NORTHWESTERN UNIVERSITY
Automated Monitoring and Inverse Analysis of a Deep Excavation in Seattle
A THESIS
SUBMITTED TO THE GRADUATE SCHOOL
IN PARTIAL FULLFILLMENT OF THE REQUIREMENTS
For the Degree
MASTER OF SCIENCE
Field of Civil and Environmental Engineering
By
Miltiadis Langousis
EVANSTON, ILLINOIS
March 2007
ii
ABSTRACT
Automated Monitoring and Inverse Analysis of a Deep Excavation in Seattle
Miltiadis Langousis
Performance monitoring of deep excavations typically includes slope inclinometers,
optical surveying of soil deformation, tiltmeters and strain gages. Current monitoring data
collection and processing requires time consuming site visits and manual data reduction
by project engineers. Development of robotic and remote access geotechnical
instrumentation conceptually allows processed data to be made available to project
engineers and contractors in “real time.”
Deep excavation design methods usually employ empirical methods and 2-
dimensional (plane strain) finite element analysis, based on soil characteristics
determined by in-situ and laboratory tests. Inaccurate predictions are often produced
when the soil input parameters are not correctly evaluated and when changes in soil
stress, due to ancillary activities, are not taken into account. A number of case histories
and numerical analysis have also demonstrated that deep excavation deflections are
greatly influenced by excavation sequence and 3-dimensional corner restraint (Finno et
al. 2007).
Inverse analysis methods can be implemented in finite element analyses for deep
excavations support design to help minimize uncertainties associated with the soil
constitutive model parameters. The inverse problem employs iterative algorithms to
update selected soil parameters based on the observed soil response and support system
performance in early excavation stages. The updated soil parameters are applied to
iii
simulations of future excavation stages to provide better soil response predictions.
Effective use of inverse analysis in finite element models requires timely collection of
soil response and excavation support system data to ensure that support design may be
updated, if needed, without causing construction delays.
This thesis focuses on the installation and performance of an automated surveying
system, a relatively new and still developing monitoring technique, at the construction of
the Olive 8 Towers in Seattle, WA. It summarizes the philosophy behind the real time
instrumentation and shows how the total station data can complement the conventional
inclinometer data. In addition, it is illustrated how soil stress changes due to ancillary
activities prior to the excavation and careful design can be used to great effect in finite
element simulations. Lastly, the observed responses from the inclinometers were used
for inverse analysis to find parameters that resulted in a good agreement with observed
data and to evaluate whether the soil stiffness input parameters had been accurately
defined in the design stage of the project.
iv
ACKNOWLEDGEMENTS
First and foremost I would like to thank my advisor Professor Richard Finno for
giving me the chance to study at Northwestern. I want to express him my sincere
gratitude for providing hours of advice, guidance and encouragement during this process.
His patience and support especially at the beginning of my academic career in the States
was remarkable and I have to thank him for that. I still believe though, that the Cubs are
more fun than the White Sox.
I also want to thank the members of my committee, Professor Charles Dowding
and Professor Jose Andrade, for their effort and their valuable suggestions and comments.
The help provided by Dave, Mat and David, from the ITI department, during the
automated monitoring set-up and the processing of the monitoring data was very valuable
and has to be acknowledged.
First of all, I would like to say I really big thank you, or as we would say in Greece
“ΕΥΧΑΡΙΣΤΩ” to all the people I know or I met here at Northwestern, either close or
distant friends of mine. Every interaction with you was really valuable for me, helped me
grow and made me the person that I am today. I truly thank you.
Thank you graduate students Izzo, Taesik (Dda), Cecilia, Young-Hoon (“doc”),
Xuxin, Sandi and the fellow “Aquitards” later known as “Earthworms” for making our
geotech group experience entertaining outside our basement (lab). I have to thank also the
people from café Ambrosia; many litters of coffee were consumed there especially when
I was writing this thesis.
I also would like to specifically thank some people that were really close to me
during these two years and we supported each other during happy or moody days. My
v
friend Wan Jei or “Evil doctor Cho” for the numerous hours we spend together in and out
of the lab; any interaction with him was priceless and made me a better person; and
Francesca or “Fra” who tolerated me the most and supported me especially during the
completion of this thesis. I also want to thank my friends in the States: Sotos, Tassos,
Kostas, Stavros, Passant, Derin, Rennaud and my friends in Greece: Dimitris, Rena,
Leonidas, Christoforos and Spyros. And last but not least I would like to thank Greg
Andrianis (with gold and capital letters); his distant help is greatly appreciated.
Finally, I want to thank the people that helped above all others in this effort, my
father Spyros, my mother Kiki and my brother Andreas. Thank you Andrea for all these
hours we spent talking on the phone solving all of my questions and listening to my
problems. Thank you mom and dad for your help and the sacrifices you made to see your
sons studying at a place so far away from you. I know you are proud for both of us.
vi
Table of Contents
Acknowledgements……………………………………………………………………...iv
Table of Contents………………………………………………………………………..vi
List of Figures…………………………………………………………………………….x
List of Tables……………………………………………………………………….......xiv
1. Introduction…………………………………………………………………………...1
2. 8th and Olive Excavation……………………………………………………...………4
2.1. Olive 8 Project Overview…………………………………………………………...5
2.1.1. Soil Geology and Stratigraphy………………………………………………6
2.1.2. Adjacent Structure Qwest Building………………………………….………8
2.1.3. Excavation Support System………………………………………………...10
2.1.3.1. Pile #6………………………………………………………………...14
2.1.3.2. Pile #13……………………………………………………………….14
2.1.3.3. Pile #18……………………………………………………………….15
2.2. GeoEngineers Finite Element Model……………………………………………...16
2.2.1. Constitutive Model ………………………………………………………...16
2.2.2. Finite Element Mesh………………………………………………………..17
2.2.3. Excavation Sequence Simulation…………………………………………..19
2.2.4. Numerical Model Predictions…………………………………....…………20
2.3. Project Instrumentation……………………………………………………………21
2.3.1. Slope Inclinometers………………………………………………………...22
vii
2.3.2. Automated Optical Surveying……………………………………………...23
2.3.2.1. Prism Location………………………………………………………..26
2.3.2.2. Leica TPS 1100 Professional Total Surveying Station……………….28
2.3.2.3. Instrument Errors……………………………………………………..30
2.3.2.4. Rigid Body Translation and Rotation Corrections……………………36
2.3.2.5. Data Acquisition Software……………………………………………41
2.3.2.6. Remote Communication Methods……………………………………42
2.4. Summary…………………………………………………………………………..44
3. Field Performance…………………………………………………………………...46
3.1. Construction Sequence…………………………………………………………….46
3.2. Wall Deformation………………………………………………………………….49
3.2.1. Slope Inclinometers………………………………………………………...49
3.2.2. Automated Optical Surveying……………………………………………...53
3.2.3. Comparison of Inclinometer and Optical Survey Data…………………….56
3.3. Summary of Performance………………………………………………………….61
3.3.1. Total Station Difficulties During Construction…………………………….63
3.3.2. The Problem of the Rigid Body Rotation Correction………………………63
3.3.3. Summary…………………………….………………………………...……64
3.4. Predicted and Observed Responses.……………………………………………….64
3.5. Summary…………………………………………………………………………..65
4. Alternate Finite Element Model And Inverse Analysis…………………………...68
4.1. Finite Element Model ……………………………………………………………..68
viii
4.1.1. Finite Element Mesh………………………………………………………..69
4.1.2. Constitutive Model and Input Parameters………………………………….72
4.1.3. Excavation Sequence……………………………………………………….77
4.2. Finite Element Results and Analyses……………………………………...………79
4.2.1. Drained Analysis…………………………………………………………...79
4.2.2. Undrained Analysis………………………………………………………...85
4.3. Analysis of Results………………………………………………………………...88
4.3.1. 3-D effects………………………………………………………………….88
4.3.2. Numerical Simulation of Tiebacks…………………………………………90
4.3.3. Summary……………………………………………………………………94
4.4. Inverse Analysis…………………………………………………………………...95
4.4.1. Procedures………………………………………………………………….95
4.4.2. Selection of Field Observations for Inverse Analysis……………………...98
4.4.3. Finite Element Simulations for Optimization……………………………..104
4.4.4. Optimization for Fully Drained Analysis…………………………………105
4.4.4.1. Optimized Parameters……………………………………………….105
4.4.4.2. Results……………………………………………………………….107
4.4.5. Optimization for Undrained Clayey Silt Analysis………………………...119
4.4.5.1. Optimized Parameters……………………………………………….119
4.4.5.2. Results……………………………………………………………….121
4.4.6. Discussion…………………………………………………………………134
4.5. Conclusions………………………………………………………………………135
ix
5. Summary and Conclusions………………………………………………………...138
References……………………………….……………………………………………..142
x
LIST OF FIGURES
Figure 2.1 Picture of the site taken by Google Earth……………………………………...4
Figure 2.2 The construction site at the beginning of the project………………………….5
Figure 2.3 Soil profile defined by boring logs…………………………………………….7
Figure 2.4 Proximity of Qwest building…………………………………………………..9
Figure 2.5 Plan view of the construction with the design sections………………………10
Figure 2.6 Design Sections………………………………………………………………13
Figure 2.7 Finite Element mesh for a typical design section…………………………….18
Figure 2.8 Predictions for the wall deflection given by Plaxis…………………………..21
Figure 2.9 Inclinometer on Pile #6………………………………………………………23
Figure 2.10 Total station on parapet wall………………………..………………………24
Figure 2.11 Location of monitoring prisms on site………………………………………25
Figure 2.12 Points screwed and glued on the wall………………………………………26
Figure 2.13 Location of monitoring points and fixed point #7…………………………..27
Figure 2.14 Location of fixed point #8…………………………………………………..28
Figure 2.15 Image of Leica TPS 1101…………………………………………………...29
Figure 2.16 Instrument axes……………………………………………………………..30
Figure 2.17 Line of sight error…………………………………………………………..31
Figure 2.18 Titling axis error…………………………………………………………….32
Figure 2.19 Vertical axis error…………………………………………………………...33
Figure 2.20 Compensator error…………………………………………………………..34
Figure 2.21 V-Index error………………………………………………………………..35
xi
Figure 2.22 Collimination error………………………………………………………….35
Figure 2.23 Rigid body translation and rotation correction……………………………...38
Figure 2.24 Hardware used for remote communication…………………………………44
Figure 3.1 Deflection for Pile #6………………………………………………………...50
Figure 3.2 Deflection for Pile #13……………………………………………………….51
Figure 3.3 Deflection for Pile #18……………………………………………………….52
Figure 3.4 Data acquired by automated surveying at Pile #6……………………………54
Figure 3.5 Displacement along the wall of the monitored points………………………..55
Figure 3.6 Comparison of inclinometer and optical survey movement towards and along
the excavation for Pile #6………………………………………………………………..58
Figure 3.7 Comparison of inclinometer and optical survey movement towards and along
the excavation for Pile #13………………………………………………………………59
Figure 3.8 Comparison of inclinometer and optical survey movement towards and along
the excavation for Pile #18………………………………………………………………60
Figure 3.9 Data lost when the instrument went out of balance………………………….62
Figure 3.11 Error in calculating the settlement when the vertical compensator was
OFF………………………………………………………………………………………64
Figure 3.12 Comparison of predicted versus observed wall deflection………………….66
Figure 4.1 Finite element mesh for typical design section………………………………71
Figure 4.2 Relative shear stress before applying the changes in soil stress history……...80
Figure 4.3 Shear stress levels (psf) after applying the changes in soil stress history……81
Figure 4-4 This study, compared to the GeoEngineers predictions and the observed
xii
deflections for Pile #6……………………………………………………………………82
Figure 4-5 This study, compared to the GeoEngineers predictions and the observed
deflections for Pile #13…………………………………………………………………..83
Figure 4-6 This study, compared to the GeoEngineers predictions and the observed
deflections for Pile #18…………………………………………………………………..84
Figure 4.7 Comparison of drained and undrained finite element analysis in contrast with
the observed and predicted displacement profiles……………………………………….87
Figure 4.8 Effects of plan dimensions on PSR (after Finno et al, 2007)………………...89
Figure 4.9. Effects of plan dimensions and depth of excavation on PSR………………..90
Figure 4.10 Deflection profiles for tiebacks and equivalent struts………………………93
Figure 4.11 Observations vs. instrument error for conventional and in-place inclinometer
at Pile #6. ………………………………………………………………………………100
Figure 4-12. Comparison of inclinometer errors……………………………………….101
Figure 4.13 Observed and stated by manufacturer instrument error…………………...102
Figure 4.14 Observed versus the calculated and optimized deflection for three optimized
parameters………………………………………………………………………………107
Figure 4.15 Objective function, RFI (%) and stiffness parameters versus the number of
iterations………………………………………………………………………………...109
Figure 4.16 Observed versus the calculated and optimized deflection for a three
optimized parameters…………………………………………………………………...111
Figure 4.17 Objective function and stiffness parameters versus the number of
iterations….……………………………………………………………………………..113
xiii
Figure 4.18 Optimization results and wall deflection versus number of optimized
parameters………………………………………………………………………………114
Figure 4.19 Shear strains (%) in function with depth and distance from the West
excavation wall………………………….……………………………………………...118
Figure 4.20 Observed versus the calculated and optimized deflection for three optimized
parameters………………………………………………………………………………121
Figure 4.21 Objective function RFI (%) and stiffness parameters versus the number of
iterations………………………………………………………………………………...124
Figure 4.22 Observed versus the calculated and optimized deflection for a three
optimized parameters…………………………………………………………………...125
Figure 4.23 Objective function RFI (%) and stiffness parameters versus the number of
iterations………………………………………………………………………………...127
Figure 4.24 Observed versus calibrated displacement deflection profiles for two and three
optimized parameters…………………………………………………………………...128
Figure 4.25 two parameter optimization results for drained and undrained Clayey Silt
analysis………………………………………………………………………………….131
xiv
LIST OF TABLES
Table 3.1 Excavation sequence during construction for Pile #6…………………………46
Table 3.2 Excavation sequence during construction for Pile #13………………………..47
Table 3.3 Excavation sequence during construction for Pile #18………………………..48
Table 3.4. Final movement recorded for monitored points……………………………...56
Table 4.1 Soil Parameters used for design……………………………………………….74
Table 4-2 Comparison of K0 values from GeoEngineers and this analysis……………..74
Table 4.3 Properties of concrete elements……………………………………………….75
Table 4.4 Stiffness of structural elements………………………………………………..76
Table 4-5. Pre-stress load for West wall tiebacks………………………………………..78
Table 4-6. Pre-stress load for East wall tiebacks………………………………………...78
Table 4-7 Representation of Tiebacks by equivalent struts……………………………...92
Table 4.8 Observations used for inverse analysis………………………………………104
Table 4-9. Results of sensitivity analysis……………………………………………….105
Table 4.10 Optimization for three parameters………………………………………….110
Table 4.11 Optimization for two parameters…………………………………………...113
Table 4.12 Overview of Stiffness Input Parameters……………………………………115
Table 4-13. Results of sensitivity analysis……………………………………………...119
Table 4.14 Optimization for three parameters………………………………………….122
Table 4.15 Optimization for two parameters…………………………………………...126
Table 4.16 Overview of Stiffness Input Parameters……………………………………129
Table 4.17 Overview of Stiffness Parameters…………………………………………..133
Chapter 1. Introduction
1
Performance monitoring of deep excavations typically includes slope inclinometers,
optical surveying of soil deformation, tiltmeters and strain gages. Current monitoring data
collection and processing requires time consuming site visits and manual data reduction by
project engineers. Development of robotic and remote access geotechnical instrumentation
conceptually allows processed data to be made available to project engineers and
contractors in “real time.”
Deep excavation design methods usually employ empirical methods and 2-
dimensional (plane strain) finite element analysis, based on soil characteristics determined
by in-situ and laboratory tests. Inaccurate predictions are often produced when the soil
input parameters are not correctly evaluated and when changes in soil stress, due to
ancillary activities, are not taken into account. A number of case histories and numerical
analysis have also demonstrated that deep excavation deflections are greatly influenced by
excavation sequence and 3-dimensional corner restraint (Finno et al. 2007). Inverse
analysis methods can be implemented in finite element analyses for deep excavations
support design to help minimize uncertainties associated with the soil constitutive model
parameters. The inverse problem employs iterative algorithms to update selected soil
parameters based on the observed soil response and support system performance in early
excavation stages. The updated soil parameters are applied to simulations of future
excavation stages to provide better soil response predictions. Effective use of inverse
analysis in finite element models requires timely collection of soil response and excavation
2
support system data to ensure that support design may be updated, if needed, without
causing construction delays.
The purpose of this study was to test the performance of an automated survey system,
a relatively new and still developing monitoring technique, summarize the philosophy
behind the real time instrumentation and show how the total station data can complement
the conventional inclinometer data. In addition, it will be illustrated how soil stress
changes due to ancillary activities and careful design can be used to great effect in finite
element simulations.
Construction of the Olive 8 Towers in Seattle, WA, provided the opportunity to
install both traditional and developmental geotechnical instrumentation. Traditional
monitoring instrumentation included slope inclinometers for the shoring wall. To
supplement the traditional data, an automated remote-access total survey station was
deployed to monitor the 3-d wall movements.
Several sets of finite element analyses were performed to evaluate the effects of soil
stresses, tieback representations and 3-d corner restraint on the predicted displacements.
Also the observed responses from the inclinometers were used for inverse analysis to find
parameters that resulted in good agreement between computed and observed data, and to
evaluate how well the soil stiffness input parameters had been defined during the design
stage of the project.
Chapter 2 presents an overview of the Olive 8 project excavation, with particular
attention to shoring wall design for the west side of the excavation. It includes the lateral
wall deflection predictions made by GeoEngineers Inc., the geotechnical consultant of the
3
project, based on numerical simulations of the excavation sequence. The project
instrumentation program is summarized, including slope inclinometers and automated
optical surveying for the shoring wall and the adjacent structures.
Chapter 3 presents the performance monitoring data collected from the
instrumentation program during the excavation. The observed west shoring wall
deflections are summarized and compared to the predictions given by the design finite
element simulation employed in design. The automated monitoring data is compared with
the inclinometer observations at the top of the west shoring wall. Finally, the difficulties
with application of the automated surveying are discussed and the corrections which was
used to process the survey data are presented.
In Chapter 4, a finite element simulation is presented that explicitly accounts for the
changes in stresses history caused by construction of the Qwest building. Stiffening at the
corners of the excavation due to 3-D effects and representation of tieback restraint in the
finite element simulation are also discussed. Results of inverse analysis of this numerical
model are presented, as are the optimized parameters that results in the best fit between the
calculated and observed deflection profiles.
Chapter 5 summarizes this thesis and presents its conclusions.
4 Chapter 2. 8th and Olive Excavation
The 8th and Olive development is a project in downtown Seattle that includes a deep
excavation to accommodate the underground parking facilities. Figure 2.1 shows the
location of the project site from Google Earth. Finite element analyses were conducted in
the design phase of this project because of the complexity of the shoring system. The city
of Seattle imposed a design constraint that the soil would not move more than 1 inch
during the excavation. The movement limitations were checked during the excavation by
instrumentation which included conventional inclinometers and optical survey points and
a robot total station, which is not widely used in practice.
Figure 2.1 Picture of the site taken by Google Earth
5
2.1. Olive 8 Project Overview
The Olive 8 tower development occupies one half of a city block bordered by Eight
Avenue on East, Olive Way on the north and an alley on the west in downtown Seattle,
Washington. The development consists of 33 stories above grade and 5 parking levels
below grade. The excavation for the underground facilities extended as much to 71 ft.
below the ground level. The urban location of the construction and the surrounding
existing buildings required a deep supported excavation. In addition, the proximity of the
Qwest building west of the excavation demanded a rather unique design for the adjacent
shoring wall. Figure 2.2 shows the construction site at the beginning of the project during
the installation of the soldier pile wall
North Wall
West Wall
East Wall
Figure 2.2 The construction site at the beginning of the project
6
For the North, South, and East shoring walls the retaining structure consisted of a
temporary soldier pile wall with wooden lagging and 5 or 6 rows of tiebacks. This design
was done by HartCrowser Inc. Their October 2000 report also included a proposed design
for the West Shoring wall.
For the West shoring wall, GeoEngineers Inc, of Redmond, Washington, proposed an
alternative design for the retaining structure on August 2005. The west wall was designed
and constructed according to their recommendations. In addition, to predict more
accurate the wall deflections they performed finite element model analyses with PLAXIS
V8.
In their design, the West shoring wall consists of a soldier pile wall with shotcrete, rather
than wood lagging, 9 to 10 rows of soil nails and 4 or 5 rows of tiebacks.
2.1.1. Site Geology and Stratigraphy
Based on boring logs presented in the HartCrowser Inc. geotechnical report (2000),
the soil generally consisted of silty sand over hard silt and clay underlain by very dense
sand. The soil samples from the boring logs were visually classified in the field and then
taken to the laboratory for further testing and classification under a relatively controlled
environment. The soil stratigraphy as defined in the boring logs is shown in Figure 2.3
and is described below.
Fill. The fill consists of loose to dense sand with variable silt and gravel content. It was
anticipated to be as thick as 10 ft. The design unit weight was γ=125 pcf, the cohesion
7
was c=200 pcf and the friction angle is φ= 38o. The SPT resistance varied from 18 to 25
blows per foot.
Silty-Sand (SM). Below the fill, a layer of dense gravelly silty sand was encountered;
horizontal slickenslides were observed in the soil samples. The density of the sand
increased with depth. The thickness of this layer ranged from 30 to 35 feet. The design
unit weight was γ=130 pcf, the cohesion was c=200 pcf with a friction angle of φ= 34o.
These fractured soils are found also in similar sites in downtown Seattle. The SPT
resistance varied from 30 to 60 blows per foot and the water content from 10 to 18 %.
Figure 2.3 Soil profile defined by boring logs (GeoEngineers Inc., 2005)
8
Clayey Silt (ML). A hard, moist clayey silt layer underlay the silty sand layer. The
thickness of the clayey silt ranged from 25 to 30 ft. The layer contained some hard lenses
and some occasional horizontal slickensides. The design unit weight was γ=125 pcf, the
cohesion was c=200 pcf and the friction angle was φ= 38o. The SPT resistance varied
from 25 to 30 blows per foot and the water content from 18 to 30 %.
Very Dense Sand (SP). The bottom of the boring showed very dense wet, gravely sand
underlying the clayey silt. The design unit weight was γ=140 pcf, the cohesion was c=0
pcf and the friction angle was φ= 40o. The SPT resistance was at least 60 blows per foot
and the water content was less than 10 %.
At the time of the drilling, ground water was not encountered in the boreholes.
2.1.2. Adjacent Structure-Qwest Building
The existing Qwest building, as shown in Figure 2.4, is located to the west of the
Olive 8 excavation. A 16 foot wide alley separates the west temporary support wall from
the Qwest building. The Qwest building has several below grade levels that extend to
approximately 52 feet below the ground surface. The exterior walls are supported on a
perimeter 7-foot-thick and 10-foot-wide strip footing. The core of the building rests on a
7-foot-thick, 90 ft x 90 ft wide mat foundation. The design bearing pressure for the Qwest
building perimeter footing and mat foundation is 10 ksf.
9
Qwest Building
Alley
Utility Vault
Figure 2.4 Proximity of Qwest building
In the alley between the Qwest building and the excavation, there are buried power and
communication utilities with 5 utility vaults. The utility vaults are located at the central
part of the west excavation wall and extend 22 feet below the ground surface.
Consequently, 3 separate sections were designed to provide lateral support for the west
wall. Figure 2.5 presents the plan view of the construction including the three design
sections.
10
Figure 2.5 Plan view of the construction with the design sections
Due to the proximity of the Qwest building, the rather conventional soldier pile and
tieback walls used for the North, South and East shoring walls were not feasible. The
shoring system for the West Wall designed by GeoEngineers consisted of soldier piles
with shotcrete lagging and soil nails in its upper reaches, with tiebacks with wood lagging
in it lower reaches.
2.1.3. Excavation Support System
This thesis focuses on the design and performance of the West shoring wall. The
three design sections for this wall are shown in Figure 2.6.
11
Soldier Piles. The soldier piles consist of W27x146 sections set in a drilled shaft which
was filled with concrete following the placement of the beam. The spacing of the soldier
piles was 8 feet, the same as that of the soldier piles used for the Qwest excavation.
Soil Nails. The soil between the Qwest building and the excavation was retained with
several rows of soil nails angled 15 degrees below the horizontal (the number of rows
depend on the wall section). The soil nails consisted of #10 rebar grouted in an 8-inch
diameter hole and are spaced at 3 feet vertically and 4 feet horizontally. Depending on the
wall section, an additional soil nail was added at an elevation approximately 30 feet
below the ground surface at the center of the 3x4 feet pattern. The length of the soil nails
was the maximum possible. The drilling stopped when the drill hit the Qwest basement
wall. The drilling for the soil nail was performed with a rotary auger, and with
subsequent grouting resulted in a load transfer of 3 kips/foot without the use of post
grouting.
Tiebacks. Below the elevation of the soil nails, tieback anchors were used to retain the
wall. The tiebacks were placed as high as possible, while maintaining a minimum 3 foot
clearance below the perimeter strip footing of the Qwest building. The maximum
declination of the tiebacks was 45 degrees below the horizontal. The tieback horizontal
spacing is 8 feet and 4 to 5 rows of tiebacks were deployed, depending on the wall
section. The bonded and unbonded length of the tiebacks is approximately 30 and 25 feet,
respectively. The maximum allowable load transfer design load was 4.5 kips/ft. The
12
drilling was performed with a rotary drill and the anchors were grouted under high
pressure with no regrouting.
As previously mentioned, in the alley between the excavation and the Qwest
building there are buried power and communication utilities in 5 utility vaults. To reduce
the movements and make the soil around the larger utility vault stiffer, grout was injected
at the sides and between of the utility vault and excavation wall after the soldier pile wall
was installed, as shown by the cross hatched area in section Pile #13.
13
Soil Nails
Tiebacks
Figure 2.6 Design Sections
13
14
2.1.3.1. Pile #6 (Design section 1)
The design section 1 refers to the wall section around pile #6 and it is shown in
Figure 2.6a. The top elevation of the fill is at +133.5 ft from the sea level. The beams
used for the soldier pile wall were 83-feet-long, W27x146 sections. Because of presence
of utility vaults, the first 19 feet of the wall are cantilever. Below the cantilevered part,
the wall is retained by 9 rows of soil nails spaced with a 3x4 feet pattern (vertical x
horizontal). After the 4th row of soil nails an extra soil nail was inserted at the center of
the 3x4 foot pattern. Tieback anchors were placed below the soil nails. For section 1, 4
rows of tiebacks are used, spaced 8 feet horizontally and 5 feet vertically. The typical un-
bonded length was 25 feet with a grouted length of approximately 30 feet. The bottom of
the excavation is at +71.0 feet. The wall extends about 20 feet below the bottom of the
excavation. Hereafter, Section 1 will be referred to as Pile # 6.
2.1.3.2. Pile #13 (Design section 2)
Design section 2 refers to the wall sections around pile #13 and includes the large
utility vault as shown in Figure 2.6b. The top elevation of the fill is at +131.5 ft over the
sea level. The beams used for the soldier pile wall were 86-feet-long, W27x146 sections.
Due to the size of the utility vault, the first 25 feet of the wall are cantilevered. Below the
cantilevered part, the wall is retained by 5 rows of soil nails spaced with a 3x4 feet
pattern (vertical x horizontal), with an additional soil nail at the center of the 3x4 foot
pattern. Tieback anchors were placed below the soil nails. For section 2, 5 rows of
15
tiebacks were used, spaced 8 feet horizontally and approximately 5 feet vertically. The
typical un-bonded length was 25 feet and the grouted length was approximately 30 feet.
The bottom of the excavation is at elevation +60.0 feet, with the wall extending 15 feet
below this elevation. It needs to be mentioned; that the strength of the soil surrounding
the utility vault in the alley, was improved by grout injections. Hereafter, Section 2 will
be called as Pile # 13.
2.1.3.3. Pile #18 (Design section 3)
The design section 3 refers to the wall section around pile #18 and is shown in
Figure 2.6c. The top elevation of the fill is at elevation +129.5 ft over the sea level. The
beams used for the soldier pile wall were 83 feet long W27x146 sections. Because of the
presence of the utility vault the first 18 feet of the wall was cantilever. Below the
cantilevered part, the wall is retained by 9 rows of soil nails spaced with a 3x4 feet
pattern (vertical x horizontal). After the 4th row of soil nails an extra soil nail was inserted
at the center of the 3x4 foot pattern. Tieback anchors follow the soil nails. For section 3,
5 rows of tiebacks are used, spaced 8 feet horizontally and 5 feet vertically. The typical
un-bonded length was 25 feet with a grouted length of approximately 30 feet. The bottom
of the excavation is at +60.0 feet. The wall extends about 15 feet below the bottom of the
excavation. Hereafter, Section 1 will be referred to as Pile # 18.
16
2.2. GeoEngineers Finite Element Model
GeoEngineers numerically simulated the excavation with the finite element
program PLAXIS V8. This program is a two dimensional (2D) finite element code
written for soil and rock mechanics analysis. One can represent the soil stratigraphy and
the structural elements as well as responses of soil and structural elements, while
numerically simulating the construction process. Since PLAXIS V8 is a 2D program, the
shoring system is modeled in plane strain conditions, where an equivalent 1-foot wide
section of the wall is analyzed with corresponding structural properties scaled by the
horizontal spacing of the structural elements.
2.2.1. Constitutive Model
The stress-strain behavior of the soil layers described in section 2.1.1 was
represented in PLAXIS V8 by the hardening-soil model (PLAXIS Manual, 2002). The
Hardening-Soil includes soil dilatancy and a volumetric yield surface that isotropically
expands due to plastic straining. Shear hardening is also included to model irreversible
plastic strain due to primary deviatoric loading.
Some features of the Hardening-Soil model are:
• Stress dependent stiffness according to a power law, defined by the power of
stress-level dependency of stiffness, m.
• Plastic straining due to primary deviatoric loading, defined by the secant stiffness
in standard drained triaxial test, E50ref.
17
• Plastic straining due to primary compression, defined by the tangent stiffness for
primary oedometer loading, Eoed,ref.
• Elastic unloading reloading, defined by the unloading and reloading stiffness
Eur,ref and the Poisson’s ratio for unloading and reloading, νur.
• Failure according to the Mohr-Coulomb model, defined by the cohesion c, the
friction angle φ and the dilatancy angle ψ.
2.2.2. Finite Element Mesh
The finite element mesh (Figure 2.7) used by GeoEngineers to model the
excavation was 134 feet high and 235 feet long. The mesh encompasses the adjacent
Qwest building and almost the entire excavation width. The left (West) and the right
(East) boundaries extend 134 feet and 100 feet away from the west excavation wall,
respectively. It has to be mentioned that the East boundary is not the symmetry axis of
the excavation since the entire width of the excavation is almost 110 ft. This mesh did not
include the west wall of the Qwest building and the east excavation wall. The lower
bound of the mesh was set at the elevation of +0.0 feet. For the mesh, 15-node triangular
elements were used. The mesh was more refined near the wall, the soil nails and the
tiebacks. The fixities specified for the mesh boundaries were zero horizontal and vertical
movements for the lower boundary of the mesh and zero horizontal movements for the
left and right boundaries. In addition, zero horizontal movements were used at the East
Qwest rigid wall.
18
Fill
Sand
Clay
Dense Sand
135 ft 100 ft
+134.0
+119.5
+84.0
+59.5
+0.0
West East
Qwest Building
Olive 8 Excavation
Figure 2.7 Finite Element mesh for a typical design section
19
2.2.3. Excavation Sequence Simulation
The construction sequence was modeled by GeoEngineers in PLAXIS V8 in a way
that approximately represented the effects of the Qwest building, the main excavation
stages, and the soil nail and tieback installations. The Qwest building was inserted in the
geometry after the initial geostatic conditions. The east rigid wall was fixed in the
horizontal direction so as to allow only vertical movement. The displacements were set to
zero after establishing the initial geostatic conditions, applying traffic loads in the alley
and installing the Qwest building.
The modeling sequence in GeoEngineers numerical simulation is:
1) Set up model geometry and calculate initial stress conditions.
2) Excavate the Qwest building and install Qwest basement wall.
3) Apply Qwest foundation loading and install Qwest basement wall
4) Install soldier pile wall
5) Excavate 2 feet below soil nail or tieback level
6) Install soil nail or tieback (including post tensioning of tiebacks)
7) Repeat steps 5 and 6 until the bottom of the excavation is reached
During the numerical simulation of the Olive 8 excavation, a rather uncommon
technique was adopted to promote the rigid block failure of the soil and the soil nails in
the alley. This technique was used only for the Upper Sand and Clayey Silt layers and is
described below.
20
Rankine theory implies that the failure plane of the soil behind a retaining wall is
oriented 45 + φ/2 from the current excavation level. As the Olive 8 excavation proceeded,
the interface friction between the soil and the soil nails was reduced by 60% at locations
higher than the point where the failure plane from each excavation level intersected the
Qwest wall. The use of this technique reduces the contribution of the soil nails to the
shoring wall stiffness and forces the retained soil to separate from the interface of the
Qwest wall. However, this technique is considered conservative because the soil nails
slide easier through the retained soil and this increases the predicted wall deflections.
2.2.4. Numerical model predictions
The results of the finite element analysis, for the 3 design sections are shown in
Figure 2.8. For Pile #6 the simulation predicted 0.8 inches of movement for the top of
the wall and 0.8 inches of deep seated movements at the bottom of the excavation. For
pile #13 the result indicate a deflection of 0.5 inches for the top of the wall and deep sited
movements of 1.2 inches. Finally for pile #18, results suggest 0.5 and 1.5 inches of
deflection for the top and the bottom of the excavation respectively. These predicted
movements exceed the limiting movement of 1 inch, which would result in a work
stoppage while the movements and required support system were revaluated. However,
GeoEngineers were satisfied with this design, and were confident that the 1 inch limits
will not be exceeded because they considered their finite element model very
conservative in terms of the wall’s stiffness, the strength of the retained soil and stress
history induced by Qwest building.
21
Pile #6
Predicted Displacement (in.)
0.0 0.4 0.8 1.2 1.6
Ele
vatio
n (ft
.)
60
70
80
90
100
110
120
130
Pile #13
Predicted Displacement (in.)
0.0 0.4 0.8 1.2 1.6
Ele
vatio
n (ft
.)
40
50
60
70
80
90
100
110
120
130
Pile #18
Predicted Displacement (in.)
0.0 0.4 0.8 1.2 1.6
Ele
vatio
n (ft
.)
40
50
60
70
80
90
100
110
120
130
Figure 2.8 Predictions for the wall deflection given by Plaxis
2.3. Project Instrumentation
An instrumentation program was established to monitor the performance of the
shoring system and to provide early detection of deflections that could potentially
damage the nearby structures. This program included installing three inclinometers along
the west wall to measure lateral movements, optical surveying of the shoring wall and
nearby structures before and during the construction, and strain gages on the soil nails
22
and load cells on the tiebacks to monitor the load transfer from the ground to the
structural components of the wall.
If the lateral movements were observed to be in excess of ½ inch between two
successive readings or if the total wall movements exceeded 1 inch, the construction of
the shoring wall would be stopped to evaluate the cause of the movement and to establish
the type and extent of remedial measures required. Based on past performance data of
excavations through similar soil conditions, typical deflections for excavations of this
height were expected to vary from 0.001H to 0.003H (¾ inch to about 2.5 inches).
Clough and O’Rourke (1991).
2.3.1. Slope Inclinometers
Three inclinometers were installed along the west wall attached to the back flange
of piles #6, #13 and #18, as shown in Figure 2.9. The casing was securely attached on
the pile every 10 feet and it was filled with water prior placing the concrete into the shaft.
All inclinometers extended to the bottom of the soldier piles. Inclinometer readings were
performed approximately once a week by GeoEngineers. The first reading was completed
prior to drilling adjacent piles.
It has to be mentioned that at this location, the inclinometers observations are
affected by the stiffness of the Soldier Pile and in common practice they should be
installed 2 to 3 feet behind the shoring wall.
23
Figure 2.9 Inclinometer on Pile #6
2.3.2. Automated Optical Surveying
Northwestern University installed an optical surveying system to monitor
automatically 3D deflections of the top of west wall at the positions where the
inclinometers were located. These same prisms were also placed in the alley behind the
wall. A Leica TPS 1101 Total Surveying Station was placed atop the parapet wall of the
rooftop at the Paramount Hotel, which was adjacent to the south excavation wall. To
access the total station, one had to have access to the Hotel rooms that connected to the
rooftop. To mount the instrument on the wall steel brackets were used. These brackets
were “π” shaped and they were tightened around the front and the back of the parapet
24
wall with the use of rubber pads. The brackets are shown in Figure 2.10. Although the
instrument was fixed on the brackets, it could not be considered as fixed since the
brackets were not screwed into the wall, so they were free to slightly translate or rotate.
A sketch of the prism locations is shown in Figure 2.11. Since this total station
could not be considered fixed in space, the prisms P-7 and P-8 were established at
locations far away from the excavation to serve as a reference baseline for the
measurements.
Figure 2.10 Total station on parapet wall
25
P7
P8
Figure 2.11 Location of monitoring prisms on site.
Readings were obtained for the monitored points 8 times a day. The first reading was at 4
a.m., about 3 hours before work started on site. Readings were taken at 7 a.m., 9 a.m.,
11a.m., 1 p.m., 3 p.m. and 5 p.m., while construction activity occurred. A final reading
was taken at 11 p.m. This cycle would be repeated for every day until the excavation was
completed. For each point, the instrument shoot at a prism, then rotated automatically
180o and the lens rotated also 180o to re-shoot the same prism. Here and on, this will be
referred as two face measurement. This technique eliminates the deviation between two
measurements on the horizontal plane.
26
2.3.2.1. Prism location
To monitor the displacements on the west wall six (6) points were used. Figure 2.12
illustrates the typical mounting of the prisms. Points P1, P3 and P5 were screwed into the
shoring wall adjacent to the piles with the inclinometers. The others were located (Figure
2-12) behind the piles at the other side of the alley and were glued on a small brick wall.
Figure 2.12 Points screwed and glued on the wall
27
Because the total station’s position could translate and rotate, displacements of the
points were calculated with respect to two fixed points that were far away from the
excavation. The first point (Point 7) was placed across of the excavation inside an alley at
a place where it could not easily seen (or maybe stolen by pedestrians). Point 8 was glued
on a building east of the excavation. The exact positions of points 1 through 7 are shown
in Figure 2-13. Point 8 is in Figure 2.14.
Point 1
Point 5
Point 2
Point 4
Point 6
Point 7
Point 3
Figure 2.13 Location of monitoring points and fixed point #7
28
Point 8
Figure 2.14. Location of fixed point #8
2.3.2.2. Leica TPS 1101 Professional Total Surveying Station
A Leica TPS 1101 Professional total surveying station, shown in Figure 2.15, was
used to monitor the displacement of eight optical prisms around the Olive 8 project. The
instrument includes monitoring firmware and robotic, automatic target recognition
capabilities which enabled repeated, automated position measurement of predetermined
points marked with prisms.
Measurement data can be accessed from storage on the instrument flash memory or
via remote communication employing the RS232 serial port of the instrument.
29
According to the manufacturer, the Leica TPS1101 is operational in temperatures
ranging from -20o C to 50o C and up to 95% humidity. Automated point measurements
can be taken within a range of 1000 meters. The stated angular accuracy of the instrument
is 1.5 seconds. The stated accuracy of the instrument is 3 mm at distances up to 300
meters.
Figure 2.15 Image of Leica TPS 1101
30
2.3.2.3. Instrument Errors
To understand the instrument errors one must to understand the instrument axes
(Leica TPS 1100 User Manual). The three axes of a total station are the vertical axis V
(standing axis), the titling axis T and the line of sight L and are shown in Figure 2.16.
Figure 2.16 Instrument axes
Instrument errors occur if the instrument differs from the following conditions:
• The line of sight L is perpendicular to the titling axis T
• The titling axis T is perpendicular to the vertical axis V
• The vertical axis is perfectly vertical
31
Line of sight error
As illustrated in Figure 2-16, the line of sight error is cause by the deviation “c”
between the optical line of sight and the perpendicular to the titling axis. This error
affects all horizontal readings and increases with steep sighting. The effect on the vertical
angle is very small and normally can be ignored. This effect can be eliminated by
precisely defining the line of sight error with the instrument’s on board calibration
function or two face measurements.
Figure 2.17 Line of sight error
Titling axis error
As shown in Figure 2.18, the titling axis error is caused by the deviation “a” of the
mechanical titling axis from the line perpendicular to the vertical axis. The titling axis
error can be observed when the telescope is moved vertically along a vertical line and the
32
crosshair moves away from the vertical line although the instrument has not turned in the
horizontal. This error increases with steep sightings, but has no effect on horizontal
sightings. The effect on the vertical angle is small and can be ignored. This effect can be
eliminated as well by precisely defining the titling axis error with the on board calibration
function, or with two face measurements.
Figure 2.18 Titling axis error
Vertical axis error
The vertical axis error (also called the standing axis error), as shown in Figure 2.19
is not an instrument error, but a set up error. It occurs if the vertical axis is not truly
plumb. The vertical axis error affects both the horizontal and the vertical angle readings
and cannot be eliminated with two face measurements. It can be avoided by carefully
leveling the instrument. The Leica TPS 1101 has two built-in axis compensators.
33
Leveling the instrument can be roughly done manually using the circular bubble and the
fine leveling to the plump line is automatically performed by the vertical compensator.
Figure 2.19 Vertical axis error
Compensator index error
The principle of the electronic compensator is quite similar to that of a circular
bubble. It consists of a small container filled with a special fluid. Due to gravity the
surface of this fluid is always horizontal. Sensors interact with the surface of this fluid to
determine the longitudinal and lateral tilt of the instrument. The horizontal and vertical
angles are then automatically corrected. As indicated in Figure 2.20, a compensator
index error occurs when the zero point of the compensator is not in the plump line. With
a dual axis compensator the index error of the compensator is divided into two
components, one alongside (longitudinal l) and the crosswise (transversal t) to the
telescope. The longitudinal index error “l” is similar to the V-index error and affects only
34
vertical angle readings. On the other hand the transversal compensator index error “t” is
similar to the titling axis error and mainly impacts the horizontal angle. The effect of both
errors can be eliminated by running the compensator calibration or by two face
measurements.
Figure 2.20 Compensator error
Vertical index error
A vertical index error “I”, as shown in Figure 2.21 exists if the 0o mark for the vertical
reading does not coincide with the instrument’s mechanical axis. The vertical index error
that affects all vertical angle readings is independent from the steepness of the shooting
point. The V-index error has no impact to the horizontal angle. The effect of the V-index
error can be avoided by precisely defining this error with the instrument’s on-board
calibration function or with two face measurements.
35
Figure 2.21 V-Index error
Collimination error
The collimination error, shown in Figure 2.22 is the angular divergence between the line
of sight and the ATR (Automated Target Recognition) camera axis. The horizontal
component of the collimination error affects the horizontal angle where as the V
component affects the vertical angle. The calibration routine allows one to set the
alignment of the centre of the camera with the optical line of sight.
Figure 2.22 Collimination error
36
2.3.2.4. Rigid Body Translation and Rotation Corrections
To evaluate the displacement of the monitored points a 3D rigid body translation
and rotations corrections were applied (L. Mcmillan, Comp136, Lecture 15 at USC,
1996). The correction procedure accounts the fact that the total station is not fixed on the
parapet wall. This transformation is only valid when all coordinate axes (x, y, z) remain
mutually perpendicular.
When the instrument was installed on the parapet, initial readings for all the
monitored points were taken. For this initial reading the total station was considered to be
at the origin of the coordinate system, i.e. at point (0, 0, 0). Thereafter, all data recorded
may have a slightly different origin because the surveying instrument slightly rotated and
translated.
The vectors that the monitored points form will be used to explain the correction
mathematically.
In section 2.3.2 it was mentioned that points 7 and 8 are far away from the
excavation and they did not move during the excavation. This means that the vectors
and should always be the same. However, in practice, due to rotations, translations
and instrument errors, a vector that is formed between two points will be different from
the one formed after the initial reading, and from all subsequent vectors. To correct these
slight changes, the vectors and have to be adjusted by a rotation and translation
matrix so that the vectors are exactly the same as that measured during the initial reading.
78V
87V
78V 87V
37
If point 7 is selected as the reference point, then the initial vector, , and the
vector at a reading i, , are:
oV78
iV78
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
−
−
−
=07
08
07
08
07
08
078
zzyyxx
V (2-1)
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
−
−
−
=ii
ii
ii
i
zz
yy
xx
V
78
78
78
78 (2-2)
In the same way the vectors from Point 7 to all the monitored points can be calculated
( and ……. and ). 071V iV71
076V iV76
The rigid body rotations and translation are shown in Figure 2-23.
38
38
Figure 2-23. Rigid Body Rotation and Translation correction
39
The unit vectors of and are: 078V iV78
078
078
0
78
^
VV
V = , where 078V is the magnitude of the initial vector. (2-3)
i
ii
VV
V78
7878
^= , where iV78 is the magnitude of the vector at a reading i. (2-4)
Let α be a new vector which is given by the cross product of:
0
78
^
78
^
0
78
^
78
^
VV
VVi
i
×
×=α (2-5)
α is a new vector which is perpendicular to the plane formed by vectors and ; it is
the vector that determines how much the current vertical axis z’ must be rotated to be
aligned with the initial vertical axis z.
078V iV78
If we take the dot product of and we get the cosine of angle θ that rotates the
current x’-y’ plane to the original x-y plane.
078V iV78
( ) 07878cos VV i •=θ (2-6)
and, ( )07878
1cos VV i •−=θ (2-7)
One must compute a rotation matrix, R, given as:
( ) ( ) ( )θθαθα cossin)()cos1( ∗∗∗ ++−= IskewsymR (2-8)
where symα is a symmetric matrix given by the product:
40
Tsym ααα *= (2-9)
and, skew(α ) is a skew matrix which is given by:
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
−
−=
00
0)(
xy
xz
yz
skewαα
αααα
α (2-10)
and I is the identity matrix.
Once R has been determined for each set of readings, the transformed vectors are given
by:
ij
Transj VRV 77 *= , where j is the number of the monitoring point (j=1, …, 6) (2-11)
To take into account the change in distance between and one must proportion the
components by:
078V iV78
iV
VScale
78
078
= (2-12)
So the final corrected vector will be,
Transj
Corrj VScaleV 77 *= , where j is the number of the monitoring point (j=1, …, 6). (2-13)
The monitored points P1 through P6 are expected to move, so any changes in the
corrected vector’s coordinates reflect the point’s displacement.
077 j
Corrj
j
j
j
VVzyx
−=⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
ΔΔΔ
(2-14)
41
The same procedure can be performed starting from Point 8. For this case the vectors are
and , with: 087V iV87
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
−
−
−
=08
07
08
07
08
07
087
zz
yy
xx
V (2-15)
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
−
−
−
=ii
ii
ii
i
zz
yy
xx
V
87
87
87
87 (2-16)
For perfect measurements the transformed and scaled matrix should be identical,
whether we use either Point 7 or Point 8 as a reference. However, there is instrument
error in each measurement. For this reason we use the average of the displacements given
by points 7 and 8.
2.3.2.5. Data Acquisition Software
Infrastructure Technology Institute (ITI) staff wrote a program code in Java to
evaluate the rotation matrix based on the theory expounded in section 2.3.2.4. The
program structure that they followed was based on work by Blackburn (2005) related to
the excavation of the Ford Center in Evanston, IL. The Java code could detect if the data
collection was successful and, if it was unsuccessful, what type of error occurred. Data
collection and remote operation errors can occur due to power loss, communication
errors, obscured prisms, and instrument disturbance.
42
In the same script, there was the ability to allow the user to alter the frequencies of
the data collection.
2.3.2.6. Remote communication Methods
In section 2.3.2.1, it was mentioned that the monitoring data could be accessed
either from storage on the instrument’s flash memory or via remote communication by
employing the RS232 serial port of the total station. In this project, the second option was
used because it provided data without human intervention. From the time the instrument
takes the readings until it is displayed on a project website, a hardware communication
system intervenes to transfer the data. Figure 2.24 shows the setup that was used on site
to obtain the data. The total station was connected through a RS232 port to a computer
that stored the monitoring data in a hard drive. A wireless and a dial up modem were
connected on this computer to provide access to the internet. Power for these instruments
was provided by an electrical line after it was transformed to the proper voltage.
The sequence of data acquisition and transfer was:
1) A reading was taken and stored in the computer via the RS232 port. During every
measurement cycle, 2 sets of measurements were taken at each target point. Each set of
measurements contained a two-face reading. Thus, after each shot to a target, the
instrument rotated 180o, the lens rotated also by 180o and re-shot at the same point. In
each measuring cycle there were 4 data points for each reported measurement.
43
2) The computer connected to the internet at preset time intervals and transferred the
recorded data to a computer at Northwestern University. The first connection attempt was
through the wireless modem.
3) If the wireless modem failed to connect, the computer tried to connect to the internet
through the dial up model.
4) If both modems failed to connect, then the computer waited until the next reading was
taken and attempt to connect again.
5) At Northwestern University, the transferred data were inserted into a software where
the 4 readings for each point were averaged, and then the displacements were computed
using the correction procedures discussed in Section 2.3.2.4.
To protect the hardware from the weather, a waterproof box was used which was
mounted on steel brackets as shown in Figure 2-24.
44
Computer
Dial up modem Wireless modem
Power supply
Figure 2.24 Hardware used for remote communication
2.4. Summary
Olive 8 project included an 70 feet deep excavation to provide space for 5
underground parking levels. The shoring system is rather unique, along the West shoring
wall, due to the proximity of the Qwest building. The maximum allowed displacement,
imposed by the city of Seattle, was 1 inch. The subsurface investigation pointed that the
soil consisted of fill, dense gravelly silty sand, hard moist clayey Silt and very dense
gravely sand (going downwards). GeoEngineers Inc. designed the West wall shoring
system, which consisted of a soldier pile wall with shotcrete lagging, 9 to 10 rows of soil
45
nails and 4 or 5 rows of tiebacks. Three separate sections were designed to provide lateral
support for the West wall, due to the presence of a large utility vault between the
excavation and the Qwest building.
Numerical analysis performed by GeoEngineers showed that the maximum
predicted wall deflections vary from 0.8 to 1.5 inches which are higher than the 1 inch
limiting displacement. However they proceeded with the proposed design because
shoring wall and the soil layers were modeled conservatively and the actual behavior was
expected to be stiffer than the numerically predicted behavior.
An instrumentation program was established to monitor the performance of the
West shoring wall. It included three conventional inclinometers to measure the lateral
wall movements, strain gages and load cells on the soil nails and tiebacks respectively.
An automated surveying system was also installed to monitor the 3D deflections of the
top of the West wall, at the positions where the inclinometers were located. The wall
movements were calculated relative to reference baseline points that were established far
away from the excavation. Readings were taken every day throughout construction until
the excavation was completed. Rigid body translation and translation corrections were
performed to account for the fact that the total station was not rigidly attached to the
parapet wall where it was mounted.
46 Chapter 3. Field Performance
3.1. Construction sequence
Tables 3.1, 3.2 and 3.3 present the construction procedure at the three design
sections along the west wall, as recorded for daily construction reports made by
GeoEngineers site personnel.
Construction Stages at Pile # 6 Stage
Number Activity Description 1 Sheet Pile Installation. 2 Excavation to +117.3 3 Soil Nails Row 2 Installation 4 Excavation to +113.4 5 Soil Nails Row 3 Installation 7 Excavation to +106.4 8 Soil Nails Rows 4, 5 and 6 Installation 9 Excavation to +101.4
10 Soil Nails Row 7 Installation 11 Excavation to +97.4 12 Soil Nails Rows 8 and 9 Installation 13 Excavation to +93.4 14 Soil Nails Row 10 Installation 15 Tieback Row A Installation 16 Excavation to +86.9 17 Tieback Row B and C Installation 18 Excavation to +75.5 19 Tieback Row D Installation 20 Excavation to +71.5
Table 3.1 Excavation sequence during construction for Pile #6
47
Construction Stages at Pile # 13 Stage
Number Activity Description 1 Sheet Pile Installation. 2 Excavation to +103.2 3 Soil Nails Rows 6 and 7 Installation 4 Excavation to +99.2 5 Soil Nails Row 8 Installation 6 Excavation to + 97.7 7 Soil Nails Row 9 Installation 8 Excavation to +93.2 9 Soil Nails Row 10 Installation
10 Tieback Row A Installation 11 Soil Excavated to +86.2 12 Tieback Row B Installation 13 Excavation to +74.2 14 Tiebacks Rows C and D Installation 15 Excavation to +65.2 16 Tieback Row E Installation 17 Excavation to +61.2
Table 3.2 Excavation sequence during construction for Pile #13
48
Construction Stages at Pile # 18 Stage
Number Activity Description 1 Sheet Pile Installation. 2 Excavation to +114.7 3 Soil Nails Rows 2 and 3 Installation 4 Excavation to + 110.5 5 Soil Nails Row 4 Installation 6 Excavation to +103 7 Soil Nails Rows 5 and 6 Installation 8 Excavation to +98 9 Soil Nails Rows 7 and 8 Installation
10 Excavation to +94.5 11 Soil Nails Row 9 Installation 12 Excavation to +92.5 13 Soil Nails Row 9 Installation 14 Tieback Row A Installation 15 Excavation to +84.5 16 Tieback Row B Installation 17 Excavation to +75.5 18 Tiebacks Rows C and D Installation 19 Excavation to +64.5 20 Tieback Row E Installation 21 Excavation to +60.5
Table 3.3 Excavation sequence during construction for Pile #18
The excavation sequence assumed in the numerical analysis made during the design
portion of the project by GeoEngineers, was described in Chapter 2.2.3. During
construction, the excavation sequence differed from that assumed in design. For example,
the specifications dictated that before installing any soil nail or tieback, the soil was not
to be excavated more than 2 feet below the installation level. However, the contractor
excavated soil below this specified depth, and occasionally two or even three rows of soil
nails were installed at the same time. In addition, the shotcrete lagging was to be placed
within 24 hours after the soil was exposed, but during construction the soldier pile wall
remained without lagging for as much as three days. While the former could be easily
49
incorporated into a finite element analysis of the excavation, the latter could only be
accounted for if a material-rate dependent model was used to represent the soil stress-
strain responses.
3.2. Wall Deformation
Slope inclinometers and automated surveying instruments were installed to monitor
the response of the wall due to the excavation process.
3.2.1. Slope Inclinometers
Pile #6: A cantilever movement profile was observed as the soil was excavated to
elev. 97.1 ft., as summarized in Figure 3-1. The maximum movement at the top of the
soldier pile wall was 0.42 inches. The reinforcing effect of soil nails is clearly in the
small lateral deflection that was observed at the level of the soil nails. As the excavation
was made to elevation +71.5 ft., small lateral movements were observed on the order of
0.1 inch. Most of the cantilever movement was obtained by the time that the excavation
level reached +97.1 ft, or about ½ the full depth of the case. Thereafter small deep seated
movements developed.
50
Pile #6
Deflection (in.)-1.0 -0.5 0.0 0.5 1.0
Elev
atio
n (ft
.)
60
70
80
90
100
110
120
130
May-25 July-20 September-7
+71.5
+97.1
+121.5
Qwest
Figure 3.1 Deflection for Pile #6
Pile #13: Pile #13 was the wall section where the large utility vault was located. In
Figure 3.2 it is observed that most of the lateral movements occurred after the excavation
depth exceeded elevation +97.0 ft. Due to the presence of the vault and the jet grouted
soil, the displacement at the top of the wall was 3 times less that the one observed at pile
#6; or approximately 0.1 inches. At this wall section the deep-seated movements were
51
slightly larger that the ones recorded at the cantilever section (0.16 versus 0.13 inches),
and were larger at a location about 5 feet above the bottom of the excavation.
Pile #13
Deflection (in.)-1.0 -0.5 0.0 0.5 1.0
Ele
vatio
n (ft
.)
50
60
70
80
90
100
110
120
130
May-25 July-20 September-7
+119.0
+97.0
+61.2
Jet-Grout
Utility Vault
Qwest
Figure 3.2 Deflection for Pile #13
52
Pile #18
Deflection (in.)
-1.0 -0.5 0.0 0.5 1.0
Elev
atio
n (ft
.)
50
60
70
80
90
100
110
120
130
May-25June-20September-7
+117.0
+94.5
+60.5
Qwest
Figure 3.3 Deflection for Pile #18
Pile #18: Response at Pile #18 is similar to that at Pile #6, as indicated in Figure
3.3. The maximum movement at the top of the soldier pile wall was 0.34 inches. At
about 10 feet above the final end of the excavation, deep-seated movements of about 0.08
inches were observed. Most of the cantilever movement developed by the time that the
excavation reached elevation +94.5 ft. After this point, small deep-seated movements
53
occurred. Clearly, in all cases the lateral movements were less than the 1 inch maximum
value targeted in design.
3.2.2. Automated Optical Surveying
As mentioned in Chapter 2, the survey prisms were placed on top of the soldier pile
wall and on the brick wall behind the alley at the sections monitored by the inclinometers.
Recorded data for Pile #6 are shown in Figure 3.4 where the displacements towards and
along the excavation, as well as the settlements are plotted versus time. The movements
parallel to the excavation are negligible. The “gaps” between the collected data are
attributed to failure of the total station to perform the point measurement which will be
discussed in section 3.3.1. The settlements recorded after July 10th do not represent the
actual response, as it will be discussed in Section 3.3.2. Figure 3.5 presents the deflection
towards the excavation for all monitored points. For Point #2, there are no data after the
2nd of June because the prism was lost due to construction activities. It is seen that the
displacements of the prisms on the brick wall (Points #2, #4 and #6) were negligible.
Table 3.4 summarizes the lateral movement recorded for the monitored points at the end
of the construction.
54
Point #1 - Pile #6
Displacement towards the excavation
4/24 5/15 6/5 6/26 7/17 8/7 8/28 9/18
Dis
plac
emen
t (in
)
-1.0
-0.5
0.0
0.5
1.0
Displacement along the excavation
4/24 5/15 6/5 6/26 7/17 8/7 8/28 9/18
Dis
plac
emen
t (in
)
-1.0
-0.5
0.0
0.5
1.0
Vertical Settlement
4/24 5/15 6/5 6/26 7/17 8/7 8/28 9/18
Settl
emen
t (in
)
-2.0
-1.0
0.0
1.0
2.0
Figure 3.4 Data acquired by automated surveying at Pile #6
55
Point #54/24 5/15 6/5 6/26 7/17 8/7 8/28 9/18
Dis
plac
emen
t (in
)
-1.0
-0.5
0.0
0.5
1.0
Point #34/24 5/15 6/5 6/26 7/17 8/7 8/28 9/18
Dis
plac
emen
t (in
)
-1.0
-0.5
0.0
0.5
1.0
Point #14/24 5/15 6/5 6/26 7/17 8/7 8/28 9/18
Dis
plac
emen
t (in
)
-1.0
-0.5
0.0
0.5
1.0
Point #24/24 5/15 6/5 6/26 7/17 8/7 8/28 9/18
Dis
plac
emen
t (in
)
-1.0
-0.5
0.0
0.5
1.0
Point #44/24 5/15 6/5 6/26 7/17 8/7 8/28 9/18
Dis
plac
emen
t (in
)
-1.0
-0.5
0.0
0.5
1.0
Point #64/24 5/15 6/5 6/26 7/17 8/7 8/28 9/18
Dis
plac
emen
t (in
)
-1.0
-0.5
0.0
0.5
1.0
Figure 3.5 Displacement along the wall of the monitored points
56
Deflection (inches)
Point No
Towards the
excavation Along the
excavation
#1 0.47 0.05 #2 - - #3 0.17 0.02 #4 -0.01 0.08 #5 0.19 -0.01 #6 0.06 0.03
Table 3.4. Final movement recorded for monitored points
3.2.3. Comparison of Inclinometer and Optical Survey Data
The location of points #1, #3 and #5 in regard with the inclinometers at piles #6,
#13 and #18, respectively, provided the opportunity to compare the lateral movement
data from automated surveys to that based on the inclinometer data.
Figures 3.6 and 3.7 compare lateral movements based on total station data at the
location of Pile #6 and Pile #13, respectively. It can be seen that the results agree quite
well throughout the entire construction period and the maximum difference is 10%.
For the section at Pile #18, Figure 3.8 shows that the total station recorded smaller
deflections than the inclinometers throughout construction. As the excavation reached
final depth, the data converged somewhat with the inclinometer data, showing slightly
larger displacements than the optical survey data. The maximum difference between the
optical survey and the inclinometer is 35% (or approximately 0.12 inches).
The total station’s accuracy is mentioned in Section 2.3.2.2 and is stated to be
±3mm (0.1 inch) at shooting distances up to 300 meters. Point 5 (at Pile #18) is located
approximately 45 meters away from the total station. Assuming that the error increases
57
linearly with distance, the error at point 5 is approximately 0.45 mm or 0.015 inches. The
inclinometer’s measurement error at the top of the shoring wall is 0.21 inches (as it will
be discussed in Section 4.2.2). This value is larger than the maximum difference between
the inclinometer and the survey data. Judging from Figures 3-6 and 3-7 and knowing the
accuracy of the Total Station the difference between the monitoring and the survey data,
at pile 18, can be caused by the inclinometer’s measurement error.
58
Displacement towards the excavation
4/24/06 5/15/06 6/5/06 6/26/06 7/17/06 8/7/06 8/28/06 9/18/06D
ispl
acem
ent (
in)
-1.0
-0.5
0.0
0.5
1.0Total Station DataInclinometer Data
Displacement along the excavation
4/24/06 5/15/06 6/5/06 6/26/06 7/17/06 8/7/06 8/28/06 9/18/06
Dis
plac
emen
t (in
)
-1.0
-0.5
0.0
0.5
1.0Total Station DataInclinometer Data
Figure 3.6 Comparison of inclinometer and optical survey movement towards and along
the excavation for Pile #6
59
Displacement towards the excavation
4/24/06 5/15/06 6/5/06 6/26/06 7/17/06 8/7/06 8/28/06 9/18/06D
ispl
acem
ent (
in)
-1.0
-0.5
0.0
0.5
1.0Total Station DataInclinometer Data
Displacement along the excavation
4/24/06 5/15/06 6/5/06 6/26/06 7/17/06 8/7/06 8/28/06 9/18/06
Dis
plac
emen
t (in
)
-1.0
-0.5
0.0
0.5
1.0Total Station DataInclinometer Data
Figure 3.7 Comparison of inclinometer and optical survey movement towards and along
the excavation for Pile #13
60
Displacement towards the excavation
4/24/06 5/15/06 6/5/06 6/26/06 7/17/06 8/7/06 8/28/06 9/18/06D
ispl
acem
ent (
in)
-1.0
-0.5
0.0
0.5
1.0Total Station DataInclinometer Data
Displacement along the excavation
4/24/06 5/15/06 6/5/06 6/26/06 7/17/06 8/7/06 8/28/06 9/18/06
Dis
plac
emen
t (in
)
-1.0
-0.5
0.0
0.5
1.0Total Station DataInclinometer Data
Figure 3.8 Comparison of inclinometer and optical survey movement towards and along
the excavation for Pile #18
61
3.3. Summary of performance
In terms of accuracy, the automated surveying recordings agree with the recordings
taken from inclinometers. However, problems were observed are related to the practical
application of the automated surveying and with the correction which was used to process
the survey data. Both of these issues are discussed in the following sections.
3.3.1. Total Station Difficulties During Construction
As described in Chapter 2 the Leica TPS 1101 has two built-in axis compensators.
The first is the horizontal compensator, which corrects the line of sight error, the tilt error
and the standing axis error (only when the vertical compensator is turned on). The second
is the vertical compensator, which measures the longitudinal and transverse tilts of the
vertical axis, and ultimately brings the instrument back to the plumb line. The working
range of this compensator is ±0.1o from the vertical axis; if the instrument diverges more
than ±0.1o, then the total station shuts down and no readings are made. During the
excavation, the working range of the vertical compensator often was exceeded due to
weather and wind conditions. To reactivate the Total Station, one had to have access to
the total station to manually re-level the vertical axis back to the plumb line.
As previously mentioned in Chapter 2, the Total Station was mounted on top of the
parapet wall on the rooftop of Paramount Hotel. Access to the instrument was limited to a
hotel room window that led to a roof top. When the instrument was out of commission,
one could only access the instrument when the room was not rented. This less that ideal
situation resulted in access being prevented for periods between three days and two
62
weeks, depending on the demand of the room. These access restrictions were the cause of
the gaps in the total station data in Figures 3.5 though 3.8
Displacement towards the excavation for pile #6
4/24/06 5/15/06 6/5/06 6/26/06 7/17/06 8/7/06 8/28/06 9/18/06
Dis
plac
emen
t (in
)
-1.0
-0.5
0.0
0.5
1.0
9 days
16 days 9 days 18 days
7 days
Figure 3.9 Data lost when the instrument went out of balance
In Figure 3.9, the discontinuities between the collected data show that the Total Station
shut down numerous times, during the excavation monitoring. However, even with these
gaps, the trends in the data were defined. As shown in Figure 3.9, these shutdown
periods did not affect much the ability to define trends in responses. However, those
delays are clearly unacceptable in ground applications and real time monitoring, where
total station data are necessary.
63
3.3.2. The Problem of the Rigid Body Rotation Correction
The frequency of the total station shutdown led to the decision to turn off the
vertical compensator in an attempt to avoid losing data. Turning off the vertical
compensator does not affect the horizontal measurements. However, the vertical angles
are affected because they are measured relative to the standing axis which differs now
from the plumb line.
As mentioned in section 2.3.2.4, the rigid body rotation correction is applicable
only when the axes of the coordinate system are perpendicular to each other. So when the
vertical compensator is disengaged, the standing axis is no longer in plumb and the
calculated settlements of the monitored points are not correct. However, the horizontal
deflections along and towards the excavation are close to the inclinometer readings
because the horizontal compensator remained active.
Change of Elevation for pile #6
4/24/06 5/15/06 6/5/06 6/26/06 7/17/06 8/7/06 8/28/06 9/18/06
Dis
plac
emen
t (m
m)
-40
-20
0
20
40
Compensator ON Compensator OFF
Figure 3.10 Error in calculating the settlement when the vertical compensator was OFF
64
In Figure 3.10 the settlements using the rigid body rotation correction (Chapter 2.3.2.4)
are presented. It is obvious that after the compensator was turned off, the scatter in the
data is large unacceptably.
3.3.3. Summary
The lateral displacements based on the automated surveying data of prisms at the
top of the wall agree well with the inclinometer data. The less than rigid connection
between the total station and the parapet throughout the construction caused the total
station to shut down numerous times during the excavation. However, in an attempt to
provide continuous readings the vertical compensator was shut off. The settlement data,
collected during this shutdown, did not reflect the actual response because the standing
axis was no longer vertical to the horizontal plane. However, the horizontal
measurements were correct since the instrument’s horizontal correction was ON.
3.4. Predicted and Observed Responses
The recorded deflection from the inclinometers is compared with the deflections
predicted by GeoEngineers in Figure 3.11 for the three design sections. The deflections
were over predicted, especially for the Upper Sand and Clay layers. However, the
patterns of observed and predicted deflection profiles are similar implying that the
shoring wall and the soil responded more stiffly than expected.
The 1 inch movement limitation specified by the city of Seattle was not exceeded at
any time or at any section of the wall. In particular, the maximum observed lateral
movement was half of this limitation. In addition, the observed deflections at the Upper
65
Sand and Clayey Silt layers were, at certain depths, even 10 times smaller that the
predictions.
During the shoring wall design, the only laboratory data available were the index
properties of the Upper Sand and the Atterberg limits of the Clayey Silt. In addition, the
only triaxial tests available were performed back in 1972 when the adjacent Qwest
building was constructed. The difference between the wall deflections, at the Upper Sand
and the Clayey Silt layers is likely to be the result of conservative selection of soil
strength and stiffness parameters. In the Chapter 4, a finite element model made after the
excavation was completed will be presented. This model accounts for changes in soil
stress caused by construction of the Qwest building. Inverse analysis was performed to
define soil parameters that result in computed lateral wall deflections that are similar to
the observed values.
66
Pile #6
Deflection (in.)
0.0 0.4 0.8 1.2 1.6
Elev
atio
n (ft
.)
60
70
80
90
100
110
120
130
Pile #13
Deflection (in.)
0.0 0.4 0.8 1.2 1.6
50
60
70
80
90
100
110
120
130
Pile #18
Deflection (in.)
0.0 0.4 0.8 1.2 1.6
50
60
70
80
90
100
110
120
Observed DeflectionPredicted Deflection
Fill
Silty Sand
Clayey Silt
Clayey Silt Clayey Silt
Silty Sand Silty Sand
Fill Fill
Dense Sand Dense Sand
Figure 3.12 Comparison of predicted versus observed wall deflection
3.5. Summary
1) The construction sequence followed differed from the sequence used in the
design.
67
2) The inclinometer observations showed the 1 inch movement limitation specified
by the city of Seattle was not exceeded. In particular, the maximum observed
movement was half of this limitation.
3) The horizontal wall deflections towards and along the excavation based on the
total station agreed well with the movement with the movement of the top of the
wall based on inclinometer data.
4) The Total Station shut down because the working range exceeded the vertical
compensator’s capacity due to less than rigid connection with the hotel’s parapet
wall.
5) The rigid body rotation correction requires all axes to be perpendicular to each
other. When the vertical compensator was turned off, the vertical axis was not
perpendicular to the horizontal plane, resulting in unacceptably large scatter in the
settlement data.
6) The design finite element predictions of the magnitude of wall deflection were
larger than that observed. But the observed and predicted displacements, along the
length of the wall, have the same pattern, implying that the shoring wall and the
soil responded more stiffly than expected.
68
Chapter 4. Finite Element Simulation and Inverse Analysis
Chapter 3 showed that the numerically-predicted deflections made during design
were larger than the observed deflections. In this chapter a numerical model is presented
that explicitly accounts for the changes in stresses history of the soil caused by
construction of the Qwest building. This model was subjected to inverse analysis where
the lateral observed wall deflections were compared to the computed values. Because the
inclinometer data were collected manually at this project, these analyses were not
performed in real time, but after the excavation was completed.
4.1. Finite Element Model
The west shoring wall design was too complex to simulate in 3-D without substantial
simplifications. Therefore the excavation was simulated in 2-D assuming that the
complexity of the support system could be reasonably modeled.
The limitations of the pre-construction numerical simulations included:
1) The stress history induced by the construction of the Qwest building was not
considered explicitly. The building was simply inserted into the finite element
mesh with horizontal fixities, approximating the rigidity of the basement wall, and
values of K0 were assigned to the soil that were less than the at-rest values.
2) The soil strength and stiffness parameters used for the analysis apparently were
conservative because the computed movements were significantly larger than
those observed at the West wall.
69
To represent the conditions at the start of the Olive 8 excavation, the finite
element simulation presented in this chapter included a staged excavation for the Qwest
building retained by a soldier pile wall and four levels of struts. After the excavation was
completed, the soldier pile wall was replaced with a rigid wall and the struts with floor
slabs. The soil strength and stiffness parameters were the same initially in both designs
with the exception of the K0 which reflected the values expected for heavily
overconsolidated soil (The K0 values are shown in section 4.1.2). Numerical simulations
were performed for all three design sections. The purpose was to compare the computed
wall deflections with the design predictions, when the effects of the Qwest building
installation are taken into account. Moreover, inverse analysis was performed to calibrate
the soil parameters that provided the best fit between the numerical and the observed wall
deflections.
It is important to note that the only soil data available when selecting the design
values were the index properties of the Upper Sand and the Atterberg limits of the Clayey
Silt, SPT values for the entire soil profile, and triaxial tests performed in 1972 when the
Qwest building was constructed. So the initial choice of the soil strength and stiffness
parameters was primarily a case of applying engineering judgment to a rather scant set of
data.
4.1.1. Finite Element Mesh
Figure 4.1 presents the finite element mesh of a typical design section. The
70
dimensions of the mesh were 900 ft x 135 ft. The left and the right mesh boundaries
extend to a distance of 5H (H defined as the excavation depth) from both Olive 8 shoring
walls. At this distance the boundaries have no influence on the movements near the wall.
15-node triangular elements were used in the mesh to represent the soil and the
Qwest mat foundation. The mesh was more refined around the wall, the soil nails and the
tiebacks. The structural elements (shoring wall, struts, tiebacks, soil nails) supporting the
Qwest building excavation were modeled as elastic materials. In addition, frictional
interface elements were placed between soil and the shoring walls, the soil nails, and the
grouted length of the tiebacks to account for relative slip between the soil and the
structural element. The concrete elements (foundations, slabs) were modeled as elastic
materials. For the bottom of the mesh, zero horizontal and vertical movement was
specified. For the left and right boundaries, only zero horizontal movement was
precluded; the soil was free to move vertically.
71
71
+134.0
+119.5
+84.0
+59.5
+ 0.0
265 ft. 135 ft. 120 ft. 380 ft.
Fill
Silty Sand
Clayey Silt
Dense Sand
Figure 4.1 Finite element mesh for typical design section
72
4.1.2. Constitutive Models and Input Parameters
The numerical analysis adopted the same soil profile as GeoEngineers used for the
design predictions. This profile is described in Section 2.1.1 and was based on the
subsurface explorations performed by HartCrowser Inc. at their report regarding the
Olive 8 excavation in 2000. While the water table was not encountered in any of the
boring logs, both drained and undrained numerical analyses were used to simulate the
behavior of the Clayey Silt. In all cases, drained conditions were assumed for the Fill and
the Silty Sand.
Drained analysis does not generate pore pressures and the finite element model
accounts for volumetric changes that occur due to compression of the voids in the soil
skeleton. This is the case for dry or fully drained soils due to high permeability or low
rate of loading.
Undrained analysis simulates constant volume condition, and thus the excess pore
water pressures. This option is used to model soils when one is interested in short term
stability and displacement estimates.
For this excavation, the subsurface investigation showed that the soil layers were not
dry but moist (10-18% water content for the Silty Sand and 18-30 % for the Clayey Silt
layer). For the case of the Silty Sand, drained analysis can represent relatively accurately
the actual material behavior. The behavior of the Clayey Silt, with its low hydraulic
conductivity, however, cannot be modeled accurately with drained analysis. The actual
73
behavior of a partially saturated, partially drained clay falls between the limits of
drained and undrained analysis. The finite element simulation was made assuming both
drained and undrained analysis for the Clayey Silt layer to compare the computed
deflections based on the two approaches.
The constitutive soil models in the analysis were either the Hardening–Soil, or the
Mohr-Coulomb models. The hardening-soil model represents the response of all soil
layers except for the jet grout around the utility vault in design section Pile #13, which is
assumed to respond as a Mohr-Coulomb material. Table 4.1 summarizes the soil
parameters used in this analysis. All the soil parameters used for the simulation were the
same with the ones used for the GeoEngineers predictions, with the exception of the
lateral stress coefficient, K0. K0 was increased by 40% for the overconsolidated Upper
Sand and the Dense Sand layers more accurately represented in in-situ conditions prior to
the excavation of the Qwest building basement. The K0 values from GeoEngineers and
this finite element analysis are compared in Table 4-2. The dilation angle, ψ, is assumed
to be zero for all the soil layers. The dilation angle is a value that measures the dilatancy
of granular soils during shearing. When ψ is equal to zero the soil expansion due to
shearing is zero.
74
Soil Parameters
Parameter Fill Silty Sand
Clayey Silt
Dense Sand Jet-Grout
Constitutive Law S/H S/H S/H S/H M/C Soil Unit Weight, γ (pcf) 125 130 125 130 145
Friction Angle φ (o) 32 38 34 40 0 Cohession,c (pcf) 100 200 200 0 7000
Lateral Stress Coefficient, K0 0.47 0.6 0.7 0.6 - Poisson's Ratio, ν 0.3 0.3 0.2 0.3 -
Dilation Angle, ψ (o) 0 0 0 0 - Soil Stiffness (ksf) refE50 600 1000 500 1500 -
Unload/Reload Stiffness (ksf) refurE 2600 3000 1600 4500 -
Oedometer Stiffness, (ksf) refoedE 500 1000 700 1500 -
Interface Reduction Factor, Rinter 0.67 0.67 0.67 1 - Reduced Interface Factor, Rinter - 0.2 0.2 - - Reference Pressure pref (atm) 1 1 1 1 1
S/H stands for the Soil Hardening Model M/C stands for the Mohr-Coulomb Model
Table 4.1 Soil Parameters used for design
Lateral Stress Coefficient, K0
Layer GeoEngineers Design
This Analysis
Fill 0.47 0.47 Upper Sand 0.38 0.66
Glaciolacustrine Silt/Clay 0.7 0.7 Lower Sand 0.36 0.6
Table 4-2 Comparison of K0 values from GeoEngineers and this analysis
Tables 4.3 and 4-4 illustrate the input parameters of the concrete and structural
75
elements, respectively. All the elements are assumed to behave as elastic materials.
Similarly to the soil parameters, all concrete elements, soil nails, tiebacks, the West
soldier pile wall and the Qwest rigid wall, have the same properties as did in
GeoEngineers design. The Qwest foundation and basement slab is represented by 15-
node elements. The Qwest shoring wall was assumed to have identical properties with the
Olive 8 West shoring wall. In addition, the struts used to retain the Qwest excavation
were defined as circular steel sections, 24 inches in diameter and 0.5 inches thick. The
vertical and horizontal spacing of the struts was 12 and 30 feet, respectively. The floor
slabs, have the same normal stiffness (EA) as the Qwest rigid wall. The East Olive 8
shoring wall consists of W24x131 sections spaced every 8 ft. The tiebacks used for the
retaining system, have the same properties with the ones used at the West Olive 8 wall.
Parameter Mat Footing
Basement Slab
Element Type 15-node element
15-node element
Unit Weight, γ (pcf) 145 145 Poisson's Ratio, ν 0.15 0.15
Young’s Modulus E (ksf) 7.2 x 105 4.32 x 105
Table 4.3 Properties of concrete elements
76
nt g
I) t)
al
t) t
e Structural ElemeBendin
Stiffness, (E(kip-ft2/f
NormStiffness, (EA)
(kips/fElemen
Typ
West Wall of Olive 8 (W27x146) 05 06 PLATE 1.42 x 1 1.24 x 1East Wall of Olive 8 (W24 x 131) 05 05 PLATE 1.1 x 1 3.1 x 1
Qwest Temporary Wall 05 05 PLATE 1.1 x 1 3.1 x 1Qwest Internal Bracing n/a 05 ANCHORS1.4 x 1
Qwest Rigid Wall 08 08 PLATE 1.89 x 1 3.5 x 1Qwest Slabs /a 08 PLATE n 3.5 x 1Utility Vault 03 05 PLATE 6.1 x 1 2.9 x 1
Soil Nails (4' by 3' spacing) /a 04 D n 6.4 x 1 GEOGRISoil Nails (4' by 3' spacing wi
additional Nail at center patter
th of
n) /a 04 D n 9.6 x 1 GEOGRI
Tieback Anchors (No Load Zone) /a 03 ANCHORSn 8.2 x 1Tieback Anchors (Grouted Body) /a 04 D n 3.5 x 1 GEOGRI
Table 4.4 Stiffness of structural elements
In PLAXIS, Plate elements are defined using a flexural rigidity, an axial stiffness,
EA, and an ultimate bending moment, EI.
Geogrids are flexible elastic elements that represent a grid or a sheet of fabric, they are
used to model soil nails, the grouted part of tiebacks and geomembranes. Geogrids cannot
sustain compressible forces, the only property is the elastic axial stiffness, EA.
Anchors are elastoplastic spring elements that are used to model struts and the non-
grouted portion of tiebacks.
For tiebacks (node-to-node anchors) it is usual to neglect any shear stresses mobilized
between the soil and the free anchor length, so the spring is connected to the wall at one
end and to the soil and fixed anchor length at the other end. A node-to-node anchor
requires specifying the axial unit stiffness, EA.
77
For struts (fixed-end anchors) the anchors are springs that are used to model a
reaction at a single point. The length is defined as the distance between the anchor
connection point and the fictitious point in the axial direction where the displacement is
assumed to be zero. A fixed end anchor requires one to specify the axial stiffness, EA,
and the out of plane spacing.
4.1.3. Excavation Sequence
The excavation sequence for each of the three design sections consists of two parts:
A) The Qwest building construction, which is the same for all three design sections.
B) The Olive 8 excavation which varies for every design section.
The Qwest building construction is simulated by following the steps:
1) Soldier pile installation, with no change in soil stresses
2) Soil excavation to a level 2 feet below each strut level
3) Strut installation
4) Repeat 2, 3 until the lowest strut is installed
5) Excavation to final depth
6) Mat foundation construction by activating the concrete elements representing the
slab and applying the design load 10 ksf for the Qwest foundation and 0.3 ksf for
the traffic loads in the alley
7) Placement of permanent Qwest structure by replacing the soldier pile wall with a
rigid concrete wall and the struts by floor slabs
78
8) Reset displacements to zero before the start of the Olive 8 excavation
The Olive 8 excavation simulation remained almost the same as described in section
3.1. In addition, it included the simulation of the east soldier pile wall and the East wall
tieback installations. The East shoring wall was retained by 6 rows of tiebacks that were
installed in a pattern of 8 ft horizontal and 10 ft vertical spacing. It was assumed that the
East shoring wall was installed at the same time as the West shoring wall. The
excavation grade was uniformly lowered across the site and the east wall tiebacks were
installed at the same time as the lateral support on the West wall that was located at the
same elevation. Tables 4.5 and 4.6 present the pre-stress load applied to each row of the
tiebacks for the West and East walls, respectively.
Pre-stress Load at West Wall Tiebacks
Tieback Row No.
Maximum Pre-stress Load
(kips)
Maximum Pre-stress Load
(kips/ft) 1 113 14.13 2 118 14.75 3 119 14.88 4 127 15.88 5 113 14.13
Table 4-5. Pre-stress load for West wall tiebacks
Pre-stress Load at East Wall Tiebacks
Tieback Row No.
Maximum Pre-stress Load
(kips)
Maximum Pre-stress Load
(kips/ft) 1 62 7.75 2 150 18.75 3 115 14.38 4 125 15.63 5 120 15.00 6 65 8.13
Table 4-6. Pre-stress load for East wall tiebacks
79
4.2. Finite Element Results and Analyses
4.2.1. Drained Analysis
Figures 4.2 and 4.3 show the computed relative shear stresses that occurred as a
result of simulating the Qwest building excavation. They represent the results of the finite
element simulation at the stage right before installing the soldier pile for the Olive 8
excavation. Figure 4.2 presents the relative shear stress as a result of wishing the Qwest
building in place and adding the foundation loads. Figure 4.3 presents the relative shear
stress after simulating the Qwest building excavation. The relative shear stress gives an
indication of the proximity of the stress point to the failure envelope and is defined as:
maxτττ =rel (4-1)
where τ is the maximum value of shear stress (radius in Mohr stress circle) and τmax is
the maximum value of shear stress for the case where Mohr circle is expanded to touch
the Coulomb failure envelope keeping the intermediate principal stress constant. A value
of zero indicates isotropic stress and a value of 1 indicates failure.
In GeoEngineers design (Figure 4-2), the low K0 values at the Silty Sand and Dense
Sand layers generally produced higher relative shear stresses. The soil in the alley, next to
the Qwest building, was not greatly affected by the Qwest building installation. This is
because the entire building was simply inserted in the mesh and the horizontal fixities at
the stiff basement wall did not allow any lateral deflection. Changes in shear stress levels
80
are also observed next to the upper portion of the Qwest basement wall and below
the Qwest foundation as a result of applying foundation loads.
As a result of simulating the Qwest excavation, the relative shear stresses decreased
at the Silty Sand and Dense Sand layers (Figure 4-3). However, in the alley next to the
Qwest building relative shear stresses near failure developed. This happened because the
Qwest shoring wall was allowed to deflect laterally. The maximum relative shear stresses
are located at the lower portion of the Qwest basement wall due to deep seated
movements that developed during the excavation, and below the Qwest foundation as a
result of applying foundation loads.
Figure 4.2 Relative shear stress: No simulation of Qwest excavation
81
Figure 4.3 Relative shear stress: With simulation of Qwest excavation.
Figures 4.4 through 4.6 compare the results of the finite element model simulations
including the Qwest excavation for all design sections with the results from
GeoEngineers (no Qwest excavation simulation) finite element predictions and the
recorded deflections from the inclinometers. In general, taking into account the
installation of the Qwest Excavation in the finite element analysis produced smaller wall
deflections than the simulation without it; however, these computed lateral deflections
were still larger than the observed soil responses. The numerically-calculated deflections
in the fill, for Pile #6 and Pile #18 reasonably agree with the observed at Figures 4.4 and
4.6. At all sections, the largest difference between the observed and the numerically-
calculated deflections occurred at the Upper Sand and the Clayey Silt layers. These layers
82
apparently responded in the field more stiffly than assumed in the finite element
model calculations; the maximum difference in responses occurred at the Clayey Silt
layer. However, the patterns of the observed wall deflections have similar shapes with the
ones calculated by the finite element analysis.
Pile #6
Displacement (in.)
-1.0 -0.5 0.0 0.5 1.0
Dep
th (f
t.)
60
70
80
90
100
110
120
130
Observed Defl.GeoEngineers This Study
Qwest
Figure 4-4. This study, compared to the GeoEngineers predictions and the observed
deflections for Pile #6
83
Pile #13
Displacement (in.)
-0.5 0.0 0.5 1.0 1.5
Dep
th (f
t.)
50
60
70
80
90
100
110
120
130
Observed Defl. GeoEngineers This Study
Jet Grout
Qwest
Utility Vault
Figure 4-5. This study, compared to the GeoEngineers predictions and the observed
deflections for Pile #13
84
Pile #18
Displacement (in.)
-0.5 0.0 0.5 1.0 1.5
Dep
th (f
t.)
50
60
70
80
90
100
110
120
Observed Disp. This StudyGeoEngineers
Qwest
Figure 4-6. This study, compared to the GeoEngineers predictions and the observed
deflections for Pile #18
The difference between the observed and the calculated displacement profiles can be
caused from several factors. First, the inclinometers were located on the piles and the
85
observed profiles are influenced by the stiffness of the soldier piles. In a plane strain
finite element analysis, the stiffness of the soldier pile wall and soil system is smoothed
when the axial stiffness (EA) and the bending moment (EI) of the piles are divided by
their out-of-plane spacing of the soldier piles. Thus, if the inclinometers are not far
enough behind the piles, the results will be influenced by the locally stiff pile section, in
contrast to the average plane strain value.
Other factors that influence the difference between the calculated and the observed
displacement profiles are, the use of drained instead of undrained analysis to model the
Clayey Silt layer, conservative choices of strength and stiffness parameters, 3-D effects,
and the way that the west wall tiebacks were modeled as anchors. All these factors will
be examined further in the following sections.
4.2.2. Undrained Analysis
To provide a lower limit at the computed response in the Clayey Silt layer, finite
element simulations were performed for all design sections with Clayey Silt assumed to
behave as an undrained material. Note that when simulating the Qwest building
excavation, the Clayey Silt was assumed to behave as a drained material because Qwest
was constructed 25 years ago and any excess pore pressures that developed would have
dissipated. While this approach is clearly an approximation, it is intended to serve as an
indicator of the stiffest response one could expect using the selected soil parameters.
Figure 4.7 compares the computed lateral deflections for drained and undrained
86
analysis with the observed and GeoEngineers predicted displacement profiles (no
Qwest excavation). Smaller deformations are seen in the Clayey Silt layer at all the
design sections when undrained analysis is used. The displacements are closer to the
observed displacement profile; however, the computed displacements still exceed the
observations. Other factors contribute to the differences in responses, as will
subsequently discussed.
87
Pile #6
Displacement (in.)
-1.0 -0.5 0.0 0.5 1.0
Dep
th (f
t.)
60
70
80
90
100
110
120
130
Pile #13
Displacement (in.)
-0.5 0.0 0.5 1.0 1.5
50
60
70
80
90
100
110
120
130
Observed Deflection Drained AnalysisGeoEngineersUndrained Analysis
Pile #18
Displacement (in.)
-0.5 0.0 0.5 1.0 1.5
50
60
70
80
90
100
110
120
130
Figure 4.7 Comparison of drained and undrained finite element analysis in contrast with
the observed and predicted displacement profiles
88
4.3. Analysis of Results
4.3.1. 3-D Effects
Figures 4.8 and 4.9 (Finno et al. 2007) show the 3-D effects in deep excavations ias
a function of the excavation depth and the dimensions of the excavation. The results of
the analyses are represented by the plane strain ratio, PSR, defined herein as the
maximum movement in the center of an excavation wall computed by 3-D analyses
divided by that computed by a plane strain simulation. L/He is the ratio of the length of
wall to the excavation depth and L/B is the ratio of the plan dimensions of the excavation,
with L being the side where movements are computed.
Figure 4.8 shows the PSR plotted versus L/B ratio. When the L/B ratio is less than
4, L/HE must be taken into consideration for determining the PSR. The Olive 8 west
shoring wall was 240 feet long and the width of the excavation was 120 feet. This defines
a ratio L/B equal to 2; the PSR varies from 0.75 to 1.0 and depends on the L/HE ratio at
any stage of the excavation.
89
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
PSR,
δ3D
/δ2D
0 1 2 3 4L/B
Figure 4.8 Effects of plan dimensions on PSR (after Finno et al. 2007)
The trends at Figure 4.9 indicate that L/HE ratios greater than 6 result in an
excavation response which has a PSR approximately equal to 1, thus suggesting that
results of plane strain and 3-D analyses will yield the same maximum wall displacement
in the center of the excavation for this geometry. Large differences between plane strain
and 3-D responses are apparent when L/He is less than 2, implying that as the excavation
gets deeper relative to its length, more restraint is provided by the sides of the excavation.
The Silty Sand Layer can be located from 13 to 50 feet below the ground surface.
This defines a ratio L/He which varies from 5 to 18. When the excavation is less than 50
ft deep, or when Silty Sand layer is being excavated, the excavation can be represented
adequately by plane strain conditions. The Clayey Silt layer is located 50 to 75 feet below
the ground surface. When excavating through the Clayey Silt layer at these depths, the
L/He varies from 3 to 5, so the excavation cannot be accurately represented by plane
90
strain conditions, especially when the excavation reaches the full depth.
All other things being equal, this means that the movements that occur when
excavating between 50 and 75 ft. are over-predicted when plane strain finite element
analysis is used to simulate the west shoring wall excavation. The stiffening effects of the
corners of the excavation are not considered in plane strain analysis which accounts for
some of the difference between the computed and observed results.
0 2 4 6 8 10L/HE
120.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
PSR,
δ3D
/δ2D
Silty SandClayey Silt
1 1
Figure 4.9. Effects of plan dimensions and depth of excavation on PSR
4.3.2. Numerical Simulation of Tiebacks
Another parameter that has to be evaluated in order to validate the difference
between the observed and the numerically calculated displacement profiles is the
numerical representation of the tiebacks in the finite element analysis.
91
Tiebacks have a three-dimensional geometry, and their representation in plane
strain analysis involves significant approximations. A question in this analysis was
whether the 2-D representation of the tiebacks brings them so close in the resulting mesh
that the stresses transmitted to the soil overlap, reducing the tieback load-bearing capacity
and producing excessive displacements.
A representation of the tiebacks as (elastic) struts approximates the field conditions
in plane strain without concentrating stresses in the soil near the anchors.
Struts that are equivalent to the tiebacks were modeled as springs that transmit axial
force. The major anchor property is the axial stiffness, EA, entered in the unit of force.
Assuming an elastic behavior, this implies the equivalence:
StrutEquivalentTieback LEA
LEA
⎟⎠⎞
⎜⎝⎛=⎟
⎠⎞
⎜⎝⎛ (4-1)
Therefore the equivalent stiffness of a strut is:
( ) StrutEquivalentTieback
StrutEquivalent LL
EAEA ⋅⎟⎠⎞
⎜⎝⎛= (4-2)
Table 4-7 illustrates the calculations for the axial stiffness of the equivalent struts. For
this calculation, the axial stiffness, EA, of the tiebacks was that used at the finite element
analysis. The spacing was assumed 8 ft, the same with the one used for the tieback
simulation. For a strut length of 60 ft, the EA for the equivalent strut, that is appropriate
for a plane strain analysis, is shown in the last column of Table 4-7.
92
Representation of Tiebacks by Equivalent Struts
Normal Tieback Stiffness
(EA)
Normal Tieback Stiffness
(EA)
Tieback Length EA/L
Length of Equal
Strut
Normal Equal Strut
Stiffness (EA)
Normal Equal Strut
Stiffness (EA)
(kips/ft) (kips) (feet) (kips/ft) (feet) (kips) (kips/ft) Row 1 8200 65600 21.1 3109 60 186540 23318 Row 2 8200 65600 16.9 3882 60 232899 29112 Row 3 8200 65600 15.1 4356 60 261355 32669 Row 4 8200 65600 15.0 4373 60 262400 32800
Stiffness and Length for the tiebacks refer to the non grouted tieback body The normalized values refer for 8 feet spacing
Table 4-7 Representation of Tiebacks by equivalent struts
Figure 4.10 presents the computed deflections at the location of Pile #6 for both
representations of the tiebacks. These simulations were performed when the Clayey Silt
was modeled as a drained material. The results show that modeling the tiebacks as
equivalent struts has little effect on the calculated deflections.
93
Pile #6
Displacement (in.)
-0.5 0.0 0.5 1.0
Dep
th (f
t.)
60
70
80
90
100
110
120
130
Equal Struts Tiebacks
Fill
Silty Sand
Clayey Silt
Figure 4.10 Deflection profiles for tiebacks and equivalent struts
4.3.3. Summary
The finite element numerical analysis that included the Qwest excavation produced a
displacement profile that was closer to the movements observed in field than that based
94
on a similar analysis without an explicit simulation of the Qwest excavation.
However, the computed deflections in the Silty Sand and Clayey Silt layers were still
larger than observed.
The stiffness of the soldier piles influenced the observed deflection profile because
the inclinometers are located on the soldier piles. In a plane strain analysis this stiffness is
smeared because the stiffness of the soldier pile is divided by the out-of-plane spacing of
the soldier pile.
Modeling the Clayey Silt layer as an undrained material reduced the calculated
deflections. However, the deflections at the bottom of the excavation still were larger
than observed. This difference can also be related to the three dimensional stiffening
effects at the corners of the excavation. For the shallower excavation through the Silty
Sand layer, the 3-D effects were insignificant and the shoring wall can be represented by
plane strain conditions. For the deeper excavation through the Clayey Silt layer, the 3-D
effects were significant, especially as the excavation reaches full depth..
Representing the tiebacks as either tiebacks or equivalent struts, had little effect on
the computed lateral deflections.
95
4.4. Inverse Analysis
Because of the differences between the observed and the computed lateral
deflections, inverse analyses were conducted to find the soil parameters that provided the
best fit to the observed lateral deflections. A significant question for this excavation case
history was whether the small observed displacements at the Silty Sand layer, were large
enough to be used for the optimization procedure. For these small lateral displacements to
be useful in inverse analysis, they must be larger than the instrument errors associated
with the inclinometer data (Rechea 2006). Two optimizations were performed for the Pile
#6 design section. Wherein the Clayey Silt was assumed to act as either drained or
undrained.
4.4.1. Procedures
Finno and Calvello (2005) and Finno and Rechea (2006) presented an inverse
analysis procedure that used construction monitoring data to update predictions of
deformations for supported excavation systems. The field observations were obtained
from inclinometers that measured the lateral movements of the soil approximately 6 ft
behind the supporting walls of the excavation throughout the construction. The
constitutive soil responses were represented by the Hardening-Soil model. Of the six
basic input parameters of that model, up to three parameters per layer were optimized
( , , ), while the other parameters were kept constant or related to the
updated value of the optimized value. This methodology was effectively used to calibrate
refE50refurE ref
oedE
96
the model of the excavation based on responses measured at the beginning stages of
construction, so that good predictions can be made of the soil lateral deformation close to
the wall at later stages. UCODE (Poeter and Hill, 1998) was used to optimize the finite
element models for several deep excavation case histories. UCODE is a universal inverse
code that can be used with any application model; it performs gradient-based inverse
modeling and estimates the parameters by calculating the values that minimize the
objective function using non-linear regression.
The difference between the calculated and the monitored quantities is evaluated
through an objective function, also known as error function; the optimum values are the
ones that minimize it. The expression for the weighted least-squares objective
( )bS ' function is:
( ) ( )[ ] ( )[ ] eebyybyybS Tωω =−−= ••
−
'' (4-3)
where:
b : is a vector containing values of the parameters to be estimated,
y : is the vector of the observations being matched by the regression,
( )by ' : is the vector of the computed values which correspond to observations,
ω : is the weight matrix reflecting the error of the measurements,
e : is the vector of residuals, y- y’(b).
This function represents a quantitative measure of the accuracy of the predictions.
97
A quantity called Relative Fit Improvement, RFI, can be employed to indicate
the percent improvement at each stage, i, compared to that based on the initial
predictions:
( ) ( )( )initial
iinitiali bS
bSbSRFI −= (4-4)
where ( )initialbS is the objective function based on the initial estimate of parameters and
( )ibS is the objective function at stage i.
Key factors for successful calibration are a reasonable representation of constitutive
responses of stress history of the soil, wall installation effects and the construction
sequence.
A sensitivity analysis was used to identify the soil parameters most likely to be
successfully calibrated. The sensitivity analysis calculates the relative importance of the
parameters being simultaneously estimated. A useful statistic parameter UCODE is the
composite scaled sensitivity, cssj, which indicates the total amount of information
provided by the observations for the estimation of parameter j, and is defined as:
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
⎟⎟⎠
⎞⎜⎜⎝
⎛⎟⎟⎠
⎞⎜⎜⎝
⎛
∂∂
∗= ∑=
b
ND
jijj
j
ij b
by
NDcss
1
2
2/1'1 ω (4-5)
where
y´i is the ith computed value,
bj is the jth estimated parameter,
98
j
ib
y∂
∂ is the sensitivity of the ith computed value with respect to the jth parameter,
and
ωij is the weight of the ith observation with respect to the jth parameter
The inverse analysis presented herein uses the inclinometer observations and
adjusts the selected parameters until the difference between the observed and calculated
quantities falls below a pre-established threshold. Herein, this threshold was either 10%
change of the objective function, or stabilized RFI over 90% for three consecutive
iterations. One factor that affects the accuracy of the inverse problem is the measurement
error, which depends on the instrument accuracy and error, field conditions and
operator’s ability. The instrument error is the only factor that can be quantified, the
others can not be defined easily but they can be minimized with good field technique.
4.4.2. Selection of Field Observations for Inverse Analysis
The manufacturer specified inclinometer measurement error is ±0.25 mm/m. The
accuracy of the measurement (in meters) versus the measurement depth, d, (in meters
also) is given by:
derror ⋅±=1000
25.0 (4-6)
The 95% confidence interval for the accuracy of a measurement is:
96.1100025.0 d
⋅±=σ (4-7)
99
The weight of each observation depends on the measurement depth and is given by:
)0001.0(11
22 dweight
⋅==
σ (4-8)
Generally, the observations employed in the analysis should to be larger than the
measurement error, and at least one order of magnitude bigger than the 95 % confidence
interval of the accuracy of the measurement (Rechea, 2006).
If all inclinometer measurements were used for the inverse analysis regardless the
measurement error, then each observed displacement is calibrated relative to one
calculated with the predetermined weight given by equation (4-6). Hence, there would be
uncertainty that enters into the parameters optimization because, when selecting the
observations for the inverse analysis, the accuracy of the observations is not considered.
Inclinometer data from Pile 6 was used for the inverse analysis. The data are the
readings taken at the end of the excavation. Recordings from earlier excavation stages
could not be used to optimize the Clayey Silt because reliable deep-seated movements
started to develop only after the excavation level reached an elevation about 2/3 (two
thirds) of the full excavation depth. Moreover, reliable observations in for Silty Sand
layer started to develop after the excavation level reached ½ of the full excavation depth.
Figure 4.11 shows the inclinometer deflection at Pile 6 versus the 95% σ accuracy of the
measurement.
It can be observed that only three readings at the Upper Sand layer are higher than
the 95% σ accuracy of the measurement. The thickness of the Upper Sand layer was 25
100
feet and the observations used were located at the first 5 feet. Within the clay layer,
the readings are larger than the instrument error and most of them can be used for the
optimization. Thus at excavation stages earlier than the final stage, not enough
observations were available for a meaningful calibration.
Pile # 6: Deflection and Instrument Error vs Elevation
Deflection And Error (in.)-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5
Ele
vatio
n (ft
.)
60
70
80
90
100
110
120
130
Deflection95% Confidence
Silty Sand
Clayey Silt
Fill
Figure 4.11 Observations vs. instrument error for conventional and in-place inclinometer
at Pile #6
A way to reduce the instrument error is to use inclinometers with accuracy greater
than the conventional inclinometers or to obtain a site specific measure accuracy. This
can be accomplished by evaluating reading taken at the start of the project before any
101
construction related to ground movement develops. Figure 4.12 compares, at Pile
6, the 95% σ accuracy of a conventional inclinometer versus a more accurate sensor. If a
more accurate sensor was used, then the number of reliable observations could be larger.
Pile # 6: Deflection and Instrument Error vs Elevation
Deflection And Error (in.)
-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5
Ele
vatio
n (ft
.)
60
70
80
90
100
110
120
130
Observed DeflectionConventional Incl. errorIn-place Incl. error
Silty Sand
Clayey Silt
Fill
Figure 4-12. Comparison of inclinometer errors
Figure 4.13 presents the 95% σ accuracy of the measurement stated by the
manufacturer versus the observed error the inclinometer at Pile 6, based on data collected
prior to the start of excavation when the wall has not started yet to deflect. The observed
error can be roughly approximated by a straight line. The equation of the line is:
error (inches) = 8.53•10-4•h (in feet) (4-9)
102
or
error (meters) = 6.38•10-5•h (in meters) (4-10)
where:
h is the height for from the deeper measurement, where the error is assumed to be equal
to zero. The measured error is about 70% less than the manufacturer’s stated value.
Pile #6: Deflection and Instrument Error vs Elevation
Observed and Stated error by Manufacturer (in)
-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5
Dep
th (f
eet)
60
70
80
90
100
110
120
130Fill
SIlty Sand
Clayey Silt
95% Confidence (Manufacturer)
Observed Error
Observed Deflectionerror (in)= + 8.13*10-4 * h (ft)
Figure 4.13 Observed and stated by manufacturer instrument error
103
Table 4.8 lists the observations in the calibration. Observations are named
after the soil layer where they were measured: F for the Fill, S for the Silty Sand layer, C
for the Clayey Silt layer and DS for the Dense Sand layer. The 95% σ measurement
accuracy, stated by the manufacturer, is calculated with equation (4-5). The 95% σ
observed measurement accuracy is calculated by (4-8). The observations used for the
optimization procedure for fully drained analysis are in bold. For these observations the
stated accuracy is used. The observations used for the optimization procedure when the
Clayey Silt is modeled as an undrained material are in bold and italic bold.
Observations marked in italic are excluded from the inverse analysis because their
magnitudes are either lower than the 95% σ measurement accuracy (stated or observed)
or the observed deflection is negative (movement away from the wall). Negative
deflections were not used because the inclinometer was located on the soldier pile. If the
inclinometer was installed 4-6 feet behind the soldier pile, negative deflections likely
would not have been observed. The observations for the Fill and the Dense Sand layers
were not used for the calibration.
104
Table 4.8 Observations used for inverse analysis
4.4.3. Finite Element Simulations for Optimizations
Two sets of inverse analysis were performed. The first section presents the inverse
analysis when the Clayey Silt is modeled as a drained material and the instrument
accuracy is the one stated by the manufacturer. The second section presents the inverse
analysis when the Clayey Silt is modeled as an undrained material and the instrument
error is the observed error at the beginning of the Olive 8 project.
105
4.4.4. Optimization for Fully Drained Analysis
4.4.4.1. Optimized Parameters
From the parameters that define the responses of the Clayey Silt and Upper Sand
layer, only two were chosen for the optimization. These are the reference values for
primary deviatoric loading , and for elastic unloading and reloading , because
they are parameters that most influence the behavior of an excavation when the
movements are “small” (Finno and Calvello 2005). A sensitivity analysis was performed,
before the optimization, to define which of the previously mentioned parameters were
more susceptible to calibrate Table 4-9 presents the results of the sensitivity analysis.
refE50refurE
Composed Scaled Sensitivities
Parameter Silty Sand
Clayey Silt
ref50E 0.748 0.94 refurE 0.343 9.92
Table 4-9. Results of sensitivity analysis
Based on the composed scaled sensitivities, of the Clayey Silt layer is the most
important parameter to be calibrated from the inverse analysis. The of the Silty Sand
and Clayey Silt layer are one order of magnitude smaller the for the Clayey Silt.
The of the Silty Sand layer has the least influence on the computed results and
therefore it was not included in the optimization procedure because the composite scaled
refurE
refE50
refurE
refurE
106
sensitivity was relatively close to zero and this parameter would likely not be
optimized.
Two optimizations were performed using inverse techniques:
1) An analysis where three parameters were calibrated simultaneously, refE50 for the
Silty Sand layer, refE50 and refurE for the Clayey Silt layer.
2) An analysis where parameters were calibrated simultaneously, refE50 for the Upper
Sand layer and refE50 for the Clay layer.
The parameters that were not optimized did not change and remained at their input
values, or were related to the updated value of the optimized value. Specifically, when
not being explicitly optimized, the unload reload stiffness was calculated in
function of as:
refurE
refE50
refurE =3* (4-11) refE50
and the and the oedometer stiffness was, in all cases, computed as: refoedE
refoedE =0.7* (4-12) refE50
For the second optimization the most important parameter, of the Clayey Silt
layer, while not optimized, it was directly related to by a factor of 3.
refurE
refE50
107
4.4.4.2. Results
Three Optimized Parameters
Wall Deflection (in.)
-1.0 -0.5 0.0 0.5 1.0
Ele
vatio
n (fe
et)
60
70
80
90
100
110
120
130Observed Deflection Before Optimization 3 Par. Optimization
SIlty Sand
Clayey Silt
Very Dense Sand
Figure 4.14 Observed versus the calculated and optimized deflection for three optimized
parameters
Figure 4.14 shows for Pile #6, the calibrated displacement profile for the Three
108
Parameter Optimization versus the inclinometer observations and the numerically
calculated profile before the optimization. The calibrated displacement profile agrees
well with the observed profile at the Fill and Clayey Silt layers. At the Silty Sand Layer,
due to lack of reliable observations at greater depths, the fit is not as good compared with
the fit at the previous two soil layers.
Table 4.10 shows the results of the inverse analysis for three optimized parameters.
The initial values of (for the Upper Sand and Clayey Silt layer) and (for the
Clayey Silt layer) were the ones used in the GeoEngineers finite element analysis, given
in Table 4.1. The initial objective function value was 5132 and the final was 138 after 24
iterations. It must be mentioned that the optimization was terminated because for the
Clayey Silt layer overcame the finite element code limitation of ≤ 20* . The
results show that the reference stiffness increased by 38.5% for the sand layer and
161.2% for the Clay layer; the unload-reload stiffness for the Clay layer increased by
1700%. was increased with a rate much higher than the rate of and eventually,
after 24 iterations, became 20 times larger than so the optimization stopped
due to the finite element code limitation.
refE50refurE
refurE
refurE refE50
refurE refE50
refurE refE50
Figure 4.15 shows the objective function and the optimized stiffness parameters
versus the number of iterations. It can be seen that when the optimization was terminated,
by the finite element code limitation, the relative fit improvement (RFI) was 97% and the
objective function was stabilized, so the optimization was close to the optimum solution.
109
The for the Silty Sand and Clayey Silt layers were not greatly affected by the
optimization in contrast to of the Clayey Silt layer, which is the most important
optimization parameter.
refE50
refurE
Iteration Number
0 5 10 15 20
RFI
(%)
0
20
40
60
80
100Iteration Number
0 5 10 15 20
Obj
ectiv
e Fu
nctio
n
0
1000
2000
3000
4000
5000
6000
Iteration Number
0 5 10 15 20
kips
/ft2
0
5000
10000
15000
20000
25000
30000E50,SandE50,ClayEur,Clayref
ref
ref
Figure 4.15 Objective function, RFI (%) and stiffness parameters versus the number of
iterations
The optimization for three parameters showed that the importance of at the
Clayey Silt layer is significant in comparison with the rest of the optimized parameters.
refurE
110
However its very large value at the end is not representative of the Clayey Silt, and
likely reflects the 3-D stiffening effect of the corner of the excavation. This kind of
response was also noted by Finno and Calvello (2005).
3 Optimized Parameters Iteration 1 Iteration 24
Layer Parameter Estimated Parameter (kips/ft2)
Objective Function
Estimated Parameter (kips/ft2)
Objective Function
Change (%)
GeoEngineers Parameters
(kips/ft2)
Silty Sand
refE50 1000 1385.3 38.5 1000
refE50 500 1306.2 161.2 500 Clayey Silt ref
urE 1600
5132
28808
138
1700.5 1600
Table 4.10 Optimization for three parameters
Two Optimized Parameters
Figure 4.16 shows the calibrated displacement profile for the Two Parameter
Optimization, the inclinometer observations and the computed profile before the
optimization for Pile 6. Again, the calibrated displacement profile agrees with the
observed profile in the Fill and Clayey Silt. In the Silty Sand Layer, due to lack of
reliable observations at greater depths, the fit is not as good compared with the fit in the
other two soil layers.
111
Wall Deflection (in.)-1.0 -0.5 0.0 0.5 1.0
Elev
atio
n (fe
et)
60
70
80
90
100
110
120
130Observed Deflection 2 Par. Optimization Before Optimization
Silty Sand
Clayey Silt
Very Dense Sand
Figure 4.16 Observed versus the calculated and optimized deflection for a three
optimized parameters
Table 4.11 summarizes the results of the inverse analysis for two optimized
parameters. The initial values for the Upper Sand and the Clayey Silt layer were refE50
112
taken from the three parameter optimization. The rest of the parameters were related
to the updated value of the optimized value by equations (4-11) and (4-12). The initial
objective function value was 2690 and the final was 154 after 13 more iterations.
Figure 4.17 shows the objective function, the RFI (%) and the optimized parameters
versus the number of iterations. This optimization was stopped when the objective
function changed less than 8% for 3 consecutive iterations. The RFI stabilized at about
93% over these three last iterations. It has to be mentioned, that when the objective
function started to stabilize, the calculated deflections profile was slightly affected by the
extra number of iterations performed. The for the Silty Sand layer was not affected
greatly by the optimization. The optimization for two parameters showed that the
importance of at the Clayey Silt layer is significant. In this optimization though,
was not optimized directly but through according to the monotonic
relationship (4-11). Both parameters were changing during the optimization and this
explains the similar calculated soil profiles at the two and the three parameter
optimizations. However the very large values at the end are not representative of the
Clayey Silt, and likely reflect the 3-D stiffening effect of the corner of the excavation. It
must be noted that for the Clayey silt is one order of magnitude higher in the two
parameter optimization and this slightly affected the calibrated deflection profile.
refE50
refurE
refurE refE50
refE50
113
Iteration Number
0 2 4 6 8 10 12 14
Obj
ectiv
e Fu
nctio
n
0
500
1000
1500
2000
2500
3000
Iteration Number
0 2 4 6 8 10 12 14
kips
/ft2
0
3000
6000
9000
12000
15000E50,SandE50,Clayref
ref
Iteration Number
0 2 4 6 8 10 12 14
RFI
(%)
0
20
40
60
80
100
Figure 4.17 Objective function and stiffness parameters versus the number of iterations
2 Optimized Parameters Iteration 1 Iteration 13
Layer Parameter Estimated Parameter (kips/ft2)
Observed Function
Estimated Parameter (kips/ft2)
Observed Function
Change (%)
GeoEngineers Parameters
(kips/ft2)
Silty Sand
refE50 1385 1150 -17.0 1000
Clayey Silt
refE50 1306 2690
12437 154
852.3 500
Table 4.11 Optimization for two parameters
114
Wall Deflection (in.)
-1.0 -0.5 0.0 0.5 1.0
Elev
atio
n (fe
et)
60
70
80
90
100
110
120
130Observed Deflection 3 parameter optimization 2 parameter optimization
Silty Sand
Clayey Silt
Very Dense Sand
Figure 4.18 Optimization results and wall deflection versus number of optimized
parameters
Figure 4.18 compares the inclinometer data with the computed deflections after the
two optimizations. The deflections in the Clayey Silt are almost identical for the two
115
optimizations and they are similar to the observed values. For the upper portion of
the wall the three parameter optimization gives slightly better results than the two
parameter optimization. This better fit is reflected in the lower final objective function
and the higher RFI for the Three Parameter Optimization. Hence, the three parameter
optimization gives a better solution, even though the simulation was terminated by the
finite element code.
Table 4.12 compares the optimized stiffness parameters for the two analyses with
their initial values.
Initial Stiffness Input Values versus the Values after the Three and Two Parameter Optimizations
3-par Optimization 2-par Optimization
Layer Parameter (kips/ft2)
Initial Value
Final Value
Initial Value
Final Value
refE50 1000 1385 1385 1150 refoedE 700 970 970 805 Silty
Sand refurE 3000 4156 4156 3450 refE50 500 1306 1306 12437 refoedE 350 914 914 8706 Clayey
Silt refurE 1600 28808 3919 37311
Table 4.12 Overview of Stiffness Input Parameters
The value in the Silty Sand increased by 38 and 11.5 % after the three and the
two parameter optimization, respectively, a reasonable result if ones considers that the
drained conditions represent the behavior of the Silty Sand layer and the 3-D cornering
refE50
116
effects are negligible. However, the stiffness parameters were optimized from
observations given by an inclinometer attached to the soldier pile.
The value in the Clayey Silt is the same order of magnitude in both
optimizations, but much larger than could be reasonably expected for this soil. The
is one order of magnitude higher in the 2 parameter optimization, than in the 3 parameter
optimization, because the most important parameter, , while optimized was directly
related to by a factor of 3.
refurE
refE50
refurE
refE50
An important factor is that the plane strain analysis conducted herein does not take
into account the 3-D stiffening effects near the corners of a deep excavation. This effect
implies that the optimized parameters in a plane strain analysis will be larger than those
that actually exist. This result was also noted by Finno and Calvello (2005) for the
Chicago-State excavation.
Finally, the stiffness parameters were optimized from observations given by an
inclinometer located right on the soldier pile. So the calibrated parameters are expected to
be greater than the actual, because they are affected by the stiffness of the soldier pile,
and not the smeared stiffness in a plane strain simulation. The slight kickback within the
Silty Sand layer would not have been measured in a inclinometer located about 6 ft
behind the soldier pile.
Figure 4.19 shows the computed shear strains (%) in the ground for displacements
induced by the excavation at pile 6. Herein, shear strain εq, refers to the invariant
117
deviatoric shear strains and is defined as:
( ) ( ) ( ) ([ ])222222 621
32
zxyzxyxxzzzzyyyyxxq εεεεεεεεεε +++−+−+−= (4-13)
In a plain strain simulation, this shear strain can be written as:
( ) ( ) ( ) ( )[ ]2222 621
32
xyxxyyyyxxq εεεεεε +++−= (4-14)
The strain components were obtained from the results of finite element calculations,
as those values that developed during excavation. Recall that displacements were zeroed
just before excavation begun. With the aid of Golden Software Surfer®, a contour plot of
the shear strain levels was generated.
The shear strains behind the Olive 8 shoring wall due to excavation are less than or
equal to 0.1%, with the exception of some localized shear strains (0.2 %) around the
bottom of the shoring wall and the lower portion of the Qwest basement wall. For this
excavation, the optimized stiffness parameters represent strains on the order of 0.1%.
Conventional laboratory soil testing without internal instrumentation allows one to
measure strains as low as 0.1%. If one does not consider the effect of the excavation of
the Qwest building, the secant modulus for the Silty Sand and the Clayey Silt layers only
can be determined only by lab tests that involve the use of research-type equipment
which is generally not available in commercial laboratories. Because no data were
available to compare computed with observed results for the Qwest building, only these
incremental strains can be evaluated. Successful calibrations thus would depend on
118
accurately defining the stresses in the ground by other means before excavating for
the Olive 8 building. (e.g. pressuremeter data to obtain “K0”)
Figure 4.19 Shear strains (%) in function with depth and distance from the West
excavation wall
119
4.4.5. Optimization for Undrained Responses of Clayey Silt
4.4.5.1. Optimized Parameters
As in the case of the drained analysis, two parameters that define the responses of
the Clayey Silt and Silty Sand were chosen for the optimization. These again were the
reference values for primary deviatoric loading , and for elastic unloading and
reloading . In this case the response of the Clayey Silt was assumed undrained. The
parameters are the same, but a penalty formulation is employed to enforce a condition of
no volume change in the layer. A sensitivity analysis was performed before the
optimization to define which of these parameters had the greatest impact on the
computed results. Table 4-8 presents the results of the sensitivity analysis.
refE50
refurE
Composed Scaled Sensitivities
Parameter Silty Sand
Clayey Silt
ref50E 0.380 3.18 refurE 0.802 9.85
Table 4-13. Results of sensitivity analysis
Based on the composed scaled sensitivities, of the Clayey Silt layer is the most
important parameter to be calibrated from the inverse analysis and of the same layer
is the second most important parameter. The of the Silty Sand layer is one order of
magnitude smaller than this value. The of the Silty Sand layer was the least influent
refurE
refE50
refurE
refE50
120
in the computed results, and was not included in the optimization procedure
because the composite scaled sensitivity was relatively close to zero and this parameter
would likely not be optimized.
Two optimizations were performed using inverse techniques:
3) An analysis where three parameters were calibrated simultaneously, refurE for the
Upper Sand layer, refE50 and refurE for the Clayey Silt.
4) An analysis where two parameters were calibrated simultaneously, refurE for the Upper
Sand layer and refurE for the Clay layer.
The parameters that were not optimized did not change and remained at their input
values, or were related to the updated value of the optimized value. Specifically, when
not being explicitly optimized, the reference stiffness was calculated in function of
as:
refE50
refurE
refur
ref EE ∗= 31
50 (4-15)
and the oedometer stiffness was, in all cases, computed as: refoedE
refrefoed EE 507.0 ∗= (4-16)
121
4.4.5.2. Results
Three Optimized Parameters
Displacement (in)
-0.5 0.0 0.5 1.0 1.5
Ele
vatio
n (ft
)
60
70
80
90
100
110
120
130
Observed DeflectionBefore OptimizationAfter Optimization
Fill
Silty Sand
Clayey Silt
Figure 4.20 Observed versus the calculated and optimized deflection for three optimized
parameters
122
Figure 4.20 shows the calibrated displacement profile for the Three Parameter
Optimization, the inclinometer observations and the computed profile before the
optimization for Pile 6. By visual inspection of the optimization results one can see that
the optimized profile does not fit with the observed.
Table 4.14 shows the results of the inverse analysis for three optimized parameters.
The initial values of (for the Upper Sand and Clayey Silt layer) and (for the
Clayey Silt layer) were the ones used in the GeoEngineers finite element analysis, given
in Table 4.1. The initial objective function value was 58317 and the last was 25273 after
9 iterations. The previous objective function values are higher than the ones found at the
optimizations for fully drained analysis. This is due to the fact that for this optimization
the observed error was used instead of the stated error. Since the observed error was
smaller then the weight of the observations derived from equation (4-9) increases and
consequently the objective function given by equation (4-3) increases also.
refE50refurE
3 Optimized Parameters Iteration 1 Iteration 9
Layer Parameter Estimated Parameter (kips/ft2)
Objective Function
Estimated Parameter (kips/ft2)
Objective Function
Change (%)
Design Parameters
(kips/ft2)
Silty Sand
refurE 3000 1803 -39.9 3000
refE50 500 277 -44.6 500 Clayey Silt ref
urE 1600
58317
5794
25273
262.1 1600
Table 4.14 Optimization for three parameters
123
The optimization was terminated because for the Clayey Silt was larger
than that imposed by the finite element code limitation of ≤ 20* . Consequently,
this optimization was terminated at an early stage and this is the reason that the calibrated
displacement profile does agree well with the observed profile. The results show that the
decreased by 40% for the Silty Sand layer and increased 262% for the Clayey Silt;
the for the Clayey Silt layer decreased by 45%. increased rapidly while
decreased and eventually, after 9 iterations, became 20 times larger than so the
optimization stopped due to the finite element code limitation.
refurE
refurE refE50
refurE
refE50refurE refE50
refurE refE50
Figure 4.21 shows the objective function, the RFI (%) and the optimized stiffness
parameters versus the number of iterations. It can be seen that when the optimization was
stopped by the finite element code limitation, so the inverse analysis had not reached an
optimum solution defined by the convergence parameters.
124
Iteration Number
0 2 4 6 8 10
Obj
ectiv
e Fu
nctio
n
0
10000
20000
30000
40000
50000
60000
Iteration Number
0 2 4 6 8 10
kips
/ft2
0
2000
4000
6000
8000
10000Eur, SandE50, ClayEur, Clayref
ref
ref
Iteration Number
0 2 4 6 8 10
RFI
(%)
0
20
40
60
80
100
Figure 4.21 Objective function RFI (%) and stiffness parameters versus the number of
iterations
125
Two Optimized Parameters
Displacement (in)
-1.0 -0.5 0.0 0.5 1.0
Ele
vatio
n (ft
.)
60
70
80
90
100
110
120
130
Observed DeflectionBefore OptimizationAfter Optimization
Fill
Silty Sand
Clayey Silt
Figure 4.22 Observed versus the calculated and optimized deflection for a three
optimized parameters
. Figure 4.22 shows the calibrated displacement profile for the Two Parameter
Optimization, the inclinometer observations and the computed profile before the
126
optimization for Pile 6. The calibrated displacement profile agrees well with that
observed in the Fill and Clayey Silt layers. In the Silty Sand layer, the fit is not as good
compared with the fit at the other two soil layers.
Table 4.15 shows the results of the inverse analysis for two optimized parameters.
The initial values for the Silty Sand and the Clayey Silt layer were those in Table
4.1, used in the GeoEngineers finite element analysis. The rest of the stiffness parameters
were related to the updated value of the optimized value by equations (4-15) and (4-16).
The value of the objective function decreased from an initial value of 65612 to 4467 after
15 iterations.
refurE
2 Optimized Parameters Iteration 1 Iteration 15
Layer Parameter Estimated Parameter (kips/ft2)
Observed Function
Estimated Parameter (kips/ft2)
Observed Function
Change (%)
GeoEngineers Parameters
(kips/ft2)
Silty Sand
refurE 3000 5319 77.3 3000
Clayey Silt
refurE 1600
65612 18840
4667 1077 1600
Table 4.15 Optimization for two parameters
Figure 4.23 shows the objective function, the RFI (%) and the optimized parameters
versus the number of iterations. This optimization was stopped when the objective
function changed less than 8% for three consecutive iterations. The RFI stabilized at
about 92% over these three last iterations. When the RFI stabilized, the calculated
deflection profile was affected little by the additional iterations performed. The value of
127
refurE for the Silty Sand layer gradually changed as the number of iterations
increased. The value of of the Clayey Silt layer increased rapidly as the number of
iterations increased.
refurE
Iteration Number
0 2 4 6 8 10 12 14
Obj
ectiv
e Fu
nctio
n
0
10000
20000
30000
40000
50000
60000
Iteration Number
0 2 4 6 8 10 12 14ki
ps/ft
20
5000
10000
15000
20000
25000
30000Eur, SandEur, Clayref
ref
Iteration Number
0 2 4 6 8 10 12 14
RFI
(%)
0
20
40
60
80
100
Figure 4.23 Objective function RFI (%) and stiffness parameters versus the number of
iterations
Figure 4.24 illustrates the updated profiles for two and three optimized parameters
versus the observed deflections. The profile given by the three parameter optimization
does agree particularly well with the observations because the calibration ended before an
128
optimum solution was reached. In the upper portions of the wall and the Clayey
Silt, the two parameter optimization provides a very good fit with the observed data in
contrast to that in the Silty Sand layer where not enough reliable observations existed to
perform the optimization.
Displacement (in)
-1.0 -0.5 0.0 0.5 1.0
Ele
vatio
n (ft
)
60
70
80
90
100
110
120
130
Observed Deflection2 par. Optimization3-par Optimization
Fill
Silty Sand
Clayey Silt
Figure 4.24 Observed versus calibrated displacement deflection profiles for two and
three optimized parameters
129
Table 4.16 presents the calibrated parameters from the two and the three
parameter optimizations assuming undrained conditions in the Clayey Silt and their initial
values. It can be seen that the results from the 3 parameter optimization differ greatly
from the results of the 2 parameter since the optimization could not finish. Because the fit
was much better in the later, hereafter the only the results of the two parameter
optimization will be discussed.
Comparison of 2 and 3 Parameter Optimization for Undrained Clayey Silt Analysis
3-par Opt.
2-par Opt.
Layer Parameter (kips/ft2)
Initial Value
Final Value
Final Value
refE50 1000 601 1713 refoedE 700 421 1199 Silty
Sand refurE 3000 1803 5139 refE50 500 277 6280 refoedE 350 194 4396 Clayey
Silt refurE 1600 5794 18840
Table 4.16 Overview of Stiffness Input Parameters
The value in the Silty Sand increased by 71% after the two parameter
optimization, a reasonable result if ones considers that the drained conditions represent
the behavior of the Silty Sand layer and the 3-D cornering effects are negligible.
However, the stiffness parameters were optimized from observations given by an
inclinometer attached to the soldier pile.
refE50
130
The value in the Clayey Silt increased by an order of magnitude but is
much larger than could be reasonably expected for this soil. The plane strain analysis
conducted herein does not take into account the 3-D stiffening effects near the corners of
a deep excavation. This effect implies that the optimized parameters in a plane strain
analysis will be larger than those that actually exist. This result was also noted by Finno
and Calvello (2005) for the Chicago-State excavation.
refurE
Finally, the stiffness parameters were optimized from observations given by an
inclinometer attached at the soldier pile. So the calibrated parameters are expected to be
greater than the actual, because they are affected by the stiffness of the soldier pile, and
not the smeared stiffness in a plane strain simulation. The slight kickback within the Silty
Sand layer would not have been measured in an inclinometer located about 6 ft behind
the soldier pile.
131
Displacement (in)
-1.0 -0.5 0.0 0.5 1.0
Ele
vatio
n (ft
)
60
70
80
90
100
110
120
130
Observed DeflectionUndrained Clayey SiltDrained Clayey Silt
Fill
Silty Sand
Clayey Silt
Figure 4.25 two parameter optimization results for drained and undrained Clayey Silt
analysis
Figure 4.25 presents the observed deflections and the calculated displacement
profiles for both two parameter optimizations. It can be seen that the calibrated profiles
132
are similar in the fill but differ slightly in the Silty Sand and Clayey Silt. In the Silty
Sand layer, when the observed error was used instead of the stated error (undrained
Clayey Silt analysis) the calculated displacement profile fitted better to the observed data.
The later is due to the fact that the number of reliable observations increased.
At the Clayey Silt layer the observed profile lies between the profiles given from the
drained and undrained analysis. For the undrained Clayey Silt response, the difference
from the observed profile is smaller. Since the number of observations used for both
optimizations is almost the same (15 instead of 14 when using the stated error), the better
fit is caused due to the undrained material response and the smaller measurement error
that produced a higher weight for each observation.
Table 4.17 presents an overview of the stiffness parameters for the two optimized
parameter calibrations when the Clayey Silt is modeled as a drained and undrained
material. For the Silty Sand layer the stiffness parameters were increased by 50% for the
later optimization. Since the observed error was used instead of the stated error, the
number of reliable observations increased and this resulted to a better fit of the optimized
to the observed profile. As mentioned in Section 4.3.1, the 3-D effects are negligible for
the Silty Sand layer; however the stiffness parameters were optimized from observations
given by an inclinometer attached on the soldier pile. So the calibrated parameters are
expected to be greater than the actual, because they encounter also for the stiffness of the
soldier which is smeared in a finite element simulation.
133
Comparison of 2 Parameter Optimizations for Drained and Undrained Clayey Silt Analysis
Drained Undrained
Layer Parameter (kips/ft2)
Initial Value
Final Value
Final Value
refE50 1000 1150 1713 refoedE 700 805 1199 Silty
Sand refurE 3000 3450 5139 refE50 500 12437 6280 refoedE 350 8706 4396 Clayey
Silt refurE 1500 37311 18840
Table 4.17 Overview of Stiffness Parameters
For the Clayey Silt layer, it can be seen that after using undrained analysis the
optimized stiffness parameters for the Clayey Silt reduced by 50%. Drained analysis
resulted long term wall deflections and the soil stiffness parameters had to increase more
to fit the calculated with the observed deep-seated displacements. The numerical
simulation of the behavior of this layer affects greatly the optimization results. However,
these calibrated stiffness parameters for the Clayey Silt do not reflect the actual
parameters that one would expect after high quality laboratory testing. The plane strain
analysis conducted herein does not take into account the apparent 3-D effects that
increase the stiffness at the corners of a deep excavation. This means that the optimized
parameters in the plane strain analysis will be larger than those that actually exist. This
result was also noted by Finno and Calvello (2005) for the Chicago-State excavation.
Finally, as previously mentioned, the optimized stiffness parameters are also affected by
134
the location of the inclinometer.
4.4.6. Discussion
The selection of observations for the inverse analyses showed that when the stated
error is used, most of the observed displacements at the Silty Sand layer are lower than
95 % of the standard deviation of the instrument error. Measurements like these are not
reliable to be used in inverse analysis because they created uncertainties in the
optimization results. If one compares the optimized and the observed profile in the Silty
Sand layer, then the lack of reliable observations created the difference between these
two displacement profiles. The manufacturer measurement error is relatively high for
conventional inclinometers and at projects where the predicted movements are small,
more accurate surveying instruments should be used, or site specific measurements
should form the basis of error estimates.
When the observed instead of the stated error was used, the number of reliable
observations increased at the Silty Sand layer and the optimization produced a better fit
for the calculated displacement profile. At the Clayey Silt layer the number of
observations used remained almost the same; however the calibrated profile slightly
improved since the observed error at this depth was very small and the weight of these
observations was increased.
Simulating the Clayey Silt as an undrained instead of drained material produced, as
expected, smaller values for the optimized parameters at this layer. However, still these
135
parameters do not reflect the actual parameters that one would expect after
extensive laboratory testing and are influenced by 3-D stiffening effects and by the
stiffness of the soldier pile which is smeared in plane strain analysis.
The shear strains behind the Olive 8 shoring wall are less than or equal to 0.1%, thus
the secant modulus can be determined only by lab tests that involve the use of research-
type equipment which is not available in commercial laboratories.
4.5. Conclusions
1) With the same set of soil parameters, the finite element numerical analysis that
included the Qwest excavation produced a displacement profile that was closer to the
movements observed in field, than a similar analysis without an explicit simulation of
the Qwest excavation.
2) When the changes in soil stresses were taken into account, the relative shear stresses,
before the Olive 8 excavation, increased behind the West excavation wall. The
calculated movements were smaller due to the soil stress relief after the Qwest
excavation.
3) Using undrained instead of drained analysis for the Clayey Silt layer resulted in
smaller computed deflection profiles.
4) The 3-D stiffening effects at the corners of the excavation are negligible in the Silty
Sand layer. In the Clayey Silt layer the 3-D effects become important especially when
the excavation reaches full depth.
136
5) The deep-seated deflections in the Clayey Silt layer are over-predicted in plane
strain analysis due to 3-D effects that increase the stiffness at the corners of a deep
excavation.
6) The observed profiles based on inclinometers attached to the soldier piles are
influenced by the stiffness of the soldier piles. In plane strain finite element analysis,
the stiffness of the soldier pile wall is smeared when dividing with the out-of-plane
spacing and behavior at the wall will be different than behind the wall.
7) Little difference was observed in computed displacement profiles when the tiebacks
were represented as equivalent struts or equivalent anchors.
8) The selection of observations for the inverse analysis showed that most of the
observed displacements at the Silty Sand layer are lower than 95 % of the standard
deviation of the manufacturer instrument error.
9) The manufacturer measurement error is relatively high for conventional inclinometers
and at projects where the predicted movements are small, more accurate surveying
instruments should be used or should be calculated from site specific measurements.
10) When optimizing the Clayey Silt layer as an undrained material the calibrated
parameters were smaller than the parameters produced when this layer was modeled
as a drained material.
11) The low shear strains behind the Olive 8 shoring wall indicate that the secant modulus
only can be determined by lab tests that involve the use of research-type equipment
137
with on specimen instrumentation which is generally not available in
commercial laboratories.
12) 3-D effects were negligible in for the Silty Sand layer and the optimized parameters
are affected by the location of the inclinometer which was next to the soldier pile. The
stiffness of the soldier pile affects the observations.
13) The calibrated stiffness input parameters for the Clayey Silt layer do not reflect the
actual parameters that one would expect after extensive laboratory testing because of
the 3-D stiffening effects.
138 Chapter 5 5. Summary and Conclusions
The Olive 8 development in Seattle, WA, included a 70 feet deep excavation to
provide space for 5 underground parking levels. The shoring system was rather unique
along the West shoring wall, due to the proximity of the Qwest building. The maximum
allowed displacement imposed by the city of Seattle was 1 inch.
GeoEngineers Inc., the project consultants, designed the West wall shoring system,
which consisted of a soldier pile wall with shotcrete lagging, 9 to 10 rows of soil nails
and 4 or 5 rows of tiebacks. Three separate sections were designed to provide lateral
support for the West wall, due to the presence of a large utility vault between the
excavation and the Qwest building. GeoEngineers performed numerical analysis to
predict the wall deflections which were found to be higher than the 1 inch limiting
displacement. However they proceeded with the proposed design because shoring wall
and the soil layers were expected to behave stiffer than the numerically-derived
predictions.
An instrumentation program was established to monitor the performance of the
West shoring wall. It included three conventional inclinometers to measure the lateral
wall movements and an automated surveying system to monitor the 3D deflections of the
top of the West wall at the positions where the inclinometers were located. The wall
movements were calculated relative to reference baseline points that were established far
away from the excavation. Rigid body translation and translation corrections were
performed to account for the fact that the total station was not rigidly attached to the
139
parapet wall where it was mounted. The lateral displacements based on the automated
surveying data of prisms agreed well with the inclinometer data. The less than rigid
connection between the total station and the parapet throughout the construction caused
the total station to shut down numerous times during the excavation.
To supplement GeoEngineers finite element analysis, a numerical model was
presented that explicitly accounted for the changes in stresses history caused by
construction of the Qwest building. Separate simulations were performed with the Clayey
Silt modeled as an undrained material and the tiebacks represented as equivalent struts.
The 3-D stiffening effects at the corners of deep excavations were evaluated for this
specific case history.
The finite element model that simulated the Qwest excavation was used in an
inverse analysis wherein the observed and the computed deflections are compared to find
the soil parameters that provided a best fit. A significant question for this excavation case
history was whether the small observed displacements in the Silty Sand layer were large
enough to be used for the optimization procedure.
The conclusions drawn from the excavation monitoring performance are:
1) The inclinometer observations showed the 1 inch movement limitation specified
by the city of Seattle was not exceeded. In particular, the maximum observed
movement was half of this limitation.
2) The horizontal wall deflections towards and along the excavation based on the
total station agreed well with the movement with the movement of the top of the
wall based on inclinometer data.
140
3) The Total Station shut down because the working range exceeded the vertical
compensator’s angular limit due to the less than rigid connection with the hotel’s
parapet wall.
4) The rigid body rotation correction requires all axes to be perpendicular to each
other. When the vertical compensator was turned off, the vertical axis was not
perpendicular to the horizontal plane, resulting in unacceptably large scatter in the
settlement data.
5) The design finite element predictions of the magnitude of wall deflection were
larger than that observed. But the observed and predicted displacements, along the
length of the wall, have the same pattern, implying that the shoring wall and the
soil responded more stiffly than expected.
The conclusions drawn from the finite element analysis conducted herein are:
1) With the same set of soil parameters, the finite element numerical analysis that
included the Qwest excavation produced a displacement profile that was closer to the
movements observed in field compared to those from a similar analysis without an
explicit simulation of the Qwest excavation.
2) Using undrained instead of drained analysis for the Clayey Silt layer resulted in
smaller computed deflection profiles.
3) The 3-D stiffening effects at the corners of the excavation are negligible in the Silty
Sand layer. In the Clayey Silt layer, the 3-D effects become important especially
when the excavation reaches full depth. The computed “deep-seated” deflections in
the Clayey Silt layer are larger than observed due to 3-D effects that increase the
141
stiffness at the corners of a deep excavation.
4) The observed profiles based on inclinometers attached to the soldier piles are
influenced by the stiffness of the soldier piles. In a plane strain finite element
analysis, the stiffness of the soldier pile wall is smeared when dividing with the out-
of-plane spacing and lateral movements of the soldier pile wall differ than these
between the piles at the wall. For inverse analysis purposes of this type wall system,
the inclinometer should be located several feet behind the wall so the localized effects
do not impact these results.
5) Little difference was observed in computed displacement profiles when the tiebacks
were represented as equivalent struts or equivalent anchors.
The conclusions drawn from the inverse analysis results are:
1) The selection of observations for the inverse analysis showed that most of the
observed displacements at the Silty Sand layer are lower than 95 % of the standard
deviation of the manufacturer instrument error.
2) The manufacturer measurement error is relatively high for conventional inclinometers
and at projects where the predicted movements are small, more accurate surveying
instruments should be used or should be calculated from site specific measurements.
3) When optimizing the Clayey Silt layer as an undrained material, the calibrated
parameters were smaller than the parameters produced when this layer was modeled
as a drained material.
4) The low shear strains behind the Olive 8 shoring wall indicate that the secant modulus
only can be determined by lab tests that involve the use of research-type equipment
142
with on-specimen instrumentation which is generally not available in commercial
laboratories.
5) The calibrated stiffness input parameters for the Clayey Silt layer should not reflect
the actual parameters that one would expect after extensive laboratory testing because
of the 3-D stiffening effects.
143
REFERENCES Blackburn, J.T. (2005). “Automated Sensing and Three-dimensional Analysis of
Internally Braced Excavations”, Ph.D. thesis, Civil and Environmental Engineering Department, Northwestern University, Evanston, Illinois.
Calvello, M. (2002). “Inverse analysis of a supported excavation through Chicago glacial
clays”, Ph.D. thesis, Civil and Environmental Engineering Department, Northwestern University, Evanston, Illinois.
Calvello, M. and Finno, R.J. (2004). “Selecting Parameters to Optimize in Model
Calibration by Inverse Analysis,” Computers and Geotechnics, Elsevier, v 31, n 5, p 411-425.
Calvello, M. and Finno, R.J. (2003). “Modeling Excavations in Urban Areas: Effects of
Past Activities,” Italian Geotechnical Journal, v 37, n 4, p. 9-23. Clough G.W. and O’Rourke T.D. (1990). “Construction Induced Movements of in-situ
Walls”. Design and Performance of Earth Retaining Structures, Proceedings of a Specialty Conference at Cornell University, ASCE, New York, 439-470.
Clough, G.W., Smith, E.M. and Sweeney, B.P. (1989). “Movement Control of
Excavation Support Systems by Iterative Design”, Current Principles and Practices, Foundation Engineering Congress, Vol. 2, ASCE, p. 869-884.
Dames And Moore (1972). “Supplemental Analysis Proposed Multi-Story Office
Building, Seventh Avenue and Pine Street, Seattle Washington”. Finno, R.J., Bryson, S. and Calvello, M. (2002). “Performance of a stiff support system in
soft clay”, Journal of Geotechnical and Geoenvironmental Engineering, v 128, n 8, p 660-671.
Finno, R.J., Blackburn, J.T. and Roboski, J.F.(2007), “Three-dimensional Effects for
Supported Excavations in Clay,” Journal of Geotechnical and Geoenvironmental Engineering, ASCE, v 133, n 1, p 30-36
Finno R.J. and Calvello, M. (2005). “Supported excavations: Observational method and
inverse modeling”, Journal of Geotechnical and Geoenvironmental Engineering, v 131, n 7, p 826-836.
Finno, R.J. and Roboski J.F. (2005), “Three-Dimensional Responses of a Tied-back
Excavation Through Clay”, Journal of Geotechnical and Geoenvironmental Engineering, v 131, n 3, p 273-282.
144
GeoEngineers Inc. (2005), “Geotechnical Engineering Shoring Design Report for the West Wall, Office Tower 8th Avenue and Olive Way, Seattle, Washington”.
HartCrowser Inc. (2000), “Geotechnical Engineering Design Study, Proposed Office Tower 8th Avenue and Olive Way, Seattle, Washington”.
Leica Geosystems (2002), User Application Manual v2.2, Leica Geosystems AG,
Heerbrugg, Switzerland. McMillan, L. (1996), “Rigid Body Transformations”, Class at UNC, Introduction to
computer graphics (Comp136) , Lecture 15. Plaxis (2002). 2D-Version 8 User Manual, Editor R.B.J. Brinkgreve, A.a. Balkems
Publidhers, Lisse. Poeter E. and Hill M. (1998). “UCODE a Computer Code for Universal Inverse
Modeling”, U.S. Geological Survey. Rechea, C.B. (2005), “Inverse Analysis of Excavations in Urban Environments” Ph.D.
thesis, Civil and Envir. Engrg. Dept., Northwestern University, Evanston, Illinois. Roboski, J.F. (2004), “Three-dimensional Performance and Analyses of Deep
Excavations” Ph.D. thesis, Civil and Envir. Engrg. Dept., Northwestern University, Evanston, Illinois.
Roboski, J.F. and Finno, R.J., (2006) “Distributions of Ground Movements Parallel to
Deep Excavations,” Canadian Geotechnical Journal, v 43 n 1, p 43-58.