the depth-averaged design undrained shear strength at the depth of 30ft below the
mudline can be calculated as:
Su(psf) Su_avg(psf)
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
0 65 0 0
111 88
20 195 20 121
286 161
40 367 40 202
Depth (ft)
Depth (ft)
447 243
60 528 60 284
906 346
80 961 80 419
1017 482
(a) (b)
47
The values of depth-averaged design shear strength in 10-foot intervals down to a depth
of 100 feet are calculated as the same manner and shown in Figure 4.1(b). While Su (z)
is the strength value at each depth, z, Su,avg (L) reflects the soil strengths from the
mudline down to the depth of penetration, L. If there is a strength jump between the
depths of 60 and 70 feet in the design undrained shear strength profile as shown in Figure
4.1(a), this strength jump is reflected in the depth–averaged shear strength at the depth of
70 feet or deeper. However, no sudden increase is found in the depth-averaged design
shear strength profile due to the averaging effect (Figure 4.1(b)).
The linear undrained shear strength profile with the intercept and slope are
commonly used to evaluate a lateral capacity of an offshore foundation. The design
undrained shear strength with depth Su (z) is replaced with the equivalent linear design
shear strength that gives the same lateral capacity with respect to the depth of penetration
(Figure 4.2). The equivalent linear shear strength is designated as follows:
where L is the depth of penetration, Su0 is the undrained shear strength at the mudline
and Su1 is the rate of strength increase per unit depth.
(a) Design undrained shear strength profile (b) Equivalent linear profile
depths of penetration ranging from 50 feet to 150 feet is calculated for all data sets. The
values of Su1 are analyzed and summarized in Section 4.3.1.
Two more metrics are chosen to represent the undrained remolded shear strength
of the soil layer: (1) the design remolded shear strength at a point, Sur (z) and (2) the
depth-averaged design remolded shear strength, Sur,avg (L) . These two metrics are similar
to those described in the previous section; however, the design remolded strengths are
used instead of the design undisturbed strengths. These metrics are chosen because they
can be used to evaluate the lower-bound capacity of foundations. The lower-bound end
bearing capacity and the lower-bound side shear capacity under the axial loading
condition are approximately proportional to Sur (z) and Sur,avg (L) , respectively.
Based on Sur (z) , the depth-averaged design undrained shear strength, Sur,avg (L) ,
is calculated as follows:
L
1
L ∫0
Sur,avg (L)= Sur (z)dz (4.3)
49
4.3 SAMPLE STATISTICS FOR METRICS
4.3.1 Sample statistics for Su (z) , Su,avg (L) and Su,linear (L)
In Section 3.2.1, three metrics related to the design undrained undisturbed shear
strength, Su (z) , Su,avg (L) and Su,linear (L) , are introduced. In this section, Su (z) ,
Su,avg (L) and Su1, (L) for the data sets in the database are analyzed and summarized. In
Figure 4.3, Su (z) and Su,avg (L) for 97 data sets are plotted as a function of depth in one
tenth-foot intervals. For a given depth at a given location, each data point in Figure 4.3
(b) represents the average of all the design undrained shear strength values down to the
depth.
The probability distribution for a random variable (e.g., Su (z) , Su,avg (L) or
Su1 (L) ) can be quantitatively described in terms of the mean and the standard deviation.
The mean represents the expected value of a random variable. The standard deviation
represents the degree of dispersions from the mean value. The mean and the standard
deviation of Su (z) and Su,avg (L) are plotted as a function of depth in one tenth-foot
intervals in Figure 4.4 and Figure 4.6, respectively. While the magnitude of the mean and
the standard deviation for Su (z) and Su,avg (L) increases with depth, the ratio of the
standard deviation to the mean decreases with increasing depth. This ratio designates the
amount of relative variability and is called as the coefficient of variation (c.o.v). The
c.o.v. values for Su (z) and Su,avg (L) decrease from the mudline to a depth of 60 feet
and then become roughly constant below as shown in Figure 4.5 and Figure 4.7,
respectively. For the depth-averaged design undrained strength, Su,avg (L) , these results
are expected due to the averaging effect; the deeper the depth, the higher the
compensation effect for layers with below and above average Su,avg (L) . The coefficients
of variation (c.o.v.) at depths from 70 feet to 110 feet are a bit larger than the coefficients
50
of variation around a depth of 60 feet for both Su (z) and Su,avg (L) . These larger
coefficients of variation result from large variability in the design undrained shear
strength at between 70 feet and 110 feet depth below the mudline due to the appearance
of a stiffer layer in many of the data sets (e.g., Figure 4.3 (a)). It is expected that there is
greater variation in design shear strengths in the upper part of the soil layer than that in
the deeper layer due to the depositional history. The hemipelagic sediments in about the
upper 60 feet were deposited by the relatively high sea level and have higher variation in
soil properties (Flemings et al. 2006) than those for the turbitites sediments in the below
about 60 feet. The sample statistics for Su (z) and Su,avg (L) are summarized in Table
and shown in Figures 4.8 and 4.9 and Tables 4.1 and 4.2. A percentile value is the value
of a random variable below which a certain percent of observation is found. For example,
5-percentile value for Su (z) at a depth of 50 feet is 0.307 ksf (Table 4.1) At a depth of 50
feet below mudline, 5 % of the Su (z) values are less than and equal to 0.307 ksf (i.e., 95-
percentile values of the Su (at 50ft) values are greater 0.307 ksf).
51
Figure 4.3: Profiles of Su (z) and Su,avg (L) in One Tenth-Foot Intervals.
52
Design undrained shear strength, Su (ksf) Coefficient of variation, c.o.v.
0.0 0.3 0.6 0.9 1.2 1.5 0.0 0.1 0.2 0.3 0.4 0.5
0 0
Depth below the mudline (ft)
100 100
Sample mean
150 150
Figure 4.4: Sample Mean and Standard Deviation for Su (z) Figure 4.5: Coefficient of Variation for Su (z)
53
Depth-averaged design shear strength, Su,avg (ksf) Coefficient of variation, c.o.v.
0.0 0.3 0.6 0.9 1.2 1.5 0.0 0.1 0.2 0.3 0.4 0.5
0 0
Depth below the mudline (ft)
Sample mean
100 100
150 150
Figure 4.6: Sample Mean and Standard Deviation for Su,avg (z) Figure 4.7: Coefficient of Variation for Su,avg (z)
54
Table 4.1: Sample Statistics for the Design Undrained Shear Strength, Su (z) .
Table 4.2: Sample Statistics for the Depth-Averaged Design Shear Strength, Su,avg (L) .
55
Design undrained shear strength, Su (ksf) Depth-averaged design shear strength, Su,avg (ksf)
0.0 0.3 0.6 0.9 1.2 1.5 0.0 0.2 0.4 0.6 0.8
0 0
100 100
95%
95%
90%
90%
5% 10% 5% 10%
150 150
Figure 4.8: Sample Mean with 5-, 10-, 90-, and 95- Figure 4.9: Sample Mean with 5-, 10-, 90-, and 95-
Percentile Values for Su (z) . Percentile Values for Su,avg (L) .
56
The mean and the standard deviation of Su1 (L) and the 5-, 10-, 90-, and 95-
percentile values are presented in Table 4.3. The c.o.v. values for Su1 (L) slightly
decrease with increasing the depth of penetration. The equivalent linear profiles of the
design undrained shear strength for the depths of penetration of 60, 90, 120 and 150 feet
are shown in. The mean value of Su1 (L) increase with depth as shown in Figure 4.10.
Table 4.3: Sample Statistics for the Slope of the Equivalent Linear Undrained Shear
Strength Profile, Su1 (L) .
57
Su (ksf)
S u0=0
30
S u1=8.11 psf/ft
Depth of Penetration (ft)
for L=60
60
S u1=8.40 psf/ft
for L=90
90
S u1=8.79 psf/ft
for L=120
120
S u1=9.29 psf/ft
for L=150
150
58
4.3.2 Sample statistics for Sur (z) and Sur,avg (L)
In Section 4.2.2, two metrics related to the design undrained remolded shear
strength, Sur (z) and Sur,avg (L) , are introduced. In this section, Sur (z) and Sur,avg (L)
for the data sets in the database are analyzed and summarized.
The mean and the standard deviation of Sur (z) and Sur,avg (L) are plotted as a
function of depth in one-tenth-foot intervals in Figure 4.11 and Figure 4.13, respectively.
While the magnitude of the mean and the standard deviation for Sur (z) and Sur,avg (L)
increase with depth, the coefficients of variation for Su (z) and Su,avg (L) decrease with
increasing depth as shown in Figure 4.12 and Figure 4.14, respectively. The sample
statistics for Sur (z) and Sur,avg (L) are summarized in Table 4.4 and Table 4.5,
respectively.
The 5-, 10-, 90-, and 95-percentile values of Sur (z) and Sur,avg (L) are
determined and shown in Figures 4.15 and 4.16 and Tables 4.4 and 4.5. Figure 4.17
shows the comparison between the mean values of Su (z) and Sur (z) , while Figure 4.18
shows the comparison between the mean values of Su,avg (L) and Sur,avg (L) . It is
interesting that the 5-percentile values for Su (z) and Su,avg (L) is much greater than the
95-percentile value for Sur (z) and Sur,avg (L) , respectively. The design undrained
remolded shear strengths, Sur (z) and Sur,avg (L) , provide a bound on variability in the
design undrained undisturbed shear strength, Su (z) and Su,avg (L) .
59
Design remolded shear strength, Sur (ksf) Coefficient of variation, c.o.v.
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.0 0.1 0.2 0.3 0.4 0.5
0 0
Sample mean
100 100
150 150
Figure 4.11: Sample Mean and Standard Deviation for Sur (z) Figure 4.12: Coefficient of Variation for Sur (z)
60
Depth-averaged design remolded shear strength, Coefficient of variation, c.o.v.
Sur,avg (ksf)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.0 0.1 0.2 0.3 0.4 0.5
0 0
Depth below the mudline (ft)
Sample mean
100 100
150 150
Figure 4.13: Sample Mean and Standard Deviation for Sur,avg (z) Figure 4.14: Coefficient of Variation for Sur,avg (z)
61
Table 4.4: Sample Statistics for the Design Undrained Remolded Shear Strength,
Sur (z) .
Table 4.5: Sample Statistics for the Depth-Averaged Design Undrained Remolded
Shear Strength, Sur,avg (z) .
62
Design remolded shear strength, Sur (ksf)
Depth-averaged design remolded shear strength,
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Sur,avg (ksf)
0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
0
Sample mean
Sample mean
Depth below the mudline (ft)
100
100
150 150
Figure 4.15: Sample Mean with 5-, 10-, 90-, and 95- Figure 4.16: Sample Mean with 5-, 10-, 90-, and 95-
Percentile Values for Sur (z) . Percentile Values for Sur,avg (L) .
63
Su and Sur (ksf) Su,avg and Sur,avg (ksf)
0.0 0.3 0.6 0.9 1.2 1.5 0.0 0.2 0.4 0.6 0.8
Depth below the mudline (ft) 0 0
Sample Sample
mean mean for
100 for Su 100 Su
Sample
mean for 5%
Sur for Su
Sample
mean for 5%
Sur for Su
95% 95% for
for Sur
Sur
150 150
Figure 4.17: Comparison between Su (z) and Sur (z) . Figure 4.18: Comparison between Su,avg (L) and Sur,avg (L) .
64
4.4 SPATIAL VARIABILITY IN DESIGN SHEAR STRENGTH
4.4.1 Introduction
In this section, possible sources for variability in design undrained shear strength
are analyzed. First, the design undrained shear strength profiles are organized by test type
(boring and jumbo piston core data versus cone penetration test data) in Figure 4.19. In
Figure 4.19(a), the blue lines are the design undrained shear strength profiles from the
boring and jumbo piston core data; while the red lines are the design undrained shear
strength profiles from the cone penetration data. A linear shear strength profile of 10 psf
per foot is also shown as a reference, which is the typical value of the rate of strength
increase for the normally consolidated clay in the Gulf of Mexico area. Figure 4.19(b)
shows a comparison of average values of design shear strength at depths of 25, 50, 75,
100, 125 and 150 feet. Each point is obtained by averaging the values of design shear
strength at a given depth from the boring data and from the cone penetration data,
respectively. A line of 45 degrees is also shown as a reference. All points are at or near
the line of 45 degrees. There is no discernable relationship found between the design
strength profiles and the type of test used to develop it.
65
Next, the design undrained shear strength profiles are organized by the time tested
(before 2000 versus after 2000) in Figure 4.20. The year 2000 is selected to divide the
whole data sets into two groups. There are 43 data sets in the “before 2000”group and 54
data sets in the “after 2000” group. In Figure 4.20(a), the blue lines represent the design
shear strength profiles from data tested before 2000, while the red lines show the design
shear strength profiles from data tested after 2000. Figure 4.20(b) shows a comparison of
average values of design shear strength at depths of 25, 50, 75, 100, 125 and 150 feet.
Each point is obtained by averaging the values of design shear strength at a given depth
from the “before 2000” group and the “after 2000” group, respectively. A line of 45
degrees is also shown as a reference. All points are at or near the line of 45 degrees, thus
is no discernable relationship is found between the design strength profiles and the era in
which it was developed.
Last, the measured test results for one site are organized by test method (field
versus laboratory test) in Figure 4.21. The measured undrained shear strength is one of
the bases for determining the design profile. The blue triangles represent the test results
from the field vane tests, while red dots are test results from laboratory tests such as the
unconsolidated undrained (UU) triaxial test, the miniature vane and the pocket
penetrometer. Figure 4.21(b) shows a comparison of the average of measured shear
strength at depths of 25, 50, 75, 100, 125 and 150 feet. Each point is obtained by
averaging the values of measure shear strength at a given depth from the field and
laboratory tests, respectively. A line of 45 degrees is also shown as a reference. Any
effect on measured data for undisturbed undrained shear strength due to test method is
not discernable from this analysis.
In summary, any recognizable variation in design undrained shear strength due to
type of test, time tested and test method is not found.
66
(a) (b)
Figure 4.19: Variability in Design Undisturbed Undrained Shear Strength Profile by Test Type.
67
(a) (b)
Figure 4.20: Variability in Design Undisturbed Undrained Shear Strength Profile by Time Tested.
68
(a) (b)
Figure 4.21: Variability in Measured Undisturbed Undrained Shear Strength –Field data versus Laboratory Data.
69
4.4.3 Analysis of spatial variability in design shear strength
graphically. Comparison between point strengths, Su (z) and Sur (z) , and depth-average
strengths, Su,avg (L) and Sur,avg (L) is shown and discussed. Relationship between
(Figures 4.22 to 4.33), which are designed to show spatial variability in the properties.
Each bubble has coordinates and size. The coordinates indicate the location of data sets
and the bubble-sizes designates the magnitude of Su (z) and Su,avg (L) . Dark dots
represent values below the field average, while light dots represent values above the field
average. The striped-patterned dot in each graph shows the magnitude of field average
with its size. The Su (z) and Su,avg (L) values at depths of 60 feet and 120 feet are
from Table 2.1 and Figure 2.16 that there are 15 sites in this study area.) The sites F, I,
and K tend to have larger values and the sites C, H, L, P and N tend to show lower values
of design undrained shear strength than the field average. The sites A, B, E, F, G, J and M
have the design shear strengths similar to the field average within a 5 % difference.
70
For a depth of 60 feet, the average values of Su,avg for each site are shown in
Figure 4.25 and the Su,avg values for each data set are presented in Figures 4.26 and 4.27.
The sites I, F, A and D tend to have larger values and the sites P, C, B and H tend to show
lower values of depth-averaged design undrained shear strength than the field average.
The sites E, G, J, K, L, M and N have depth-averaged design shear strengths similar to
the field average within a 5 % difference.
For a depth of 120 feet, the average values of Su for each site are shown in Figure
4.28 and the Su values for each data set are presented in Figures 4.29 and 4.30. The
sites D, B, A, and M tend to have larger values and the sites H, I, E, L, K, N and J tend to
have lower values of design undrained shear strength than the field average. The site F
has design shear strengths similar to the field average within a 5 % difference. The
penetration depth for site C and P is shorter than 120 feet below the mudline.
For a depth of 120 feet, the average values of Su,avg are shown in Figure 4.31 and
the Su,avg values for each data set are presented in Figures 4.32 and 4.33. The sites A, E,
F and I tend to have larger values than the field average, and the sites H, L, E and M tend
to show lower values of depth-averaged design undrained shear strength than the field
average. The sites B, E, J, K and N have depth-averaged design shear strengths similar to
the field average within a 5 % difference.
In summary, recognizable effect on variability in Su and Su,avg due location
(both horizontal and vertical) is found, and Su and Su,avg show different variation from
values at depths of 60 feet and 120 feet are presented and analyzed.
For a depth of 60 feet, the site average values of Sur are shown in Figure 4.34
and the Sur values for each data set are presented in Figures 4.35 and 4.36 The sites G,
E, F, H and J tend to have larger values and the sites B, L, C, A, P and M tend to show
lower values of design undrained remolded shear strength than the field average. The
sites D, I and K have the design remolded shear strengths similar to the field average
within a 5 % difference.
For a depth of 60 feet, the site average values of Sur,avg are shown in Figure 4.37
and the Sur,avg values for each data set are presented in Figures 4.38 and 4.39. The sites
H, G, E, J, F, I and D tend to have larger values and the sites P, B, L, C and K tend to
show lower values of depth-averaged design remolded shear strength than the field
average.
For a depth of 120 feet, the site average values of Sur are shown in Figure 4.40
and the Sur values for each data set are presented in Figures 4.41 and 4.42. The sites G,
D, K and F tend to have larger values and the sites L, M, I and H tend to have lower
values of design undrained shear strength than the field average.
For a depth of 120 feet, the site average values of Sur,avg are shown in Figure
4.43 and the Sur,avg values for each data set are presented in Figures 4.44 and 4.45. The
sites G, F, K, E, H and J tend to have larger values and the sites L, B, M and D tend to
show lower values of depth-averaged design undrained shear strength than the field
average.
72
In summary, recognizable effect on variability in Sur and Sur,avg due location
(both horizontal and vertical) is found, and Su and Su,avg show similar horizontal
higher variation in Sur , there is a high probability that we will have a larger variation in
Sur,avg .
The relationship between the undisturbed shear strength and the remolded shear
strength is analyzed graphically. Figure 4.48 shows comparisons of the design
undisturbed shear strength, Su and the design remolded shear strength, Sur at depths of
73
60 feet and 120 feet, respectively. Figure 4.49 shows comparisons of the depth-average
design undisturbed shear strength, Su,avg and the depth-averaged design remolded shear
strength, Sur,avg at depths of 60 feet and 120 feet, respectively. The undisturbed strengths
and the remolded strengths have a minor effect on one another; this explains using the
design remolded strength instead of the design value of sensitivity. It is important to note
that the minimum values of the undisturbed shear strength are greater than the maximum
values of the disturbed shear strength. For example, all values of Su at a depth of 60 feet
are greater than 0.3 ksf (300psf) and all values of Sur at a depth of 60 feet are smaller
than 0.3 ksf. This result would be strong support for the existence of the lower-bound for
the design shear strength of soil.
74
Figure 4.22: Spatial Variability in Design Undrained Undisturbed Shear Strength at z=60 feet (Site Average).
75
Figure 4.23: Spatial Variability in Design Undrained Undisturbed Shear Strength at z=60 feet (Each Data Set).
76
Figure 4.24: Spatial Variability in Design Undrained Undisturbed Shear Strength at z=60 feet (Each Data Set).
77
Figure 4.25: Spatial Variability in Depth-Average Design Undrained Shear Strength at z=60 feet (Site Average).
78
Figure 4.26: Spatial Variability in Depth-Average Design Undrained Shear Strength at z=60 feet (Each Data Set).
79
Figure 4.27: Spatial Variability in Depth-Average Design Undrained Shear Strength at z=60 feet (Each Data Set).
80
Figure 4.28: Spatial Variability in Design Undrained Undisturbed Shear Strength at z=120 feet (Site Average).
81
Figure 4.29: Spatial Variability in Design Undrained Undisturbed Shear Strength at z=120 feet (Each Data Set).
82
Figure 4.30: Spatial Variability in Design Undrained Undisturbed Shear Strength at z=120 feet (Each Data Set).
83
Figure 4.31: Spatial Variability in Depth-Average Design Undrained Shear Strength at z=120 feet (Site Average).
84
Figure 4.32: Spatial Variability in Depth-Average Design Undrained Shear Strength at z=120 feet (Each Data Set).
85
Figure 4.33: Spatial Variability in Depth-Average Design Undrained Shear Strength at z=120 feet (Each Data Set).
86
Figure 4.34: Spatial Variability in Design Remolded Shear Strength at z=60 feet (Site Average).
87
Figure 4.35: Spatial Variability in Design Remolded Shear Strength at z=60 feet (Each Data Set).
88
Figure 4.36: Spatial Variability in Design Remolded Shear Strength at z=60 feet (Each Data Set).
89
Figure 4.37: Spatial Variability in Depth-Average Design Remolded Shear Strength at z=60 feet (Site Average).
90
Figure 4.38: Spatial Variability in Depth-Average Design Remolded Shear Strength at z=60 feet (Each Data Set).
91
Figure 4.39: Spatial Variability in Depth-Average Design Remolded Shear Strength at z=60 feet (Each Data Set).
92
Figure 4.40: Spatial Variability in Design Remolded Shear Strength at z=120 feet (Site Average).
93
Figure 4.41: Spatial Variability in Design Remolded Shear Strength at z=120 feet (Each Data Set).
94
Figure 4.42: Spatial Variability in Design Remolded Shear Strength at z=120 feet (Each Data Set).
95
Figure 4.43: Spatial Variability in Depth-Average Design Remolded Shear Strength at z=120 feet (Site Average).
96
Figure 4.44: Spatial Variability in Depth-Average Design Remolded Shear Strength at z=120 feet (Each Data Set).
97
Figure 4.45: Spatial Variability in Depth-Average Design Remolded Shear Strength at z=120 feet (Each Data Set).
98
0.4 at z=60 ft
R2 = 0.38
0.3
Su, avg (ksf)
0.2
0.1
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Su (ksf)
1.0
at z=120 ft
0.8
R2 = 0.07
Su, avg (ksf)
0.6
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Su (ksf)
Figure 4.46: Design Undrained Shear Strength at a point versus Depth-Averaged Design
Undrained Shear Strength.
99
0.25
0.20
R2 = 0.67
0.15
Sur (ksf)
0.10
0.05
at z=60 ft
0.00
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
Sur,avg (ksf)
0.40
0.30
R2 = 0.52
Sur (ksf)
0.20
0.10
at z=120 ft
0.00
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Sur,avg (ksf)
Figure 4.47: Design Remolded Shear Strength at a point versus Depth-Averaged Design
Remolded Shear Strength.
100
0.4
at z=60 ft
0.3
Sur (ksf)
0.2
0.1
R2 = 0.30
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Su (ksf)
1.0
at z=120 ft
0.8
0.6
Sur (ksf)
0.4
0.2
R2 = 0.17
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Su (ksf)
Figure 4.48: Design Undrained Undisturbed Shear Strength versus Design Undrained
Remolded Shear Strength.
101
0.20
at z=60 ft
0.15
Sur,avg (ksf)
0.10
0.05
R2 = 0.37
0.00
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
Su,avg (ksf)
0.50
at z=120 ft
0.40
Sur,avg (ksf)
0.30
0.20
0.10
R2 = 0.22
0.00
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Su,avg (ksf)
102
4.4.4 Analysis of spatial correlation
the horizontal line from the pole (Figure 4.50). Second, the shortest distance from any
river along the coast, d all , (Figure 4.51) is evaluated for all data sets. Third, distance off
the coast line, d coast , is the shortest distance from the coast line (Figure 4.51 and Figure
4.52). Forth, a water depth, WD , for each data set is evaluated. Finally, distance along
and away from the continental shelf line ( a shelf , d shelf ) are considered because deltaic
depocenters of ancestral Mississippi River are near the transition zone between
continental shelf and slope (Winker and Stancliffe 2007), which is typically identified by
water depth of about 1000 ft.
The depth-averaged design shear strengths, Su,avg for all data sets are sorted by
rM , θM , d all , d coast , WD , a shelf , and d shelf for depths of 10, 20, 30, 40, 50, 60, 70, 80,
90 and 100 feet. The Su,avg for each data set is then colored to indicate whether the value
103
is greater, less than or similar to the average value (Figures 4.54 and to 4.57). The dark
blue indicates that the value of Su,avg for a given location is less than 90 % of the
average value for a given depth, while the light blue represents the value is greater than
110% of the average value for a given depth. The values of Su,avg within an average ±
10% difference are colored with blue. It is expected that no noticeable correlation in
Su,avg was found at the shallow depth which results from the recent sedimentation
process (i.e., Hemipelagic deposits), while the values of Su,avg in the layer at about 60
feet or deeper tend to be larger than the average when the data location is closer to the
sources (i.e., the smaller the factors, rM , θM , d all , d coast , a shelf , and d shelf , the higher the
values of Su,avg ). The turbidities in the layer at about 60 feet or deeper were deposited by
the relatively low sea level and thus the soil properties vary with distance from the
shoreline at the low sea level (Figure 2.1, Anderson et al. 2004). It seems that water depth
has little effect on spatial correlation in Su,avg between data points. Spatial variation in
Su,avg at depths of 60 and 120 ft based on ( a shelf , and d shelf ) coordinates is shown in
Figures 4.58 and 4.59. Two importance conclusions can be drawn from Figures 4.58 and
4.59. First, there is spatial correlation in Su,avg due to the distance from the sources of
sediments (e.g., in an area of high strength, all locations tend to have higher strength).
Second, there is more spatial correlation in the lower layer than the upper layer. It is
consistent with the depositional models (i.e., turbitites versus hemipelagic deposits). The
turbitites in the deeper layer were deposited by the relative low sea level and their
properties are a function of the distance from the depositional sources during the
lowstand of sea level (Figure 2.1, Anderson et al. 2004).
104
Figure 4.50: Polar Coordinate System.
Figure 4.51: Rivers along the Coast and Contour of Distance off the Coast.
105
Figure 4.52: Distance off the Coast.
Figure 4.53: Distance along The Shelf Line and Distance off the Shelf Line.
106
Figure 4.54 Spatial Correlation based on Distance from the Mouth of Mississippi River.
107
Figure 4.55: Spatial Correlation based on Distance from Rivers and Distance off the
Coast.
108
Figure 4.56: Spatial Correlation based on Water Depth.
109
Figure 4.57: Spatial Correlation Based on Distance along the Continental Shelf Line and
Distance off the Continental Shelf Line.
110
100
Greater than 1.1*average
Su,avg at 60 ft Average +- 10%
Less than 0.9*average
Distance off the shelf line (mi)
75
50
25
0
100 200 300 400 500 600 700 800 900 1000
Figure 4.58: Spatial Correlation in Depth-Average Design Shear Strength at z=60 feet.
100
Greater than 1.1*average
Su,avg at 120 ft Average +- 10%
Less than 0.9*average
Distance off the shelf line (mi)
75
50
25
0
100 200 300 400 500 600 700 800 900 1000
Figure 4.59: Spatial Correlation in Depth-Average Design Shear Strength at z=120 feet.
111
4.5 SUMMARY
decrease with increasing depth from 0.37 and 0.35, respectively, at the mudline to 0.12 at
a depth of 150 feet. The average value of the coefficient of variation for Su and Su,avg
at depths between 50 to 150 feet is 0.15 and 0.12, respectively. The coefficients of
variation for the design undrained remolded shear strength ( Sur ) and the depth-averaged
design remolded shear strength ( Sur,avg ) are greater than those for Su and Su,avg ;
however, the values of standard deviation for Sur and Sur,avg are smaller than the those
for Su and Su,avg . The average value of the coefficient of variation for Sur and Sur,avg
112
value of design undrained remolded shear strength at each depth. Spatial correlation
based on distance from the sources and water depth is investigated. The values of depth-
averaged design shear strength sorted by distance along and away from the coast shows
greater spatial correlation and less variation in the deeper turbitites than the hemipelagic
deposits (about the upper 60 feet)
113
Chapter 5 Analysis of Spatial Variability in Caisson Capacity
5.1 INTRODUCTION
In this section, spatial variability in design axial capacity due to soil is presented
and analyzed. For design axial capacity calculations, the friction factor of 0.8 and the
bearing capacity factor of 9 were assumed, and the design undrained strength at a point
and the depth-averaged design shear strength were used. For design lower-bound axial
capacity calculations, the friction factor of 1.0 and the bearing capacity factor of 9 were
assumed, and the design remolded undrained strength at a point and the depth-averaged
design remolded shear strength were used. The axial capacities analyzed in this section
are only soil contribution, thus a weight of a suction caisson is not included.
The design axial capacities for 60- and 120-foot caissons are shown in Figure 5.1.
Both the total axial capacity due to soil, side shear and end bearing, and the contribution
of the reverse end bearing are shown. In Figure 5.1, some data sets may not have any
values of capacity because the design undisturbed undrained shear strength data are not
114
available for those data sets in that depth of penetration. The sample means, standard
deviations, and coefficients of variation are given in Table 5.1. The sample standard
deviation and the coefficient of variation were not calculated for Groups 4 and 5 because
the number of samples in those groups is less than three. Spatial variability in axial
capacity is presented in the bubble graphs (Figures 5.2 to 5.3).
For a 60-foot long caisson, the average value of axial capacity for the entire field
is 801 kips with the coefficient of variation (c.o.v.) of 0.10. Site I yielded the highest
axial capacity of 949 kips. Because Site I has only one data set, it is not possible to draw
conclusions about the variability within the site. The sites in the west side shows lower
capacity than the field average, and the sites in the middle of the field have higher
capacity than the field average. The sites in the east of the field do not show strong
correlation in the design axial capacity.
For a 120-foot long caisson, the average value of axial capacity for the entire field
is 6394 kips with the c.o.v. value of 0.11. The sites in the west half of the fields show
lower capacities with lower variance (the c.o.v. value of 0.09), than the field average,
while the east half tend to have higher capacities with larger variation (the c.o.v. value of
0.12) than the field average.
The coefficients of variation for the design axial capacity across the fieldare 0.10
and 0.11 for 60- and 120-foot caissons, respectively. The coefficient of variation for the
reverse end bearing is higher (0.14 for both 60- and 120-foot caissons) than the
coefficient of variation for the side shear capacity (0.11 for 60-ft caisson and 0.12 for
120-foot caissons). Gilbert et al.(1999) reported coefficient of variance of 0.07 for 300-
foot long and 3.5-foot diameter pile in normally (primarily) consolidated clay. The
coefficients of variation for this project are slightly larger than the reported value by
Gilbert et al. (1999) because the caissons analyzed are shorter than 300 feet and there is
115
less averaging of variations than driven piles with depth. In addition the caissons
analyzed have larger diameters, meaning that the design end bearing capacity contributes
more to the design axial capacity. Therefore, the coefficient of variation is larger.
116
0
500
1000
1500
0
2500
5000
7500
10000
A-B1 A-B1
A-B2 A-B2
A-B3 A-B3
A-B4 A-B4
A-J1 A-J1
A-C1 A-C1
B-B1 B-B1
B-B2 B-B2
B-B3 B-B3
B-B4 B-B4
B-B5 B-B5
B-B6 B-B6
C-J1 C-J1
C-J2 C-J2
C-J3 C-J3
C-J4 C-J4
C-J5 C-J5
C-J6 C-J6
D-B1 D-B1
D-B2 D-B2
D-B3 D-B3
E-B1 E-B1
E-B2 E-B2
E-B3 E-B3
E-B4 E-B4
F-B1 F-B1
F-B2 F-B2
F-B3 F-B3
F-B4 F-B4
F-B5 F-B5
F-B6 F-B6
F-B7 F-B7
F-B8 F-B8
G-B1 G-B1
H-B1 H-B1
I-B1 I-B1
J-B1 J-B1
J-B2 J-B2
J-B3 J-B3
J-B4 J-B4
117
K-C2 K-C2
K-C3 K-C3
K-C4 K-C4
K-C5 K-C5
K-C6 K-C6
L=60ft D=10ft
K-C7
L=120ft D=20ft
K-C7
K-C8 K-C8
K-J1 K-J1
K-J2 K-J2
K-J3 K-J3
K-J4 K-J4
K-J5 K-J5
K-J6 K-J6
K-J7 K-J7
K-J8 K-J8
K-J9 K-J9
K-J10 K-J10
L-B1 L-B1
L-B2 L-B2
L-C1 L-C1
M-B1 M-B1
M-C1 M-C1
M-C2 M-C2
M-C3 M-C3
M-C4 M-C4
M-C5 M-C5
M-C6 M-C6
M-C7 M-C7
M-C8 M-C8
M-C9 M-C9
M-C10 M-C10
M-C11 M-C11
M-C12 M-C12
M-C13 M-C13
N-B1 N-B1
N-B2 N-B2
N-B3 N-B3
N-B4 N-B4
End Bearing (kips)
Skin Friction (kips)
N-B5 N-B5
N-B6 N-B6
N-C1 N-C1
N-C2 N-C2
Skin Friction (kips)
End Bearing (kips)
N-C3 N-C3
N-C4 N-C4
N-C5 N-C5
N-C6 N-C6
N-C7 N-C7
N-C8 N-C8
10850000 10850000
10650000 10650000
A
B
C
10450000 10450000
D
E
H
10250000 10250000
F
G
Northing (ft)
.
10050000 I 10050000
y (ft)
9850000
M 9850000
200000 ft J
L
9650000 9650000
Average-all site
801 kips
N 9450000
9450000
9250000 9250000
970000 1170000 1370000 1570000 1770000 1970000 2170000 2370000 2570000 800000 1000000 1200000 1400000 1600000
x ft) x (ft)
Figure 5.2: Spatial Variability in Axial Capacity for 60-ft Caisson (Site Average Value).
118
10850000 10850000
10650000 10650000
B A
D
10450000 10450000
E
H
10250000 10250000
G F
Northing (ft)
.
Northing (ft)
10050000 10050000
I K
M
9850000 9850000
L J
N
9450000 9450000
9250000 9250000
970000 1170000 1370000 1570000 1770000 1970000 2170000 2370000 2570000 800000 1000000 1200000 1400000 1600000
Easting (ft) Easting (ft)
Figure 5.3: Spatial Variability in Axial Capacity for 120-ft Caisson (Site Average Value).
119
5.2.2 Spatial variability in design lower-bound axial capacity
For the design lower-bound axial capacity due to soil, the sample means, standard
deviations, and coefficients of variation are given in Table 5.2. For a 60-foot long
caisson, the design lower-bound axial capacity ranges from 230 to 539 kips (an average
of 383 kips) and the ratio of the design lower-bound axial capacity to the design axial
capacity ranges from 0.32 to 0.65 (average of 0.47). For a 120-foot long caisson, the
design lower bound axial capacity ranges from 1416 to 3698 kips (average of 2759 kips)
and the ratio of the design lower-bound axial capacity to the design axial capacity ranges
from 0.26 to 0.61 (average of 0.42).
Coefficients of variation for the total lower-bound axial capacity are 0.17 and 0.20
for 60- and 120-foot caissons, respectively. These values are higher than the coefficients
of variation for the total design axial capacity because, while the standard deviation is
smaller, the mean is even smaller.
Figure 5.4 compares the design lower-bound axial capacity and the design axial
capacity and Figure 5.5 shows the ratio of the design lower-bound axial to the design
axial. There is minor correlation between the design lower-bound axial and the design
axial capacity. Even if we have higher lower-bound axial capacity for given caisson
dimensions at a given location, this does not mean we have higher axial capacity. It is
important to note that the minimum values of the design axial capacity are greater than
the maximum values of the design lower-bound capacity. This result would be strong
support for the existence of the lower-bound for the design axial capacity due to soil.
120
Table 5.2: Statistics Relating to Design Lower-Bound Axial Capacity.
Length=60 feet, Diameter=10 feet
121
600
Lower-bound axial capacity (kips)
L=60ft D=10ft
400
2
R = 0.2552
200
0
0 200 400 600 800 1000 1200
6000
Lower-bound axial capacity (kips)
L=120ft D=20ft
4000
2000
2
R = 0.196
0
0 2000 4000 6000 8000 10000 12000
Figure 5.4: Lower-Bound versus Design Axial Capacities for 60- and 120-ft Caissons.
122
Figure 5.5: Ratio of Lower-Bound to Design Axial Capacity for 60- and 120-ft
Caissons.
123
5.3 SPATIAL VARIABILITY IN DESIGN LATERAL CAPACITY
125
0
500
1000
1500
2000
2500
0
5000
10000
15000
20000
25000
A-B1
A-B1 A-B2
A-B2 A-B3
A-B3 A-B4
A-B4 A-J1
A-J1 A-C1
A-C1 B-B1
B-B1 B-B2
B-B2 B-B3
B-B3 B-B4
B-B4 B-B5
B-B5 B-B6
B-B6 C-J1
C-J1 C-J2
C-J2 C-J3
C-J3 C-J4
C-J4 C-J5
C-J5 C-J6
C-J6 D-B1
D-B1 D-B2
D-B2 D-B3
D-B3 E-B1
E-B1 E-B2
E-B2 E-B3
E-B3 E-B4
E-B4 F-B1
F-B1 F-B2
F-B2 F-B3
F-B3 F-B4
F-B4 F-B5
F-B5 F-B6
F-B6 F-B7
F-B7 F-B8
F-B8 G-B1
G-B1 H-B1
H-B1 I-B1
I-B1 J-B1
126
K-C1 K-C1
K-C2 K-C2
K-C3 K-C3
K-C4 K-C4
K-C5 K-C5
K-C6 K-C6
L=120ft D=20ft
L=60ft D=10ft
K-C7 K-C7
K-C8 K-C8
K-J1 K-J1
K-J2 K-J2
K-J3 K-J3
K-J4 K-J4
K-J5 K-J5
K-J6 K-J6
K-J7 K-J7
K-J8 K-J8
K-J9 K-J9
K-J10 K-J10
L-B1 L-B1
L-B2 L-B2
L-C1 L-C1
M-B1 M-B1
M-C1 M-C1
M-C2 M-C2
M-C3 M-C3
M-C4 M-C4
M-C5 M-C5
M-C6 M-C6
M-C7 M-C7
M-C8 M-C8
M-C9 M-C9
M-C10 M-C10
M-C11 M-C11
M-C12 M-C12
M-C13 M-C13
N-B1 N-B1
N-B2 N-B2
N-B3 N-B3
N-B4 N-B4
N-B5
Lateral Capacity (kips)
N-B5
N-B6 N-B6
N-C1
N-C1
Lateral Capacity (kips)
N-C2
N-C3
N-C2
N-C4 N-C3
N-C5 N-C4
N-C6 N-C5
N-C7 N-C6
N-C8 N-C7
N-C8
10850000 10850000
10650000 10650000
B A
C
10450000 10450000
D
10250000 H
10250000 F
G
Northing (ft)
.
Northing (ft)
10050000 10050000
I
K
M
9850000 9850000
J
L
N
9450000 9450000
9250000 9250000
970000 1170000 1370000 1570000 1770000 1970000 2170000 2370000 2570000 800000 1000000 1200000 1400000 1600000
Easting (ft)
Easting (ft)
Figure 5.7: Spatial Variability in Lateral Capacity for a 60-ft Long Caisson (Site Average).
127
10850000 10850000
10650000 10650000
B A
10450000 10450000
D
10250000 10250000
H G F
Northing (ft)
.
Northing (ft)
10050000 10050000
I K
M
9850000 9850000
J
L
N
9450000 9450000
9250000 9250000
970000 1170000 1370000 1570000 800000 1000000 1200000 1400000 1600000
Easting1770000
(ft) 1970000 2170000 2370000 2570000 Easting (ft)
Figure 5.8: Spatial Variability in Lateral Capacity for 120-ft Caisson (Site Average).
128
1200
L=60ft D=10ft
Axial capacity (kips)
800
2
R = 0.80
400
0
0 400 800 1200 1600 2000 2400
Lateral capacity (kips)
12000
L=120ft D=20ft 2
R = 0.74
Axial capacity (kips)
8000
4000
0
0 4000 8000 12000 16000 20000 24000
Lateral capacity (kips)
Figure 5.9: Design Axial Capacity Versus Design Lateral Capacity for 60- and 120-ft Caissons.
129
5.4 SPATIAL VARIABILITY IN TOTAL CAPACITY OF A SUCTION CAISSON UNDER COMBINED
LOADING
In this section, spatial variability in total design capacity due to soil under a loading angle
of 30 degrees, which is common for a mooring system such as a taut leg system(Figure 5.10) in
offshore structures, is presented and analyzed.
The sample means, standard deviations, and coefficients of variation are given in Table
5.4. Spatial variability in the total combined capacity is presented in Figures 5.11 and 5.12. For a
60-foot long caisson, the average of lateral capacity for the entire field is 10,804 kips with the
c.o.v. value of 0.14. For a 120-foot long caisson, the average of lateral capacity for the entire
field is 1,236 kips with the c.o.v. value of 0.14. Spatial variation in total capacity is similar to
spatial variability in the lateral capacity. When we divide the entire filed roughly into three big
regions, the sites in the west region have lower lateral capacity, while the sites in the center show
higher lateral capacity than the field average. The sites in the east region do not show strong
correlation in the design total caisson capacity with higher variation (c.o.v. of 0.18) than the field
average.
130
Table 5.4: Statistics Relating to Combined Loading Capacity.
Axial Lateral
Number Total capacity, F
capacity, V capacity, H
of data
Sample Sample Sample mean Sample mean
sets c.o.v.
mean (kips) stdev. (kips) (kips) (kips)
Group1 17 1259 241 0.19 629 1090
Group2 8 1066 228 0.21 533 924
Group3 8 1378 84 0.06 689 1194
Group4 2 1266 84 0.07 633 1096
Group5 1 1618 - - 809 1401
Group6 23 1285 109 0.08 643 1113
Group7 17 1206 118 0.10 603 1045
Group8 14 1147 75 0.07 574 994
All site 90 1236 172 0.14 618 1070
Axial Lateral
Number Total capacity, F
capacity, V capacity, H
of data
Sample Sample Sample mean Sample mean
sets c.o.v.
mean (kips) stdev. (kips) (kips) (kips)
Group1 12 11577 1710 0.15 5788 11026
Group2 4 10652 900 0.08 5326 9225
Group3 8 13039 542 0.04 6520 11292
Group4 2 9686 2740 0.28 4843 8388
Group5 1 11965 - - 5983 10362
Group6 16 11056 535 0.05 5528 9575
Group7 15 9881 1481 0.15 4940 8557
Group8 14 9687 511 0.05 4844 8389
All site 72 10804 1524 0.14 5402 9357
131
Figure 5.11: Spatial Variability in Total Capacity under Combined Loading for a 60-ft Long Caisson (Site Average).
132
Figure 5.12: Spatial Variability in Total Capacity under Combined Loading for 120-ft Caisson (Site Average).
133
5.5 SUMMARY
134
Chapter 6: Model for Design Shear Strength of Soil
6.1 INTRODUCTION
This chapter presents the development of models to predict five metrics of soil:
the design undrained shear strength at a point; the depth-averaged design shear strength;
the equivalent linear profile of design strength; the design undrained remolded shear
strength; and the depth-averaged design remolded shear strength without having a site-
specific investigation. The models are developed for the geologic setting in a deepwater
offshore field in the Gulf of Mexico, which is described in Section 2.3. Thus, the use of
these models is restricted to this or very similar geologic settings. In this chapter, the
models are presented and the method of calibration is explained. Next, calibration results
and examples are presented.
expected value for the soil property if an actual investigation was performed at that
particular location. The standard deviation represents the uncertainty in this predicted soil
property. The terms “generic” and “unconditioned” are used to designate the predicted
design properties at a location relatively far from any existing data points. The generic
135
profile describes the properties of the entire field as a whole; it is not spatially
conditioned on local data points.
The model can be developed to take into account the new location’s proximity to
one or more nearby data points. For a new location, j, in the area of an existing data point,
i , the conditional forms for the first and second moments of the equations are as follows
(e.g., Ang and Tang, 1975):
and μ X j are the mean values from the unconditioned model; σ Xi and σ Xj are the
standard deviation values from the unconditioned model; xi is the known value at
location i; and ρXi ,X j is the correlation coefficient between values at i and j.
If the marginal probability distribution for the conditional variable is the same
form as that for the unconditional variable and if it is fully characterized by its first two
moments, then the cumulative distribution function is as follows (Journel and Huijbregts
1978):
( )
v
FX j Xi ( x j xi ) = g F xi , location i , location j, μ X j Xi , σ X j Xi , φF (6.3)
v
gF ( ) is a model for the cumulative distribution function; φF is a vector containing
model parameters that describe g F ( ).
136
The general form of the conditional model when there are one or more nearby
data points can be expressed as:
uuv v
μ x j |xi = μ X j + CijCii−1 (x i − μ Xi ) (6.4)
where:
⎡ ρi1 ,i1 σi1 σi1 : ρi1 ,in σi1 σin ⎤
⎢
Cii = ⎢ ... ...
⎥
⎥ and Cij = {ρ j,i1 σ jσi1 }
. . . ρ j,i n σ jσin .
⎢ρ σ σ : ρin ,in σin σin ⎥⎦
⎣ in ,i1 in i1
Curved brackets are used here to denote vectors; square brackets are used to denote
matrices. Note that the correlation coefficient, ρ , for each element in matrices Cij and
Cii changes depending on the distance between the two locations comprising the
element.
The correlation coefficient ( ρ ) indicates how strong the linear relationship is in
values of the design strength between location j and location i . It has a maximum
absolute magnitude of 1.0 which indicates a perfect correlation in design shear strengths
between j and location i . It has a minimum absolute magnitude of 0.0 which indicate no
relationship in design strengths between j and location i. The correlation coefficient can
be expressed as:
v
ρxi ,x j = g R ( x i , x j location i , location j, φR ) (6.6)
v
where g R ( ) is a model for the correlation coefficient, and φR is the model parameter
that describe g R .
137
6.2.2 Model calibration
This section describes the procedure for model calibration. The objective of
v
calibration is to estimate the model parameters, φ , which describes the developed
models based on the data in the database. The maximum likelihood method (e.g.
Benjamin and Cornell 1970, and Wang 2002) is used to calibrate the models. A detailed
description of principle and application of the maximum likelihood method can be found
in Benjamin and Cornell 1970; Melsa and Cohn 1978; and Wang 2002. The calibration
v
process involves finding the set of model parameters, φ , that maximizes the value of the
v
likelihood function with the data set, xi , in the database (Gilbert, 1999). The natural
⎛ dFX j Xi ( x j xi ) ⎞
( (( ) ( )))
v r v r n
ln L x j | φ | xi | φ = ∑ ln ⎜ ⎟ (6.7)
⎜ d x ⎟
i =1
⎝ i ⎠
v
where FX X,i
( x j xi ) is the cumulative distribution function for a value at j given i. and φ
j
is a vector containing one set of model parameters. For example, the natural logarithm of
the likelihood function for a normal and lognormal distribution, which is most commonly
used in a civil engineering application, is as follows, respectively:
⎛ ⎛ x j | xi − μ x j |xi ⎞
2
⎞
(( ))
v r v r n
1
ln L ( x j | φ) ( x i | φ) = ∑ − ⎜ ln(2πσ 2x j |xi ) + ⎜ ⎟ ⎟ (6.8)
2⎜ ⎜ σ x j |xi ⎟ ⎟
i =1
⎝ ⎝ ⎠ ⎠
⎛ ⎛ ⎞
2
⎞
(( ))
v r v r ln( ) − ln( )
( )
n
1⎜ x | x x
ln L ( x j | φ) ( x i | φ) = ∑ − ln 2π ln(1 + δ x j |xi )( x j | xi ) + ⎜ ⎟ ⎟
2 2 j i median,x |x
j i
2 ⎜ ⎜ ln(1 + δ 2
) ⎟ ⎟
i =1
⎝ ⎝ x j |x i ⎠ ⎠
(6.9)
138
where δ x j |xi is the coefficient of variation and xmedian, x j | xi is the median, they are
expressed as:
μ x j |xi
xmedian, x j | xi = (6.11)
1 + δ 2x j |xi
where μ x j |xi and σ x j |xi are the mean and standard deviation of the conditioned model
6.3.1 INTRODUCTION
Based on the basic form of the model described in Section 6.2.1, equations for the
mean and the standard deviation for the generic model, the mean, the standard deviation
and the horizontal and vertical correlation distance for the conditioned model are chosen
to describe the models for five metrics of the design shear strengths of soil, Su , Su,avg ,
Su1 , Sur and Sur,avg , respectively. The calibration results are summarized.
The design undrained shear strength at a point Su can be predicted using the
model of SU (L) , which is quantitatively described in terms of the mean, μSU (L) , and the
standard deviation, σSU (L) :
139
⎧⎪φC4 + φC5 L for L < 100ft
σSU (L) = ⎨ (6.13)
⎪⎩ φC6 for L ≥ 100ft
where L is the depth of penetration in feet, μSU (L) and σSU (L) are in units of ksf. Each
equation is selected based on the actual values. For example, from Figure 4.4 the mean of
Su is a function of a depth and modeled using a polynomial equation as Equation (6.12).
Likewise, from Figure 4.4, the standard deviation of Su increases with depths from 0 to
100 feet and becomes constant at depths of 100 feet or deeper and is modeled as Equation
(6.13).
For a new location, X, conditioned on existing data points, Y1 , Y2 ,…, Yn the
conditional forms of the equations for the mean and the standard deviation are as follows:
−1
v v
μ X|Y for SU (L) = μSU (X) + CXY CYY (Su,Y − μSU (Y) ) (6.14)
−1
σ X|Y for SU (L) = σSU (X) 2 − CXY C YY CTXY (6.15)
The correlation coefficient tends to decrease with distance between any two
locations (Figure 6.1).
− τ /θ
ρ = (e ) (6.16)
where τ is the distance separating j and i n ’s, and θ is the correlation distance. From
Chaper 3, the Su values show the different spatial (horizontal) variation with depth. The
− τH / θH − τ V / θV
ρ = (e )(e ) (6.17)
140
where τH and τV are the horizontal and vertical distance, respectively, separating j and
i n ’s, and θH and θV are the horizontal and vertical correlation distances, respectively.
beyond which there is no further correlation of a physical property associated with that
point. (e.g., Mela and Louie 2001). In other words, data points within the correlation
distance tend to resemble one another. For example, if a new location is horizontally
141
within the horizontal correlation distance ( θH ) of an existing data point, and the existing
data point indicates a higher design shear strength than the field average, then the new
location will also probably have a higher shear strength than the field average. Similarly,
if the known value of design shear strength at the end of a boring is lower than the
generic prediction, then at a point vertically within the vertical correlation distance ( θV )
below the end of the boring, the value of undrained shear strength will probably be lower
than the generic prediction.
The horizontal and vertical correlation distance of the predicted design undrained
strength at point, SU (L) is modeled as follows:
where θH and θV are in unit of feet. The horizontal correlation distance is modeled as
a function of penetration depth because for the study area, correlation in Su tend to
increase with depth. The φCi variables are parameters that describe the model of SU (L) .
Estimates for these nine model parameters were developed using the method described in
Section 6.2.2. The results are presented in Table 6.1.
μ σ θH θV
φC1 φC 2 φ C3 φC 4 φ C5 φC6 φC7 φC8 φ C9
0.0039 0.0078 0.0000034 0.005 0.00135 0.14 8.4 4.3 3.2
ksf ksf/ft ksf / ft 2
ksf ksf/ft ksf ln(ft) ln(ft)
142
6.3.3 Models for the depth-averaged design undrained shear strength ( Su,avg )
σ S U ,A V G ( L ) = φ A + φ A L
4 5
(6.21)
where L is the depth of penetration in feet, μSU ,AVG (L) and σSU ,AVG (L) are in units of ksf.
conditional forms of the equations for the mean and the standard deviation are as follows:
−1
v v
μ X|Y for SU ,AVG (L) = μSU ,AVG (X) + CXY CYY (Su _ avg,Y − μSU ,AVG (Y) ) (6.22)
−1
σ X|Y for SU ,AVG (L) = σSU ,AVG (X) 2 − CXY CYY CTXY (6.23)
The following equations are chosen for the correlation coefficient and the horizontal and
vertical correlation distance of the predicted depth-averaged design undrained strength,
SU,AVG (L) :
− τH / θH − τ V / θV
ρ = (e )(e ) (6.24)
The φAi variables are parameters that describe the model of SU,AVG (L) . Estimates for
these eight model parameters have been developed using the method described in Section
6.2.2. The results are presented in Table 6.2
143
Table 6.2: Model parameters for SU,AVG (L) model
μ σ θH θV
φA 1
φA 2
φA 3
φA 4
φA 5
φA 6
φA 7
φA 8
The SU (L) and SU,AVG (L) models are plotted with penetration depth are shown
in Figure 6.3 and the correlation distances for SU (L) and SU,AVG (L) are summarized in
Table 6.3. The horizontal correlation distances for both models increase with penetration
depth, while the vertical correlation distances are constant with penetration depth. The
correlation distances for SU,AVG (L) are greater than those for SU (L) .
Su or Su,avg (ksf)
0.0 0.5 1.0 1.5
0
30
Penetration depth (ft)
60
μ SU
90
144
Table 6.3: Correlation distance for SU (L) and SU,AVG (L) model
6.3.4 Models for the equivalent linear profile of design shear strength ( Su1 )
The slope of equivalent linear shear strength profile ( Su1 ) can be predicted using
the model of SU1 (L) , which is quantitatively described in terms of the mean, μSU1 (L) , and
the standard deviation, σSU1 (L) :
The correlation coefficient, and the horizontal and vertical correlation distance for
the predicted slope of equivalent linear shear strength profile, SU1 (L) are modeled as
follows:
− τH / θH − τ V / θV
ρ = (e )(e ) (6.29)
φD5
θH for SU1 (L) = eφD 4 + Le (6.30)
where the φDi variables are parameters that describe the model of SU1 (L) . Estimates for
these six model parameters have been developed by using the method described in
Section 6.2.2. The results are presented in Table 6.4
145
Table 6.4: Model parameters for SU1 (L) model
μ σ θH θV
φD1 φD2 φ D3 φD4 φ D5 φ D6
7.4 0.012 1.05 8.5 4.5 3.4
psf/ft psf/ft2 psf/ft ln(psf/ft) ln(psf/ft)
The SU1 (L) model is plotted with penetration depth are shown in Figure 6.4 and
the correlation distances for SU1 (L) are summarized in Table 6.4. The horizontal and
vertical correlation distances for SU1 (L) are similar to those for those for SU,AVG (L) .
Su1 (psf/ft)
0 2 4 6 8 10 12 14 16 18 20
0
30
μSU1
Penetration Depth (ft)
60
90
μSU1-σSU1 μSU1+σSU1
120
150
146
6.3.5 Models for the design remolded shear strength at a point ( Sur )
The design undrained remolded shear strength at a point Sur can be predicted
using the model of SUR (L) , which is quantitatively described in terms of the mean,
μSUR (L) , and the standard deviation, σSUR (L) :
where L is the depth of penetration in feet, μSUR (L) and σSUR (L) are in units of ksf.
conditional forms of the equations for the mean and the standard deviation are as follows:
−1
v v
μ X|Y for SUR (L) = μSUR (X) + C XY CYY (FSur ,Y − μSUR (Y) ) (6.34)
−1
σ X|Y for SUR (L) = σSUR (X) 2 − C XY CYY CTXY (6.35)
The following equations are chosen for the correlation coefficient, and the horizontal and
vertical correlation distance of the predicted design remolded shear strength at point,
SUR (L) :
− τH / θH − τ V / θV
ρ = (e )(e ) (6.36)
where the φEi variables are parameters that describe the model of SUR (L) . Estimates for
these six model parameters have been developed using the method described in Section
6.2.2. The results are presented in Table 6.6.
147
Table 6.6: Model parameters for SUR (L) model
μ σ θH θV
φE1 φE 2 φE3 φE 4 φE 5 φE6
0.002 0.0029 0.0045 0.00072 8.7 3.55
ksf ksf/ft ksf ksf/ft ln(ft) ln(ft)
6.3.6 Models for the depth-averaged design remolded shear strength ( Sur,avg )
where L is the depth of penetration in feet, μSUR ,AVG (L) and σSUR ,AVG (L) are in units of ksf.
conditional forms of the equations for the mean and the standard deviation are as follows:
−1
r v
μ X|Y for SUR ,AVG (L) = μSUR ,AVG (X) + CXY CYY (Su,avg,Y − μSUR ,AVG (Y) ) (6.41)
−1
σ X|Y for SUR ,AVG (L) = σSUR ,AVG (X) 2 − CXY CYY CTXY (6.42)
The following equations are chosen for the correlation coefficient, and the horizontal and
vertical correlation distance of the predicted depth-averaged remolded shear
strength, SUR,AVG (L) :
− τH / θH − τ V / θV
ρ = (e )(e ) (6.43)
148
θV for SUR ,AVG (L) = e φF 6 (6.45)
The φFi variables are parameters that describe the model of SUR,AVG (L) . Estimates for
these six model parameters have been developed using the method described in Section
6.2.2. The results are presented in Table 6.7.
μ σ θH θV
φF1 φF2 φF3 φF3 φF5 φF6
0.007 0.0014 0.0045 0.00031 8.9 3.6
ksf ksf/ft ksf ksf/ft ln(ft) ln(ft)
The SUR (L) and SUR,AVG (L) models are plotted with penetration depth are
shown in Figure 6.5 and the correlation distances for SUR (L) and SUR,AVG (L) are
summarized in Table 6.3. The correlation distances for SUR,AVG (L) are greater than those
Table 6.8: Correlation distance for SUR (L) and SUR,AVG (L) model
149
Sur or Sur,avg (ksf)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
0
30
Penetration depth (ft)
60
μSUR
90
μsur,avg
120 μSUR+σSUR
μSUR-σSUR
μsur,avg-σsur,avg μsur,avg+σsur,avg
150
Figure 6.5: Plot for FSur (L) and FSur,avg (L) model
6.3.7 Models for the depth-averaged design shear strength ( Su,avg ) with anisotropic
coordinates
In Section 4.4.4, spatial correlation due to distance from the source of deposits
was analyzed. One important conclusion from those analyses is spatial (horizontal)
correlation could be different depending on the direction. In this section, the horizontal
correlation between data points is modeled using the ( a shelf , d shelf ) coordinates to see if
there is any anisotropy in horizontal variation in Su,avg .
Recall from Equation 6.24 in Section 6.3.3 that the correlation coefficient for
SU,AVG (L) is modeled based on the horizontal and vertical distance. For a given depth,
the horizontal distance τH for Equation (6.24) is the distance separating two data points
150
With the ( a shelf , d shelf ) coordinates, the correlation coefficient SU,AVG (L) is
modeled as follows:
where τa shelf and τdshelf are the separation distances (Figure 6.6 (b)), and θashelf and θdshelf
are the correlation distances based on ( a shelf , d shelf ) coordinates. The correlation distances,
θashelf and θdshelf , are modeled as follows:
where φA9 and φA10 are model parameters that describe the model of the correlation
coefficient for SU,AVG (L) based on ( a shelf , d shelf ) coordinates. Estimates for these model
parameters have been developed using the method described in Section 6.2.2. The results
are presented in Table 6.9.
Table 6.9: Model Parameters for the Correlation Coefficient for SU,AVG (L) Based on
( a shelf , d shelf ) Coordinates.
θashelf θdshelf
φA9 φA10
SU,AVG (L)
5.3 4.9
151
(a) (b)
The correlation distances for SU,AVG (L) based on a shelf and d shelf for depths of
60, 90 and 120 feet are summarized in Table 6.10 and the isotropic horizontal correlation
distance for SU,AVG (L) is also shown for comparison. The correlation distances for both
a shelf and d shelf increase with penetration depth and the correlation distance for a shelf is
greater than that for d shelf for depths of 60, 90, and 120 ft. The horizontal isotropic
Table 6.10: Isotropic versus Anisotropic Horizontal Correlation Distance for SU,AVG (L) .
Isotropic Anisotropic
L(ft)
θH (ft) θa shelf (ft) θdshelf (ft)
60 11400 12000 8100
90 14100 18000 12100
120 16800 24000 16100
152
6.4 EXAMPLES: PREDICTED DESIGN SHEAR STRENGTH PROFILE
Suppose we wanted to predict the design shear strength profile for depths ranging
from 0 to 100 feet at a location relatively far from any existing data points. The mean and
standard deviation of the predicted design undrained strength at a point, SU , are
calculated according to Equations (6.12) and (6.13). For example, the mean and standard
deviation of SU at a depth of 90 feet:
μSU (90) = [0.0039 ksf + 0.0078 ksf / ft(90 ft) + 0.0000034 ksf / ft 2 (90 ft) 2 ] = 0.7334 ksf
The mean and standard deviation of the generic depth-averaged design shear
strength SU,AVG are calculated according to Equations (6.20) and (6.21). For example,
the mean and standard deviation of SU,AVG at a depth of 90 feet:
μSU,AVG (90) = [0.01418 ksf + 0.00363 ksf / ft(90 ft) + 3.6x10-6 ksf / ft 2 (90 ft)2 ]
= 0.3699 ksf
σSU ,AVG (90) =[0.0112 ksf + 0.00039 ksf/ft(90 ft)] = 0.0463 ksf
Results of the predicted genetic shear strength at a point and depth averaged shear
strength for depths of 0 to 100 feet are shown in Figure 6.7.
153
Shear strength (ksf)
0.0 0.2 0.4 0.6 0.8 1.0
0
20 Su_avg
Su, point
Depth below mudline (ft)
40
60
80
100
feet from the mudline are known for a given location P-J1 (Figure 6.8) and we want to
extrapolate these Su and Su,avg profiles to the depth of 90 feet.
The mean and standard deviation of spatially conditioned SU and SU,AVG can be
calculated according to Equations (6.14), (6.15), (6.22) and (6.23), respectively. If there
are n existing data points, dimensions of matrices CYY and CXY are [n × n] and [1 × n] ,
154
introducing the threshold value ρ* . If the values of ρX,Y* are very small (meaning that
i
the existing data points, Yi* , are far from the new data point, X), values of SU and
SU,AVG at existing data points Yi* have almost no effect on values of SU and SU,AVG at
a new location X. Suppose that only k out of n elements of ρX,Yi are greater than the
coefficient values with the new data point (at the depth of 90 feet at P-J1 in this example)
is greater than 0.05, as shown in Figure 6.9. The results of this example are presented in
Figure 6.10. Red dots and lines at depths ranging from 65feet to 90 feet represent
extrapolated (predicted) profiles of design shear strength at point ( Su ) and depth-
Figure 6.8: Example-Existing Design Profiles for the P-J1 Data Set.
155
Figure 6.9: Example-Correlated Data Points.
156
6.4.3 Example: Ways of depicting extrapolated design profiles
We have separate models for the point values and the depth-average values (e.g.,
SU (L) versus SU,AVG (L) , and SUR (L) versus SUR,AVG (L) ) because the models for the
point values cannot reproduce the depth-averaged values, meaning that there are higher-
order relationship in the data that the models with the first two moments does not capture.
One possible solution to capture this high-order spatial behavior in the data is to develop
an advance model for the point values that include higher order moments (e.g., third,
fourth, fifth, etc. cross moments). The difficult for the model with higher order moments
is that it is much more complicated to develop and to use. Another possible solution is to
develop separate models for the point and the depth-averaged values, which is selected
for this study. This approach is more practical; however, understanding of the different
scales between the models is required. The different resolution in the two models is
explained with the following example.
There could be two different extrapolated Su profiles: one is extrapolated based
on existing data of Su as shown in Figure 6.11(a); the other is evaluated from the
extrapolated Su,avg as shown in Figure 6.11 (b). Figure 6.11 (a) is the same as Figure
6.10 (a), while the extrapolated profile for Su in Figure 6.11 (b) is calculated from the
extrapolated Su _ avg profile in Figure 6.10 (b).
At this particular location (P-J1), the design Su profile at 0 to 25 feet is very low,
and then increases rapidly from 25 feet to 60 feet. Because the extrapolated Su , based on
existing data of Su , is directly conditioned on the value in that profile (e.g., Su at 60
feet), it is simply an extension of the Su profile. On the other hand, the extrapolated
profile for Su is not expected to increase as rapidly as the existing Su profile at depths
ranging from 40 to 60 feet below the mudline if it is obtained from the extrapolated
157
Su,avg profile which is conditioned on the existing Su,avg profile value. Thus, there is a
discontinuity at 60 feet between the existing Su profile and the extrapolated Su profile.
a jump in the design undrained shear strength at about 60 feet below the mudline. If
depths of the jumps are 56 feet and 64 feet, the two Su profiles would not correlate with
one another. For these same borings, the depth-averaged design shear strength at 65 feet
in both profiles will reflect the jumps at about 60 feet and two Su,avg profiles will
correlate more strongly with one another. It is important to understand these different
scales when the side shear and end bearing capacity of a suction caisson are evaluated by
depicting all of these extrapolated profiles. The local, point information is important for
end bearing, while the regional, depth-averaged information is important for side shear.
158
Therefore, the extrapolated Su based on the existing Su (e.g., Figure 6.10(a)) is
recommended to predict the design side shear; and the extrapolated Su,avg based on the
existing Su,avg (e.g., Figure 6.10(b)) is recommended to predict the design end bearing
capacity.
6.5 SUMMARY
Geostatistical models has been developed and calibrated for predicting the design
shear strengths at locations where no site-specific investigation was performed. The
vertical correlation distance for the design undrained shear strength a point is
approximately 25 feet and is constant with depth. The horizontal correlation for the
design undrained shear strength a point distance increases with depth, with a value of
4,400 feet at the mudline and a value of 15,500 feet at a depth of 150 feet. The spatial
correlation structure for the depth-averaged undrained shear strength has larger
correlation distances both horizontally and vertically. The vertical correlation distance is
30 feet and the horizontal correlation distance increases from 6,000 feet at the mudline to
19,500 feet at a depth of 150 feet for the depth-averaged undrained shear strength. The
separate models for the point values and the depth-averaged values are developed to
capture the higher spatial correlation in the data. Understanding the different scales that
these two models include is important. The predicted values of design shear strength at
point is recommended for the design axial end bearing capacity and the predicted values
of depth-average design shear strength is recommended for the design axial side shear
capacity. The use of these models is restricted to this study area or very similar geologic
settings.
159
Chapter 7 Example Implementation for Suction Caisson Capacity
7.1 INTRODUCTION
This chapter presents examples to show how to implement the models for the
design shear strengths developed in Chapter 6 to predict the design capacity of a suction
caisson including axial, lateral and combined capacity.
The average unit side shear and the unit end bearing is directly proportional to the
depth-average design shear strength and the design shear strength at a point, respectively.
The predicted average unit side shear, FS (L) , and the predicted unit end bearing FB (L)
where SU,AVG (L) is the model for the depth-averaged design shear strength, SU (L) is
the model for the design shear strength at a point, α is a dimensionless friction
coefficient that must be less than or equal to 1.0 and N c is a bearing capacity factor.
The α value is assumed to be 0.8 throughout the depth of penetration, which is the
typical value used in design of suction caissons for normally consolidated clay in Gulf of
Mexico. The bearing capacity factor of 9 is chosen as recommended in API (2003).
160
Two separate random variables, FS (L) and FB (L) , are combined into one
composite variable, the predicted design axial capacity, Vmax . The mean capacity, μ Vmax ,
and the standard deviation of axial capacity, σ Vmax are calculated as follows:
2
⎛ πD 2 ⎞ ⎛ πD 2 ⎞
(σ ) ( )
2 2
( πDL ) + 2 ( πDL ) ⎜
2
σ Vmax = FS (L) +⎜ ⎟ σFB (L) ⎟ σ FS (L) ⋅ σFB (L) ⋅ρFS ,FB (7.4)
⎝ 4 ⎠ ⎝ 4 ⎠
where D is the diameter of a suction caisson, L is the depth of penetration, and ρFs ,FB is
locations, P-J1 and A-B1 as examples. The coefficient of variation for Vmax increases
slightly as ρFS ,FB increases from 0 to 1. Therefore, uncertainty in the composite variable,
FS (L) and FB (L) . This assumption is reasonable because the design shear strength at a
point is not highly correlated with the depth-average shear strength as shown in Figure
4.46. (the R 2 value for the linear relationship between s u and s u,avg is 0.38 and 0.07
161
Figure 7.1: Uncertainty in Axial Capacity versus Correlation between FS (L) and
FB (L) .
Recall from Section 3.2.2 that the lower-bound axial capacity can be calculated in
the similar manner to the design axial capacity. The design undrained remolded shear
strength at a point and the depth-averaged design remolded shear strength are used
instead of the design undisturbed shear strengths. The predicted lower-bound average unit
side shear, FS,LB (L) , and the predicted lower-bound unit end bearing FB,LB (L) can be
designated as:
where SUR,AVG (L) is the model for the depth-averaged design remolded shear strength,
SUR (L) is the model for the design remolded shear strength at a point, α is a
162
dimensionless friction coefficient that must be less than or equal to 1.0 and N c is a
bearing capacity factor. The α value is assumed to be 1.0 throughout the depth of
penetration and N c is assumed to be 9.0.
Two separate random variables, FS,LB (L) and FB,LB (L) , are combined into one
composite variable, the predicted design lower-bound axial capacity, Vmax,LB . The mean
capacity, μ Vmax,LB , and the standard deviation of axial capacity, σ Vmax,LB are calculated as
follows:
2
⎛ πD 2 ⎞ ⎛ πD 2 ⎞
( ) ( )
2 2
( πDL ) + 2 ( πDL ) ⎜
2
σ Vmax,LB = σFS,LB (L) +⎜ ⎟ σFB,LB (L) ⎟ σ FS,LB (L) ⋅ σ FB,LB (L) ⋅ρFS,LB ,FB,LB
⎝ 4 ⎠ ⎝ 4 ⎠
(7.8)
where D is the diameter of a suction caisson, L is the depth of penetration, and ρFS,LB ,FB,LB
is the correlation coefficient between FS,LB (L) and FB,LB (L) . Uncertainty in Vmax,LB is
obtained based on correlation between FS,LB (L) and FB,LB (L) because the design
respectively.
163
the minimum value of zero. However, the lower-bound capacity is generally greater than
zero so that the conventional lognormal distribution does not capture the existence of the
lower-bound capacity (Najjar 2005).) Najjar (2005) recommended a mixed lognormal
distribution as a simple and practical alternative for distribution of capacity that account
for the lower-bound capacity effectively. A mixed lognormal distribution has a finite
probability for capacities at the lower bound, while it has a continuous probability density
function for capacities greater than the lower-bound (Figure 7.2). The finite probability,
p, at the lower bound corresponds to the probability that is less than and equals to the
lower bound in the conventional lognormal distribution.
In this study, the mixed lognormal distribution is selected for a design axial
capacity to account for its lower-bound because a mixed lognormal distribution provides
the closest approximation based on results from three pile-load test databases compared
with conventional lognormal, beta, truncated lognormal probability distributions (Najjar
2005). A mixed lognormal distribution is convenient to use because the mean and the
standard deviation are approximately the same as for a conventional lognormal
distribution unless the lower-bound capacity is too close to the mean.
164
7.3 MODEL FOR DESIGN LATERAL CAPACITY
In Section 4.2.1, the equivalent linear profile of design undrained strength was
evaluated using a spreadsheet program for lateral capacity developed for this study. The
equivalent linear profile gives the same lateral capacity as the design undrained shear
strength. The design lateral capacity can be predicted using the predicted equivalent
linear profile. The model for the slope of equivalent linear shear strength profile, SU1 (L)
depth, the design lateral capacity will also follow a lognormal distribution.
7000
L/D=6 L=90 ft
6000
5000
L=80 ft
Lateral capacity (kips)
4000
L=70 ft
3000
2000 L=60 ft
L=50 ft
1000
0
0 2 4 6 8 10 12 14
Su1 (psf/ft)
165
7.4 COMBINED CAPACITY
7.5 EXAMPLES
7.5.1 Example: generic side capacity for a 90-ft long, 15-ft diameter caisson
Suppose we wanted to predict the design side shear capacity for a 90-ft long, 15-ft
diameter caisson at a site relatively far from any existing investigation locations. The
mean and standard deviation of FS (L) are calculated based on Equation (7.1). From the
example in Section 5.4.1, μSU ,AVG (90) = 0.3699 ksf and σSU ,AVG (90) = 0.0463 ksf .
The mean and standard deviation of design axial side capacity are found by multiplying
the length and circumference of the caisson:
The same results with the generic end bearing and the generic design axial capacity for
90-ft penetration from the spreadsheet program are shown in Figure 7.4.
The lower-bound side shear capacity is predicted in the similar manner. The mean
and standard deviation of the depth-averaged design remolded shear strength
( SUR ,AVG (L) ) can be obtained from Equations (6.39) and (6.40):
μSUR ,AVG (90) = 0.007 ksf + 0.0014 ksf / ft × 90ft = 0.133 ksf
Figure 7.4: Example: Generic Side Capacity for a 90-ft Long, 15-ft Diameter Caisson.
167
Then, the lower-bound average unit side shear, FS,LB (L) are calculated based on Equation
(7.5):
μ FS,LB (90) = α ⋅μSUR ,AVG (90) = 1.0 × 0.1330 = 0.1330 ksf
The mean and standard deviation of the lower-bound design axial side capacity are found
by multiplying the length and circumference of the caisson:
Finally, the coefficient of variation for lower-bound axial side capacity is:
The generic axial capacity and the generic lower-bound axial capacity for a 15-ft
diameter caisson for depths of penetration from 5 to 90 feet are shown in Figure 7.5.
The probability density function (PDF) and the cumulative distribution function
(CDF) for the generic design axial capacity of a 90-foot long, 15-foot diameter caisson
are shown in Figure 7.6. The probability that the generic design axial capacity is less than
the generic lower-bound axial capacity for a 90-foot long, 15-foot diameter caisson is
extremely low. Therefore, the probability distribution for the generic design axial
capacity (a mixed lognormal distribution) is essentially the same as a lognormal
distribution.
168
Axial capacity (kips)
0 500 1000 1500 2000 2500 3000
0
10 Generic
20
Generic-Lower Bound
Depth of penetration (ft)
30
40
50
60
70
80
90
Figure 7.5: Example: Predicted Axial and Lower-Bound Axial Capacity (Generic) for a
15-ft Diameter Caisson.
Figure 7.6: PDF and CDF for Generic Design Axial Capacity (L=90ft, D=15ft).
169
7.5.2 Example: Horizontally conditioning the profile on a nearby data point
Suppose that a new platform is to be located at the site “A”, and we want to
predict the design average side shear ( FS ) for a 90-ft long caisson. A plan view of the area
and the magnitudes of the FS values for 90-ft penetration at each of the existing data sets
are presented in Figure 7.7. The platform is to be constructed at X N location which is
due west of the A-B1 data set, with the horizontal distance of 6000 feet:
τH = 6000 feet
From an example in Section 7.5.1, μ FS (90) = 0.2959 ksf and σ FS (90) = 0.0370 ksf. From
Figure 7.7, FS,Y = 0.3751 ksf. Applying Equations (6.4) and (6.5):
There is zero uncertainty in FS at the existing data point. As the new location moves
farther and farther from the existing data point, the uncertainty increase back to the
generic value. The reduction in uncertainty is expressed as the ratio of σX|Y to σgeneric
170
Figure 7.7: Plan View of the Site A(Numbers are Fs values in ksf for 90-ft penetration).
Figure 7.8: Reduction in Uncertainty versus Distance from the Existing Data Point.
171
7.5.3 Example: Vertically conditioning the profile on an overlying data point
Suppose A-B1 was drilled to only 90ft below the mudline and we are interested in
using a longer suction caisson at this site, say 120 feet. The vertical separating distance
is:
τV = 30 feet
θV = e3.4 ln(ft)
= 30 feet
1/ 2
σ X|Y = ⎡⎣σ Fs(120) 2 (1-ρ2 ) ⎤⎦ (7.10)
From an example in Section 7.5.1, μ Fs(90) = 0.2959 ksf and σFs(90) = 0.0370 ksf. From
Figure 7.7, FS,Y = 0.3751 ksf.
μSU ,AVG (120) = [0.01418 ksf + 0.00363 ksf / ft(120 ft) + 3.586x10-6 ksf / ft 2 (120 ft)2 ]
= 0.5014 ksf
σSU ,AVG (120) =[0.0112 ksf + 0.00039 ksf/ft(120 ft)] = 0.0580 ksf
172
Based on Equation (7.1):
μ FS (120) = 0.8 × 0.5014 =0.4011 ksf
σ FS (120) = 0.8 × 0.580 = 0.0464 ksf.
μ X|Y = 0.4011 ksf + (0.367)(0.0464 ksf/0.0370 ksf )(0.3751 ksf –0.2959 ksf)=0.4376 ksf
Suppose we are interested in using a 90-foot long, 15-foot diameter caisson at the
location of the P-J1 data set. Recall from Section 6.4.2 that the P-J1 location was drilled
to only 65 feet below the mudline (Figure 6.8). The proposed model allows the option of
spatially conditioning the generic Fs(90) on the multiple existing data points. From
Section 6.4.2, the design side shear at nineteen adjacent data points are correlated with
the design side shear at the new location (in this example, the depth of 90 feet at the P-J1
location) with correlation coefficient values greater than 0.05. Therefore, the dimensions
v v
of CXY , CYY and (FS, Y − μFs ( B)) are [1 x 19], [19 x 19] and [19 x 1], respectively, in
−1
σ X|Y for FS (90) = σ FS (90) 2 − C XY CYY CTXY = 0.0343 ksf
The uncertainty reduction ratio is 0.926 from σgeneric = σ Fs(90) = 0.0370 . The mean and
standard deviation of axial side capacity are found by multiplying the length and
circumference of the caisson:
173
μ Qside = (0.2579 ksf )(90 ft)(π 15 ft) = 1094 kips
The results including the end bearing and the design axial capacity for 90-ft penetration
from the spreadsheet program are shown in Figure 7.9.
The predicted side shear capacity at P-J1 for 90-ft penetration (1094 kips) is less
than the generic side shear capacity for 90-ft penetration (1255 kips from Section 7.5.1);
however, is greater than the generic lower-bound side shear capacity for 90-ft penetration
(564 kips from Section 7.5.1).
Figure 7.9: Example: Spatially Conditioned Side Capacity for a 90-ft Long, 15-ft
Diameter Caisson
174
The extrapolated axial capacities for a 15-ft diameter caisson for depths of
penetration from 60 to 90 are shown in Figure 7.10 with the extrapolated lower-bound
axial capacity, the generic axial capacity, the generic lower-bound capacity and the
existing design axial capacity. As shown in the figure below, the extrapolated axial
capacities at P-J1 are between the generic lower-bound and the generic axial capacity.
The PDF and the CDF for the extrapolated design axial capacity of a 90-foot long,
15-foot diameter caisson are shown in Figure 7.11. The probability that the extrapolated
design axial capacity is less than the extrapolated lower-bound axial capacity is extremely
low. Therefore, the probability distribution for the extrapolated design axial capacity (a
mixed lognormal distribution) is essentially the same as a lognormal distribution.
10 Existing Design
Exsisting Design-LB
20 Extrapolated - Mean
Extrapolated-Mean-LB
Depth of penetration (ft)
30 Generic
Generic-Lower Bound
40
50
60
70
80
90
175
Figure 7.11: PDF and CDF for Extrapolated Design Axial Capacity (L=90ft, D=15ft).
176
Chapter 8: Reliability-Based Design Approach
8.1 INTRODUCTION
The design reliability for a suction caisson is defined as the probability that the
load on the suction caisson does not exceed the capacity over the design life. There is
uncertainty in both load and capacity of suction caissons. To account for this uncertainty,
the capacity, R, and the load, S, are modeled as random variables with the main
descriptors such as μ R , μ S , σ R and σ S . The first moments, μ R and μ S are the
expected minimum values for the resistance and maximum load over the design life, and
the second moments, σ R and σ S , are the standard deviations of the capacity and the
load over the design life. The coefficients of variation ( Ω R and ΩS ) are the ratio of the
standard deviation to the expected value and account for uncertainty in the capacity and
the load over the design life ( Ω R = σ R / μ R and Ω S = σ S / μ S ). The design reliability for a
177
where R is the resistance of the suction caisson, S is the load on the suction caisson, and
Pf is the probability of failure. If R and S are statistically independent, the probability of
∫
Pf = P(R < S) = FR (s)fS (s)ds = Φ(-β)
0
(8.2)
where FR (s) is the cumulative distribution function of the capacity evaluated at the load
s; fS (s) is the probability density function of the load S; Φ ( ) is the standard normal
function; and β is the reliability index. The relationship between reliability and the
reliability index is shown in Table 8.1. Typical values for the reliability index for
offshore platforms range from 2.0 to 4.0.
178
8.3 DESIGN RELIABILITY ASSESSMENT
In cases where the capacity, R, and the load, S, are both lognormally distributed,
the design reliability can be expressed as closed form solution as follows:
⎛ ⎞
ln ( FSmedian )
Reliability = Φ (β) = Φ ⎜ ⎟ (8.3)
⎜⎜ ⎟⎟
⎝ (
ln (1+ Ω 2R )(1+ ΩS2 ) ) ⎠
where FSmean is the mean factor of safety defined as the ratio of the mean capacity to the
mean load; FSmedian is the median factor of safety defined as the ratio of the median
capacity, rmedian to the median load, s median . Equation (8.3) can be expressed with the first
two moments of the probability distributions for the load and the capacity as follows (Wu
et al., 1989):
⎛ ln ( FS ⎞
median )
Re liability = 1 − Φ (−β) ≅ Φ ⎜ ⎟ (8.4)
⎜ Ω2 + Ω2 ⎟
⎝ R S ⎠
This approximation has been made in typical reliability analyses for offshore foundations
(Tang and Gilbert, 1993).
A mixed lognormal distribution was chosen to model design axial capacities with
the lower-bound in Chapter 6. In mixed lognormal distributions, capacities greater than
the lower-bound follow the conventional lognormal distribution, while capacities at the
lower bound has a finite probability that corresponds to the probability that the capacity is
179
less than or equal to the lower bound in conventional lognormal distribution (Figure 7.2).
In cases where the load, S, is lognormally distributed and the capacity, R follows a mixed
lognormal capacity, closed form solutions for the probability of failure are obtained based
on the theorem of total probability:
Reliability = P(S < R) = P(S < R | r = rLB )P(R = rLB ) + P(S < R | r > rLB )P(r > rLB ) (8.5)
∞
P(S < R | r > rLB )P(r > rLB ) = ∫ P(S < R | r)f
rLB
R (r)dr
∞ ⎡ ⎛ ⎞⎤ ⎡ ⎛ ⎞ ⎤
2
ln(r) − ln(s ) 1 1 ln(r) − ln(r )
∫
= ⎢Φ ⎜ median ⎟ ⎥
exp ⎢ − ⎜ median ⎟ ⎥
dr
⎢ ⎜ 2 ⎟ ⎥ 2 ⎢ 2⎜ 2 ⎟ ⎥
rLB ⎣ ⎝ ln(1 + ΩS ) ⎠ ⎦ r 2π ln(1 + Ω R ) ⎢⎣ ⎝ ln(1 + Ω R ) ⎠ ⎥⎦
Equation (8.5) can be expressed in terms of the ratio of the lower bound to the median
capacity and the median factor of safety by normalizing the capacity and the load by the
median capacity, rmedian :
⎛ rLB 1 ⎞ ⎛ r ⎞
⎜ ln( r ) − ln( ) ⎟ ⎜ ln( LB ) ⎟
FSmedian rmedian
Re liability = Φ ⎜ median ⎟Φ⎜ ⎟
⎜ ln(1 + ΩS )2 ⎟ ⎜ 2 ⎟
ln(1 + Ω R )
⎜ ⎟ ⎜ ⎟
⎝ ⎠ ⎝ ⎠
⎛ r 1 ⎞ ⎡ ⎛ r ⎞ ⎤
2
ln(
⎜ r ) − ln( ) ⎢ ⎜ ln( ) ⎥
∞ FSmedian ⎟ 1 1 rmedian ⎟ ⎥ ⎛ rLB ⎞
∫
+ rLB Φ ⎜ median
⎟ exp ⎢ − ⎜ ⎟ d⎜ (8.6)
⎜ ln(1 + ΩS2 ) ⎟ r 2 ⎢ 2 ⎜ ln(1 + Ω 2 ) ⎟ ⎥ ⎝ rmedian ⎟⎠
rmedian
⎜ ⎟r 2 π ln(1 + Ω R ) ⎢ ⎜ R
⎟ ⎥
⎝ ⎠ median ⎣ ⎝ ⎠ ⎦
The first part of Equation (8.6) can be solved analytically, while the second part requires
numerical integration to solve. In this research, the engineering calculation software,
180
MathCAD was used to solve numerical integration. The magnitude of the effect of the
lower-bound capacity on reliability is illustrated in Figure 8.1. The coefficients of
variation, Ω R of 0.4 and ΩS of 0.2 are assumed for the chart, which are the typical
values for offshore applications. From the chart, the lower-bound capacity has a
significant effect on reliability when the ratio of lower bound to median capacity is
greater than 0.5. For example, the reliability index increases from 2.5 for the case of no
lower-bound capacity to 4 for the case where the ratio of lower-bound to median capacity
is 0.7 when the median factor of safety is the typical value of 3.0.
For the study area, the ratio of the lower bound to median capacity (Figure 8.2)
from the generic case is about 0.4, and the ratio of the design remolded to design
undisturbed shear strength at depths of 60 and 120 feet ranges from 0.26 to 0.65 with the
average of 0.45 (Figure 5.5). Therefore, it is expected that increase in the reliability due
to a lower-bound capacity is not significant for the study area.
181
Axial capacity (kips)
0 1000 2000 3000 4000 5000
0
Generic-Median
Generic-Mean
30 Generic-Mean-(LB)
Depth of penetration (ft)
60
90
120
150
In the current version of design codes, the design factors related to the resistance
(e.g., the resistance factor in Load Resistance Factor Design and the factor of safety I
Working Stress Design) were developed implicitly assuming that the soil boring is drilled
at the site of the foundation. If a soil boring is not drilled at the site of the foundation,
these factors need to be adjusted to achieve the same level of reliability. This additional
source of uncertainty is called the spatial variability in resistance and is expressed by a
coefficient of variation, δ R .
182
The additional partial safety factor in the WSD format (or the partial resistance
factor in the LRFD format), which accounts for the spatial variability in resistance can be
determined by setting the reliability indices with and without spatial variability equal to
one another.
If the load R and capacity S are assumed independent random variables with
lognormal distributions, the reliability index with no spatial variability is obtained from
Equation (8.4). The corresponding reliability index including spatial variability in
resistance is:
ln(FSmedian ⋅ FSδR )
β= (8.7)
Ω 2R + ΩS2 + δ R2
where FSδR is the additional partial factor of safety that is introduced to achieve the
same level of reliability in the WSD (Working Stress Design) format. Equations (8.4) and
(8.7) are set equal to one another and then the target reliability index is:
ln(FSmedian ) ln(FSmedian ⋅ FSδR )
β t arg et = = (8.8)
Ω 2R + ΩS2 Ω R2 + ΩS2 + δ 2R
Similarly, the additional partial resistance factor that accounts for a lower bound
is obtained as follows:
183
The design chart that relates FSδR to δR is shown in Figure 8.3. This design chart was
created using Ω R of 0.4 and ΩS of 0.2, which are the typical values for offshore
applications. As the distance from existing data points increases, the magnitude of spatial
variability increases and thus the required value for FSδR increases, which results in a
more conservative suction caisson design. For the study area, the spatial coefficient of
variation δ R for the generic design axial capacity for penetration depths ranging from 60
to 150 ft is about 0.10 (Figure 8.4). The penetration depths of 60 to 150 feet are the
typical for a caisson design in an offshore application. From Chapter 3, δ R for design
axial, lateral and total capacity for 60- and 120-foot caissons for the entire field ranges
from 0.10 to 0.14. Therefore, it is expected that decrease in the reliability due to spatial
variability in resistance is not significant for the study area.
1.25
β_target 2.0
β_target 2.5
ΩR =0.4
1.20
β_target 3.0
ΩS=0.2
β_target 3.5
β_target 4.0
1.15
Spatial FS, Fs δR
1.10
1.05
1.00
0.00 0.05 0.10 0.15 0.20
Spatial c.o.v, δR
Figure 8.3: Partial Safety Factor versus Spatial Variability in Capacity (R and S are
modeled with a lognormal distribution).
184
Spatial Coefficient of Variation, δR
0.0 0.1 0.2 0.3 0.4 0.5
0
30
Depth of penetration (ft)
60
90
120
150
185
⎛ rLB 1 ⎞
⎜ ln( r ) − ln( )⎟ ⎛ r
ln( LB )
⎞
FSmedian FSδR ⎜ r ⎟
Re liability = Φ ⎜ ⎟Φ⎜ ⎟
median median
⎜ ln(1 + ΩS2 ) ⎟ ⎜ ln(1 + Ω 2 + δ2 ) ⎟
⎜ ⎟ ⎜ R R
⎟
⎝ ⎠ ⎝ ⎠
⎛ r 1 ⎞ ⎡ ⎛ r ⎞ ⎤
2
ln( ) − ln( ) ⎢ ⎜
∞
⎜ r FSmedian FSδ ⎟ 1 1
ln(
rmedian
) ⎟ ⎥ ⎛ r ⎞
+ rLB Φ ⎜
∫ ⎟ exp ⎢ − ⎜ ⎟ ⎥ d ⎜ LB ⎟
median R
⎜ 2
ln(1 + ΩS ) ⎟ r 2 2 ⎢ 2 ⎜ ln(1 + ΩR + δR ) ⎟ ⎥ ⎝ rmedian ⎠
2 2
rmedian
⎜ ⎟r 2π ln(1 + ΩR + δR ) ⎢ ⎜ ⎟ ⎥
⎝ ⎠ median ⎣ ⎝ ⎠ ⎦
(8.11)
The effect of spatial variability in capacity on the reliability index with a lower-
bound capacity is shown in Figure 8.5. The design chart that relates FSδR to δR with a
lower-bound capacity is shown in Figure 8.6. This design chart was created using βt arg et
of 3.0, ΩR of 0.4 and ΩS of 0.2, which are the typical values for offshore applications.
As the uncertainty in resistance increases, the reliability index decreases and this
decreasing effect reduces as the ratio of the lower bound to median capacity increases.
The reliability accounting for uncertainty due to spatial variability is typically not a large
impact on a design. For example, consider the typical case in the study area where the
ratio of lower bound to median capacity is about 0.5 and the spatial coefficient of
variation in resistance, δR , is less than 0.2. From Figure 8.6, the partial additional safety
factor is less than 1.025 and therefore, decrease in capacity due to spatial variability is
less than 2.5 %.
186
6
Ω R =0.4
Ω S=0.2
5
Reliability Index
δR=0
3 δR=0.1
2
δR=0.2
δR=0.3
1
0 0.2 0.4 0.6 0.8 1
Ratio of Lower-bound to Median Capacity
1.30
Ratio of Lower-Bound
to Median Capacity
βtarget = 3.0
1.25
Ω R =0.4 0.0
Ω S=0.2
1.20 0.1
Spatial FS,FSδR
1.15
0.3
1.10
0.4
1.05
0.5
0.6 0.7
0.8
1.00 0.9
0 0.05 0.1 0.15 0.2 0.25 0.3
Spatial c.o.v., δR
Figure 8.6: Partial Safety Factor versus Spatial Variability in Capacity (R is modeled
with a mixed lognormal distribution).
187
8.5 DESIGN EXAMPLES
8.5.1 Example: a site far from any existing data point (generic)
Suppose that we want to design a 90-foot long and 15-foot diameter suction
caisson for a site relatively far from any existing investigation location. From the
example and Figure 7.4 in Section 7.5.1, the mean and standard deviation for the design
side shear capacity and the design end bearing are:
The net weight of caisson, W ' , is assumed to be 270 kips. Assuming no correlation
between Qside and Qend , the mean and standard deviation in ultimate axial capacity can
be calculated as:
Typical values of ΩS and ΩR are 0.2 and 0.4, respectively. The ΩR value for this
example can be reduced because the weight of suction caisson has no uncertainty.
Suppose that β t arg et is 4.0 and the load and capacity are modeled with a lognormal
188
exp(βt arg et Ω'2R + ΩS2 + δ2R ) exp(4.0 0.3602 + 0.202 + 0.09482 )
FSδR = = =1.0441
exp(βt arg et Ω'2R + ΩS2 exp(4.0 0.3602 + 0.202
Without Accounting
for Spatial Variability
75 in Resistance
Accounting for
Spatial Variability
in Resistance
Caisson Length (ft)
80
without Lower Bound
85 Accounting for
Spatial Variability
in Resistance
90 with rLB/rmedian=0.4
95
100
Figure 8.7: Design Axial Capacity versus Caisson Length for Generic Site (D=15ft)
189
8.5.2 Example: P-J1 site with nearby data points
Recall from Section 7.5.4 that we are interested in design a 90-foot long and 15-
foot diameter suction caisson at P-J1 location. From the example and Figure 7.9 in
Section 7.5.4, the mean and standard deviation for the design side shear capacity and the
design end bearing are:
The net weight of caisson, W ' , is assumed to be 270 kips. Assuming no correlation
between Qside and Qend , the mean and standard deviation in total capacity can be
calculated as:
Typical values for ΩS and Ω R are assumed to be 0.2 and 0.4, respectively. The
Ω R value for this example can be reduced because the weight of suction caisson has no
uncertainty.
Suppose that β t arg et is 4.0 and the load and capacity are modeled with a lognormal
190
exp(βt arg et Ω'2R + ΩS2 + δ2R ) exp(4.0 0.3562 + 0.202 + 0.0977 2 )
FSδR = = =1.0472
exp(βt arg et Ω'2R + ΩS2 exp(4.0 0.3562 + 0.202
Table 8.2: Example-Extrapolated Axial Capacity and Partial Factor of Safety for a 15-
ft Caisson ( ΩS =0.2, Ω R =0.4, βt arg et =4.0)
191
Axial capacity (kips)
0 500 1000 1500 2000 2500 3000
0
10 Existing Design
Exsisting Design-LB
20 Extrapolated - Mean
Extrapolated-Mean-LB
Depth of penetration (ft)
Generic
30
Generic-Lower Bound
Extrapolated Factored
40
50
60
70
80
90
Without
Accounting
75 for Spatial
Variability
in Resistance
Caisson Length (ft)
80
Accounting for
85 Spatial Variability
in Resistance
with rLB/rmedian=0.5
90
Accounting for
95 Spatial Variability
in Resistance
without Lower Bound
100
193
Chapter 9: Conclusions
9.1 CONCLUSIONS
The following major conclusions have been developed from this research:
1. Based on the design undrained shear strength profiles for the past 20 years in this
Gulf of Mexico deepwater area, the design undrained shear strength varies
194
spatially but is not dependent on the time the site investigation was performed or
the methods used to perform the site investigation.
2. The mean value for the design undrained shear strength increases approximately
linearly with depth at a rate of 8.5 psf/ft.
3. The standard deviation in the design undrained shear strength increases from the
mudline to a depth of about 100 feet and then keeps almost constant from that
point on to a depth of 150 feet. The corresponding coefficient of variation for the
design undrained shear strength decreases from about 0.35 at the mudline to 0.12
at a depth of 150 feet.
4. The correlation distance for the design undrained shear strength depends on
direction and depth below the mudline. The vertical correlation distance is
approximately 25 feet. The isotropic horizontal correlation distance increases with
depth, with a value of 4,400 feet (0.8 miles) at the mudline and a value of 15,500
feet (2.9 miles) at a depth of 150 feet.
5. The mean value for the depth-averaged design undrained shear strength increases
approximately linearly with depth at a rate of 4.2 psf/ft.
6. The standard deviation for the depth-averaged undrained design undrained shear
strength is two times smaller than that for the undrained shear strength at a point.
However, the coefficient of variation values for the depth-averaged undrained
shear strength are similar to those for the point undrained shear strength at any
given depth except 65 to 120 feet below the mudline where the ratio of coefficient
of variation values for the depth-average to the point shear strength is about 0.8 .
7. The spatial correlation structure for the depth-averaged undrained shear strength
has larger correlation distances both horizontally and vertically. This result means
that a spatial correlation model of the point shear strength based on only the first
195
two moments (means and covariances) is not adequate in capturing fully the
spatial correlation in undrained shear strength. Therefore, separate models for the
point and depth-averaged design undrained shear strengths are required if they are
constrained to the second-moment formulations. The vertical correlation distance
is 30 feet and the horizontal correlation distance increases from 6,000 feet (1.1
miles) at the mudline to 19,500 feet (3.7miles) at a depth of 150 feet for the depth-
averaged undrained shear strength.
8. The spatial correlation structure in the horizontal direction is better fit with an
anisotropic model, where the correlation distance is larger along the continental
shelf than it is transverse to the shelf. This result is significant in that it
demonstrates that the depositional forces, which would move approximately
transverse to the shelf, are an important factor causing spatial variations in the
undrained shear strength. The hemipelagic deposits (about the upper 60 feet)
shows greater variation and less spatial correlation in depth-averaged shear
strength than the deeper turbities which was deposited by a relatively low sea
level.
9. The mean value for the design undrained remolded shear strength at a point and
the depth-averaged design undrained remolded shear strength increase
approximately linearly with depth at a rate of 3 psf/ft and 1.5 psf/ft, respectively.
10. The standard deviation for the design undrained remolded shear strength is about
30 % smaller than that for the undrained undisturbed shear strength. However, the
coefficient of variation values for the design undrained remolded shear strength
are two times greater than those for the design undrained undisturbed shear
strength.
196
11. The spatial correlation structure for the design undrained remolded shear strength
has larger vertically and smaller horizontally. The vertical correlation distance is
approximately 35 feet and the horizontal correlation distance is about 6,000 feet
(1.1 miles)
12. The minimum value of design undrained undisturbed shear strength at each depth
is greater than the maximum value of design undrained remolded shear strength at
each depth. This result is significant because it indicates that the lower-bound
design axial capacity can be used to bound spatial variations in the ultimate
design axial capacity, which limits the impact of spatial variability in achieving a
reliable design.
13. The design undrained remolded shear strength has a minor correlation with the
design undrained undisturbed shear strength ( R 2 <0.37), indicating that the
remolded shear strength cannot be reliably estimated from the undisturbed shear
strength at that location.
14. The depth-averaged undrained shear strength is highly correlated with the
equivalent linear undrained shear strength for lateral capacity, indicating that the
lateral capacity is approximately proportional to the depth-averaged shear
strength. If a suction caisson yields higher axial capacity than the field average,
then there is a higher chance it will also have lateral capacity higher than the field
average.
15. The reliability accounting for uncertainty due to spatial variability is typically not
a large impact on a geotechnical design. The required increase in the safety factor
is less than about 10 % (the partial additional safety factor of 1.1) for the study
area.
197
16. The use of the models developed in this study is restricted to the study area or
very similar geologic settings. However, the design methodology is general and
can be applied to other fields.
In future work, the following activities are recommended to build upon this
research:
1. Instead of developing models for the point values and the depth-averaged values,
a better model for the point values with 3rd, 4th, 5th, etc. cross moments needs to
be developed to capture the higher-order relationships in the point values.
2. In future calibration efforts, a parameter calibration process needs to be updated
by an advanced technique based on Bayesian approach that can account for the
uncertainty in the estimates so that the model parameters are treated as random
variables.
198
Appendix. Spreadsheet Programs Developed
In this appendix, spreadsheet programs developed for this study are presented.
factor with depth, and the bearing capacity factor. Outputs from the program are the
depth-averaged shear strength, the unit side shear, the depth-averaged unit side shear, the
unit end bearing, the side shear capacity, the end bearing capacity, the effective caisson
weight and the total axial capacity with caisson length. The calculation worksheet for this
program is shown in the figure below.
199
A2. SPREADSHEET PROGRAM FOR LATERAL CAPACITY OF SUCTION CAISSON
The lateral capacity of a suction caisson in clay can be estimated using the
methodology described in Section 3.3. Inputs to the program are dimensions of a suction
caisson, the undrained shear strength with depth, the soil unit weight with depth, the α
factor with depth, and the load attachment depth. Outputs from the program are the depth
to the center of rotation and the ultimate lateral capacity of a suction caisson at a given
length and a given load attachment point. The calculation worksheet for this program is
shown in the figure below.
The design model (described in Section 3.3) for this spreadsheet program is the
same as that for Fall16 which is better known as SAIL and developed by Aubeny et al.
200
(2000, 2003). The UT lateral capacity program can accommodate both the linear and
non-linear profiles of submerged unit weight of soil and the undrained shear strength,
while only a linear profile of undrained shear strength and an average value of submerged
unit weight of soil can be used for Fall 16. The UT lateral capacity program uses
trapezoidal rule to calculate the integral with the greater resolution (by discretizing the
entire caisson length with a tenth-foot interval).
The lateral capacity at the given location is calculated for three cases: 1) using
FALL16 with a linear profile, 2) using the UT program with a linear profile, and 3) using
the UT program with an actual profile. All inputs, except the actual profiles of design
shear strength and unit weight of soil, are shown in the table below.
Nomenclature
Lf = Length of caisson
D = Diameter of caisson
Su0 = Intercept of linear shear strength profile
Su1 = Slope of linear shear strength profile
γ' = Average effective unit weight of soil
Li = Load attachment depth at center
Hmax = Lateral capacity
Lo = Depth to center of rotation
201
The actual design profiles for all data sets are included in the UT program. The
results of lateral capacity analyses are shown in the table below. The calculated lateral
capacities from the UT program with an actual profile are about the same as those from
FALL16 and the UT program with a linear profile. One of advantages of the UT lateral
program is ease of calculations because the user does not need to fit a linear design
profile in order to calculate design capacity. It also will provide a better estimate of
spatial variations in lateral capacity for non linear design shear strength profiles. It should
be mentioned that the UT program does not deal with inclined load capacity and it is only
for pure lateral capacity.
202
References
Altuntas, C. (2002) “Development of a statistical model for laterally loaded pile analysis
to reduce uncertainty in an offshore field” Master thesis, University of Texas at
Austin, 210 pp.
American Petroleum Institute (1995), Recommended practice for design and analysis of
station keeping systems for floating structures, API-RP-2SK.
Andersen, K.H. and Jostad, H.P. (1999), "Foundation Design of Skirted Foundations and
Anchors in Clay," Proceedings of the Offshore Technology Conference, OTC
Paper No. 10824, Houston.
Anderson, J.B., Rodriguez, A., Abdulah, K., Banfield, L.A., Bart, P., Fillon, R.,
McKeown, H., and Wellner, J., (2004): "Late Quaternary stratigraphic evolution
of the northern Gulf of Mexico: a synthesis." Special Publication, Society of
Sedimentary Research.
Ang, A. H-S. and Tang, W. H. (1975) “Probability Concepts in Engineering Planning and
Design, Volme I – Basic Principles”, John Wiley & Sons, Inc.
Ang, A. H-S. and Tang, W. H. (1984) “Probability Concepts in Engineering Planning and
Design, Volume II - Decision, Risk and Reliability”, John Wiley & Sons, New
York.
ASTM D1587 (2008) “Standard Practice for Thin-Walled Tube Sampling of Soils for
Geotechnical Purposes”, American Society for Testing and Materials.
ASTM D2573 (2008) “Standard Test Method for Field Vane Shear Test in Cohesive
Soil”, American Society for Testing and Materials.
203
ASTM D2850 (2003) “Standard Test Method for Unconsolidated-Undrained Triaxial
Compression Test on Cohesive Soils”, American Society for Testing and
Materials.
ASTM D3441 (2005) “Standard Test Method for Mechanical Cone Penetration Tests of
Soil”, American Society for Testing and Materials.
ASTM D3550 (2001)Standard Practice for Thick Wall, Ring-Lined, Split Barrel, Drive
Sampling of Soils”, American Society for Testing and Materials.
ASTM D4648 (2005) “Standard Test Method for Laboratory Miniature Vane Shear Test
for Saturated Fine-Grained Clayey Soil”, American Society for Testing and
Materials.
ASTM D4823 (1995) “Standard Guide for Core Sampling Submerged, Unconsolidated
Sediments”, American Society for Testing and Materials.
ASTM D5778 (2007) “Standard Test Method for Electronic Friction Cone and Piezocone
Penetration Testing of Soils”, American Society for Testing and Materials.
ASTM D6282 (1998) “Standard Guide for Direct Push Soil Sampling for Environmental
Site Characterizations”, American Society for Testing and Materials.
ASTM D6528 (2007) “Standard Test Method for Consolidated Undrained Direct Simple
Shear Testing of Cohesive Soils”, American Society for Testing and Materials.
Aubeny, CP, Han, SW., Murff , JD.(2003) “Inclined load capacity of suction caissons”.
International Journal for Numerical and Analytical Methods in
Geomechanics,2003;27:1235-1254.
Aubeny, C.P., Moon, S.K., and Murff, J.D. (2001). “Lateral undrained resistance of
suction caisson anchors,” International Journal of Offshore and Polar
Engineering, Vol. 11, No. 2, pp. 95-103.
Aubeny, C.P. and Murff, J.D. (2003), “Simplified limit solutions for undrained capacity
of suction anchors,” Int. Symposium on Deepwater Mooring Systems, Houston,
Proc., ASCE, pp. 76-90.
204
Bea, R.G., Jin, Z., Valle, C., and Ramos, R. (1999) “Evaluation of Reliability of Platform
Pile Foundations”, Journal of Geotechnical and Geoenvironmental Engineering,
ASCE, Vol. 125, No. 8, pp. 696-704.
Benjamin, J. R. and Cornell, C. A., (1970) Probability, Statistics, and Decision for Civil
Engineers, McGraw-Hill, Inc., New York.
Bolton, M.D. (1993) “What are Partial Factors For?”, Proceedings of the International
Symposium on Limit State Design in Geotechnical Engineering, Vol. 3, pp. 565-
583.
Chen, W. F., (1975). Limit Analysis and Soil Plasticity, Elsevier Publishing Co.,
Amsterdam, The Netherlands.
Cho, G.C., Lee, J.-S. and Santamarina, J.C. (2004), “Spatial variability in soils: high
resolution assessment with electrical needle probe”, J. Geotech. Geoenviron. Eng.
130 pp. 843–850
Choi, Y. J., Gilbert, R. B., Ding, Y., Zhang, J. (2006) ‘Reliability of mooring systems for
floating production systems’, Final Report for Minerals Management Service,
Offshore Technology Research Center, College Station, Texas, 90 pp.
Clukey, E.C., Aubeny, C.P. and Murff, J.D. (2003), “Comparison of analytical and
centrifuge model tests for suction caissons subjected to combined loads,” Proc.
OMAE, Paper 37503.
Curray, J.R., 1960 “Sediment and history of Holocene transgression, continental shelf,
northwest Gulf of Mexico”, in F.P. Shepard, F.B. Phleger, and T.H. van Andel,
eds., Recent sediments, northwest Gulf of Mexico: Tulsa, American Association
of Petroleum Geologists, pp. 221-266.
Earth Observatory (2009) “Sediment in the Gulf of Mexico” Earth Obsetory online
magazine, EOS Project Science Office, NASA Goddard space flight center,
Posted November 13, 2009.
205
Ellwood, B. B; Balsam, W. L.; and Roberts, H. H. (2006) “Gulf of Mexico sediment
sources and sediment transport trends from magnetic susceptibility measurements
of surface samples” Marine Geology, Volume 230, Issues 3-4, 24 August 2006,
Pages 237-248
Finley, C. A. (2004) “Designing and analyzing test Programs with censored data for civil
engineering applications,” Ph.D. Dissertation, The University of Texas at Austin,
253 pp.
Fenton, G.A., (1994), “Error evaluation of three random field generator”, ASCE Journal
of Engineering Mechanics, 120(12), 2487-2497.
Flemings, P.B., Behrmann, J.H., John, C.M., and the Expedition 308 Scientists, (2006).
Proc. IODP, 308: College Station TX (Integrated Ocean Drilling Program
Management International, Inc.).
Gambino, S. J (1998), “Pile Capacity Models for an Offshore Field,” Master thesis,
University of Texas at Austin, 176 pp.
Gambino, S. J. & Gilbert, R.B. (1999), “Modeling Spatial Variability in Pile Capacity for
Reliability-Based Design,” Analysis, Design, Construction and Testing of Deep
Foundations, Roesset Ed., ASCE Geotechnical Special Publication No. 88, 135-
149.
Gilbert, R.B., Fulvio, T, Silva, C.T., and Maidment, D.R (2006), “Visualizing
Uncertainty with Uncertainty Multiples,” Proceedings, GeoCongress 2006:
Geotechnical Engineering in the Information Technology Age pp. 1-6
206
ISFOG 2005, Frontiers in Offshore Geotechnics, Perth, Western Australia, 317-
324.
Gilbert, R.O. (1987). “Statistical Methods for Environmental Pollution Monitoring”. John
Wiley& Sons, Inc., New York.
Gore, R.H. (1992) "The Gulf of Mexico. Pineapple Press", Inc. Sarasota Florida. 384 p.
Horsnell, M.R., and Toolan, F.E. (1996) “Risk of Foundation Failure of Offshore Jacket
Piles”, Proceedings, Offshore Technology Conference, OTC 7997, pp. 381-392.
Jeanjean, P.; Andersen, K. H.; and Kalsnes, B. (1998) “Soil Parameters for Design of
Suction Caissons for Gulf of Mexico Deepwater Clays”, Offshore Technology
Conference, 4 May-7 May 1998, Houston, Texas
Journel, A. G. and Huijbregts, Ch. J. (1978) Mining Geostatistics, Academic Press Inc.,
San Diego.
Luke, A.M., Rauch, A.F., Olson, R.E., and Mecham, E.C. (2003). “Behavior of suction
caissons measured in laboratory pullout tests,” Proceedings of the 22nd
International Conference on Offshore Mechanics and Arctic Engineering,
OMAE03, Cancun, Mexico, June 8-13, Vol. 3, pp. 839-847.
Matlock, H., 1970. “Correlations for Design of Laterally Loaded Piles in Soft Clay”,
Proc. Offshore Technology Conference, OTC Paper 1204, Houston.
207
Matlock, H. and Reese, L. C., (1960), “Generalized Solutions for Laterally Loaded
Piles”, Jrnl. of Soil Mechanics and Foundation Engineering, ASCE, Vol.86, No.
SM5.
Minerals Management Service. Gulf of Mexico OCS Region. (May 2000). Gulf of
Mexico Deepwater Operations and Activities. Environmental Assessment. U.S.
Department of the Interior, Minerals Management Service, Gulf of Mexico OCS
Region, New Orleans. Figure III-1.
Neubecker, S.R. and Randolph, M.F. (1995), “Profile and frictional capacity of
embedded anchor chains,” Journal of Geotechnical Engineering, ASCE, 121 (11),
787-803.
Olson, R. E., Rauch, A. F., Gilbert, R. G., Tassoulas, J. L., Aubeny, C. P. and Murff, J.
D., (2001), “ Toward the Design of New Technologies for Deep-Water Anchors”,
Proceedings ISOPE, Stavanger.
O’Neill, M. P., Bransby, M. F., and Randolph, M. F. (2003), “Drag anchor fluke-soil
interaction in clays”, Canadian Geotechnical Journal, Vol. 40, pp. 78-94.
Ostermeier, R. M.; Pelletier, J. H.; Winker, C. D.; Nicholson, J. W.; Rambow, F. H.; and
and Cowan,K. M. (2000) “Dealing with Shallow-Water Flow in the Deepwater
Gulf of Mexico”, Offshore Technology Conference, 1-4 May 2000, Houston,
Texas
208
Quiros,G.W., Little, R.L. (2003) “Deepwater Soil Properties and Their Impact on the
Geotechnical Program” Proceedings, Offshore Technology Conference, Houston,
Texas, OTC 15262, 1-11.
Quiros, G.W., Young, A.G., Pelletier, J.H. and Chan, J.H.-C. (1983), “Shear strength
interpretation for Gulf of Mexico clays”, Geotechnical Practice in Offshore
Engineering, ACSE Conf. pp. 144–165.
Randolph, M.F. and House, A.R. (2002), “Analysis of suction caisson capacity in clay,”
Proceedings of the Offshore Technology Conference, OTC paper 14236, Houston.
Rice, J. A. (1995), Mathematical Statistics and Data Analysis, Second Edition, Duxbury
Press, Belmont, California.
Richardson,M. D., Muzi, E. Troiano, L. and Miaschi, B. (1991), “Sediment Shear Waves:
A Comparison of In Situ and Laboratory Measurements”, Book:The
Microstructure of fine-grained sediments, from mud to shale By Richard Harold
Bennett, William R. Bryant, Matthew H. Hulbert, Chapter 44, 403-415.
Rodriguez, S.G.H., Henriques, C.C.D., Ellwanger, G.B., and de Lima, E.C.P. (1998)
“Structural Reliability with Truncated Probability Distributions”, Proceedings of
the International Conference on Offshore Mechanics and Arctic Engineering,
OMAE, July, 5-9, 1998, Lisbon, Portugal, 6 pp.
Ruinen, R. and Degenkamp. (2001), “Anchor selection and installation for shallow and
deepwater mooring systems,” Proc., ISOPE, Stavanger.
Stahl, B., Aune, S., Gebara, J. M. and Cornell, C. A. (1998) ‘Acceptance criteria for
offshore platforms’, Proceedings, Conference on Offshore Mechanics and Arctic
Engineering, OMAE98-1463.
Tang. W.H. and Gilbert, R.B (1989), “Statistics of Spatial Average in a Random Two-
State Medium” International Symposium on Methods of Stochastic Mechanics
and Application, Urbana, IL, Vol 6, 271-282.
Tang, W. H. and Gilbert, R. B.: "Case Study of Offshore Pile System Reliability," Res.
Report for Project PRAC 89-29, API (1992) 156.
209
Tang, W. H. (1987), “Updating anomaly statistics—single anomaly cases.” Struct.
Safety, 4/2, 151–163.
Tang, W. H., and Halim, I. (1987). “Updating geological anomaly statistics based on
boring log information.” Structural Research Series Report, Dept. of Civ. Engrg.,
Univ. of Illinois at Urbana, Urbana, Ill
Tang, W. H., and Quek, S.T., (1986), “Statistical model of boulder size and fraction”, J
Geotech Engng ASCE 112 1 pp. 79–90.
Templeton, J.S. (2002), “The role of finite elements in suction foundation design
analysis,” Proceedings of the Offshore Technology Conference, OTC Paper
14235, Houston.
Thorne, C. P. (1998), “Penetration and load capacity of marine drag anchors in soft clay”,
Journal of Geotechnical and Geoenvironmental Engineering, Vol. 124, No. 10,
pp. 945-953.
Tribus, M. (1969), Rational Descriptions, Decisions and Designs, Pergamon, New York.
Turner, R.E. (1999) "Inputs and outputs of the Gulf of Mexico". Chapt. 4, In: Kumpf, H.,
Steidinger, K. and Sherman, K. (eds.) The Gulf of Mexico large marine
ecosystem; assessment, sustainability and management. Blackwell Science, Inc.
704 pp.
Valdez-Llamas, Y.P. (2003), “Spatial Variability of the Marine Soil in the Gulf of
Mexico” Proceedings, Offshore Technology Conference, Houston, Texas, OTC
15266, 1-6.
Vanmarcke. E.H, (1983), “Random fields: Analysis and Synthesis”, The MIT Press
(1983).
Vivatrat, V., Valent, P. J. and Ponterio, A. (1982), “ The influence of chain friction on
anchor pile design”, Proceedings of theOffshore Technology Conference, OTC
4178, Houston.
Vryhof Anchors (1999), Anchor Manual 2000, Krimpen ad Yssel, The Netherlands.
Wang, D. (2002) ‘Development of a method for model calibration with non-normal data’,
Ph.D. Dissertation, The University of Texas at Austin, 265 pp.
210
Whitman, R. V. (1984) ‘Evaluating calculated risk in geotechnical engineering’, Journal
of Geotechnical Engineering, ASCE, 110 (2), 145-188.
Wu, T.H., Tang, W.H., Sangrey, D.A., and Baecher, G. (1989) ‘Reliability of offshore
foundations’, Journal of Geotechnical Eng., ASCE, 115 (2), 157-178.
Young, A.G., Honganen, C.D., Silva, A.J. and Bryant, W.R. (2000), “Comparison of
geotechnical properties of large-diameter long cores and borings in deep water
Gulf of Mexico”, Proc., Offshore Technol. Conf. 12089
Vryhof Anchors. (2000). “Vryhof Anchor Manual”, Vryhof Anchors bv, The Netherlands
Wilde, B., Treu, H., Fulton, T. (2001), “Field testing of suction embedded plate anchors”,
Proceedings, ISOPE.
211
Vita
Jeong Yeon Cheon was born in Gwangju, Korea in 1976 as the youngest child of
Il Young Cheon and Keum Nyeon Goh. After graduating Daekwang women’s high
school located in Gwangju, Korea in 1994, she enrolled at Korea University in Seoul,
Korea in 1995 and majored in civil engineering. She earned a Bachelor of Engineering
degree in February 1999. Right after her graduation, she began graduated studies at the
Korea University at Seoul, Korea and specialized in geotechnical engineering under
supervision of Dr. In-Mo Lee. After earning an M.S. degree in February 2001, she
worked for ESCO Consultant Co., Ltd., Seoul, Korea for 5 years as an engineer. She
entered the University of Texas at Austin in the fall of 2006 to begin work on her Ph.D.
degree in geotechnical engineering in under supervision of Dr. Robert B. Gilbert.
212