Professional Documents
Culture Documents
A Comparison of One-Dimensional, Two-Dimensional, and Three-Dimensional Settlement Analyses of The Tower of Pisa
A Comparison of One-Dimensional, Two-Dimensional, and Three-Dimensional Settlement Analyses of The Tower of Pisa
A Comparison of One-Dimensional, Two-Dimensional, and Three-Dimensional Settlement Analyses of The Tower of Pisa
A thesis submitted by
Andrew J. Klettke
Master of Science
in
TUFTS UNIVERSITY
August, 2009
i
UMI Number: 1468186
INFORMATION TO USERS
The quality of this reproduction is dependent upon the quality of the copy
submitted. Broken or indistinct print, colored or poor quality illustrations and
photographs, print bleed-through, substandard margins, and improper
alignment can adversely affect reproduction.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if unauthorized
copyright material had to be removed, a note will indicate the deletion.
______________________________________________________________
ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106-1346
A Comparison of 1D, 2D, and 3D Settlement Analyses of the Tower of Pisa
by
Andrew J. Klettke
Abstract:
Pisa by one, two, and three-dimensional computer models. The Tower of Pisa has
been an important settlement case study for more than a century. The Tower
models and more recently, two and three-dimensional finite element analysis.
The recent work provides a large amount of subsurface data used in these
not accounting for shear induced deformations. In 2D, a plane strain analyses can
model the Tower’s tilt but the magnitude of the settlement must be calibrated for
the difference between the infinite strip plane strain loading and the actual circular
analyses are comparable to the findings of the calibrated 2D analysis and provide
and inclinations.
ii
ACKNOWLEDGEMENTS
I would like to thank Jesus Christ for all the gifts He has given me,
thesis. I would like to thank my parents, Neal and Susan Klettke, as well my
entire family for their continued support in the entire process. I would also like to
especially thank the Powelson and Oien families for their help during the time
student like myself and being a consistent, strong advisor on my project. I would
like to thank my committee members, Christopher Swan and David Shields, for
their time invested in attending meetings and reviewing my thesis. I would like to
thank my fellow Tufts University CEE graduate students for their support,
especially Ms. Elsbeth Hearn for her assistance in teaching me some of the
software programs used in this thesis. I would like to acknowledge the Tufts
University CEE faculty as well as the staff, including Grace Clemence and Laura
Sacco.
iii
TABLE OF CONTENTS
Abstract ii
Acknowledgements iii
Table of Contents iv
1.1) Purpose 1
1.2) Scope 3
2.1) Construction 5
2.2) Inclination History 6
2.3) Stabilization Methods 10
2.4) Subsurface Conditions 12
2.5) Soil Properties Data 16
2.6) Causes of Inclination 20
2.7) Estimated Settlement 21
2.8) Previous Settlement Calculations 22
3.1) Winsaf-I 32
3.1.1) Winsaf-TR 33
3.2) One-Dimensional Profile 34
3.3) Methods and Parameters 36
3.3.1) Assumptions 36
3.3.2) Settlement Parameters 37
iv
3.3.3) Consolidation Rate Parameters 40
3.3.4) Loading Sequence 42
3.4) Results 43
3.4.1) Primary Consolidation 44
3.4.2) Secondary Compression 44
3.5) Discussion 45
3.6) Winsaf Verification 47
3.6.1) Previous 1D Calculations 47
3.6.2) Olson Method 48
4.1) Plaxis 2D 50
4.2) Methods and Parameters 51
4.2.1) Soil Models 52
4.2.2) Soil Behavior Types 53
4.2.3) Parameters 54
4.3) Axisymmetric Analysis 57
4.3.1) Initial Conditions 58
4.3.2) Calculation Phases 59
4.3.3) Axisymmetric Results 60
4.4) Axisymmetric Verification 61
4.4.1) Mesh Fineness 62
4.4.2) Mesh Size 63
4.4.3) Excess Pore Pressure Verification 64
4.4.4) Plastic-Drained Analysis 66
4.4.5) φ = 0 Analysis 67
4.5) Plane Strain Analysis 68
4.5.1) Plane Strain Symmetric Loading 69
4.5.2) Plane Strain Non-Symmetric Loading 70
4.5.3) Plane Strain Calibration 73
v
Chapter 5: Three-Dimensional Analysis – Plaxis 3D 77
5.1) Plaxis 3D 77
5.2) Methods and Parameters 78
5.3) Symmetric Loading 78
5.3.1) Initial Conditions 79
5.3.2) Calculation Phases 82
5.3.3) Symmetric Loading Results 83
5.3.4) Mesh Fineness 84
5.4) Non-Symmetric Loading 85
5.4.1) Non-Uniform Loading Results 87
References 114
Appendix 117
vi
LIST OF TABLES
vii
LIST OF FIGURES
viii
Tower 46
3.7. Previous Settlement Calculations – Center of Tower 48
3.8. Olson Time-Rate Settlement Calculation 49
ix
6.5. 2D Axisymmetric 6-Noded vs. 3D Uniformly Distributed Center
Settlement 97
6.6. 2D Plane Strain Calibrated vs. 3D Non-Symmetric Loading 98
6.7. Inclination Angle vs. Time 99
x
Chapter 1: Purpose and Scope
1.1 Purpose
walls where soil and structural interfaces exist can become rather involved with
has brought about considerable advances in these analyses, which includes the use
capabilities has also progressed with technology. Modeling that once was only
available in two dimensions can now simulate full three dimensional behavior.
model requires far more approximations than a detailed three dimensional model
of the same problem. It would be expected that the results of the three
dimensional model be more accurate than those of the one dimensional model.
Each added level of complexity in a model should ideally provide an added level
dimensions should enhance the quality of the model with each added dimension.
A comparison of the results of one, two, and three dimensional calculations of the
same problem could reveal how much accuracy and quality each added dimension
brings.
1
The leaning tower of Pisa is a famous historical landmark. From the time
its construction began in August 1173 until the 1990s, the tower inclined
southward nearly 5.5o (Burland, 2003). At that point, fears of a collapse led to the
thorough investigation of the subsurface conditions was conducted and the soils
beneath the tower have been well characterized. The Tower of Pisa also has
the settlement of the soils beneath the Tower of Pisa are a good candidate for
important settlement case study for much of the past century. As early as 1934,
when Karl Terzaghi proposed a time-differential settlement curve, the Tower has
programs chosen for this study include Winsaf1 (1D), Plaxis 2D v.8.6, and Plaxis
2
settlements as a sum of the vertical strain. An accompanying program, Winsaf-
includes far more input features including soil interfaces, boundary conditions,
analysis of foundation structures (Plaxis 3D, 2007). Many of the features are
similar to those found in Plaxis 2D, only calculations and graphics are performed
1.2 Scope
This thesis compares computations for the settlement of the Tower of Pisa
in one, two, and three dimensions. The goal is to provide a comparison of the
3
settlement and inclination and provides a historical background of the Tower
of Pisa.
6.0 Discussion – discusses and compares the results of each analysis and presents
4
Chapter 2: Literature Review and Background Information – Tower of Pisa
Finite element analysis has freed the geotechnical engineer from basing
1999). In order to perform an acceptable finite element analysis, the history and
the inclination and the results of previous model attempts. From prior studies,
there is a great deal of subsurface data and modeling information available. The
2.1 Construction
(Costanzo, 1994). Construction of the Tower began on the 9th of August 1173. In
1178, after three and a half stories had been built, work was interrupted. The
it is likely that soils beneath the foundation of the Tower would have experienced
After nearly a century, work on the tower resumed in 1272. At this point,
the strength of the soil beneath the tower had increased due to consolidation. By
1278, construction reached the seventh cornice when work again was stopped.
Political or military reasons are cited as likely causes and not the concern for
5
Tower stability. Again, a Tower collapse would have been almost certain if work
Work on the bell chamber began in 1360 and the Tower was finally
completed in 1370. Figure 2.1 provides a timeline for the Tower construction. At
made to the masonry layers that the lean of the tower to the south was significant
(Burland, 2003). The center line of the eighth floor changed and the fact that the
axis of the Tower is not perfectly straight both indicate southward inclination
(Figure 2.2).
It is known that the layers of masonry display relative inclinations and that
the center line of the tower is not straight (Costanzo, 1994). In 1991, Burland
reconstructed the history of the tilt of the Tower of Pisa. The reconstruction is
6
based upon the theory that the masons made continuous leveling adjustments
the tilt history shown in Figure 2.3. The values displayed in Table 2.1 until 1758
are highly qualitative, while since the first plumb line measurement in 1817, the
During the first phase of construction, it is apparent the Tower was subject
the south (Potts, 2000). As the tower reached the seventh cornice in 1278,
inclination was about 0.103o. The tilt increased to 1.611o southward over the next
90 years. After the completion of the bell tower in 1370, the inclination
surmise that leaning instability is responsible for this dramatic change. Another
7
possible explanation is with the already slight southward lean, the addition of the
bell chamber brought a very high stress concentration to the south side of the
S 6
o
1990 - 5.469
4
Inclination Angle ( )
O
0
1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
Initial Load
(1173-1178)
-1
N
FIGURE 2.3. Deduced Inclination History
8
In 1817, Cresy and Taylor recorded the first plumb line measurement of
the Tower. In 1838, a walkway was excavated around the base of the Tower
known as the catino. With the water table only 1 m to 2 m below the ground
surface, the excavation caused an inflow of water on the south side carrying
quantities of soil with it. The operation is believed to have caused an inclination
of 0.25o to 0.5o and brought the tower very close to collapse (Burland, 2000).
measure the inclination angle of the Tower and was found to be 5.246o (Croce,
1981). This method was used to record the angle until 1934 when a Girometti-
Bonechi’s inclinometer was installed in the tower. The readings from the
inclinometer were added to the measured value from Pizzetti’s method in 1934.
They were used as the primary inclination measurement tool until remediation
techniques on the Tower began in the early 1990s. Figure 2.4 provides a plot of
9
Some significant events were recorded after the installation of the
by grout injections into the foundation. In the late 1960’s and early 1970’s
pumping from the lower sands caused an approximate 41 seconds (0.0114o) of tilt
(Potts, 2000). Fifteen 12 m deep boreholes in 1953 and borings conducted by the
Polvani Committee in the late 1960s also generated small jumps in the inclination
angle. This shows how sensitive the Tower of Pisa is to disturbances of the
Cathedral of Pavia, caused the closing of Tower of Pisa to the public (Heiniger,
1995). By this time, the inclination of the tower was 5.469o and continually
increasing at a rate of 6 arc seconds (0.0017o) per year (Potts, 2000). Two
problems threatened the stability of the Tower at this point. First, the hollow
cylindrical structure of the Tower caused concern for the strength of the concrete
masonry (Figure 2.5). The second involved the stability of the soil foundation.
around the first and second levels of the Tower. The tendons significantly
reduced the buckling failure risk of the thin walls (Burland, 1998). Studies of
elevations around the outside of the Tower suggested a load applied to the
masonry at the foundation on the north side would help counteract the continual
lean. After a computer analysis, a temporary concrete ring was installed around
10
the base of the Tower and lead ingots were placed on the ring in calculated
A small diameter hollow stem auger inserted under the north end of the Tower,
was used to remove small volumes of the Horizon A sandy silt. Figure 2.6 shows
the tower. However, they present many complexities in modeling the settlement
of the tower. As a result, they will be left out of the modeling comparison.
11
FIGURE 2.6. Adopted Soil Extraction Technique (Burland, 2003)
The subsurface profile of the Tower has been well characterized due to the
examination of the soils foundation of the Tower during the major remediation
investigate methods to stabilize the Tower of Pisa (Lo Presti, 2003). As early as
1907, the first documented investigation was headed by Mario Canaveri. In 1965,
the Polvani Committee, appointed by the Ministry of Public works, divided the
soil beneath the Tower of Pisa into distinct three units: Horizons A, B, and C
1992 provided laboratory results from some of the most current sampling and
testing techniques. Soils collected from this study were looked at in great detail.
The cone penetration tests and laboratory tests performed on samples from
standard tube and large diameter thin-walled samplers give a thorough description
of the soil. A simplified profile with soil descriptions from the efforts of the
12
Polvani Committee is shown in Figure 2.7 (Mitchell, 1977). Figure 2.8 provides a
north-south cross section created by Rampello and Callisto (1998) based on all
silt, clay and sand. It can be broken into four layers. After 3 feet of topsoil and
fill, Horizon A consists of a silty layer over a sand layer. The silty layer can be
divided into a clayey silt (A1’) and a silty sand (A1”). Layer A2 is a gray sand
with a roughly constant thickness of 2.0 m. Above A2 at the north end of the
Tower exists mostly the A1” layer with thin lenses of A1’ material. To the south,
the A1’ material predominates over the A1” material. Figure 2.9 presents the
percentage of A1’ (clayey silt) in Horizon A to the north and south of the Tower.
13
Figure 2.8 Detailed Soil Profile (Rampello, 1998)
14
Much of the subsoil is composed of clayey deposits found in Horizon B.
This upper “Pancone” clay is highly plastic grey-blue silty clay and can be broken
into three sublayers (B1, B2 and B3) (Rampello, 1998). Layer B2 is a lower
The intermediate clay (B4, B5) and intermediate sand (B6) layers are
similar to the deposits of Horizon A. The B4, B5 layer is a stiff, silty clay with
less plasticity and higher unit weight than the upper clay. A 2.4-ft thick layer of
The lower clay (B7, B8, B9, B10) is a medium plastic silty clay. Sublayer
B7 can actually be divided into an upper part (B7’) with more clay and a lower
part (B7”) with less plasticity. The remaining sublayers (B8, B9, B10) are all
slightly silty dense gray sand that extends to depths of 65 to 70 meters. Further
15
estuarine clay, grey-blue marine clay, and littoral sediment of gravel and sand (Lo
Presti, 2003).
data on the soils below the Pisa Tower from the past century and puts emphasis
and stress history are all addressed. Much of the data in the Lo Presti et al (2003)
paper is also presented in a Tower of Pisa subsoil study by Rampello and Callisto
distinguish the sand and clay layers. Laboratory testing performed on standard
tube and large diameter thin-walled samplers revealed the clay deposits are
are provided including compressibility values higher than those found by the
the parameters of these studies will be used in the settlement calculations of this
thesis.
Saturated unit weight, specific gravity, and water content for the soil
layers are available from Lo Presti et al (2003) and shown in Figure 2.10.
Multiple symbols are used to differentiate Laval, mechanical piston and Osterberg
samplers.
16
Figure 2.10. Unit weight, Specific Gravity, Water Content (Lo Presti, 2003)
shown in Figure 2.11 are from Rampello and Callisto (1998). Again, different
symbols indicate different samplings techniques and the vertical lines are drawn
in by the authors as proposed values. The compression index values, Cc, tend to
be slightly higher for samples obtained with the Laval sampler. The ratio of
swelling index, Cs, to Cc falls in the range of 0.1-0.2. The values for permeability,
k, show some scatter for the different methods used. The secondary compression
index, Cα, ranges from 0.002 to 0.04 with the highest values observed in the upper
clay.
Drained shear strength and friction angles from triaxial compression tests
inferred from drained and undrained tests on undisturbed samples under both Ko
17
FIGURE 2.11. Cc, Cs, k, and Cα Values (Rampello, 1998)
TABLE 2.2. Effective Cohesion and Friction Angle (Lo Presti, 2003)
(1977) in Figure 2.12. Results were obtained from the Polvani Committee and
average values have been selected as the vertical lines drawn in the figure.
18
FIGURE 2.12. Coefficient of Consolidation (Mitchell, 1977)
shown in Figure 2.13 were inferred from oedometer tests using the Casagrande
Higher OCR values exist in Horizon A and the intermediate clay. The upper and
19
2.6 Causes of Inclination
A single clear explanation for the cause of the lean of the Tower of Pisa
does not exist. Rather, the inclination is believed to have been caused by a
the initial cause of tilt of the tower. As alluded to in Section 2.4, there is a distinct
special variation in the layer of soil 5-8 m below the surface. Silty sand
predominates to the north while mostly soft clayey silt is present under the south
side of the Tower. The clayey silt is much more compressible and deformed more
The soft, upper clay layer, called “Pancone” clay has contributed greatly
to the overall settlement of the tower. As shown in Figure 2.8, a depression exists
in the upper layers of Horizon B. The overall Tower load caused the clay to
eccentric weight of the Tower continued to affect its lean. As seen in Figure 2.3,
the inclination of the Tower grew rapidly following completion of the bell
chamber (1370-1550). There are multiple theories for this escalated rotation.
insufficient soil stiffness causes failure. It happens usually in tall structures where
a small lean creates an overturning moment greater than the resisting moment in
the soil due to the same rotation (Potts, 2000). The analogy of children building
block towers on soft carpet is used to explain the phenomenon. No matter how
20
perturbation will cause rapid inclination (Burland, 2000). It is also possible the
high stress concentrations on the south side brought the foundation material to a
state of failure (Bai, 2008). This would be the case where insufficient soil
strength would cause a bearing capacity failure. It is quite possible the very soft
ground below the Tower in combination with the 19.6 m diameter base and 58.2
Terracina, Mitchell, Soga, Burland, Potts and Bai are among those who have
made such settlement calculations which are discussed in Section 2.8. However,
Although, there are some estimated final settlements reported in the literature.
Based on the findings of the Commision of the Italian Ministry of Public Works
published in 1971, Mitchell et al (1979) estimate the final settlement of the Tower
to be approximately 140 cm on the north side, 220 cm in the center, and 310 cm
on the south side. Also reported in figures of the Bai et al (2008) paper are
observed final settlement values of 201 cm to the North and 390 cm to the South.
A linear interpolation between the two points produces an estimated final Center
settlement of 296 cm. The source of the Bai et al (2008) measurements is unclear.
21
et al (1977) and Bai et al (2008). Therefore, for comparison purposes in this
the lean of the Tower. Terzaghi (1934) attributed the cause to the differential
settlement of the soft highly plastic clays beneath the foundation. This resulted in
analysis was based upon limited subsurface data. Even so, the results in Figure
differential settlement) between the highest and lowest points of the base of the
Tower versus time. The final differential settlement of about 155 cm, assuming
currently not available and information gathered is mostly based on the figures
presented.
22
FIGURE 2.14. Non-Symmetric Soil Pressure from Leaning Tower (Terzaghi, 1934)
inclination, the difference in soil consistency from the north to the south and the
from the continual incline of the Tower. While the mean contact pressure is
known to be about 490 kPa, the paper attempts to reconstruct the history of the
inclination and estimates the soil pressures in 1962 to be 50 kPa at the North and
23
930 kPa at the South side. This extreme distribution has strengthened the soils to
the south and reduced their compressibility. The Tower continues to settle on the
south side though because pressures are still increasing. The curves calculated by
Terracina (1962) for settlement at the north, center, and south are shown in Figure
2.16.
FIGURE 2.16. Estimated Tower of Pisa Soil Pressures and Settlement (Terracina, 1962)
dimensional settlement of the Tower of Pisa. The analysis had the benefit of
utilizing the comprehensive studies carried out by the Polvani Commission (1965-
1970) and published by the Italian Ministry of Public Works (1971). The curve
constructed in Figure 2.17 is based on dividing the subsoil into an upper 7 m thick
sand zone and a 30 m thick clay beneath the sand. Immediate settlements of the
sand were assumed to be total settlements while immediate settlements of the clay
were calculated using linear elastic theory with consideration given to the effects
of local yielding. Both were assumed to occur at the time of load application.
The consolidation of the clay layer was assumed to begin at the middle of each
24
dimensional conditions using relationships developed by Davis and Poulos (1968)
for circular foundations. Secondary compression was added by dividing the clay
into upper and lower layers and forming reasonable Cα values based on the water
at the point when the primary consolidation curve slope became less than 26
cm/log cycle. The calculation generated an ultimate settlement at the center of the
original authors. Cambefort (1978) suggested much of the recent settlement has
occurred due to increase in the pumping rates from the deep sand layer and
approached the current problem of inclination using the results of more recent
25
cm and suggested, contrary to the author’s conclusions, that a bearing capacity
failure did occur within the upper sands due to their volcanic silt and clayey sand
properties. Nathan (1978) also contested the original author’s conclusion that the
including drilling and lowering of the water table are known to have caused
elevation of the water table with respect to the time period 1967-1977 and
explained its affects on the angle of tilt. While the recent settlement, after 1967,
may be neglected if compared to the overall total, the estimated final settlement
from the literature at the center of the foundation should be 230-260 cm and
greater than the 180 cm presented by Mitchell et al (1977). Leonards (1979) also
argued that, consistent with Terzaghi (1934), the estimated final settlement at the
center of the Tower is at least 240 cm. Leonards (1979) contended the results of
the penetration tests in the upper sand layer favor a random distribution of
compressibility and not one of directional bias which Mitchell et al (1977) uses to
settlement on the south side. In addition, it cannot be certain any of the floors
beyond the third story were constructed exactly horizontal and any record of
tilting versus time is not reliable until the 1800s. Leonards (1979) felt the best
theoretical concept for secondary compression used by the author was not
adequate for the soft clays beneath the Tower. From Figure 2.18 based on
26
cm. These findings favor a local shear failure in the upper clay stratum. In
response to each of the discussions, the original authors reexamined the Italian
Commission report to find estimated final settlements of 140 cm at the north, 220
cm at the center and 310 cm at the south side of the Tower. The discrepancy
As far as the bearing capacity failure theory supported by Nathan (1978) and the
local shear failure by Leonards (1979), Mitchell et al (1979) reiterates that (p.
1364), “no one can state the cause of the leaning with certainty.” Also, it was
(1977) curves.
Soga and Mitchell (1995) to first evaluate the complex stress conditions and then
used those results to estimate the time-dependent deformation of the soils. The
study used the finite element program FEAP for the three dimensional analysis
and the soil parameters from the explorations summarized by Calabresi in 1993.
27
The results of the analysis significantly under estimated the actual tilt of the
Tower. The study found an elasto-plastic finite element analysis provided more
performed on Pisa clays indicated larger deformations than those predicted by the
Cam clay model used in the analysis, suggesting creep played a significant role in
rotation, the actual loading stress path of the Tower should be followed rather
model of the Tower for stabilization purposes was performed in the late 1990s. In
order to calibrate a model to simulate the history of tilting, Burland and Potts
(1994) developed a hypothesis that with the construction of each story, the
masons attempted to bring the center line of the Tower back over the center of the
indicates the Tower oscillated in different directions for many years until a
significant tilt of 1.6o southward was present in 1272. Each analysis used the
finite element program ICFEP using a constitutive soil model that is non-linear
elastic/work hardening plastic. The analyses were carried out using the Modified
Cam Clay Model with fully coupled consolidation. The soil parameters used
were those reported by Calabresi et al (1993) and Lancellota and Pepe (1990).
Initial efforts by Burland generated a two dimensional plane strain model. The
mesh extended 100 m to all sides of the Tower and all layers were modeled as
28
layer of slighty more compressible material to the south. The analysis
inclination occurred. The results of the model are shown in Figure 2.19,
believed to be due to the plane strain formulation of the analysis. The results of
the inclination analysis matched the deduced values fairly well. When running
which was much closer to the estimated final range of values. The entire purpose
FIGURE 2.19. 2D Plane Strain Settlement and Inclination Calculations (Burland, 2003)
Bai et al (2008) model the Tower three dimensionally using the effects of
traditional secondary compression, beyond the scope of this thesis. The finite
plasticity model was used in the analyses. Initially, the problem was modeled two
29
either side of two rigid plates that served as the foundations of the Tower. The
soil parameters were based on values from the previous studies and the loading
was based on the corresponding actual construction stages. The results showed
that creep is important in the upper layers of Horizons A and B. The final
settlement calculated (in 1990) is 330 cm. The settlement curve shown in Figure
2.20 corresponds to an inclination curve in Figure 2.21 and a final angle of 3.4o
when accounting for creep. A sensitivity analysis performed found that the
present angle of 5.469o could not be achieved simply by changing the parameters
to the considerable tilt between 1370 and 1550. The three dimensional analysis
was performed both with and without incorporating the Tower structure. Without
the structure, a more complex load increment scheme was used and incorporated a
more compressible clayey silt layer thinning to the north. The analysis that
included creep calculated an ultimate settlement of 313 cm at the north and 415
modeling with the Tower structure, eight weaker elements were introduced to the
load and overturning moment. The Tower structure was modeled as linear elastic
material with densities correlating to the actual weight of the Tower. The results
was presented.
30
FIGURE 2.20. Plane Strain Settlement Curve (Bai, 2008)
The efforts of these investigations relate to this thesis in that there are
comparisons between the calculations made in one, two and three dimensions.
From the early one dimensional investigations of Terracina (1962) and Mitchell et
are able to roughly model the inclination. These models, though, require more
input parameters and, unless calibrated, overestimate settlement due to their plane
strain formulation. The three dimensional efforts do not require much more input
than the two dimensional models and are superior in terms of computing final
angle.
31
Chapter 3: One-Dimensional Analysis
settlement at the center of the Tower. An analysis at the center will be compared
uses the computer software program Winsaf. Soil properties obtained from
previous investigations at the Tower described in Section 2.5 are used for this
analysis.
3.1 Winsaf-I
vertical design loads acting on a soil profile created as a set of horizontal layers.
and are based upon the prescribed surface loading. The options for computing the
settlement include change in void ratio from stress changes, change in void ratio
from compression and rebound curves, vertical strain from stress changes, and
volumetric change from stress. A typical input page is shown in Figure 3.1.
From these inputs, Winsaf-I computes an ultimate primary settlement for the
analysis. The secondary settlement used for this analysis is made with an
total for primary because the number of log cycles is dependent upon when the
2
Prototype Engineering Inc., Winchester, MA
32
end of primary consolidation occurs. The exact timing of the end of primary in
the soils beneath the Tower is unknown and, if specified incorrectly, would cause
3.1.1 Winsaf-TR
consolidation times based on Terzaghi’s theory (1925). The program can analyze
systems. Figure 3.2 is a typical input page for a double layered system. Based on
33
FIGURE 3.2. Winsaf-TR Consolidation Parameters Input
2.8. The settlement program Winsaf-I is able to analyze a maximum of ten layers.
The profile to be analyzed was split into eight layers (Figure 3.3) based on the
compressible. They include the Clayey Silt (A1), the Upper Clay (B1, B2, B3),
the Intermediate Clay (B4, B5) and the Lower Clay (B7, B8, B9, B10). The
ground surface elevation was set at +3.0 m. The base of the Tower was placed at
an elevation of 0.0 m and the water table was set below the surface at an elevation
of +1.8 m.
soil layers were subdivided using the Winsaf-I multiple sublayer option. A
increased until the change in settlement was not significant. The number of
sublayers used is summarized with the Winsaf-I soil parameters in Table 3.1.
34
Top of
Soil Type γ (kN/m3) 2 Compressible Sublayers eo 3 CR 4 RR 5 SR 6 Cα 7
Layer (m) 1 sat
A1 - Clayey Silt 0.0 18.94 Yes 5 0.84 0.122 0.014 0.014 0.004
B1,B2,B3 - Upper
-7.4 16.94 Yes 6 1.45 0.348 0.058 0.058 0.015
Clay
B4,B5 - Intermediate
-17.8 19.68 Yes 1 0.75 0.169 0.024 0.024 0.004
Clay
B6 - Intermediate
-22.0 19.11 No 1 - - - - -
Sand
B7,B8,B9,B10 - Lower
-24.4 18.88 Yes 3 0.95 0.266 0.042 0.042 0.010
Clay
(1) Figure 2.8 - Detailed Soil Profile, Rampello and Callisto (1998)
(2) Figure 2.10 - Unit Weight, Lo Presti et al (2003)
(3) eo = (Gs*wn)/S, Gs, wn, S = 100% - Figure 2.10, Lo Presti et al (2003)
(4) CR = Cc/(1+eo), Cc - Figure 2.11, Rampello and Callisto (1998)
(5) RR = SR
(6) SR = Cs/(1+eo), Cs - Figure 2.11, Rampello and Callisto (1998)
(7) Figure 2.11, Rampello and Callisto (1998)
TABLE 3.1. Winsaf-I Parameters
+3.0
Fill
0.0
A1 - Clayey Silt
-5.4
-7.4 A2 - Silty Sand
B1,B2,B3 - Upper
Clay
-17.8
B4,B5 - Intermediate
Clay
-22.0
B6 - Intermediate
-24.4 = Pervious
Sand
B7,B8,B9,B10 -
Lower Clay
-37.0
C - Lower Sand
FIGURE 3.3. Simplified Subsurface Profile
35
3.3 Methods and Parameters
method was chosen to approximate the Tower’s hollow cylindrical structure seen
in Figure 2.5. Also, since maximum past pressures are known, settlement was
For the Winsaf-TR time-rate analysis, the soil profile was broken into
three clay sections based on the locations of the A2 and B6 sand layers to allow
for top and bottom drainage for each layer system. The A1-Clayey Silt drained
into the Fill and A2 layers. The B7, B8, B9, B10-Lower Clay drained into the B6
and C sand layers. Both of these were modeled as single layer systems with top
and bottom drainage. The B1, B2, B3-Upper Clay and B4, B5-Intermediate Clay
were modeled together as a double layered system. The A2 and B6 layers served
as top and bottom drainage routes for the double layered system. The excess pore
3.3.1 Assumptions
for the center of the Tower. Results obtained for the north and south sides of the
Tower would not reflect the varying soil layers and the eventual non-uniform
loading. The one dimensional analysis cannot account for an applied overturning
moment associated with a leaning tower. A crude analysis of the effect of the
varying the profile and applied pressure to represent the conditions at the north
and south sides. However, the Winsaf software is not designed for such analysis.
36
Additionally, the calculation assumes the thicknesses of the soil layers are
constant. Most notably, the A1 layer is considered a single layer and not divided
as A1’ and A1” as shown in Figure 2.8. In reality, the clayey silt predominates
the A1 layer on the south side (Mitchell, 1977). The uniform thicknesses of a one-
The Tower is not a solid cylinder, but rather a hollow cylinder with a
helicoidal staircase leading to the bell chamber at the top of the structure
structure with a circular footing that carries an equally distributed load. This
assumed in the analysis that for the 5, 6, and 10 year periods of construction,
uniform ramp loading occurred. The actual stages of construction are unknown
Historically there are several other factors that are cited as potentially
affecting the Tower settlement. In 1838, a walkway around the base of the Tower
However, with such a shallow water table, inflow of water occurred during
excavation. In 1935, repeated angled drilling into the foundation of the Tower
took place (Burland, 1998). Both of these instances are believed to have caused
37
In order to remain consistent in the analysis process, the major
assumptions of this section, unless specified otherwise, will be carried over to the
The subsoil at the Tower of Pisa has been well studied as discussed in
parameters used in this analysis and their sources can be found in Section 2.5.
The parameters selected for the Winsaf-I settlement calculation are shown
in Table 3.1. As mentioned in Section 3.2, the layers and their thicknesses are a
collected from Figure 2.10 (Lo Presti, 2003). The clay layers are compressible,
while the sand and fill layers are assumed incompressible. The values for initial
void ratio, eo, are calculated from water content, wn, and specific gravity, Gs,
gathered from Figure 2.10 (Lo Presti, 2003) using the equation:
G s ⋅ wn
eo = (Eq. 3.1)
S
Compression ratio values, CR, are found using compression index, Cc,
values from Figure 2.11 (Rampello, 1998) and eo values in the following
equation,
Cc
CR = (Eq. 3.2)
1 + eo
38
Likewise, swelling ratios, SR, were found in a similar manner. Swelling
index, Cs, values were obtained from Figure 2.11 (Rampello, 1998) and were used
Cs
SR = (Eq. 3.3)
1 + eo
Recompression Ratios, RR, were set equal to SR since there is no available data
The values for maximum past pressure, σ’vm, used to define stress history
as a function of depth are shown in Table 3.2. Values are calculated as a product
of effective overburden stress and average OCR values at the same depth
estimated from Figure 2.13 (Lo Presti et al, 2003). Effective stresses, σ’, are
where total stress, σ, is calculated from the unit weight of the soil and pore
Effective stress and maximum past pressure are plotted versus depth in Figure 3.4.
Maximum
Elevation
Total Stress, σ Pore Pressure, u Effective Stress, Past
Layer Top of Layer γsat (kN/m3) 1
OCR
5
(kPa) 2 (kPa)
3
σ'vo (kPa) 4 Pressure, σ'
σ vm
(m) 6
(kPa)
Fill 3.0 20.00 0.00 0.00 0.00 - -
A1- Clayey Silt 0.0 18.94 57.60 17.66 39.94 4.07 162.56
A2 - Upper Sand -5.4 18.07 159.88 70.63 89.24 2.38 211.95
B1,B2,B3 - Upper Clay -7.4 16.94 196.02 90.25 105.76 2.00 211.53
B4,B5 - Intermediate Clay -17.8 19.68 372.19 192.28 179.92 1.25 224.90
B4,B5 - Intermediate Clay -17.8 19.68 372.19 192.28 179.92 2.00 359.83
B6 - Intermediate Sand -22.0 19.11 454.85 233.48 221.37 2.00 442.74
B7,B8,B9,B10 - Lower Clay -24.4 18.88 500.71 257.02 243.69 1.25 304.61
C - Lower Sand -37.0 20.52 738.60 380.63 357.97 1.25 447.47
(1) Figure 2.8 - Detailed Soil Profile, Rampello and Callisto (1998)
(2) σ = γsat*Depth
(3) u = γwater*Depth Below Water Table (Elevation +1.8m)
(4) σ'vo = σ - u
(5) Figure 2.13, Lo Presti et al (2003)
(6) σ'vm = σ'vo*OCR
TABLE 3.2. Maximum Past Pressures
39
σ vm (kPa)
σ, σ'
0 100 200 300 400 500
0.0
A1
-5.0
A2
-15.0
Depth (m)
-20.0 B4,B5
B6
-25.0
-30.0
B7,B8,B9,B10
-35.0
C
-40.0
FIGURE 3.4. Effective Stress and Maximum Past Pressure vs. Depth
The parameters for the consolidation rate portion of the analysis are
displayed in Table 3.3. As mentioned in the Section 3.3, the subsoil is divided
into three separate layer systems where each system has top and bottom drainage.
Each layer has an initial thickness that decreases with each stage of settlement.
The initial pore pressure, uo, is the product of the hwater at mid-depth and the γwater
(9.81 kN/m3). The ultimate settlement for each loading stage was computed using
Winsaf-I.
compute settlement over time. The values of cv for the Horizon B layers are
Lo Presti et al (2003), the B6 intermediate sand layers are very similar to the
40
deposits of Horizon A. As a result, the cv found for the B6 layer was used for the
Fill
2 1
cv - 15.77m /yr
2
Load Thickness uo (kN/m ) ρultimate (cm)
Number (m) 2 3
A1 - Clayey Silt
1 5.40 26.5 19.75
2 5.20 25.5 9.89
3 5.10 25.0 1.39
A2 - Upper Sand
2 1
cv - 0.85 m /yr
2
Load Thickness uo (kN/m ) ρultimate (cm)
B1,B2,B3 - Upper Number (m) 2 3
Clay
1 10.40 51.0 54.13
2 9.86 48.4 35.45
3 9.50 46.6 5.31
2 1
cv - 3.15 m /yr
Intermediate Clay
1 4.20 20.6 1.39
2 4.19 20.6 0.57
3 4.18 20.5 0.09
B6 - Intemediate Sand
2 1
cv - 1.26 m /yr
2
Load Thickness uo (kN/m ) ρultimate (cm)
B7,B8,B9,B10 - Number (m) 2 3
Lower Clay
1 12.60 61.8 2.77
2 12.57 61.7 2.4
3 12.55 61.6 0.72
C - Lower Sand
secondary compression index, Cα, were taken from Figure 2.11 from Rampello
and Callisto (1998). The data is from multiple studies that used different sampling
techniques. An average Cα value was selected for each of the four compressible
layers. The values for the Upper Clay layer, which is considered the softest, were
found to be almost a full order of magnitude greater than the other layers.
41
3.3.4 Loading Sequence
using three separate runs. For each calculation, the ultimate settlement was found
first using Winsaf-I. Then for the consolidation portion, Winsaf-TR was used to
The first calculation simulates the construction of three and a half stories
years. The second phase, construction up to the seventh cornice, brings the total
load to 137280 kN over 6 years. Consolidation following the second phase lasts
for 82 years. The final construction phase represents the completion of the
construction with the addition of the bell tower. It occurred over 10 years
consolidation until the end of primary is reached. In the analysis, the end of
primary is assumed to occur at 90 percent of the total settlement of the final load,
secondary consolidation began (in the year 1390, 217 years after initial
construction). The calculation for secondary consolidation ran from this point to
the year 1990, a period 600 years. Secondary compression was neglected during
the first two phases of construction since the purpose of the one-dimensional
analysis is to provide a simple calculation for the settlement of the Tower. The
incorporation of secondary into the first and second phases would complicate the
calculation.
42
3.4 Results
The ultimate settlement under the center of the Tower using Winsaf was
computed to be 152 cm. Figure 3.5 is a plot of the settlement versus time
produced by Winsaf. Also shown is the estimated final settlement under the
center of the Tower which ranges from 220-296 cm. As discussed in Section 2.7,
sequence.
150
Tower Load (MN)
100
50
0
1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
-50
-100
1990 - 152 cm
-150
-200
-250 ACTUAL
CENTER
-300
Settlement (cm)
-350
-400
-450
-500
FIGURE 3.5. 1D Computed Settlement – Center of Tower
From the initial construction load of the first three floors and the 99 years
until the next phase, computed settlement was 78 cm. The construction up to the
seventh floor that took 6 years and the ensuing 82 year consolidation was
43
phase of the bell tower accounted for an additional 8 cm of settlement. At the end
the construction and primary consolidation, the soil was computed to compress an
The Tower of Pisa construction was completed over 600 years ago. It is
reasonable to assume then that the soil beneath it has reached the end of primary
(1977) for each of the compressible layers and assuming 90 percent settlement to
consolidation to be 30 years after the final load application, 217 years after the
For this analysis, the secondary compression was assumed not to begin
until the end of primary consolidation. This assures that secondary compression
occurs at constant effective stress after excess pore pressures have dissipated
(Ladd, 1973). As noted is Section 3.1, secondary compression was not computed
with the Winsaf program. The formula for secondary compression is given by:
t
ρ sec = Cα ⋅ H ⋅ Log ( ) (Eq. 3.5)
tp
where Cα is the secondary compression index, H is the thickness of the soil layer,
t is the total time, and tp is the time for primary consolidation. Secondary
compression began around the year 1390 and computed 18 cm of settlement over
44
3.5 Discussion
be noted as potential causes for the difference. The first entails the neglect of
phase and 2 cm for the third phase. By including these shear induced immediate
into each phase, meaning about half of the time from the 99 years of the first
phase and the 88 years from the second phase could have contributed to
the first two phases would have complicated the simple calculation intended by a
45
150
50
0
1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
-50
-100
Winsaf
1990 - 152 cm
-150
1990 - 204 cm
-200 Winsaf + Shear Induced Settlement
-250 ACTUAL
CENTER
-300
Settlement (cm)
-350
-400
-450
-500
FIGURE 3.6. 1D Computed Shear Induced and Consolidation Settlement – Center of Tower
points for the consolidation parameters Cc and Cs. The data presented in Figure
2.11 is compiled from three different studies using three different techniques.
Lancellota and Pepe (1990) summarize data collected using thin walled piston
Laval. The points associated with Calabresi (1993) are a collection of data from
all previous investigations that have been reanalyzed and used to derive a new set
of soil parameters (Rampello, 1998). With all the variation, it is difficult to assess
which set is the best. Parameters used in this study and the Cc and Cs input into
Winsaf are the proposed values given by Rampello and Callisto (1998). The large
variation is one possible source for the under calculation of Tower settlement.
Higher Cc values in the A1 or B1, B2, B3 layers would lead to more settlement.
46
3.6 Winsaf Verification
Tower of Pisa have been made and are presented in Section 3.6.1 for comparison
Terracina (1962) and Mitchell et al (1977) are among the engineers who
the Tower. These analyses were performed lacking more sophisticated sampling
devices used for the geotechnical information gathered in the 1990s and 2000s.
Figure 3.7 shows each curve from the past analyses compared to the results of this
investigation. The Terracina (1962) curve does not take into account the lapses in
the Tower’s construction and ultimately surpasses the Mitchell et al (1977) and
ultimate settlement of the Mitchell et al (1977) curve is 173 cm. Both the Winsaf
and the Mitchell et al (1977) curves compute settlements less than the estimated
final center settlement and both show rapidly accelerating settlements at the times
et al (1977) curve has a much steeper slope after these loads are added while the
Winsaf curve is more gradual. This is likely due to that fact Mitchell et al (1977)
47
assume shear-induced instantaneous consolidation of some of the layers, whereas
the Winsaf curve applies the same consolidation rate to each layer. The Winsaf
150
Tower Load (MN)
100
50
0
1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
-50
-100
Winsaf
-150
Mitchell et al (1977)
-250 ACTUAL
CENTER
-300
Settlement (cm)
-350
-400
-450
-500
FIGURE 3.7. Previous Settlement Calculations – Center of Tower
series of step loads and applying the theory for instantaneous loading to each step.
The calculation is based upon ultimate settlement, Su, and computing the time
cv ⋅ t
T= (Eq. 3.6)
H2
where H is the drainage distance and t is the time since the load was applied.
The result of the method is compared to the Winsaf analysis in Figure 3.8.
The settlement curves generated are nearly identical to those made in Winsaf.
48
Both curves indicate rapid initial settlements followed by a tapering of the
settlement 30-40 years after each loading increment. The agreement of the curves
The ultimate settlement of about 133 cm is reached by the year 1390 and
secondary compression was assumed to start at this time. The same secondary
1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
0
-20
-40
-60
Settlement (cm)
Winsaf
-100
-120
-140
-160
FIGURE 3.8. Olson Time-Rate Settlement Calculation
49
Chapter 4: Two-Dimensional Analysis – Plaxis 2D
This analysis will use the finite element analysis program, Plaxis 2D. The
the tilt of the Tower. The 2D analysis will consider both, axisymmetric and plane
strain models. The axisymmetric model can accurately represent the circular
foundation loading the Tower exerts on the underlying soils and generate a
computed average settlement, but cannot model the inclination. The plane strain
model can establish a well defined cross-section and will be able to account for
the tilt but requires a calibration for the difference between the infinite strip plane
4.1 Plaxis 2D
The finite element software package Plaxis 2D v.8.6 (Plaxis 2D) was used
soil conditions and boundaries, the program can perform detailed computations
and output total displacements, stresses, dissipation of pore pressures and soil
deformation. The calculations are made using robust numerical procedures using
ability to model the effects of loads, displacements, walls, plates, anchors, and
geotextiles on soils. It can generate a finite element mesh with five standard
options for coarseness, described as very coarse to very fine. Refinement of the
mesh can occur at specific clusters or line segments to create finer meshes. Plaxis
50
2D also introduces the use of constitutive soil models. Several soil models are
and Hardening Soil models. Within each soil model, layers can be considered
In the Calculation mode, calculation phases include plastic, consolidation and phi-
viewed at any nodes within the cross-section. Each node can be examined for
total horizontal and vertical displacements, strains, total and effective stresses,
simulate the mechanical behavior of soils. Each soil model requires a different set
Winsaf-I, three different calculations were made for each of the three different
run by establishing multiple phases. Each phase can specify a calculation type
51
4.2.1 Soil Models
The constitutive soil models available are intended to capture the essential
features of soil and rock behavior. For this analysis the Mohr-Coulomb (MC)
model is chosen for the sand and fill layers. The Soft Soil Creep (SSC) model is
rock behavior. The sands beneath the Tower of Pisa do not contribute much to
the foundation settlement. Therefore, a basic soil model like the MC is suitable.
Friction angle, φ, Cohesion, c, and Dilatancy Angle, ψ. The values E and ν affect
soil elasticity, while φ and c govern soil plasticity. Young’s Modulus is the basic
stiffness modulus based on the initial slope of the stress-strain curve of a soil.
Poisson’s ratio can be defined as the ratio of horizontal to vertical strain and for
Plaxis 2D purposes usually falls into the 0.3-0.4 range. Cohesion, or cohesive
strength, often pertains to clays. In the case of sand c is typically equal to zero. A
nominal c should be of at least 0.2 kPa should be used (Plaxis 8.2, 1998) to
improve the numerical performance of the program. The friction angle plays a
large role in determining the shear strength of the soil as calculations are made
based on the Mohr-Coulomb failure criterion. The dilatancy angle defines the
angle at which compression increases the volume of a dense sand. For the sands
The SSC model is a more advanced model than the MC. The clays
beneath the Tower are an integral part of the settlement analysis and such an
52
advanced model is appropriate. The SSC model is intended for near normally
The input parameters needed for the SSC model include: compression
index, Cc, swelling index, Cs, secondary compression index, Cα, initial void ratio,
eo, and the φ, c, and ψ parameters common to the MC model. The first four
parameters listed are used by Plaxis 2D to compute the basic stiffness. The final
three serve as failure parameters as they do in the MC model. The Cc and Cs are
derived from the void ratio versus log effective stress plots. The Cα can be found
using the change in strain over the change in logarithmic time. Values for eo can
and specific gravity. The c, φ, and ψ are the same as described in the MC model.
soil conditions. Since pore pressures greatly influence soil response, the presence
of water is important. The two behavior types chosen in the analysis are drained
permeability. Therefore, drained behavior is chosen for the sand and fill layers
excess pore pressures. In undrained layers, the program accounts for the stiffness
of water in addition to the soil strength and stiffness. It is reasonable then to use
the undrained behavior for the relatively slow draining clay layers.
53
4.2.3 Parameters
necessary to select the appropriate soil parameters. A description of the basic soil
parameters used in this Plaxis 2D analysis and their sources can be found in
Section 2.5. To remain consistent with the one dimensional analysis, the common
input parameters as well as the thicknesses of the soil layers remained the same.
The compression index, Cc, swelling index, Cs, and void ratio, eo, values are the
same as those used to find the Compression Ratio, CR, and Swelling Ratio, SR
The input parameters introduced in section 4.3.2, while not seen in the one
dimensional analysis, are necessary to use the appropriate soil models. The
majority of the values are obtained from the same primary sources. The entire
The literature does not provide any values for Unsaturated Unit Weights
because the water table rests at an elevation of about +1.8 m and all of the critical
layers are assumed to be fully saturated. The program will not function if
parameters are left blank so for the analysis reasonable unsaturated unit weights
are taken as 2 kN/m3 less than their corresponding saturated unit weights. The
only affected layer is the fill with an assumed unsaturated unit weight of 18
54
Initial
Saturated Unsaturated Horizontal Vertical Young's Secondary Friction Dilatancy
Material Behavior Poisson's Compression Swelling Void Cohesion, Overconsolidation
Soil Layer Unit Weight, Unit Weight, Permeability, Permeability, Modulus, Compression, Angle, ϕ ( Angle, ψ
Model Type 3 1 3 2 Ratio, ν 4 Index, Cc 3 Index, Cs 3 Ratio, eo cref (kPa) 6 Ratio, OCR 7
γsat (kN/m ) γunsat (kN/m ) kx (m/day) 3 ky (m/day) 3 Eref (kPa) 4 Cα 3
5
o 6
) (o)
Mohr-
Fill Drained 20.00 18.00 1 1 22000 0.3 - - - - 1 34 0 -
Coulomb
Soft Soil
A1 - Clayey Silt Undrained 18.94 16.94 0.00086 0.00086 - - 0.225 0.026 0.0035 0.84 6.8 34 0 3.23
Creep
Mohr-
A2 - Upper Sand Drained 18.07 16.07 1 1 13000 0.3 - - - - 1 34 0 -
Coulomb
B6 - Intermediate Mohr-
Drained 19.11 17.11 1 1 18000 0.3 - - - - 1 34 0 -
Sand Coulomb
Mohr-
C - Lower Sand Drained 20.52 18.52 1 1 100000 0.3 - - - - 1 34 0 -
Coulomb
55
The permeability values used for the clay layers are from Figure 2.11
(Rampello, 1998). The input values are identical to the proposed values of the
figures but converted to m/day. The sand and fill layers are considered to be
highly permeable, therefore, the permeability values for each are assumed to be 1
m/day. These values correlate well with those used in the numerical analysis by
Potts and Burland (2000). Permeability is assumed to be the same in the vertical
Young’s Modulus and Poisson’s Ratio apply only to the layers modeled
with the MC soil model, the sands and fill. The values are obtained from the soil
With the exception of the very dense sand of Horizon C, the Young’s Modulus
values used fall in the typical range for Loose Sand according to the Elastic
Constants of Various Soils (FHWA, 1999). All Poisson Ratio’s are set to 0.3. As
recommended by the Plaxis 2D manual, the value should be in the range of 0.3-
c = 0 for each of the sand layers. However, Plaxis 2D recommends values of 0.2
value of 1 kPa in the sand layers. Plaxis 2D uses effective soil parameters and the
values for effective cohesion in the clay layers should be relatively small. Values
for each clay layer are taken from Table 2.2 (Lo Presti, 2003) and are in the range
of 1-10 kPa. The values for both the sand and clay layers friction angles are also
56
taken from Table 2.2 (Lo Presti, 2003). Values for all dilatancy angles are
ratio (OCR) values. The values used (Table 4.2) are consistent with those used in
the Winsaf calculation. However, in Winsaf the values are input based on the
OCR values at the top and bottom of each layer. The Plaxis 2D values are those
taken at the midpoint of each layer, which are straight line averages of the top and
Winsaf Plaxis 2D
OCR OCR OCR
Layer
(Top) (Bottom) (Midpoint)
Fill - - -
A1- Clayey Silt 4.07 2.38 3.23
A2 - Upper Sand - - -
B1,2,3 - Upper Clay 2.00 1.25 1.63
B4,5 - Intermediate Clay 2.00 2.00 2.00
B6 - Intermediate Sand - - -
B7,8,9,10 - Lower Clay 1.25 1.25 1.25
C - Lower Sand - - -
TABLE 4.2. Plaxis 2D Overconsolidation Ratios
reasonably compute settlement beneath the center of the Tower. The varying soil
layers beneath the Tower will have to be modeled as layers of uniform thickness
induced from the weight of the portion of the Tower overhanging the south edge
57
4.3.1 Initial Conditions
After establishing the necessary soil parameters, the mesh used in the
analysis can be created. This includes defining the geometry of the soil profile,
the placement of the Tower, and the size and fineness of the mesh. The soil layers
thickness consistent with the 1D Winsaf analysis. The Tower was represented as
an evenly distributed vertical load acting on a plate with a 9.8 m radius and
negligible weight acting as the base of the Tower. A 3.0 m long vertical plate
(also of negligible weight) acted as the outside wall. The coarseness and size of
the mesh were determined in mesh fineness and mesh size analyses discussed in
Sections 4.4.1 and 4.4.2. The results of the analyses supported the use of a 60 m
wide mesh with a very fine mesh fineness with the layer A1 and B1, B2, B3 soil
clusters refined one time each. The mesh, shown in Figure 4.1, consists of 2622
58
Following the generation of the mesh, the initial conditions are defined
within Plaxis 2D. This involves the elevation of the phreatic surface, flow
stresses. Consistent with the literature and one dimensional analysis, a general
vertical boundaries of the soil layers as well as the horizontal boundary beneath
the C (Lower Sand) layer are assumed to be open and free to drain. Therefore, no
horizontal effective stress to vertical effective stress using Jaky’s formula for Ko
(1-sinφ) and the unit weight of the soil layers. The Ko values can be defined by
the user, but for the case of the Tower those calculated by the program are
considered acceptable.
which types of calculations are performed, the types of loading, and the
construction stages. In the case of the Tower of Pisa, the calculation type
saturated clay-like soils over time. The use of consolidation calculation type is
59
To represent the load of the Tower, a distributed load system is applied.
In Plaxis 2D distributed loads are applied in terms of force per area. The load can
The corresponding pressures can be calculated from the weights during the
Tower begins in Plaxis 2D, a level surface of fill at elevation +3.0 is assumed.
Upon construction, fill is excavated to a depth of three meters and replaced with
plates intended to simulate the base and outside wall of the Tower foundation.
The parameters of the plate material are given in Table 4.3. The weight of the
plate material is small enough in comparison to the load of the Tower that it can
single run. For the simplest case, the construction can be broken into six phases,
three loadings followed by three consolidations. Table 4.4 summarizes the stages
Tower
Actual Number Calculation Tower
Load No. Actual Stage Pressure
Year of Years Type Weight (kN)
(kPa)
Construct 3 1/2
1 1173-1178 5 Consolidation 94800 314.20
Stories
2 Consolidation 1178-1272 94 Consolidation 94800 314.20
Construct up to
3 1272-1278 6 Consolidation 137280 454.99
7th Story
4 Consolidation 1278-1360 82 Consolidation 137280 454.99
Construct Bell
5 1360-1370 10 Consolidation 144530 479.02
Chamber
6 Consolidation 1370-1990 620 Consolidation 144530 479.02
60
4.3.3 Axisymmetric Results
The axisymmetric analysis uses the parameters of Table 4.1 and the
loading phases defined in Table 4.4. The mesh used is shown in Figure 4.1. The
4.2. The Plaxis 2D settlement of 260 cm falls in the range of the estimated actual
loading increment. At the completion of Tower construction, the curve tapers off
significantly indicating much of the settlement occurred during the first 200 years.
This suggests the end of primary consolidation has been reached and much of the
Tower settlement over the final 600 years has been due to secondary compression.
150
Tower Load (MN)
100
50
0
1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
-50
-100
-150
-200
-250 ACTUAL
CENTER
1990 - 260 cm
Settlement (cm)
-300
-350
-400
-450
-500
FIGURE 4.2. 2D Axisymmetric Settlement – Center of Tower
checks were made concerning the definition of the geometry and input settings.
61
These checks include a mesh fineness study, a mesh size study, a pore pressure
nodes to operate a mesh on. The objective is to create a mesh fine enough to
generate an accurate result, yet coarse enough to not waste time performing
options are available for global coarseness: very coarse, coarse, medium, fine, or
very fine. Within each of these levels of coarseness, mesh refinement can occur
a cluster. In this way, a concentration of nodes can be placed in areas where more
deformation is expected. For the analysis of the Tower of Pisa, it was found a
global coarseness of very fine along with refinement of soil layer clusters A1 and
B1, B2, B3 would generate the most accurate result while using a reasonable
number of approximately 21000 nodes. As shown in Figure 4.3, the data points to
the far left represent different levels of global coarseness. The data points to the
right represent results from selective soil layer clusters that have been refined
brings about an increase in nodes, but not much change in computed settlement.
62
Number of Nodes
0 5000 10000 15000 20000 25000 30000 35000 40000 45000
230
235
240
Settlement (cm)
245
250
255
260
265
270
FIGURE 4.3. Plaxis 2D Mesh Fineness Analysis
small mesh constrains soil movements. Figure 4.4 plots mesh width versus
displacement for the axisymmetric mesh size analysis. Figure 4.4 shows larger
settlement for small mesh width. The frictionless vertical boundaries in the
smaller mesh width increase the intensity of the vertical stress (greater than
Boussinesq stresses) at depth below the Tower. For this analysis, a 60 m wide
cross-section was selected as the optimal width, more than six times larger than
the 9.8 m wide Tower. Figure 4.5 shows a shaded rendering of the total
63
400
380
360
340
Settlement (cm)
320
300
280
260
240
220
200
0 10 20 30 40 50 60 70 80 90 100
Mesh Width (m)
FIGURE 4.4. Plaxis 2D Axisymmetric Mesh Size Analysis
calculation type consolidation. This accounts for the fact that consolidation of
soils was occurring even during the construction phases. One of the concerns
with a program like Plaxis 2D is how it handles the application of the load in
64
consolidation calculation. Excess pore pressures will develop in saturated soils
under the application of a load. By investigating the excess pore pressures in the
clay layers during construction, it can be determined whether or not the Tower
load in Plaxis 2D is being properly applied. As shown in Figure 4.6, spikes occur
at the years 1173, 1272, and 1360, the same years each of the construction phases
pressures dissipate as they should. Also noted, the largest excess pore pressures
occur after the second loading in the B1, B2, B3 clay layer (Figure 4.7). This all
80
70
60
Excess Pore Pressure (kPa)
50
40
A1
B1,B2,B3
B4,B5
B7,B8,B9,B10
30
20
10
0
1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
FIGURE 4.6. Excess Pore Pressure in Clay Layers Versus Time
65
FIGURE 4.7. Final Excess Pore Pressures
drained and the clay layers as undrained. These selections were based upon
recommendations of the Plaxis 2D users manual. Plaxis 2D also requires all input
undrained shear deformation in the Plaxis 2D analysis, the following check was
made. All soil parameters were kept the same, but the previously undrained clay
layers were changed to drained and all calculation phases were changed from
While not exactly the same, the fully drained plastic analysis suggests that lower
shear strength during undrained shear deformation is not a large contributor in the
likely coming from soil that is slightly weaker in undrained conditions. This can
66
be considered a good verification that Plaxis 2D is behavior type undrained for
4.4.5 φ = 0 Analysis
Plaxis 2D. In each of the behavior type undrained clay layers, the values for φ
were set to 1 (Plaxis 2D cannot compute both φ = 0 and ψ = 0). Values for c were
estimated from Rampello and Callisto, 1998 and summarized in Table 4.5. The
failure occurs due to either φ or c being too low. Similar to the conclusions of
Burland et al (2003), this suggests the Tower was close to an undrained failure
(but remained standing likely due to the actual partial drainage). Upon further
investigation, the undrained shear strength data from Rampello and Callisto
These test results are highly influenced by sample disturbance and measure
strengths that are probably too low. The results of the φ = 0 analysis indicate it
cannot be ruled out whether or not undrained shear is a significant factor in the
67
Cohesion, Friction
Soil Layer
cref (kPa) Angle, ϕ ( o )
Fill 1 34
A1 - Clayey Silt 60 1
A2 - Upper Sand 1 34
B1,B2,B3 - Upper Clay 50 1
B4,B5 - Intermediate Clay 125 1
B6 - Intermediate Sand 1 34
B7,B8,B9,B10 - Lower Clay 110 1
C - Lower Sand 1 34
TABLE 4.5. φ = 0 Analysis - Undrained Strength Parameters
center of the Tower, it does not model the Tower’s inclination. Inclination can be
modeled using the plane strain formulation available in Plaxis 2D. Within the
plane strain model it is possible to model varying thicknesses in the soil layers as
infinite strip footing which generates larger soil pressures at depth than the
circular symmetric model and will over estimate settlements. The results of the
plane strain analysis will require a calibration for non-plane strain conditions.
The detailed soil profile presented in Figure 2.8 was used as a basis for the
cross-section of the Plaxis 2D plane strain model. The profile was created based
on available borehole and CPT data. The greatest difference in this profile from
the simplified profile of the axisymmetric analysis is the division of the A1 layer
into A1’-Clayey Silt and A1”- Silty Sand. Layer A1’ is more prevalent to the
south and tapers out to the north. Table 4.6 shows the properties associated with
these layers. The lower clays (B4, B5, B7, B8, B9, B10) and sands (B6, C) do not
contribute as much as A1 and B1, B2, B3 to the settlement or lean of the Tower,
and therefore are generally horizontal (Burland 2003). The model of Figure 4.8
68
attempts to represent pre-Tower conditions based on the geometry of Figure 2.8.
The bowl shaped depression beneath the Tower is replaced with straight line
interpolations of soil layers from the north to south sides (Figure 4.8).
Unsaturated Saturated Unit Horizontal Vertical Young's Secondary Friction Dilatancy
Material Behavior Poisson's Compression Swelling Initial Void Cohesion, Overconsolidation
Soil Layer Unit Weight, Weight, γ sat Permeability, Permeability, Modulus, Eref Compression, Angle, ϕ Angle, ψ
Model Type 3 Ratio, ν Index, Cc Index, Cs Ratio, eo cref (kPa) Ratio, OCR
γ unsat (kN/m ) (kN/m3) kx (m/day) ky (m/day) (kPa) Ca (o) (o)
Soft Soil
A1' - Clayey Silt Creep
Undrained 16.94 18.94 0.00086 0.00086 - - 0.25 0.026 0.0035 0.84 60 34 0 3.23
Soft Soil
A1" - Sand and Silt Creep
Undrained 16.94 18.94 1 1 - - 0.15 0.026 0.0035 0.84 60 34 0 3.23
Fill
A1” A1’
A2
B1,B2,B3
B4,B5
B6
B7,B8,B9,B10
C
FIGURE 4.8. Plaxis 2D Plane Strain Cross Section
The plane strain analysis was first performed using symmetric loading. As
mentioned in Section 2.6, one hypothesis from the literature suggests that the
loading in the initial plane strain analysis exhibits a very small differential
settlement. After the first construction phase, the north end settles 2 cm more
than the south end, which actually causes a northward inclination of 0.06o. This is
consistent with the Potts and Burland (2000) model in that during the early stages
of loading, before an overturning moment was applied, the model did show a
69
small inclination to the north. It is believed that consolidation of the thin northern
end of the tapered layer occurred more rapidly than the southern end. Eventually,
the final computed settlement is 365 cm at the north and 366 cm at the south
differential subsurface conditions are not a major cause of the Tower’s present
day tilt. The symmetric load alone applied in this initial plane strain analysis is
not representative of the Tower of Pisa structure and the overturning moment
associated with it. A further investigation of the tilting Tower and the moments it
100
50
0
1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
-50
-100
-150
ACTUAL
NORTH
-200
ACTUAL
-250 CENTER
Settlement (cm)
-300
SOUTH 366 cm
-400
-450
-500
FIGURE 4.9. 2D Plane Strain Symmetric Loading
Much of the inclination of the Tower may be due to the load and
and overturning moments, the inclinations shown in Table 2.1 were used to
70
calculate the non-symmetric load at various times throughout the Tower’s
existence. The history of the Tower’s inclination is discussed in Section 2.2. The
and area of the Tower. The overturning moment, σoverturn, is then calculated using
the equation,
M ⋅c
σ overturn = (Eq. 4.1)
I
where M is the moment, c is the distance from center to the extreme edge of the
Tower, and I is the moment of inertia for the circular base of the Tower. Moment
values associated with inclination angle are obtained from Costanzo et al (1994).
The calculated σoverturn is then added to the stresses at the south edge of the Tower
and subtracted from the north edge. Stresses calculated from these inclinations
approximation can be made of the actual loads experienced by the Tower during
71
the specific years. Several sources in the literature support this linear distribution
σnorth = 65.7 kPa and σsouth = 942.7 kPa (Figure 2.14). Terracina (1962) estimates
soil pressures σnorth = 50 kPa and σsouth = 930 kPa. Both sets correlate fairly well
three curves shown in Figure 4.10 for locations beneath the north, center and
south ends of the Tower. Figure 4.11 is a shaded rendering of the total
displacements. The results are far greater than the estimated actual settlements
presented by Mitchell et al (1979) and Bai et al (2008) due to the plane strain
analysis model. Thus, the plane strain model requires a calibration as discussed in
150
Tower Load (MN)
100
50
0
1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
-50
-100
-150
ACTUAL
NORTH
-200
-250 ACTUAL
CENTER
ACTUAL
-350 SOUTH
Settlement (cm)
CENTER 392 cm
-400
-500
72
FIGURE 4.11. 2D Plane Strain Total Displacements
Figure 4.12 compares the vertical stresses under uniform strip and uniform
circular loadings. By comparing the two figures it is evident the vertical stresses
under a strip load are greater than those of a circular load. The differences in
these stresses are the cause of the large differences in displacements computed by
73
FIGURE 4.12. Vertical Stress Under Strip and Circular Loading (Som and Das Figure 5.17,
2003)
with uniform soil layers generates a settlement at the center of 260 cm. In order
to produce a proper scaling factor, a new plane strain model was created. Like the
axisymmetric model, this model did not have variation in the A1 layer and had the
same uniform layers, size and loading sequence. The settlement at the center of
the plane strain model was computed to be 380cm. The scaling factor, SF, can be
calculated as follows,
ρ AXISYMMETRIC 260cm
SF = = = 0.683 (Eq. 4.2)
ρ PLANE _ STRAIN 380cm
where the settlements are collected from the same geometry point in both models
beneath the center of the footing. By applying the scaling factor to the results of
the non-symmetric loading solution, the curves for the north, center and south
74
Tower Load (MN)
150
100
50
0
1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
-50
-100
-150
ACTUAL
NORTH
ACTUAL
-250 CENTER 268 cm CENTER
ACTUAL
-350 SOUTH
-400
Settlement (cm)
-450
-500
the Tower from the plane strain analysis is 268 cm. The settlement from the
axisymmetric model is 260 cm. This a good check that the scaling factor is
producing reasonable results as the differences between the values are likely due
to the varying soil thicknesses and the subdivision of the A1 layer into A1’ clayey
silt and A1” silty sand introduced in the plane strain analysis. While applying the
scaling factor reduces the settlement, it also reduces the computed angle of
calculated by:
ρ south − ρ north
Inclination = sin −1 ( ) (Eq. 4.3)
Diameter
Prior to the scaling factor, the computed inclination angle is 4.84o which is
reduced to 3.30o after calibrating the settlements. The actual inclination reported
75
by Costanzo et al (1994) prior to remediation efforts is 5.469o (Table 2.1). The
plane strain formulation can model differential settlement. However, even after
76
Chapter 5: Three-Dimensional Analysis – Plaxis 3D
like that of the Tower of Pisa. A 3D analysis can also consider radial drainage as
well as stresses and displacements in all three directions without the need of any
calibration.
5.1 PLAXIS 3D
rock materials (Plaxis 3D, 2007). Based upon input soil conditions and the
structure and the soil. Plaxis 3D contains the same basic features of Plaxis 2D but
also introduces many new choices for structures including beam, piles, and floors.
In Plaxis 3D, the input phase is a plan view of the x, z plane. Boreholes are used
to define soil layers and work planes are horizontal planes used to create
geometry point, lines, structures and loads. In the Plaxis 3D mesh generation, the
connected with 8-node quadrilateral elements in the vertical direction. The mesh
refinement procedures as well as the soil models and behavior types are consistent
with Plaxis 2D. The outputs of Plaxis 3D can be examined for total horizontal
77
and vertical displacements, strains, total and effective stresses, active and excess
pore pressures.
procedures very similar to the two dimensional analysis. For example, in the
Plaxis 3D analysis the Mohr-Coulomb model will remain as the soil model for the
sand and fill layers. The Soft Soil Creep model will be used to model the
compressible clay layers. These soil models used are discussed in depth in
Section 4.2.1.
The soil behavior types selected in Plaxis 3D are also the same as those of
the Plaxis 2D analysis. The sand and fill layers will be modeled as drained type
behavior and the clay layers will be modeled as undrained. These selections are
Table 5.1, are the exact same parameters used in the Plaxis 2D axisymmetric
analysis. The sources and a basic description of these soil parameters can be
found in Section 2.5. Explanations for the calculation of parameters not taken
to attempt to model the center settlement of the Tower. In the same manner, as a
78
preliminary calculation of the 3D analysis, a uniform circular loading over
The soil parameters and soil layer thickness for the Plaxis 3D analysis are
the same as those determined in the Plaxis 2D analysis. The mesh, however, is
area. The mesh size was established as 120 m (width) x 120 m (length) x 43 m
(depth). As the Plaxis 2D axisymmetric mesh size analysis used a mesh that
extended 60m from the center of the Tower, the boundaries of the Plaxis 3D mesh
are also to be 60 m from the center. The layer thicknesses and the height of the
mesh are the same. The fineness selection of the mesh was also different. In a
three dimensional finite element analysis, the number of nodes and elements is
critical as execution times can easily exceed 24 hours, but if the mesh is too
coarse, the results will be inaccurate. The procedure for mesh fineness is
discussed in Section 5.3.4. The mesh used in the uniform distributed load
analysis was originally established with a global coarseness of coarse. In the +3.0
m work plane, a 60 m x 60 m box was centered around the Tower. The cluster
within the box was refined 2 times. The mesh created was composed of 19250
elements and 51998 nodes and is shown in Figure 5.1. In Plaxis 2D, the load of
the Tower was replicated by excavating the fill, installing plates of negligible
weight as a base and exterior wall, and adding a distributed load equal to the
weight of the Tower to the top of the plate. To match the Plaxis 2D loading, the
79
Saturated Unsaturated Horizontal Vertical Young's Secondary Friction Dilatancy
Material Behavior Poisson's Compression Swelling Initial Void Cohesion, Overconsolidation
Soil Layer Unit Weight, Unit Weight, Permeability, Permeability, Modulus, Compression, Angle, ϕ Angle, ψ
Model Type 3 1 3 2 3 3 4 Ratio, ν 4 Index, Cc 3 Index, Cs 3 3 Ratio, eo 5 cref (kPa) 6 Ratio, OCR 7
γsat (kN/m ) γunsat (kN/m ) kx (m/day) ky (m/day) Eref (kPa) Cα (o)6 (o)
Mohr-
Fill Drained 20.00 18.00 1 1 22000 0.3 - - - - 1 34 0 -
Coulomb
Soft Soil
A1 - Clayey Silt Undrained 18.94 16.94 0.00086 0.00086 - - 0.225 0.026 0.0035 0.84 6.8 34 0 3.23
Creep
Mohr-
A2 - Upper Sand Drained 18.07 16.07 1 1 13000 0.3 - - - - 1 34 0 -
Coulomb
B6 - Intermediate Mohr-
Drained 19.11 17.11 1 1 18000 0.3 - - - - 1 34 0 -
Sand Coulomb
Mohr-
C - Lower Sand Drained 20.52 18.52 1 1 100000 0.3 - - - - 1 34 0 -
Coulomb
80
input in the Plaxis 3D analysis required a slightly different technique. A 19.6 m
diameter pile with no interfaces or thickness was installed on +3.0 m work plane
and was extruded down to the 0.0 m work plane. The wall element of the pile
was given the same properties as the Plaxis 2D wall. The fill material within the
pile was excavated and a plate was installed at elevation 0.0 m. A uniformly
distributed load equal to the weight of the Tower was then placed on the plate.
The wall and plate material parameters are shown in Table 5.2.
lower and outer boundaries were allowed to be open and free to drain. The
81
calculation of Ko values to generate initial pore pressures and stresses still uses
Jaky’s formula for Ko (1-sinφ) and the unit weight of the soil layers.
After generating the finite element mesh, the calculation phases must be
construction. The distributed load system defined in the input also is activated.
The distributed load is applied in terms of force per area. The load can be applied
over the circular area within the pile wall and can be changed during various
construction stages. The weight of the Tower at the end of each the construction
stage is known. Corresponding pressures are calculated and applied using the
performed in a single run. For the uniformly distributed load case, the
construction can be broken into six phases, each period of construction followed
Tower
Phase Actual Number Calculation Tower
Actual Stage Pressure
No. Year of Years Type Weight (kN)
(kPa)
Construct 3 1/2
1 1173-1178 5 Consolidation 94800 314.2
Stories
2 Consolidation 1178-1272 94 Consolidation 94800 314.2
Construct up to
3 1272-1278 6 Consolidation 137280 455.0
7th Story
4 Consolidation 1278-1360 82 Consolidation 137280 455.0
Construct Bell
5 1360-1370 10 Consolidation 144530 479.0
Chamber
6 Consolidation 1370-1990 620 Consolidation 144530 479.0
82
5.3.3 Symmetric Loading Results
Table 5.1, the loading phases defined in Table 5.3, and the mesh shown in Figure
presented in Figure 5.2. The final value of 240 cm from the 3D analysis is in the
range of the estimated final center settlement. The curve indicates rapid initial
much of the settlement occurred during the first 200 years. The shallow slope of
the curve around the year 1400 suggests the end of primary consolidation has
been reached and much of the Tower settlement over the final 600 years has been
basic inputs of Plaxis 3D are appropriate and the analysis can move forward.
150
Tower Load (MN)
100
50
0
1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
-50
-100
-150
-200
-300
-350
-400
Settlement (cm)
-450
-500
83
5.3.4 Mesh Fineness
fineness study was conducted. Figure 5.3 is a plot of number nodes versus total
many more options for refinement than in Plaxis 2D. Particularly with a
consolidation analysis, the global coarseness of the mesh cannot be too fine,
otherwise too many nodes will exist before any refinement begins. The algorithm
used in this mesh refinement study started with a globally coarse mesh. From
there a 60 m x 60 m box in plan view encompassed the extents of the Tower and
the cluster within the box was refined multiple times. The leveling of the curve in
Figure 5.3 indicates the optimum number of nodes for a mesh of this size is
between 35000 and 50000. Ideally a mesh of this size and fineness will be used
Number of Nodes
0 10000 20000 30000 40000 50000 60000 70000 80000
100
120
140
Settlement (cm)
160
180
200
220
240
260
FIGURE 5.3. Plaxis 3D Mesh Fineness
84
5.4 Non-Symmetric Loading
Tower. The Tower of Pisa leans due to a combination of reasons, namely its tall,
slender structure with a shallow foundation resting over soft, highly compressible
reconstruction, the Tower oscillated in random directions until near the end of the
second stage of construction when a distinct southward lean was apparent and the
of the actual Tower it was modeled as a series of five line loads representative of
(Table 2.1). The loads were placed on a thin, stiff plate element in the Plaxis 3D
model. This is similar to the load application in the Plaxis 2D Plane Strain Non-
center of the circle, with two line loads 4.9 m and 8.33 m to the north and two line
loads 4.9 m and 8.33 m to the south ends (Figure 5.4). The loads were calculated
by first finding the differential stress between the north and south ends using the
inclination angles. A straight line interpolation between the two stress values was
taken and used to determine what the stresses would be at 1/8, 3/8, 5/8, and 7/8
across the diameter of the Tower. The average stresses between these points were
then translated into line loads in dividing by the length of the east-west line they
acted over. The calculated loads are presented in Table 5.4. Each historical
85
FIGURE 5.4. Non-Symmetric Distribution of 5 Line Loads
86
inclination angle required a new of loading and construction phase, which resulted
The total loads and moments represented by the series of line loads were
calculated to ensure proper values were being applied. As seen in Table 5.4, the
percent. This is due to the integration of the distributed soil pressure into only
five segments. The pressures at the outermost points of the circle are slightly
underestimated and thus not enough moment is being generated and the actual
uniform pressure across a circle, the only way to improve the moment is to
integrate the area with more line loads. Considering the calculation already
consists of 11 phases, more line loads greatly increase the setup time of the
problem. As a result, five line loads are selected as an optimum tradeoff between
The mesh for the non-symmetric loading analysis was created using the
same algorithm used to find the optimal mesh in the uniform loading case. The
mesh was composed of 13882 elements and 37602 nodes, with an average
element size of 3.55 m. As shown in Figure 5.5, the computed north (201 cm)
and center settlement (250 cm) fall in the range of the estimated final settlements.
The computed south settlement (299 cm) is less than the estimate final range. By
assuming a rigid foundation, the final inclination of 2.85o was calculated using
Equation 4.2. This is well below the final measured 5.469o inclination. The
87
shapes of the curves of Figures 5.5 indicate rapid initial settlement. Differential
settlement is not evident until after the end of the second construction phase.
Shortly after that the slope of the settlement curves level off, indicating a long
100
50
0
1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
-50
-100
-150
ACTUAL
NORTH
NORTH - 201 cm
-200
CENTER - 250 cm
-250 ACTUAL
CENTER
SOUTH - 299 cm
-300
ACTUAL
-350 SOUTH
Settlement (cm)
-400
-450
-500
FIGURE 5.5. 3D Non-Symmetric Loading Settlement
Figures 5.6, 5.7, and 5.8 are some of the final outputs produced in the
the foundation looking from the west. The shaded portion represents the base of
experienced by the soil beneath the Tower. The majority of deformation occurs
directly adjacent to the south side of the Tower. Figure 5.8 is a shaded north-
88
N S
FIGURE 5.6. 3D Tower Base Inclination
89
FIGURE 5.8. 3D Non-Symmetric Loading North-South Cross Section Total Displacements
90
Chapter 6: Comparison of 1D, 2D and 3D Computations
dimensional settlement computation at the Tower of Pisa where the soil pressure
applied at the surface does not dissipate with depth. In Winsaf-I, this can be done
model with the outermost vertical boundary 9.8 m from the center. In Plaxis 3D,
a one-dimensional model is defined where soil only exists beneath the extents of
the Tower and the mesh boundaries do not extend beyond it. The depth in all
three cases was set to 43 m involving all the same soil layers. Each analysis
applied the same stress of 479.0 kPa, representative of the final load of the Tower
of Pisa. The same soil parameters established in Chapters 3, 4, and 5 were used
for this verification analysis. The fineness of the meshes in the Plaxis 2D and
Plaxis 3D cases were the same as those found in the mesh fineness studies for
each program.
The results for the 1D, 2D, and 3D one dimensional computations are
318.8 cm, 314.3 cm, and 314.2 cm respectively and are shown in Figure 6.1. The
91
settlements are larger than those found in the center settlement analyses but only
because stress was applied constantly with depth. The high level of agreement of
these results indicates that the software programs are consistent. From this check,
and Plaxis 3D are likely to produce a different settlement than Winsaf. The
supports this notion. Also, because Winsaf does not incorporate secondary
compression, in Plaxis 2D and Plaxis 3D the Cα of the Soft Soil Creep model was
set very low to 0.0001 for the analysis (Plaxis calculations cannot function with a
Cα=0). While some secondary compression may be incorporated into the final
92
displacement, a value of 0.0001 is small enough that it can be considered
negligible.
point directly beneath the center of the Tower of Pisa. In the 1D analysis (Section
3.4), the only curve generated was for beneath the center of the Tower. In the 2D
analysis (Section 4.3.3) the curve for center settlement was produced in a
preliminary uniformly distributed load analysis (Section 5.3.3). Figure 6.2 shows
150
Tower Load (MN)
100
50
0
1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
-50
-100
WINSAF - 152 cm
-150
-200
PLAXIS 3D - 240 cm
-250 ACTUAL
PLAXIS 2D - 260 cm CENTER
-300
-350
Settlement (cm)
-400
-450
-500
FIGURE 6.2. 1D, 2D, 3D Center Settlement
During the first construction stage (1173-1272), the curves are very
93
flattening of the curve. Also, the portions of the curves after the year 1400 are
1990 are very similar. The major difference exists during the second phase of
and Plaxis 3D during this time is greater than Winsaf-I by about 50 cm. This
two and three dimensional loading. In the one dimensional case, the loading can
only cause movement in the vertical direction. In two and three dimensions, soil
movement can be both horizontal and vertical. Figure 6.3 shows a shaded
analysis, while Figure 6.4 shows cross sections of the Plaxis 3D uniform loading
case. The horizontal displacements are especially evident in the B1, B2, B3 layer,
that the Tower is close to a bearing capacity failure. This coincides with the
Burland et al (2003) study that the Tower would have experienced an undrained
bearing capacity failure had construction continued past the 7th story.
B1, B2, B3
94
B1, B2, B3
(a)
B1, B2, B3
(b)
While the neglect of horizontal displacements is the main cause for the
smaller settlements in the 1D analysis, there are a few other potential causes. The
basic compression functions of WINSAF-I may be less accurate than the more
95
advanced soil models in Plaxis 2D and Plaxis 3D, Soft-Soil Creep and Mohr-
performed using the Winsaf-TR program and largely based on input values for
parameters may affect the slope portion of the curves, although not the ultimate
settlement.
While the 1D analysis produces significantly less settlement than both the
noticeably less than the 2D computation. The reasons for this lie in mesh fineness
and density of nodes. In Plaxis 2D, 15-noded elements were used. Due to the
of nodes (21193) and average element size (0.992 m) of the 2D analysis signify a
much higher density of nodes than the 3D analysis (nodes - 51998, average
element size – 3.18 m). While gaining the ability to model problems in three
dimensions, there is a sacrifice in accuracy due to the lack in the number of nodes
As another check between Plaxis 2D and Plaxis 3D, the 6-noded option
was selected for the 2D axisymmetric case since Plaxis 3D only uses 6-noded
elements. This decreased the number of nodes to 5353 (average element size
remained the same). The resulting curve is compared to the Plaxis 3D curve in
96
Figure 6.5. The Plaxis 2D 6-noded element final settlement of 239 cm is much
closer to the Plaxis 3D 240 cm. This suggests further that the disparity in Figure
100
50
0
1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
-50
-100
-150
2D 6-Noded
-200 3D Distributed Load
-250 ACTUAL
CENTER
Settlement (cm)
-300
-350
-400
-450
-500
FIGURE 6.5. 2D Axisymmetric 6-Noded vs. Plaxis 3D Uniformly Distributed Center Settlement
analyses.
The 2D and 3D results plotted in Figure 6.6 are from the Plaxis 2D plane
strain calibrated analysis and the Plaxis 3D Non-Uniform Loading. The curves
are in good agreement, with the 2D settlements slightly greater than the 3D
values. This is likely due to the node density as discussed in Section 6.2. In
Plaxis 3D, there is a sacrifice of mesh refinement for the ability to model a
97
complete three dimensional problem. Also, potential for error in the 2D
both curves show settlements that are low on the south side and high on the north
side of the Tower. This suggests that the tilts computed by Plaxis 2D and Plaxis
100
50
0
1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
-50
3D NORTH
-100 3D CENTER
3D SOUTH
2D CALIBRATED NORTH
2D CALIBRATED CENTER
-150 2D CALIBRATED SOUTH
Load ACTUAL
NORTH
NORTH
-200
-250 ACTUAL
CENTER
CENTER
-300 SOUTH
ACTUAL
-350 SOUTH
Settlement (cm)
-400
-450
-500
As described in Section 2.7 there has been some debate over the reliability
final settlements from the 2D and 3D analyses with the actual ranges may not be
the best way to evaluate if the proper amount of differential settlement is being
are compared to the measured inclination angles used to calculate the overturning
98
moment applied in the analyses. Once again the Plaxis 2D results (3.30o) are
slightly closer to the actual value (5.469o) than the 3D results (2.85o), but both
S
6
o
1990 - 5.469
4
Measured (Burland, 1991)
2D o
1990 - 3.30
3D
o
)
3
Inclination Angle (
o
1990 - 2.85
0
1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
-1
N
Discretization error
99
Some of the error associated with the underestimate may lie in the validity of the
when the first plumb line measurement of inclination was made, the tilt of the
Tower was based upon the changes in the angles in the levels of masonry.
measurements. The largest variance between the inclination curves and the
measured curve lies in the period between the years 1370 and 1550. This is
similar to the findings of Bai et al (2008). Neither the analyses of this thesis nor
the Bai et al (2008) work can model the 3 degree tilt increase during this period.
Bai et al (2008) suggests it is possible the tilt from 1370 to 1550 is due to external
cylinder. Both the 2D and 3D analyses modeled the circular foundation of the
Tower without the 4.5 m diameter hole in the center. This assumption was made
to simplify the input process and to remain consistent with the 1D analysis. In
doing so, a notable amount of the load was incorrectly concentrated at the center
of the Tower. Moving the load towards the outer wall, while not affecting the
center of gravity or moment, would increase the contact pressures as there is less
area to distribute the load. Increasing the contact pressures would likely increase
differential settlement. However, attempting to model the hole in the center of the
100
especially in 3D. Representing the load of the hollow Tower with a series line
discussed in Section 2.2 and shown in Figure 2.2, the central axis of the Tower is
not straight (but rather more banana-like) due to leveling adjustments were made
considerable time and effort. Investing such effort may not be worthwhile
considering other issues such as soil parameters, mesh fineness, and construction
history may also affect the settlement computations. It should be kept in mind
that modeling is a process (Barbour, 2004) and before becoming too complex, all
Nodes are assigned to all the layers within the large areas or volumes of the
meshes. The nodes within each layer are given the same soil properties. As
shown in Section 2.5, there is large scatter in the soil property values for soils
beneath the Tower of Pisa. For this thesis, in layers with large amounts of scatter,
proposed average values were used. So, a higher refinement of the mesh along
with a more accurate assignment of soil properties would likely provide more
accurate computations.
Noted by Rampello and Callisto (1998), the upper clay found in the B1,
B2, B3 layers has shown to be very sensitive material. Sampling disturbance may
101
While high quality samples were used, there may continue to be and
this thesis, a different grouping of parameters is required for each set of analyses.
the computations.
Software Package Winsaf Plaxis 2D Plaxis 3D Software Package Winsaf Plaxis 2D Plaxis 3D
Secondary
Layer Thickness x x x x x x
Compression, C α
Saturated Unit
x x x Initial Void Ratio, eo x x x
Weight, γsat
Unsaturated Unit
x x Cohesion, cref x x
Weight, γunsat
Permeability, kx and
x x Friction Angle, ϕ x x
ky
Young's Modulus,
x x Dilatancy Angle, ψ x x
Eref
Stress History, OCR
Poisson's Ratio, ν x x x x x
or σ'v m
Compression, CR Coefficient of
x x x x
or C c Consolidation, cv
Recompression, RR Plate and Wall
x x x
or C r Parameters
Swelling, SR or C s x x x TOTAL 9 15 15
TABLE 6.1. Input Parameter Comparison
The parameters used in each software package are designated with an “x”
in the column in Table 6.1. Winsaf required a total of 9 input parameters, while
Plaxis 2D and Plaxis 3D each required 15. In total there are 7 parameters
Winsaf is clearly simpler than Plaxis 2D and Plaxis 3D. This is consistent with
102
on site and performing appropriate laboratory testing. It is important to consider
how the parameters are obtained. For example, of the Winsaf parameters, 7 of 9
can be found from the traditional laboratory consolidation test. They include:
Secondary Compression (Cα), Initial Void Ratio (eo), Maximum Past Pressure
(σ’vm), and Coefficient of Consolidation (cv). In Plaxis 2D and Plaxis 3D, a much
smaller percentage, 6 of 15, can be found with the consolidation test. These
parameters are: Compression Index (Cc), Swelling Index (Cs), Cα, eo, permeability
friction angle, specified by the Soft Soil Creep model of the Plaxis programs,
require more testing. The Winsaf program is thus, less intensive in the amount of
Each software program has a different setup associated with it. A more
complex setup usually means more setup time but should provide more detailed
results. The details of the setup associated with each program are described in the
respective chapters. The order of magnitude of time for the setup, calculation,
and verification of each software program, beyond the time necessary to present
103
The simplest of the setups for the Tower of Pisa settlement analyses was
within Winsaf. The Winsaf computation was split into two programs, Winsaf-I
and Winsaf-TR. The Winsaf-I setup consisted of inputting the number of soil
layers and thicknesses, the soil parameters, the stress history of the soil, and
consolidation parameters and the results of the Winsaf-I calculation. Given all the
parameters, the entire setup takes less than an hour to complete and the
results with previous calculations and applying a different method for the
consolidation rate. Verification time is about a day and no mesh refinement study
is involved.
The Plaxis 2D problem setup was more involved than Winsaf. In Plaxis
Selecting a properly sized and refined mesh takes a considerable amount of effort
as it is a process of trial and error. Given all parameters and settings, the setup of
within an hour. For the mesh fineness, mesh size and other verification analyses,
The formulation of the Plaxis 3D model was very similar to Plaxis 2D,
only in three dimensions. The model is first defined in plan view and then
generated, first in the 2D plan view and then in 3D. Again, much effort is
104
required to find the proper mesh fineness. The calculation phases are then
prescribed with the initial conditions defined in the first phase. The entire setup
process can take hours. The duration of the computation may last 24 hours.
These long execution times lead to a lengthy verification process which can take
at least a week depending on the capability and quantity of the computers utilized.
Beyond the length of time for setup, the interpretation process must also
analysis does not require multiple analyses involving axisymmetric and plane
calibration. While the agreement of results in Figures 6.6 and 6.7 are good
indication the calibration is performed correctly, it also means that the 2D result is
105
Chapter 7: Summary and Conclusions
7.1 Summary
A settlement analysis for the Tower of Pisa has been performed using 1D,
2D, and 3D computation methods. The 1D analysis was completed using the
computation program Winsaf. The 2D and 3D analyses utilized the finite element
programs Plaxis 2D and Plaxis 3D. Each added level of dimension introduced
new capabilities but also more complexity. Each analysis has also shown to have
limitations.
measured final settlements, and measured angles of inclination over the past eight
centuries. The settlement of the Tower has occurred due to the compressible soils
beneath the foundation. The remediation efforts of the latter part of the 20th
century to stabilize the Tower have provided a well detailed description of the
engineers to investigate the mechanisms that triggered the extreme settlement and
inclination of the Tower. All of these items make the Tower of Pisa a great case
and necessary soil property parameters were collected from various sources of
106
produced a result that underestimated the settlement beneath the center of the
Tower and was unable to properly model the Tower’s differential settlement
method in the software package Plaxis 2D. Computation features of the program
included automated mesh generation, advanced constitutive soil models, and soil
behavior drainage conditions. The soil property parameters were gathered from
model to compute the settlement beneath the center of the Tower. Several
estimated final center settlement. The 2D analysis was used to model differential
circularly symmetric problem like the Tower of Pisa, this required a calibration
process from plane strain to axisymmetric conditions. In the plane strain analysis
it was possible to create non-uniformly thick soil layers, such as the A1’-Clayey
Silt lens beneath the Tower. However, it was found the overturning moment
produced from the ever increasing inclination of the Tower had a far greater affect
overturning moment generated by the tilt of the Tower and input into the
107
a ratio between the results of the axisymmetric analysis and a new plane strain
analysis with uniformly thick soil layers. The calibrated results were in better
agreement with the estimated final settlement but still underestimated the
The 3D analysis also made use of the finite element method with the
software program Plaxis 3D and involved many input procedures common with
the 2D analysis. The same sets of soil property parameters, soil models and
drainage behavior types from 2D were used in 3D. The observation in the 2D
analysis that non-uniform soil layers have little affect on differential settlement
was applied to the 3D analysis in keeping all soil layers of uniform thickness. By
modeling the problem in 3D, the calibration process of the 2D analysis was
eliminated. A mesh refinement study was performed to find the optimum number
distributed load analysis computed a settlement beneath the center of the Tower
that was within the range of the estimated final settlement. From there a 3D
historical inclination of the Tower by means of a series of five line loads applied
over a rigid plate foundation. The magnitudes of the line loads were based upon
their location along the north-south plane of the foundation and the overturning
found reasonable results for settlement at the north, center and south locations of
108
7.2 Conclusions
one, two, or three dimensions. The purpose of this thesis was to investigate how
the results of one, two, and three dimensional computations of the settlement of
Tower of Pisa compare to one another. The computations were evaluated on how
well the results compared with actual estimated values, the complexity of setup
process, and the length of time associated with setup and execution. The
conclusions reached for each dimensional analysis are summarized and discussed
below:
time, although, the settlement results were the least accurate. The
shear displacements.
Tower weight and the center of gravity moving horizontally was the
109
• The 3D analysis produces results that agree well with the 2D analysis
for 2D.
the actual settlement of the Tower. The 1D analysis was unable to account for the
varying stress distribution generated by the leaning Tower and therefore, could
results, the performance of the calibrated 2D analysis provided outputs that agreed
best with the estimated final settlements. The computed settlement of 260 cm
beneath the center of the Tower was within the range of estimated final settlement
of 220 cm to 296 cm. The 2D analysis also provided the best approximation of
the differential settlement which occurred beneath the base of the Tower. The
associated inclination angle of 3.30o, although low, was closest to the value of
5.469o measured in 1990. The results of the 3D analysis were similar to those
found in 2D. A settlement of 240 cm for the Tower’s center settlement fell in the
estimated final range. The inclination angle associated with the differential
settlement though, was 2.85o and less accurate than the 2D analysis.
110
In terms of the complexity of program operations, the 1D analysis was the
parameters. It also took the least amount of effort to setup and execute a
calculation. The 2D and 3D analyses were fairly similar in program setup time,
with the 3D analysis requiring more time and effort for the execution and
verification. Both analyses required the same number of input parameters. The
geometry, mesh generation, and the definition of calculation phases. While the
2D analysis took less execution time, the lack of calibration necessary in the 3D
were made in order to generate a scaling factor used to calibrate results of a plane
uncertainty not present in the 3D analysis. Thus, the setup and execution of the
evaluate the time associated with the mesh fineness and verification analyses that
ensure a proper computation is being made. This aspect is much less time
intensive in the 2D analysis than in 3D, and therefore makes 2D a very viable
alternative.
compute differential settlements that are less than the actual measured settlements
of the Tower. This may be because of errors in the measured settlements and
inclinations. Throughout the analysis, the results are compared to estimated final
111
the Tower in the year 1173 are unknown. If a more refined settlement history
existed, the process of evaluating the accuracy of the analyses would be much
more definitive. Other sources for the underestimate of settlement include the
assumption to treat the hollow cylindrical Tower as a complete circle, how the
error, and variation in the soil properties. Also, as suggested in the literature, the
large 3o increase in tilt from years 1370 to 1550 could be a result of external
in the Tower of Pisa. Many documents concerning the Tower’s subsurface, the
soil parameters, the Tower’s construction and the inclination history are available.
For these reasons, the Tower of Pisa was considered a good choice for a case
study.
complex shaped be analyzed. In some ways the shape of the Tower was
beneficial but also provided limitations. The Tower is actually a hollow cylinder,
but for the analysis, it was modeled as having a solid circular base. This circular
formation was easy to model in all three dimensions, which made it reasonable to
make settlement comparisons. However, the circularly symmetric Tower does not
112
been rectangular, a cross, or L-shaped the 1D analysis would have been far more
uniform structures would have been displayed. The circularly symmetric Tower
provided a level playing field, but also inhibited some of the potential of the more
sophisticated programs.
important to stay current and verify its place in the state of practice. If further
113
REFERENCES
Andreson, L., Edgers, L. and Jostad, H.P. (2008). “Capacity Analysis of Suction
Anchors in Clay by PLAXIS 3D Foundation”, PLAXIS Bulletin, Vol. 24,
pp. 5-9.
Bai, J., Chan, D., and Morgenstern, N. (2008). “Numerical Analysis of Time-
Dependent Behaviour for the Leaning Tower of Pisa”, Japanese
Geotechnical Society, Soils and Foundations, Vol. 48, No. 2, pp. 207-220.
Calabresi, G., Rampello, S., and Callisto, L. (1993). “The leaning Tower of Pisa:
Geotechnical characterization of the tower’s subsoil within the framework
of the critical state theory.” Universita degli studi di Roma “La Sapienza”,
Dipartimento di Ingegneria Strutturale e Geotecnia, Studi e Ricerche
Croce, A., Burghignoli, A., Calabresi, G., Evangelista, A., and Viggiani, C.
(1981). “The Tower of Pisa and the Surrounding Square: Recent
Observations”, 10th International Conference on Soil Mechanics and
Foundation Engineering (ICSMFE), Stockholm, Vol. 3, pp. 61-70.
114
Davis, E.H. and Poulos, H.G. (1968). “The Use of Elastic Theory Settlement
Prediction Under Three-Dimensional Conditions,” Geotechnique, London,
England, Vol. 18, No. 1, pp. 67-91.
Duncan, J.M. (1999). “Applying the Finite Element Method to Practical Use in
Geotechnical Engineering”, Civil Engineering Practice, Vol. 14, No. 2, pp.
75-80.
Lambe, T.W. and Whitman, R.V. (1969). “Soil Mechanics” p. 553., John Wiley
and Sons, Inc., New York.
Mitchell, J.K., Vivatrat, V., and Lambe, T.W. (1977). “Foundation Performance
of Tower of Pisa”, ASCE, Journal of the Geotechnical Engineering
Division, Vol. 103, No. GT3, pp. 227-249.
115
Mitchell, J.K., Vivatrat, V. and Lambe, T.W. (1979). Closure of “Foundation
Performance of Tower of Pisa”. ASCE, Journal of Geotechnical
Engineering Division, Vol. GT11, pp. 1363-1365.
PLAXIS 8.2. (1998). Finite element code for soil and rock analysis, version 8.2,
R.B.J. Brinkgreve and P.A. Vermeer, eds., Rotterdam, The Netherlands.
PLAXIS 3D Foundation Version 2 (2007). Finite element code for soil and rock
analysis, version 2, R.B.J. Brinkgreve and W.M. Swolfs, eds., Rotterdam,
The Netherlands.
Soga, K. and Mitchell, J.K. (1995). “Numerical and experimental studies related
to the deformation history of the Tower of Pisa”, Compression and
Consolidation of Clayey Soils. Yoshinkuni and Kusakabe, Balkema,
Rotterdam.
Son, N.N. and Das, S.C. (2003). “Theory and Practice of Foundation Design”, p.
517, PHI Learning Pvt. Ltd., New Dehli, India.
Terzaghi, K., (1934). “Die Ursachen der Schiefstellung des Turms von Pisa,” Der
Bauingenieur, No. ½, 1934, pp. 1-4.
116
APPENDIX
117
Plaxis 2D Runs
118
Plaxis 3D Runs