Académique Documents
Professionnel Documents
Culture Documents
John Krahn
SIGMA/W
Table of Contents
Table of Contents
1
Introduction ............................................................. 1
1.1
Applications.................................................................................. 1
Deformation analysis............................................................. 1
Embankment/excavation construction .................................. 3
Excess pore-water pressures................................................ 4
Soil-structure interactions...................................................... 4
Consolidation analyses ......................................................... 6
Introduction .................................................................................. 9
2.2
2.3
2.4
2.5
Page i
Table of Contents
SIGMA/W
2.6
2.7
Introduction ................................................................................ 39
3.2
3.3
3.4
Plasticity..................................................................................... 44
3.5
3.6
3.7
3.8
3.9
3.10
3.11
Units........................................................................................... 67
Meshing ................................................................ 69
4.1
Introduction ................................................................................ 69
Page ii
SIGMA/W
4.2
Table of Contents
4.3
Regions...................................................................................... 77
Region types ....................................................................... 78
Region points ...................................................................... 78
Region properties ................................................................ 80
4.4
4.5
4.6
4.7
4.8
Infinite elements......................................................................... 97
Infinite elements in SIGMA/W ............................................. 98
4.9
Page iii
Table of Contents
SIGMA/W
5.2
5.3
5.4
5.5
5.6
5.7
Page iv
SIGMA/W
Table of Contents
5.8
5.9
6.2
6.3
6.4
6.5
6.6
7.2
7.3
7.4
7.5
8.2
Page v
Table of Contents
SIGMA/W
8.4
8.5
10
9.2
9.3
9.4
10.2
10.3
10.4
10.5
10.6
10.7
10.8
10.9
10.10
10.11
Page vi
SIGMA/W
Table of Contents
11
11.2
Convergence............................................................................ 233
Graphing convergence...................................................... 236
12
11.3
11.4
11.5
11.6
11.7
12.2
12.3
12.4
12.5
12.6
12.7
13
12.8
12.9
13.2
13.3
Page vii
Table of Contents
SIGMA/W
13.5
13.6
13.7
13.8
13.9
13.10
13.11
13.12
13.13
13.14
13.15
Page viii
SIGMA/W
Table of Contents
14
14.2
14.3
14.4
14.5
14.6
14.7
14.8
Page ix
Table of Contents
15
SIGMA/W
Theory................................................................. 369
15.1
15.2
16
15.3
15.4
15.5
15.6
16.2
References................................................................... 393
Index ............................................................................ 397
Page x
SIGMA/W
Chapter 1: Introduction
Introduction
SIGMA/W is a finite element software product that can be used to perform stress
and deformation analyses of earth structures. Its comprehensive formulation makes
it possible to analyze both simple and highly complex problems. For example, you
can perform a simple linear elastic deformation analysis or a highly sophisticated
nonlinear elastic-plastic effective stress analysis. When coupled with other GEOSLOPE software products, it can also model the pore-water pressure generation
and dissipation in a soil structure in response to external loads using either a fully
coupled or un-coupled formulation. SIGMA/W has application in the analysis and
design for geotechnical, civil, and mining engineering projects.
1.1
Applications
Page 1
Chapter 1: Introduction
SIGMA/W
30
Exaggeration 25X
28
Tank
El
26
24
ti
(
22
20
)
18
16
14
12
10
8
6
4
2
0
Tank
Elevation (metres)
26
24
22
30
20
20
18
16
14
10
12
10
8
Page 2
SIGMA/W
Chapter 1: Introduction
Embankment/excavation construction
Finite elements can be added or removed from the finite element mesh to simulate
the construction of fill placement or excavation. You can identify elements to be
activated or deactivated at various stages, making it possible to simulate the
process over time. Figure 1-3 shows a deformed mesh after placement of one lift,
and Figure 1-4 shows the closure around a tunnel in anisotropic linear-elastic
material.
+2
+3
+3
+3
+3
+3
+3
+3
+3
+2
+2
+2
+2
+3
+2
+2
+2
+2
+2
+2
+2
+2
+2
+2
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+2
+1
+3
+2
+2
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
Page 3
Chapter 1: Introduction
SIGMA/W
40
60
80
100
120
160
140
100
120
60
80
40
140
Infinite Elements
160
Infinite Elements
200
260
Soil-structure interactions
SIGMA/W can accommodate soil-structure interaction problems by including
structural elements in two-dimensional plain strain analyses. These elements can
be beam elements which have flexural stiffness, or bar elements which have only
axial stiffness with no flexural stiffness. These structural elements are particularly
useful when analyzing cases such as sheet-pile walls. Figure 1-6 illustrates the
deformation for such a case, and Figure 1-7 shows the associated moment
distribution in the sheet-pile wall.
Page 4
SIGMA/W
Chapter 1: Introduction
Distance
20
15
10
0
-300
-200
-100
100
200
300
400
Moment
Page 5
Chapter 1: Introduction
SIGMA/W
Consolidation analyses
SIGMA/W can be used together with SEEP/W to perform a fully-coupled
consolidation analysis. When these two integrated products are run simultaneously,
SIGMA/W calculates the deformations resulting from pore-water pressure changes
while SEEP/W calculates transient pore-water pressure changes. This procedure is
used to simulate the consolidation process in both saturated and unsaturated soils.
A fully-coupled analysis is required to correctly model the pore-water pressure
response to an applied load. In certain cases, the pore-water pressure increase
under an applied load can be greater than the applied load. This phenomenon is
known as the Mendel-Cryer effect. Figure 1-8 shows a SIGMA/W analysis of a
saturated triaxial sample with an applied lateral load of 100 kPa. The initial porewater pressure before loading is zero. Figure 1-9 shows the pore-water pressure
response with time at the center of the sample. The pore-water pressure rises to
about 110 kPa, (110% of the applied load), before it gradually decreases.
Page 6
SIGMA/W
Chapter 1: Introduction
Pore-Water Pressure
150
100
50
0
0
100
200
300
400
500
600
Time
Page 7
Chapter 1: Introduction
SIGMA/W
Page 8
SIGMA/W
2.1
Introduction
Page 9
SIGMA/W
This chapter discusses the what, why and how of the numerical modeling
process and presents guidelines on the procedures that should be followed in good
numerical modeling practice. It should be noted now that some of the examples in
this chapter pertain to stress-deformation analysis while others relate to seepage
analysis. While the actual type of analyses may differ, the concepts illustrated
apply to all engineering analyses.
2.2
Page 10
SIGMA/W
Z (m)
0.7
Fine Sand
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
2.2
2.4
X (m)
The fact that mathematics can be used to simulate real physical processes is one of
the great wonders of the universe. Perhaps physical processes follow mathematical
rules, or mathematics has evolved to describe physical processes. Obviously, we
do not know which came first, nor does it really matter. Regardless of how the
relationship developed, the fact that we can use mathematics to simulate physical
processes leads to developing a deeper understanding of physical processes. It may
even allow for understanding or discovering previously unknown physical
processes.
Page 11
SIGMA/W
Numerical modeling has many advantages over physical modeling. The following
are some of the more obvious advantages:
Page 12
SIGMA/W
2.3
Ground
profile
Empiricism,
Precedent
Soil
behaviour
Modeling
The soil behavior component includes laboratory tests, in situ tests and field
measurements. The ground profile component basically involves site
characterization: defining and describing the site conditions. Modeling may be
conceptual, analytical or physical.
Page 13
SIGMA/W
Page 14
SIGMA/W
Genesis / geology
Ground
Profile
Site investigation,
ground description
Empiricism,
precedent,
experience,
risk management
Soil
Behaviour
Modeling
Idealization followed by
evaluation. Conceptual
or physical modeling,
analytical modeling
2.4
Why model?
The first reaction to the question, why model? seems rather obvious. The
objective is to analyze the problem. Upon more thought, the answer becomes more
complex. Without a clear understanding of the reason for modeling or identifying
what the modeling objectives are, numerical modeling can lead to a frustrating
experience and uncertain results. As we will see in more detail in the next section,
it is wrong to set up the model, calculate a solution and then try to decide what the
results mean. It is important to decide at the outset the reason for doing the
modeling. What is the main objective and what is the question that needs to be
answered?
The following points are some of the main reasons for modeling, from a broad,
high level perspective. We model to:
Page 15
SIGMA/W
compare alternatives,
Quantitative predictions
Most engineers, when asked why they want to do some modeling, will say that
they want to make a prediction. They want to predict the deformation under a
footing, for example; or the time for a contaminant to travel from the source to a
seepage discharge point; or the time required from first filling a reservoir until
steady-state seepage conditions have been established in the embankment dam.
The desire is to say something about future behavior or performance.
Making quantitative predictions is a legitimate reason for doing modeling.
Unfortunately, it is also the most difficult part of modeling, since quantitative
values are often directly related to the material properties. The extent of
deformation under a footing, for example, is in large part controlled by the
stiffness of the soil, which is itself controlled by the amount of confining stress.
While it is realistic to obtain accurate stress distributions beneath the footing, we
should realize that the accuracy of the deformation is only as good as our
knowledge and confidence in our stiffness parameters.
Carter et al. (2000) presented the results of a competition conducted by the German
Society for Geotechnics. Packages of information were distributed to consulting
engineers and university research groups. The participants were asked to predict
the lateral deflection of a tie-back shoring wall for a deep excavation in Berlin.
During construction, the actual deflection was measured with inclinometers. Later
the predictions were compared with the actual measurements. Figure 2-5 shows the
best eleven submitted predictions. Other predictions were submitted, but were
considered unreasonable and consequently not included in the summary.
There are two heavy dark lines superimposed on Figure 2-5. The dashed line on
the right represents the inclinometer measurements uncorrected for any possible
base movement. It is likely the base of the inclinometer moved together with the
base of the wall. Assuming the inclinometer base moved about 10 mm, the solid
heavy line in Figure 2-5 has been shifted to reflect the inclinometer base
movement.
At first glance one might quickly conclude that the agreement between prediction
and actual lateral movement is very poor, especially since there appears to be a
Page 16
SIGMA/W
Page 17
SIGMA/W
Deflection (mm)
-60
-50
-40
-30
-20
-10
10
0
16
20
12
24
computed
28
measured
32
-60
-50
-40
-30
-20
-10
Compare alternatives
Numerical modeling is useful for comparing alternatives. Keeping everything else
the same and changing a single parameter makes it a powerful tool to evaluate the
significance of individual parameters. For modeling alternatives and conducting
sensitivity studies it is not all that important to accurately define some material
properties. All that is of interest is the change between simulations.
Consider the example of a cut-off wall beneath a structure. With SEEP/W it is easy
to examine the benefits obtained by changing the length of the cut-off. Consider
Page 18
SIGMA/W
two cases with different cut-off depths to assess the difference in uplift pressures
underneath the structure. Figure 2-6 shows the analysis when the cutoff is 10 feet
deep. The pressure drop and uplift pressure along the base are shown in the left
graph in Figure 2-7. The drop across the cutoff is from 24 to 18 feet of pressure
head. The results for a 20-foot cutoff are shown in Figure 2-7 on the right side.
Now the drop across the cutoff is from 24 to about 15 feet of pressure head. The
uplift pressures at the downstream toe are about the same.
The actual computed values are not of significance in the context of this
discussion. It is an example of how a model such as SEEP/W can be used to
quickly compare alternatives. Secondly, this type of analysis can be done with a
rough estimate of the conductivity, since in this case the pressure distributions will
be unaffected by the conductivity assumed. There would be no value in carefully
defining the conductivity to compare the base pressure distributions.
We can also look at the change in flow quantities. The absolute flow quantity may
not be all that accurate, but the change resulting from various cut-off depths will be
of value. The total flux is 6.26 x 10-3 ft3/s for the 10-foot cutoff and 5.30 x 10-3 ft3/s
for the 20-foot cutoff, only about a 15 percent difference.
6.2592e-003
Page 19
SIGMA/W
Cutoff - 20 feet
25
25
20
20
Cutoff 10 feet
15
10
5
0
30
15
10
5
50
70
90
110
Distance - feet
0
30
50
70
90
110
Distance - feet
Page 20
SIGMA/W
Base Case
thinner
high
low
Thickness
of Growth Medium
bare surface
Transpiration
low
high
deep
Hydraulic Conductivity
of Compacted Layer
shallow
Root Depth
low
Decreasing
Net Percolation
high
Hydraulic Conductivity
of Growth Medium
Increasing
Net Percolation
Page 21
SIGMA/W
the SEEP/W engineering book, describes in detail the nature of the hydraulic
conductivity (or permeability) functions. For this example, relative conductivity
functions such as those presented in Figure 2-10 are sufficient. At low suctions
(i.e., near saturation), the coarse material has a higher hydraulic conductivity than
the fine material, which is intuitive. At high suctions, the coarse material has the
lower conductivity, which often appears counterintuitive.
Fine
Coarse
Coarse
Fine
OR
Material to be
protected
Material to be
protected
1.00E-05
Conductivity
Coarse
Fine
1.00E-06
1.00E-07
1.00E-08
1.00E-09
1.00E-10
1
10
100
1000
Suction
After conducting various analyses and trial runs with varying rates of surface
infiltration, it becomes evident that the behavior of the cover system is dependent
on the infiltration rate. At low infiltration rates, the effect of placing the fine
Page 22
SIGMA/W
material over the coarse material results in infiltration being drained laterally
through the fine layer, as shown in Figure 2-11. This accomplishes the design
objective of the cover. If the precipitation rate becomes fairly intensive, then the
infiltration drops through the fine material and drains laterally within the lower
coarse material as shown in Figure 2-12. The design of fine soil over coarse soil
may work, but only in arid environments. The occasional cloud burst may result in
significant water infiltrating into the underlying coarse material, which may result
in increased seepage into the waste. This may be a tolerable situation for short
periods of time. If most of the time precipitation is modest, the infiltration will be
drained laterally through the upper fine layer into a collection system.
So, for an arid site the best solution is to place the fine soil on top of the coarse
soil. This is contrary to what one might expect at first. The first reaction may be
that something is wrong with the software, but it may be that our understanding of
the process and our general thinking is flawed.
A closer examination of the conductivity functions provides a logical explanation.
The software is correct and provides the correct response given the input
parameters. Consider the functions in Figure 2-13. When the infiltration rate is
large, the negative water pressures or suctions will be small. As a result, the
conductivity of the coarse material is higher than the finer material. If the
infiltration rates become small, the suctions will increase (water pressure becomes
more negative) and the unsaturated conductivity of the finer material becomes
higher than the coarse material. Consequently, under low infiltration rates it is
easier for the water to flow through the fine, upper layer soil than through the
lower more coarse soil.
Fine
Coarse
Page 23
SIGMA/W
Fine
Coarse
This type of analysis is a good example where the ability to utilize a numerical
model greatly assists our understanding of the physical process. The key is to think
in terms of unsaturated conductivity as opposed to saturated conductivities.
Numerical modeling can be crucial in leading us to the discovery and
understanding of real physical processes. In the end the model either has to
conform to our mental image and understanding or our understanding has to be
adjusted.
1.00E-04
1.00E-05
Conductivity
Coarse
Fine
1.00E-06
1.00E-07
Intense
Rainfall
1.00E-08
Low to Modest
Rainfall
1.00E-09
1.00E-10
1
10
100
1000
Suction
Page 24
SIGMA/W
This is a critical lesson in modeling and the use of numerical models in particular.
The key advantage of modeling, and in particular the use of computer modeling
tools, is the capability it has to enhance engineering judgment, not the ability to
enhance our predictive capabilities. While it is true that sophisticated computer
tools greatly elevated our predictive capabilities relative to hand calculations,
graphical techniques, and closed-form analytical solutions, still, prediction is not
the most important advantage these modern tools provide. Numerical modeling is
primarily about process - not about prediction.
The attraction of ... modeling is that it combines the subtlety of human judgment with the
power of the digital computer. Anderson and Woessner (1992).
2.5
How to model
Page 25
SIGMA/W
=0
=0
u=0
Z=2m, = 10
kN/m3
0
20
20
=-u
Z=10m, = 20 kN/
m3
220
120
100
20
15
10
0
20
40
60
80
100
Y-Effective Stress
The hand calculation together with the SIGMA/W output can now be used to judge
the validity of the computed results. If there is no resemblance between what is
expected and what is computed with SIGMA/W then either the preliminary mental
picture of the situation was not right or something has been inappropriately
specified in the numerical model. Perhaps the pressure boundary condition was not
specified on the ground surface in addition to the definition of the water table. The
water table controls pore-water pressures in the soil, but does not add a weight load
Page 26
SIGMA/W
at the ground surface. Both must be specified and this error would be obvious by
comparing expectations to computed results. Any differences ultimately need to be
resolved in order for you to have any confidence in your modeling. If you had
never made a preliminary guess at the solution then it would be very difficult to
judge the validity of the numerical modeling results.
Another extremely important part of modeling is to clearly define, at the outset, the
primary question to be answered by the modeling process. Is the main question the
stress distribution or is it deformation? If your main objective is to determine the
stress distribution, there is no need to spend a lot of time on establishing an
advanced soil model, a more simple linear-elastic model is adequate. If on the
other hand your main objective is to estimate deformation, then a greater effort is
needed in determining the appropriate stress-strain model.
Sometimes modelers say I have no idea what the solution should look like - that is
why I am doing the modeling. The question then arises, why can you not form a
mental picture of what the solution should resemble? Maybe it is a lack of
understanding of the fundamental processes or physics, maybe it is a lack of
experience, or maybe the system is too complex. A lack of understanding of the
fundamentals can possibly be overcome by discussing the problem with more
experienced engineers or scientists, or by conducting a study of published
literature. If the system is too complex to make a preliminary estimate then it is
good practice to simplify the problem so you can make a guess and then add
complexity in stages so that at each modeling interval you can understand the
significance of the increased complexity. If you were dealing with a very
heterogenic system, you could start by defining a homogenous cross-section,
obtaining a reasonable solution and then adding heterogeneity in stages. This
approach is discussed in further detail in a subsequent section.
If you cannot form a mental picture of what the solution should look like prior to
using the software, then you may need to discover or learn about a new physical
process as discussed in the previous section.
Effective numerical modeling starts with making a guess of what the solution should look
like.
Other prominent engineers support this concept. Carter (2000) in his keynote
address at the GeoEng2000 Conference in Melbourne, Australia, when talking
about rules for modeling, stated verbally that modeling should start with an
Page 27
SIGMA/W
estimate. Prof. John Burland made a presentation at the same conference on his
work with righting the Leaning Tower of Pisa. Part of the presentation was on the
modeling that was done to evaluate alternatives and while talking about modeling
he too stressed the need to start with a guess.
Simplify geometry
Numerical models need to be a simplified abstraction of the actual field conditions.
In the field the stratigraphy may be fairly complex and boundaries may be
irregular. In a numerical model the boundaries need to become straight lines and
the stratigraphy needs to be simplified so that it is possible to obtain an
understandable solution. Remember, it is a model, not the actual conditions.
Generally, a numerical model cannot and should not include all the details that
exist in the field. If attempts are made at including all the minute details, the model
can become so complex that it is difficult and sometimes even impossible to
interpret or even obtain results.
Figure 2-16 shows a stratigraphic cross section (National Research Council Report
1990). A suitable numerical model for simulating the flow regime between the
groundwater divides is something like the one shown in Figure 2-17. The
stratigraphic boundaries are considerably simplified for the finite element analysis.
As a general rule, a model should be designed to answer specific questions. You
need to constantly ask yourself while designing a model, if this feature will
significantly affects the results. If you have doubts, you should not include it in the
model, at least not in the early stages of analysis. Always start with the simplest
model.
Page 28
SIGMA/W
The tendency of novice modelers is to make the geometry too complex. The
thinking is that everything needs to be included to get the best answer possible. In
numerical modeling this is not always true. Increased complexity does not always
lead to a better and more accurate solution. Geometric details can, for example,
even create numerical difficulties that can mask the real solution.
Start simple
One of the most common mistakes in numerical modeling is to start with a model
that is too complex. When a model is too complex, it is very difficult to judge and
interpret the results. Often the result may look totally unreasonable. Then the next
question asked is - what is causing the problem? Is it the geometry, is it the
material properties, is it the boundary conditions, or is it the time step size or
something else? The only way to resolve the issue is to make the model simpler
and simpler until the difficulty can be isolated. This happens on almost all projects.
It is much more efficient to start simple and build complexity into the model in
stages, than to start complex, then take the model apart and have to rebuild it back
up again.
A good start may be to take a homogeneous section and then add geometric
complexity in stages. For the homogeneous section it is likely easier to judge the
validity of the results. This allows you to gain confidence in the boundary
conditions and material properties specified. Once you have reached a point where
the results make sense, you can add different materials and increase the complexity
of your geometry.
Page 29
SIGMA/W
Another approach may be to start with a steady-state analysis even though you are
ultimately interested in a transient process. A steady-state analysis gives you an
idea as to where the transient analysis should end up: to define the end point. Using
this approach you can then answer the question of how does the process migrate
with time until a steady-state system has been achieved.
It is unrealistic to dump all your information into a numerical model at the start of
an analysis project and magically obtain beautiful, logical and reasonable
solutions. It is vitally important to not start with this expectation. You will likely
have a very unhappy modeling experience if you follow this approach.
Do numerical experiments
Interpreting the results of numerical models sometimes requires doing numerical
experiments. This is particularly true if you are uncertain as to whether the results
are reasonable. This approach also helps with understanding and learning how a
particular feature operates. The idea is to set up a simple problem for which you
can create a hand calculated solution.
Consider the following example taken from a seepage analysis. You are uncertain
about the results from a flux section or the meaning of a computed boundary flux.
To help satisfy this lack of understanding, you could do a numerical experiment on
a simple 1D case as shown in Figure 2-18. The total head difference is 1 m and the
conductivity is 1 m/day. The gradient under steady state conditions is the head
difference divided by the length, making the gradient 0.1. The resulting total flow
through the system is the cross sectional area times the gradient which should be
0.3 m3/day. The flux section that goes through the entire section confirms this
result. There are flux sections through Elements 16 and 18. The flow through each
element is 0.1 m3/day, which is correct since each element represents one-third of
the area.
Another way to check the computed results is to look at the node information.
When a head is specified, SEEP/W computes the corresponding nodal flux. In
SEEP/W these are referred to as boundary flux values. The computed boundary
nodal flux for the same experiment shown in Figure 2-18 on the left at the top and
bottom nodes is 0.05. For the two intermediate nodes, the nodal boundary flux is
0.1 per node. The total is 0.3, the same as computed by the flux section. Also, the
quantities are positive, indicating flow into the system. The nodal boundary values
on the right are the same as on the left, but negative. The negative sign means flow
out of the system.
Page 30
12
15
11
14
10
13
18
21
24
27
30
17
20
23
26
29
16
19
22
25
28
1.0000e-001
3.0000e-001
1.0000e-001
SIGMA/W
A simple numerical experiment takes only minutes to set up and run, but can be
invaluable in confirming to you how the software works and in helping you
interpret the results. There are many benefits: the most obvious is that it
demonstrates the software is functioning properly. You can also see the difference
between a flux section that goes through the entire problem versus a flux section
that goes through a single element. You can see how the boundary nodal fluxes are
related to the flux sections. It verifies for you the meaning of the sign on the
boundary nodal fluxes. Fully understanding and comprehending the results of a
simple example like this greatly helps increase your confidence in the
interpretation of results from more complex problems.
Conducting simple numerical experiments is a useful exercise for both novice and
experienced modelers. For novice modelers it is an effective way to understand
fundamental principles, learn how the software functions, and gain confidence in
interpreting results. For the experienced modeler it is an effective means of
refreshing and confirming ideas. It is sometimes faster and more effective than
trying to find appropriate documentation and then having to rely on the
documentation. At the very least it may enhance and clarify the intent of the
documentation.
Model only essential components
One of the powerful and attractive features of numerical modeling is the ability to
simplify the geometry and not to have to include the entire physical structure in the
model. A very common problem is the seepage flow under a concrete structure
with a cut-off as shown in Figure 2-19. To analyze the seepage through the
Page 31
SIGMA/W
foundation it is not necessary to include the dam itself or the cut-off as these
features are constructed of concrete and assumed impermeable.
Another common example is the downstream toe drain or horizontal under drain in
an embankment (Figure 2-20). The drain is so permeable relative to the
embankment material that the drain does not contribute to the dissipation of the
head (potential energy) loss through the structure. Physically, the drain needs to
exist in the embankment, but it does not need to be present in a numerical model. If
the drain becomes clogged with fines so that it begins to impede the seepage flow,
then the situation is different and the drain would need to be included in the
numerical model. With any material, the need to include it in the analysis should
be decided in the context of whether it contributes to the head loss.
Another example is the downstream shell of a zoned dam as illustrated in Figure
2-21. Often the core is constructed of fine-grained soil while the shells are highly
permeable coarse granular material. If there is a significant difference between
core and shell conductivities then seepage that flows through the core will drip
along the downstream side of the core (usually in granular transition zones) down
to an under drain. If this is the case, the downstream shell does not need to be
included in the seepage analysis, since the shell is not physically involved in the
dissipation of the head loss. Once again the shell needs to exist physically, but
does not need to be included in the numerical seepage model.
Page 32
SIGMA/W
10
15
20
25
30
35
40
Figure 2-21 Head loss through dam core with downstream shell
Page 33
SIGMA/W
Page 34
SIGMA/W
Page 35
SIGMA/W
2.6
Page 36
SIGMA/W
2.7
Closing remarks
Page 37
SIGMA/W
Page 38
SIGMA/W
Chapter 3: Fundamentals
3.1
Introduction
SIGMA/W is a powerful finite element method tool that can be used to model a
wide range of stress-strain problems. In its appearance and usability it is much like
other products in GeoStudio. For example, geometry and finite element meshes or
soil regions common to seepage or slope analysis can be used directly in
SIGMA/W. However, there are certain fundamental differences between
SIGMA/W and the other models. Foremost of these differences is that SIGMA/W
uses an incremental load formulation. This means that you enter a step change
condition (e.g. an applied surface load) and it computes step changes in parameters
such as strain and pore-water pressure. It does not compute the actual end
condition unless you have established in-situ conditions prior to loading.
All of the GeoStudio Engineering books have a chapter focusing on Modeling
Tips. The Modeling Tips chapters in the other books generally cover items specific
to enhancing the usefulness of the program from a model set up and data
interpretation perspective. However, due to SIGMA/Ws uniqueness, this book has
this chapter titled Fundamentals and Practical Modeling Considerations and it is
placed near the front of the book in hopes it is read and understood prior to reading
the rest of the book. This entire chapter will be most useful to novice users. More
advanced users will find the first part a good review, but not entirely necessary. All
users should ensure they read and understand the second part of this chapter.
This chapter has two main parts: the first few sections deal with an overview of
fundamental concepts related to stress, volume change and strength of materials. A
detailed theoretical discussion of various soil strength models is given in the
Theory chapter. This discussion will focus more on ensuring there is a basic
understanding of important parameters affecting volume change and soil strength
so that the reader appreciates at a deeper level why SIGMA/W was formulated a
certain way and what it is formulated to do.
The second several sections of this chapter will focus on how the fundamental
concepts have been put into the model and how the model can be used in various
unique and creative ways to solve real problems.
Page 39
Chapter 3: Fundamentals
SIGMA/W
A key concept that will be developed and stressed throughout this chapter is that
SIGMA/W is a very powerful tool for investigating the serviceability of
engineered soil systems. It is not a tool for predicting stability of these systems.
Serviceability deals with obtaining confidence that a design will function as
intended. The numerical soil models available in SIGMA/W can help show that
engineered systems exposed to various applied external or internal loads will not
fail.
Stability, on the other hand, deals with quantification of a factor of safety for an
engineered system. However, in order to obtain a factor of safety it is necessary to
have solved the numerical models to some point past failure so that the stress
condition after failure can be compared to the allowable stress condition at failure.
The problem with this is that the soil models in a small strain stress-deformation
analysis are not defined at stress conditions that exceed failure. Therefore, a safety
factor can not be computed.
It is important to realize this, so that you can begin the modeling and design
process by looking for an engineered system that works! If you generate results
that indicate failure, two things should come to mind: 1) that the results are likely
meaningless and, 2) that you are reminded the true objective is to design a working
system, not one that fails. In other words, model to design a system that is
serviceable.
3.2
m
V
where:
Page 40
SIGMA/W
Chapter 3: Fundamentals
Some caution should be exercised when working with the term density. Density
relates to mass per unit volume, which should not be confused with weight
density denoted by the symbol, . The weight density is also called the unit
weight and it has units of gravitational force (N, lbf) per unit volume. Since
SIGMA/W is formulated to deal with displacement in response to forces, it makes
use of the unit weight form of density.
The bulk, or total density of a soil refers to the mass of bulk soil which includes
soil particles, water and air in a given volume. The bulk density will therefore
change throughout a domain if there is a change in water content throughout the
domain. There are other useful ways of quantifying density: the dry density,
saturated density and buoyant density.
Dry density is the mass of dry soil per unit volume and is commonly referred to
when dealing with compaction of earthworks. The dry density is related to the bulk
density of the soil according to:
d =
1+ w
where:
w
The saturated density, sat, describes the bulk density of a soil when all voids are
filled with water and the submerged density is simply the saturated density minus
the density of water as follows:
' = sat w
where:
As with the unit weight of the bulk soil (including water weight), the submerged
unit weight can be a useful parameter to use in SIGMA/W. The submerged unit
weight is simply the total unit weight minus the unit weight of water as follows:
' = w
Page 41
Chapter 3: Fundamentals
SIGMA/W
where:
w
Further discussion on the use of submerged unit weight is given in the Analysis
Types chapter discussion on in-situ analysis, and in the chapter on Pore-water
Pressures.
3.3
Soil is used as a fill material in many engineering projects and in order to obtain
satisfactory serviceability of the project, compaction (densification) of the fill is
often required. Compaction of a loose material has several advantages from a
design perspective, as it can be used to increase shear strength, decrease
compressibility, and decrease permeability. In order to know how much
compaction effort is required in order to achieve the desired soil property, it is
useful to conduct a laboratory compaction test on a representative soil sample from
the site. The objective of the test is to determine, for a given compactive effort, the
optimum moisture content that will result in the maximum dry density. The
laboratory test used most often is the Standard Proctor Test, introduced by R.R.
Proctor in 1933.
Compaction involves application of mechanical energy in order to reduce the air
voids in a soil with little or no reduction in water content. At low moisture
contents, the soil particles are surrounded by thin films of water which tend to keep
the grains separated when compacted. When more water is added, the soil grains
move together more easily which results in a reduction of the air voids and an
increase in bulk (total) density. At some point, making the soil too wet prior to
applying the compactive effort will result in a reduction in density. The water
content that corresponds with the maximum dry density is termed the optimum
water content. Figure 3-1 shows the water contents and maximum dry densities for
a single soil type subjected to different compactive efforts. Inset in the figure are
illustrations of the orientation of soil fabric at different states. Generally, soils are
quantified based on whether they are compacted dry of optimum water content, at
the optimum water content, or wet of optimum water content. The usefulness of
this classification in terms of SIGMA/W will be discussed shortly.
Page 42
SIGMA/W
Chapter 3: Fundamentals
0
10
ati
tur
sa
on
Increasing
compactive
effort
Figure 3-1 Dry density, water content and compactive effort (after Holtz and
Kovacs, 1981)
Notice that adding more effort does not result in elimination of all air voids in a
soil. It does result in reducing the required optimum water content and in
increasing the maximum dry density. This type of a test comparison is very useful
for helping assess the economic benefits of applying less water or with more
compaction effort or visa versa.
Compaction and strength
Knowing the compactive effort and water content applied to an engineered soil can
lead to a better understanding of its anticipated behavior under loading. For clay
type soils, it is recognized that for a fixed compactive effort, as more water is
added, the soil structure becomes increasingly oriented (refer to points A, B and C
in Figure 3-1). That is, the clay minerals become less flocculated and more
dispersed. Dry of optimum, the structure or soil fabric is considered flocculated,
while wet of optimum, it is considered dispersed, or oriented. As the compactive
effort is increased, the amount of orientation is increased (refer to points C and D
in the figure).
Page 43
Chapter 3: Fundamentals
SIGMA/W
A full discussion of the effects on soil properties due to changing water content
and compactive effort is beyond the scope of this book. In general, however, it is
important to understand that soils compacted dry of optimum are likely to have
higher strengths than those compacted wet of optimum. In addition, hydraulic
permeability at a constant compactive effort decreases with increasing water
content and reaches a minimum at about the optimum water content. If compactive
effort is increased, the permeability decreases because the void ratio decreases.
Understanding the anticipated behavior based on these observations can help to
make the model set up and definition easier and the interpretation of output data
more meaningful. For example, if you are solving a consolidation analysis and you
are getting delayed response in dissipation of pore-water pressures within an
engineered structure, you may want to adjust the applied unit weight (a function of
total density) and make a corresponding change to the hydraulic permeability
function applied in SEEP/W. This type of fine tuning soil properties is not usually
recommended. However, you should understand from this discussion that there is a
connection between how things are actually built in the field and what the soil
properties can be.
3.4
Plasticity
SIGMA/W does not require, as input, any of the Atterberg limit parameters.
However, as they are fairly simple to ascertain, it is useful to discuss their
significance from a practical perspective. If you know the soil plastic and liquid
limits, then you can anticipate the soils response and better understand what you
are asking the model to do.
The Atterberg Limits were developed in 1911 as a practical means of describing
the plastic behavior of clays. While there are several classifications, the most
relevant to this discussion are the Plastic Limit, the Liquid Limit and the Liquidity
Index.
The liquid and plastic limits are both water content by weight values expressed in
decimal form. The plastic limit, PL, is the water content at which point the soil
starts to behave plastically when loaded. The liquid limit, LL, is the water content
when the soil just starts to behave as a viscous fluid when loaded. The liquidity
index is defined as:
LI =
w PL
LL PL
Page 44
SIGMA/W
Chapter 3: Fundamentals
where:
w
If the LI is between 0 and 1, the soil will behave plastically, and if it is greater than
1, the soil will behave like a viscous fluid when sheared. Figure 3-2 compares
anticipated stress-strain behavior for a range of liquid and plastic limits. It is clear
that as the plastic limit is approached, the potential for larger strains in response to
small changes is shear stress. The figure clearly shows that for soils approaching or
exceeding the liquid limit, the amount of strain in response to small load changes
can be infinite.
This is important to understand. If you are asking the model to compute strain for a
condition where the soil is at a state between its plastic and liquid limits, then there
is chance for large strains. The SIGMA/W formulation is a small strain
formulation, and often it can not converge on a reasonable solution when strains
become too large. In some cases, a solution can be reached, but it can have no
meaning. It is up to you to understand the behavior you expect from the soil and to
model a stress condition that will lead to a meaningful response. If you want a
structure to be useful in its application, then you must design (and model) it for
conditions that are within is range of serviceability.
Page 45
Shear stress
Chapter 3: Fundamentals
SIGMA/W
w < PL
w approaches PL
w approaches LL
w > LL
Strain
Figure 3-2 Stress-strain and Atterberg Limits
3.5
Volume change
v = i zi
i
Page 46
SIGMA/W
Chapter 3: Fundamentals
where:
z
i,n
So, in this case, the total stress equals (2m*10 kN/m3) + (10m*20 kN/m3), or 220
kN/m2 = 220 kPa.
Z=2m, = 10 kN/m3
Z=10m, = 20 kN/m3
The effective stress is different than the total stress because the presence of water
in between the soil grains provides a force that can either tend to push the particles
apart or pull them closer together; as would be the case if the water in the soil was
in a capillary state. In equation form, the effective stress is:
v' = v u
where:
u
Page 47
Chapter 3: Fundamentals
SIGMA/W
Based on the example illustrated in Figure 3-3, the vertical effective stress in the
soil element would be computed as the total stress, 220 kPa, minus the pore-water
pressure (= 12m*10 kN/m3 = 120 kPa) = 100 kPa.
The effective stress is less than the total stress which is, in effect, stating that the
water in the soil is taking up some of the weight (body) load of the soil and thereby
reducing the stress in the soil.
Compressibility and consolidation
Compressibility and consolidation deal with how the soil changes in response to a
change in its loading. When a load is first applied on a soil, it will compress, or
change volume due to:
In general, we neglect the deformation of soil grains and the compressibility of air
and water because they are small compared with the third point.
In a coarse material, the squeezing of water from the voids can occur quite quickly
which results in a rapid compression of the soil under loading. However, in a fine
material, the rate of compression is directly a function of the hydraulic
permeability of the soil and, therefore, the compression is time dependent. This
phenomenon is called consolidation.
Consider Figure 3-4 where a submerged 1m2 footing with a 50 kN load is applied
to the previously illustrated soil element. Two important points can be made using
this illustration.
When the load is first applied, there is a change in total stress on the soil of 50
kN/m2 or 50 kPa. Therefore, the total stress in the soil element is now 270 kPa.
When the load is first applied, there is no time for the finer grained soil particles to
shift around in order to assume the new load so the pore-water carries the load and
its pressure increases by an amount equal to the applied new stress. Therefore, the
pore-water pressure at the soil element increases by 50 kPa and the new effective
stress is computed to be 270 kPa minus 170 kPa = 100 kPa; which is the same as
the starting effective stress. If there is no change in effective stress, there is no
change in soil volume.
Page 48
SIGMA/W
Chapter 3: Fundamentals
50 kN load
Z=2m, = 10 kN/m3
Z=10m, = 20 kN/m3
Because this is a fine grained soil, we know that with time, the excess pore-water
pressure will dissipate back to the pre-load value. When this happens, the effective
stress will be computed as the new total stress minus the original pore-water
pressure, or 270 kPa 120 kPa = 150 kPa. The effective stress has increased which
means that the soil must change volume or consolidate.
No change in effective stress means no volume change. An increase in effective stress
means consolidation or shrinkage. A decrease in effective stress means swelling or heave.
Sometimes it is useful to plot individual effective stress parameters at the start and
end of a load step. Consider Figure 3-5 which shows water pressure response to the
loading discussed in Figure 3-4. The new total stress load here is considered to be
applied instantly which results in an equal increase in pore-water pressure at the
start of the consolidation phase.
Page 49
Chapter 3: Fundamentals
Z=2
SIGMA/W
= u
Pressure
dissipation over
time
Z=10
u end
u start
We can use this same approach and simultaneously consider the calculation of all
parameters for the two cases described above. Figure 3-6 shows the calculation of
all stresses and pressures for the two examples as a function of depth to the soil
element. The top of the figure shows the calculation of effective stress for the insitu case (prior to the load step) while the bottom of the figure shows what the
stresses are just after the footing load is applied and after consolidation has
occurred.
There are several key points to understand from this figure:
The step load from the footing instantly increases the pore water
pressure because the soil has not had time to dissipate the pressure.
The pressure increase is equal to the newly applied load.
Page 50
SIGMA/W
Chapter 3: Fundamentals
=0
Z=2
=0
u=0
20
20
=-u
Z=10
220
120
100
=0
Z=2
Pressure
dissipation
over time = 0
u=0
70
70
Increase in
effective stress
with time
= 50
Z=10
=-u
270
120
170
u decreases
100
150
increases
Figure 3-6 Stress and pressure profiles for two example cases
A concept that should be getting more evident at this point is that we can not
generally compute the actual stresses in the ground at any given point in time
simply based on the applied loads at that time. We must consider how we have
changed the external conditions at any given time and then compute the change in
the grounds response. For modeling purposes, we can, and should, think of our
effective stress equation as:
v' = v u
Page 51
Chapter 3: Fundamentals
SIGMA/W
where the delta symbol indicates a step change in each parameter. From our earlier
examples, the change in total stress, , equals 50 kPa and the change in porewater pressure, u, is also 50 kPa. Therefore, the change in effective stress just
after loading is equal to zero, and, there is no volume change.
This concept of changes in loads, or incremental loading, is discussed in more
detail later in this chapter. It is a critical concept to understand.
Zero change in pore-water pressure
What happens if there is no pore-water pressure change during a change in total
stress? If this is the case then u = 0 and our effective stress equation becomes:
v' = v 0
or
v' = v
In this case, the change in effective stress is equal to the total stress. By always
thinking of the stresses in the soil in terms of effective stress, our rule about
volume change only occurring with a change in effective stress always applies. It
will get confusing if you start thinking of volume change in response to changes in
total stress some of the time, and in terms of effective stress at other times. Always
understand what the pore-water pressure changes are (or are not), and think in
terms of effective stress.
Ko horizontal and vertical stresses
We know that with depth the water pressure acts equally in both the horizontal and
vertical directions. This is rarely the case for soils. We can express the ratio of
horizontal to vertical stress in the ground as:
h = K v
where:
K
This is not generally a valuable parameter to use because there can be variations in
the water content and total stress in the soil which means K is not constant. A
Page 52
SIGMA/W
Chapter 3: Fundamentals
h' = K 0 v'
where:
KO
K0 =
where:
h' = h u
h = h' + u
h = 50 + 120 = 170kPa
Page 53
Chapter 3: Fundamentals
SIGMA/W
Notice that the horizontal total stress is not equal to one half of the total vertical
stress of 220 kPa. It would be an error to apply the K0 value to the total stress. The
only way to get the correct total stress is to follow the procedure outlined above.
Having said that, the K0 hand calculation method only applies to an in-situ analysis
where the ground surface is horizontal and the vertical effective stress can readily
be computed as a function of depth. For the case where the ground is sloping and
the water table is not horizontal it is necessary to use a more advanced analysis.
This is discussed further in the Analysis Types chapter of this book.
3.6
Strength parameters
H = x Sin
A = Sin
T= x1
A=
A = Cos
V = y Cos
If the forces in on each face are broken into horizontal and vertical components
and converted to stresses by dividing them by the cross sectional area upon which
they act, two equations in terms of , and are developed. If these equations are
each set to zero, they can be squared and added together. The end result is a single
equation for a circle with a radius of:
Page 54
SIGMA/W
Chapter 3: Fundamentals
( x y )
2
and with its center at:
( x + y )
, 0
The vertical and horizontal planes in the soil element have no shear force acting on
them so they are, by definition, principal planes. Therefore, the forces acting on
them are principal forces. An important point to remember here is that principal
stresses act on planes where the shear strength is zero. Another point to note is that
the stress with the largest magnitude is called the major principal stress (denoted
1), and the lower value stress is the minor principal stress (denoted 3).
If the information from the summation of forces equations can be used to construct
a circle on the shear-normal stress plane as shown in Figure 3-8. Using this circle,
the applied normal and shear stress along any potential failure plane in a soil
element can be determined. In this case, the shear and normal stress point is shown
for a plane inclined at an angle, , from horizontal.
( , )
1
Figure 3-8 Mohr circle of stress
Page 55
Chapter 3: Fundamentals
SIGMA/W
'
c'
3'
1'
3'
1'
Figure 3-9 shows a Mohr Coulomb failure envelope for a clay type soil. Two Mohr
circles are plotted in this figure and a line is drawn tangent to the two circles such
that it also intersects the vertical shear stress axis. The point of intersection of the
tangent line provides the effective cohesion parameter and the angle of inclination
of the tangent line provides the effective friction angle. All of the parameters
illustrated in the figure have the apostrophe symbol, which indicates that they
are effective stress based values and not total stress based values. They are
effective stress based because the pore-water pressure in the sample was measured
during the testing and subtracted from the total stress values applied in the test.
Once the data from the shear tests has been graphed, as shown in Figure 3-9, it is
possible to create the equation that represents the data. We see that the tangent line
is a straight line with a y-intercept. Therefore, the equation of the line, termed the
Mohr Coulomb failure criteria, is given by:
Page 56
SIGMA/W
Chapter 3: Fundamentals
where the subscript, f, denotes the shear and normal stress condition in the soil
element at failure.
Triaxial testing
Entire textbooks can be written on the subject of performing and interpreting data
from the various types of triaxial tests on various types of soil. This type of a
discussion is far too advanced in this context. It is important, however, to
distinguish between two main types of test that can be carried out in a triaxial cell.
Undrained tests
Undrained tests can be performed on undisturbed as well as disturbed samples of
clay and silt, and disturbed samples of sand or gravel. The sample is allowed to
consolidate under a desired confining pressure and then sheared in axial
compression.
In an un-drained test, the pore-water in the sample is not allowed dissipate, so the
pressure builds up in response to changes in total stress. The changes in stress
result from application of an applied strain to the top of the sample. So, for
example, if the applied strain results in a principal major deviator stress of 50 kPa,
the water pressure in the soil should also increase by an equal amount.
The undrained test can be performed with or without pore-water pressure
measurements. If pore-water pressure is not measured, the results must be
expressed in terms of total stress. The field conditions that a soil consolidates
under are usually quite different than the conditions in the laboratory so the results
of this test, when expressed in terms of total stress, can only be applied in a limited
way to the field.
The results of a total stress triaxial test would look like those illustrated in Figure
3-10.
Page 57
Chapter 3: Fundamentals
SIGMA/W
=0
cu
1 3
If the pore-water pressure is measured during loading, the rate of applied strain
must be slow enough to allow the pressures to equalize and be measured. With the
measurement of pore-water pressure, the results can be expressed in terms of
effective stress and the values of c and can then be applied to a wider range of
field applications. The results of an undrained effective stress triaxial test were
illustrated previously in Figure 3-9.
The undrained test results are useful when a soil structure has had a long time to
come to equilibrium and then a sudden change in total stress is applied with
limited drainage. This may be the case for application of fill or rapid drawdown of
a reservoir.
Note, in SIGMA/W, if you use undrained total stress soil property parameters then
an estimate of the pore-water pressure response can be obtained using the porewater pressure parameters A and B. These are discussed in more detail in the
following section and in the Theory chapter.
Drained tests
A drained test can be performed on all types of samples, whether disturbed or not.
It is prepared in the same manner as an undrained test sample, but during the
loading stage, the water is allowed to drain further. The applied rate of strain must
be slow enough to prevent the development of excess pore-water pressures; in
other words, it should remain at the state it was in after initial consolidation prior
to loading.
Page 58
SIGMA/W
Chapter 3: Fundamentals
At completion of testing, the Mohr circle failure envelope will look very similar to
that shown for the undrained effective stress analysis, which makes sense because
if the soil is permitted to drain, then the change in pore-water pressure will be zero
and the change in effective stress will equal the change in total stress.
The drained test results are quite useful to use when the long term seepage
dependent consolidation is an issue. However, it is not an easy test to conduct
because it is important to ensure that pressures are not building up in the sample. In
addition, for low permeability soils, it can take a long time to allow drainage which
means a very slow rate of applied strain is required.
Soil stiffness Youngs modulus
Regardless of how the testing is carried out, the objective is to determine the
strength parameters (described above) and also the stress-strain behavior. The
stress-strain behavior is the actual constitutive model used in the solution of the
partial differential equations. The strength parameters are only useful for telling if
any given soil element has exceeded its yield point. Further discussion on the
usefulness of strength parameters in stress-deformation modeling is included in the
section on stability and serviceability below.
Consider Figure 3-11, which shows the stress-strain behavior for three confining
loads during a strength test. As the confining stress increases, so does the slope of
the deformation curve. It is the slope of these curves that relates the amount of
strain in response to an applied stress. In equation form, the slope is given by:
Ei =
or:
= Ei
which is better known as Hookes Law.
The slope, E, is known as Youngs Modulus and it is NOT constant for any given
soil. This modulus, which represents the stiffness of the soil, is dependent on the
effective confining stress, and it can change throughout the entire problem domain
depending on the stresses throughout the domain. So, the parameter we are solving
in SIGMA/W is strain, and in order to know strain we must know the stress and the
stiffness. However, the stiffness is itself a function of stress. This circular logic is
Page 59
Chapter 3: Fundamentals
SIGMA/W
Stress, (13)
what makes the solution a non-linear process, and it is why we have to iterate to
reach a converged result.
Strain,
Figure 3-11 Stress-strain behavior for various confining loads during testing
3.7
As we have been discussing in the previous few sections, the presence of water
under pressure can affect both how we obtain material properties and how they are
used in the solution. Following common geotechnical practice, material properties
can be specified in SIGMA/W using either effective stress parameters for analyses
of drained soils, or total stress parameters for undrained soils. In materials
specified using total stress parameters, undrained pore-water pressure changes can
be computed from total stress changes using pore-pressure parameters A and B. If
you want to model a situation using effective stress parameters with pore-water
pressure change, then you must do a fully coupled consolidation type analysis
where pore-water pressures are computed simultaneously using SEEP/W. An
uncoupled consolidation analysis does not compute pore-water pressure changes
due to changes in total stress, because the pore-water pressure changes are
computed independent of total stress changes in an external program such as
SEEP/W, SIGMA/W, VADOSE/W or QUAKE/W.
Page 60
SIGMA/W
Chapter 3: Fundamentals
When you specify the material properties as effective stress parameters, then
SIGMA/W will take the existing total stress, subtract the specified pore-water
pressure and arrive at an effective stress. This effective stress is then the value used
to obtain the soil stiffness in a non-linear soil model. The solver then computes the
change in total stress due to loading and it ADDS back on the original pore-water
pressures in order to report a new effective stress at the end of the load step. So, in
every case where effective stress parameters are specified and the analysis is not
fully coupled, the change in effective stress will always equal the change in total
stress.
In every case where effective stress parameters are specified and the analysis is not fully
coupled, the change in effective stress will always equal the change in total stress.
When you specify material properties as total stress parameters, you have the
option to specify A and B pore-water pressure parameters. If you specify valid A
and B pore-water pressure parameters, then SIGMA/W will compute a change in
pore-water pressure based on those parameters. Consider the equation for porewater pressure change using A and B parameters shown below. A full discussion
of this equation is provided in the Theory chapter.
+ 2 + 3
u = 1
+
3
( 1 2 ) + ( 2 3 ) + ( 3 1 )
2
where:
B, and
A.
From this equation, it is clear that any generated pore-water pressure is due to a
stress change consisting of two components; namely, u, which is a function of
changes in total stresses, and u, which is due to changes in deviatoric stress. It
can be quite common to neglect a change in pore-water pressure along shearing
planes so the A parameter can be set as zero.
Use of the B parameter requires a bit of thinking. Often, it is desired to have the
pore-water pressure response be equal to a change in total stress during loading.
This can be achieved by setting the B value to 1 in a fully saturated soil and then
Page 61
Chapter 3: Fundamentals
SIGMA/W
also setting Poissons ratio to a value of 0.49 (ideally 0.5, but for numerical
reasons it is fixed at 0.49 as a maximum value). With this Poissons ratio, the
average in all three direction will equal the applied and then u will also
equal .
The pore-pressure parameters A and B are defined in SIGMA/W using functions.
B is defined as a function of the pore-water pressure and A is defined as a function
of the deviatoric stress. The B function can be defined for both positive and
negative pore-water pressures, which makes it possible to analyze both saturated
and unsaturated soil conditions. Example functions of these parameters are shown
in Figure 3-12 and Figure 3-13 .
In a load-deformation analysis, materials with properties specified using effective
stress parameters and materials specified using total stress properties can be
intermixed. However, in a fully coupled consolidation analysis using SEEP/W
only materials with effective stress parameters are allowed.
Page 62
SIGMA/W
Chapter 3: Fundamentals
It is possible to specify a B function, but no A function. This means that your porewater pressure change will result purely from volumetric strain and not from any
deviatoric stress change as discussed above. No pore-water pressure changes will
be computed if both parameter A and parameter B are undefined in a total stress
analysis.
3.8
Incremental formulation
SIGMA/W is formulated for incremental analysis. For each time step, incremental
displacements are calculated for the incremental applied load. These incremental
values are then added to the values from the previous time step and the
accumulated values are reported in the output files. In equation form:
Page 63
Chapter 3: Fundamentals
SIGMA/W
can be added or subtracted such that at the end of the loading sequence, the ground
stresses reflect the desired conditions.
Consider for example that you want to set up some initial conditions such that the
total horizontal stress at a depth of 10 m is 200 kPa , while the total vertical stress
is only 100 kPa. Assume, for now, that the unit weight of the soil is 20 kN/m3. It is
not common in reality to have the horizontal stress be greater than the vertical
stress, but this example makes a good point of illustrating how some creative
thinking using incremental load steps can work to your advantage.
There are at least two ways to establish the desired in-situ stress conditions. In the
first approach:
1. Set the Ko value to be 2 and run an In-situ 1 type of analysis. This
would get the desired result in one step.
2. However, to illustrate a point, the same conditions could be
established by the following two step approach:
Step 2: Change the unit weight to a value of -10 kN/m3 and give
Poissons ratio a value of zero. Solve a load / deformation analysis
using the results of step 1 as the initial conditions for step 2. The result
of this load step is to subtract 100 kPa from the vertical stresses and 0
kPa from the horizontal stress.
In both cases, the end result is a horizontal stress of 200 kPa and a vertical stress of
100 kPa.
Obviously, if you had a choice, you would choose the one step approach to get the
desired result. The point of this example is that the incremental formulation allows
to you add and subtract loads, such that you end up with the ground stress
conditions you think exist in the field.
Body forces
The body force is the self weight of the soil and it must be included in an analysis
if actual stresses in the ground are desired, or if a non-linear soil model requires
that the actual stresses are known so that the correct soil stiffness values can be
obtained. The body force can be left out of a linear-elastic soil model analysis,
Page 64
SIGMA/W
Chapter 3: Fundamentals
however, you must realize that any stresses reported are only incremental stresses
due to the added load, and not the load plus the weight of the soil itself.
Full details of how the body loads are computed are given in the Theory chapter.
For now, it is only necessary to realize that the body load applied in the model is
simply a function of the unit weight specified and the size of the finite element. In
other words, a finite element shape one meter square with the assumed one meter
unit thickness would have one cubic meter volume. The body load applied at each
node in the element would then have a proportion of the unit weight multiplied by
the elemental volume.
Using the incremental approach, the unit body force is generally only applied when
an element is included for the first time during an analysis. However, if there is a
numerical advantage to apply the body load over several load steps, then it is
possible to set this as an option.
If the body forces were applied in an in-situ analysis, and then the results from the
in-situ analysis are used as the starting condition for a different analysis, you must
turn off the body forces in the new analysis. It does not know they were applied
previously, so it will apply them again in the first loading step if valid body force
data exists in the file. This point is often over-looked by users. Hopefully, if you
understand how the incremental loading formulation works, you will check what
you are asking the model to do.
3.9
As stated in the introduction to this chapter, SIGMA/W is a very powerful tool for
investigating the serviceability of engineered soil systems. It is not a tool for
predicting stability of these systems.
Serviceability deals with obtaining confidence that a design will function as
intended. Serviceability can tell you how a soil will move towards an unstable state
(e.g. going excessively plastic), but it does not tell you, or quantify for you, what
that state of instability is. Stability, on the other hand, deals with quantification of a
factor of safety for an engineered system. It deals with applied loads and resisting
forces. It says nothing about how a soil moves from one state to the other. A limit
equilibrium tool such as SLOPE/W should be used where factors of safety are the
primary concern.
In this chapter we have discussed some fundamental soil properties and how they
can be measured, computed and interpreted in the model. One of the more
Page 65
Chapter 3: Fundamentals
SIGMA/W
important concepts introduced was that the strength properties of a soil are a
measurement of the failure shear strength under different principal stresses. This
has significant implications in a numerical stress-deformation model, because it is
very likely that there are several failure planes throughout the geometry, and that
they are at various orientations to the principal stresses. Given the highly varied
combination, and add the fact that there is no way to quantify a soil strength once it
has failed, how can a quantification of stability be made?
In a limit equilibrium analysis, the potential failure plane is specified. With this
information, it is possible to determine what the shear strength is along that plane
given principle stresses acting at that point. The mobilized shear strength can be
divided by the resisting shear strength to determine a stability factor greater or
less than 1.0. This can not be done in stress-deformation analysis.
3.10
Modeling progression
A finite element analysis consists of two steps. The first step is to model the
problem, while the second step is to formulate and solve the associated finite
element equations. Modelling involves designing the mesh, defining the material
properties, choosing the appropriate constitutive soil model, and defining the
boundary conditions. SIGMA/W can formulate and solve the finite element
equations. The modelling, however, must be done by the user; that is, the user
must design an acceptable mesh, select the applicable soil properties, and control
the boundary conditions. SIGMA/W cannot make judgments; this is your
responsibility as the user.
Good modelling techniques require practice and experience. To assist you with this
part of the analysis, this chapter presents some general modelling guidelines. The
information presented is not an exhaustive statement on the "how-to" of modelling,
but instead provides suggestions on how to model various conditions and how to
outline conditions that may lead to difficulties.
One of the most important rules to follow in finite element modelling is to progress
from the simple to the complex. When you include all the possible complexities at
the start of an analysis, it often becomes difficult to interpret the results,
particularly when the results are unrealistic. Moving from the simple to the
complex makes it easier to pinpoint the reasons for unrealistic results. It is
therefore good practice to first define a simplified version of the problem and then
add complexity in stages.
Page 66
SIGMA/W
Chapter 3: Fundamentals
In finite element modelling, it is also important that the results obtained are of a
form similar to results obtained from simple hand calculations. It is easier to make
this judgment if you start with a simplified version of the problem.
For example, a good method of moving from a simple analysis to a complex
analysis is to start with a linear-elastic model. These results provide a reference
with which to compare more complex nonlinear results, and they are also a means
of checking that the boundary conditions have been applied correctly. Once you
are satisfied that this part of the analysis is correct, you can move onto the
nonlinear analysis with more confidence.
3.11
Units
Any set of units can be used in a SIGMA/W analysis. However, the units must be
used consistently throughout the analysis. For example, if the geometry is defined
in meters, then deformation is in meters.
Units must be selected for length, force and unit weight. The unit weight of water
is set when the units of length are selected. Table 3-1 shows examples of consistent
sets of units.
Table 3-1 Examples of consistent sets of units
Property
Units
Metric
Imperial
Geometry
meters
feet
pcf
F/L
kN/m
kN/m
pcf
Cohesion
F/L
kPa
pcf
Pressure
F/L2
kPa
psf
Force
kN
lbs.
kPa
psf
E (modulus)
F/L
F/L
The unit weight of water must be entered correctly for whichever system of units
you may have chosen. This unit weight is used in SIGMA/W to convert hydraulic
heads into pressures. It is also used to calculate a limiting confining stress in some
nonlinear constitutive models. For example, if the unit weight of water is kN/m3,
the minimum principal stress is limited to 1 kPa in the Elastic-Plastic Model when
the internal friction angle is greater than 0. If the unit weight is 62.4 pcf, this
Page 67
Chapter 3: Fundamentals
SIGMA/W
Page 68
SIGMA/W
Meshing
4.1
Introduction
Chapter 4: Meshing
Page 69
Chapter 4: Meshing
SIGMA/W
The purpose of this chapter is to introduce some of the basic concepts inherent in
meshing and outline some procedures which must be followed when developing a
mesh. An understanding of these fundamentals is vital to proper discretization.
Much of this chapter is devoted to describing the meshing systems and the features
and capabilities available in GeoStudio. In addition, there are also discussions on
the selection, behavior and use of various element types, sizes, shapes and patterns.
A summary of practical guidelines for good meshing practice are also outlined.
4.2
Element fundamentals
Element nodes
One of the main features of a finite element is the nodes. Nodes exist at the corners
of the elements or along the edges of the elements. Figure 4-1 and Figure 4-2 show
the nodes, represented as black dots.
The nodes are required and used for the following purposes:
The nodes are used to connect or join all the elements within a
domain. All elements with a common node are connected at that node.
It is the common nodes between elements that ensure compatibility,
which is discussed in further detail below.
All finite element equations are formed at the nodes. All elements common to a
single node contribute to the characteristics and coefficients that exist in the
equation at that node, but it is the equation at the node that is used to compute the
primary unknown at that node. In other words, the seepage equation is developed
for each node and the material properties which are used within the equations are
contributed from the surrounding elements.
There can be multiple finite element equations developed at each node depending
on the degrees of freedom. In a SIGMA/W analysis there are up to three degrees of
freedom at each node, and the number of finite element equations to be solved is
Page 70
SIGMA/W
Chapter 4: Meshing
equal to the number of nodes used to define the mesh multiplied by the number of
degrees of freedom at each node. In a 2D stress-deformation analysis, there are two
degrees of freedom at each node displacement x and displacement y.
Consequently, the number of equations for the whole domain is equal to two times
the number of nodes. In a coupled consolidation analysis there are three degrees of
freedom at each node displacement x, displacement y and pore-water pressure.
For a coupled consolidation analysis the total number of equations required to
solve the problem is three times the number of nodes. For a stress-deformation
analysis with structural beam elements, there is an additional rotational degree of
freedom at beam nodes only.
Since the number of finite element equations is related to the number of nodes, the
number of nodes in a problem is one of the main factors in the computing time
required to solve for the primary unknowns.
Field variable distribution
In a finite element formulation it is necessary to adopt a model describing the
distribution of the primary variable within the element (e.g., displacement). The
distribution could be linear or curved.
For a linear distribution of the primary unknown, nodes are required only at the
element corners. The two nodes (points) along an edge are sufficient to form a
linear equation. Figure 4-1 illustrates this situation. Elements with nodes existing
at the corners are referred to as first-order elements.
The derivative of the primary unknown with respect to distance is the gradient. For
a linear distribution the gradient is consequently a constant. In the context of a
stress-deformation formulation the primary unknown is the displacement. The
derivative of displacement with respect to distance is the gradient (or strain) and
the gradient is therefore constant within a first order element.
Page 71
Chapter 4: Meshing
SIGMA/W
With three nodes defined along an edge, we can write a quadratic equation
describing the distribution of the primary unknown within the element.
Consequently the distribution of the primary unknown can be curved as shown in
Figure 4-2. The derivative of the quadratic displacement distribution results in a
linear gradient distribution. Elements with three or more nodes along an edge are
referred to as higher order elements. More specifically, an element with three
nodes along an edge is known as a second-order element.
Higher order elements are more suited to problems where the primary unknowns
are vectors as in a stress-deformation analysis (deformation x and y). When the
primary unknown is a scalar, there is often little to be gained by using higher-order
elements. Smaller first-order elements can be as effective as larger higher-order
elements. This is discussed in more detail in the meshing guidelines at the end of
this chapter.
Element and mesh compatibility
Element and mesh compatibility are fundamental to proper meshing. Elements
must have common nodes in order to be considered connected, and the distribution
of the primary unknown along an element edge must be the same for an edge
common to two elements.
Consider the illustration in Figure 4-3. Element numbers are shown in the middle
of the element and node numbers are presented beside the nodes. Even though
elements 4, 5 and 6 appear to be connected to elements 7, 8 and 9, they are actually
not connected. Physically, the elements would behave the same as the two element
groups shown with a physical separation on the right side of Figure 4-3. Common
nodes are required to connect the elements as shown in Figure 4-4. Node 11, for
example, is common to Elements 5, 6, 8 and 9.
Page 72
SIGMA/W
Chapter 4: Meshing
Mixing elements of a different order can also create incompatibility. Figure 4-5
shows 4-noded quadrilateral elements connected to 8-noded elements. Elements 1
and 2 are 8-noded elements while Elements 3 to 10 are 4-noded first-order
elements. The field variable distribution in Element 1 along edge 9 to 11 could be
curved. In Elements 3 and 4 the field variable distribution between 9 and 10 and
between 10 and 11 will be linear. This means the field variable distributions
between Elements 1 and 2 are incompatible with the field variable distributions in
Elements 3 to 6.
The meshing algorithms in GeoStudio ensure element compatibility within
regions. A special integer-based algorithm is also included to check the
compatibility between regions. This algorithm ensures that common edges between
regions have the same number of elements and nodes. Even though the software is
very powerful and seeks to ensure mesh compatibility, the user nonetheless needs
to be careful about creating adjoining regions. The illustration in Figure 4-3 can
also potentially exist at the region level. At the region level, region points need to
be common to adjoining regions to ensure compatibility.
16
15
20
14
13
19
12
11
10
9
15
23
18
31
39
22
26
17
30
43
38
23
42
19
13
21
25
16
29
46
37
22
41
33
Page 73
47
34
10
17
24
20
14
48
35
11
18
44
36
21
27
7
1
32
12
8
2
40
28
9
3
24
45
Chapter 4: Meshing
SIGMA/W
8
3
12
6
7
2
11
20
12
15
8
10
4
16
19
11
14
7
18
10
13
17
13
12
11
10
9
18
6
5
4
3
17
16
15
14
23
10
9
8
7
22
21
20
19
The integer programming algorithm in GeoStudio seeks to ensure that the same
number of element divisions exist between points along a region edge. The number
of element divisions are automatically adjusted in each region until this condition
is satisfied. It is for this reason that you will often notice that the number of
divisions along a region edge is higher than what was specified. The algorithm
computes the number of divisions required to achieve region compatibility.
Numerical integration
In a finite element formulation there are many integrals to be determined, as shown
in the Theory chapter. For example, the integral to form the element characteristic
matrix is:
Page 74
SIGMA/W
Chapter 4: Meshing
[ B ] [C ][ B ] dv
t
For simple element shapes like 3-noded or 4-noded brick (rectangular) elements, it
is possible to develop closed-formed solutions to obtain the integrals, but for
higher-order and more complex shapes it is necessary to use numerical integration.
GeoStudio uses the Gauss quadrature scheme. Basically, this scheme involves
sampling the element characteristics at specific points known as Gauss points and
then adding up the sampled information. Specific details of the numerical
integration in GeoStudio are presented in the Theory Chapter.
Generally, it is not necessary for most users to have a comprehensive
understanding of the Gauss integration method, but it is necessary to understand
some of the fundamentals since there are several options in the software related to
this issue and some results are presented at the Gauss sampling points.
The following table shows the options available. Use of the defaults is
recommended except for users who are intimately familiar with numerical
integration and understand the significance of the various options. The integration
point options are part of the meshing operations in GeoStudio. In the table you will
see two comments related to special cases when using the elastic-plastic soil
model. The reasons for this are discussed in the meshing guidelines section later in
this chapter.
Element Type
Integration Points
Comments
4-noded quadrilateral
Default
8-noded quadrilateral
4 or 9
4 is the default
3-noded triangle
1 or 3
3 is the default
6-noded triangle
Default
Some finite element results are computed at the Gauss sampling points. GeoStudio
presents the results for a Gauss region, but the associated data is actually computed
at the exact Gauss integration sampling point. Even though a Gauss region is
displayed, the data is not necessarily constant within the region.
With the View Element Information command, you can click inside an element
and the nearest Gauss region is displayed together with all the associated
information. The number of Gauss regions within an element is equal to the
number of Gauss integration points used in the analysis.
Page 75
Chapter 4: Meshing
SIGMA/W
Page 76
SIGMA/W
Chapter 4: Meshing
Ideal
G ood
Acceptable
Poor
Unacceptable
Figure 4-6 Element shapes and relative performance
The influence of element shapes is addressed both in the discussion below and a
later discussion on structured and unstructured meshing.
4.3
Regions
GeoStudio uses the concept of regions to define the geometry of a problem and to
facilitate the discretization of the problem. The attraction of regions is that they
replicate what we intuitively do as engineers and scientists to illustrate concepts
and draw components of a system. To draw a stratigraphic section, for example,
we intuitively draw the different soil types as regions.
The utilization of regions offers all the advantages of dividing a large domain into
smaller pieces, working and analyzing the smaller pieces, and then connecting the
smaller pieces together to obtain the behavior of the whole domain, exactly like the
concept of finite elements. Generally, all physical systems have to be broken down
Page 77
Chapter 4: Meshing
SIGMA/W
into pieces to create, manage and control the whole body. A vehicle is a good
example. It is made up of numerous pieces joined together to make a useful means
of transportation.
A collection of highly adaptive individual pieces that can be joined together makes
it possible to describe and define almost any complex domain or physical system.
Such an approach is more powerful and can be applied to a wider range of
problems than any system that attempts to describe the whole domain as a single
object.
Further evidence of the advantages of the region approach is the acceptance in
industry of object-oriented technology, which has been widely applied in software
design and development.
In GeoStudio, regions can be thought of as super elements even though they do not
behave like conventional finite elements. They are the host or parent of the smaller
conventional elements. From a usability point of view it is obviously much easier
to draw and describe a few big elements or regions than to individually draw and
describe each finite element.
Region types
Regions may be simple straight-side shapes like quadrilaterals or triangles or a
non-regular, multi-sided polygon. Figure 4-7 illustrates a domain constructed using
one quadrilateral and two triangular regions. Figure 4-8 shows a single multi-sided
polygonal region defined using 10 points.
Region points
Regions have points like elements have nodes. The points make it possible to join
regions and ensure there is continuity between the regions. Points common to
different regions means the regions are joined at these points.
Page 78
SIGMA/W
Chapter 4: Meshing
The points can be selected and moved to modify the shape and position of regions,
which provides for great flexibility in making adjustments and alterations to a
problem definition.
15
16
14
17
13
11
12
Page 79
Chapter 4: Meshing
SIGMA/W
Points are also required in order to join regions of different sizes and to control the
meshing for specific purposes. Figure 4-9 shows a homogeneous soil region with a
concrete footing region. The foundation region is made up of Points 11, 12, 13, 17
and 14. The footing region is made up of Points 14, 17, 16, and 15. Points 14 and
17 are common to both regions and therefore the two regions are properly joined
and connected along this edge. In addition, Point 17 ensures that an element node
will be created and will exist at the edge of the footing, which is required for
proper meshing.
Region properties
When a region is defined, it is restricted to:
However, after a region is generated, any individual element within that region can
be altered manually using the Draw Element Property command. If the region is
regenerated intentionally or because an adjacent region required a regeneration,
then any manually changed element properties will be set back to the default
region values.
4.4
Structured mesh
Figure 4-10 presents what is known as a structured mesh. The elements are ordered
in a consistent pattern. This illustration has five regions. The same mesh could
have been created with two regions, but the same number of points would have
been required. All rectangular regions could be one region, but intermediate points
(4, 5 and 8) would be required to achieve the same mesh.
Unstructured mesh
The diagram in Figure 4-11 shows the same section as in Figure 4-10, but this time
with an unstructured mesh. In this case the mesh is automatically created using
Delaunay triangulation techniques. Only one region is required, but now the only
Page 80
SIGMA/W
Chapter 4: Meshing
user control is the number of divisions along edges of the region. The shape and
positions of the elements within the region is controlled by the algorithm.
Changing the number of division along any region edge can result in a completely
different mesh.
One of the great attractions of unstructured meshing is that almost any odd-shaped
region can be meshed. This meshing simplicity, however, has some numerical and
interpretation consequences as discussed in more detail in the Structured versus
Unstructured Section below. Automatic mesh generation with unstructured
elements is NOT the answer to all meshing problems!
9
10
25
24
21
23
22
Triangular regions
GeoStudio has a special structured pattern for triangular regions. The next figure
(Figure 4-12) shows a typical triangular region with a structured mesh. The
Page 81
Chapter 4: Meshing
SIGMA/W
elements are a mixture of squares, rectangles, trapezoids and triangles. The use of
this pattern is fairly general, but it does have some limitations and restrictions.
2
Figure 4-12 Triangular region
It is useful to think of the region as having three sides: a short side, an intermediate
length side and a long side. The algorithm attempts to sort the edges so that the
sides go from the shortest to the longest in a counter-clockwise direction. In this
example, the shortest side is 3-1, the intermediate 1-2 and the longest 2-3.
The meshing algorithm works best when the number of divisions is controlled on
the shortest and intermediate sides. To retain the even pattern shown in Figure
4-12, the number of divisions should be defined on the shortest side first and then
on the intermediate side. The number of divisions on the intermediate side can be
an even multiple of the number on the shortest side. In the above example, the
shortest side has 5 divisions and the intermediate side can have 10, or 2 times that
of the shortest side. The algorithm works best and gives the best structured mesh if
the numbers of divisions on the longest side are left undefined allowing the
algorithm to compute the appropriate number of divisions.
If a triangular region is mixed in with other more general regions, GeoStudio will
attempt to ensure mesh compatibility. Sometimes, however, it may not be possible
to adhere to the requirements for generating a structured mesh in a triangular
region and then GeoStudio will substitute an unstructured mesh.
Generally, the triangular structured pattern functions best when it is used with
other structured mesh regions, although this is not a strict requirement.
Transfinite meshing
Much of the structured meshing in GeoStudio is based on transfinite mapping
techniques. Without going into detail, this mapping technique creates a mesh that
Page 82
SIGMA/W
Chapter 4: Meshing
The transfinite mapping technique provides for great flexibility in that the sides
can have almost any curved or jagged line-segmented shape. However, the system
needs what in GeoStudio are referred to as corners.
Quadrilateral regions must have four corners and triangular regions must have
three corners. The mesh in Figure 4-14 has four corners defined, even though there
are 10 points used to define the region. In this case the corners are at the natural
corners of the rectangle, but this is not always necessary as we will see a little later.
The second essential requirement for transfinite mapping is that the number of
element edges between corners must be the same on opposing sides of the region.
In this example, the number of element edges between Points 13 and 16 must be
the same as between Points 11 and 17, and the number of element edges between
Points 11 and 13 must be equal to the number of edges between Points 16 and 17.
Page 83
Chapter 4: Meshing
11
SIGMA/W
19
18
17
12
20
13
14
15
16
In Figure 4-14 the number of divisions along the bottom horizontal and top curved
sides have the same number of divisions and the two vertical sides have the same
number of divisions.
Figure 4-15 further illustrates transfinite mapping. In this case the corners are
Points 1, 4, 5 and 8. Edge 1-8 is opposite edge 4-5 and must have the same number
of divisions. Edge 1-2-3-4 is opposite edge 5-6-7-8. The total number of divisions
along 1-2-3-4 must be the same as along 5-6-7-8.
GeoStudio makes a best guess at the corner locations when you draw the regions,
but the algorithm does not always identify the user-intended corner positions. One
of the features of regions is that the corner positions can be moved using the mouse
as necessary to achieve the intended mesh configuration.
1
Page 84
SIGMA/W
Chapter 4: Meshing
Figure 4-17 Meshing opening with ring of elements around the opening
Page 85
Chapter 4: Meshing
SIGMA/W
Regions with openings have various parameters to control the opening and mesh
characteristics. Descriptions of all the parameters are in the software online help.
The important point here is that GeoStudio has special regions to assist with
meshing special conditions.
Surface regions
At the ground surface conditions change in response to the climate and climatic
conditions can change dramatically over short periods of time. For example, the
ground maybe highly desiccated near the surface on a hot day before a
thunderstorm. In a short period of time, the soil changes from being very dry to
being saturated. Another example may be penetration of frost from the ground
surface. To numerically deal with rapid and dramatic boundary changes it is
necessary to have fine discretization near the ground surface. GeoStudio has a
special procedure for constructing a surface mesh. Figure 4-18 illustrates a surface
mesh placed over the surface of a larger region. The surface mesh capability is also
invaluable for discretizing features such as engineered soil covers over waste
material, which may consist of several relatively thin layers of soil.
The ability to construct a surface mesh is available in VADOSE/W, SEEP/W and
TEMP/W. In SEEP/W the surface mesh is used to tell the solver that it should
track seepage face flows and infiltration events for any unit flux boundary
condition. As a result, water that does not immediately infiltrate the ground is not
considered lost from the analysis, but is allowed to pond and build up a positive
pressure head in any user-defined low points along the surface. In VADOSE/W
and TEMP/W, the surface region is use for application of climate boundary
conditions. The other GeoStudio modules cannot be used to construct a surface
mesh, but once the surface mesh has been created it will exist in all the other
modules. Consequently, if a surface mesh has been created for a particular
analysis, the surface mesh will also be part of all other analyses, since GeoStudio
uses only one mesh definition within a single data file.
Once the main soil profile has been meshed, a special Draw Surface Region
command can be used to build up a single or multi layer region along all or part of
a ground surface. Parameters such as the soil type and individual layer geometry
are defined and a quadrilateral element mesh with vertically oriented nodes is
automatically built on top of the existing ground region. The structure of the mesh
will ensure optimum numerical stability during the solution.
Quadrilateral elements are much better for modeling ground surface processes
because the primary unknown gradients are usually steeper in a direction
Page 86
SIGMA/W
Chapter 4: Meshing
Surface mesh regions have special viewing options. Consider the two meshes
illustrated in Figure 4-19. The left diagram shows a surface region mesh without
all the cluttering details as illustrated on the right. When many thin elements are
located in a close proximity to each other, they can appear unclear when viewed
from a far away scale. By optionally turning off the surface mesh details a clearer
image of the structure of the near surface soil layers can be viewed.
Figure 4-19 Surface region mesh with details off (left) and on (right)
Page 87
Chapter 4: Meshing
SIGMA/W
Figure 4-20 is another illustration of this optional viewing concept. The left
diagram in the figure shows the detailed mesh and soil layers across the 0.75m
thick surface region and the right diagram leaves the details out, but still shows the
layer colors. A couple of additional key points can be made in regards to the
figure. Notice that bottom two elements of the left diagram are the same soil type
as the main underlying soil. This is a good mesh design strategy that being to
have the bottom most layer of the surface mesh be made of the same soil as the
existing ground. Consider if the bottom layer of the surface soil was VERY
different from the underlying soil. If a fine spaced mesh were placed directly on
top of the different underlying soil then the numerical integration of material
properties at the common mesh node between the two soils would be less accurate
because of the influence of the large element area from the material below the
common nodal point. By having the bottom layer of the surface region the same as
the underlying soil, the element shapes are very similar in size and aspect at the
common nodal point between the two very contrasting soils.
The second point to node from Figure 4-20 is that in the right diagram the nodes
that are located at the interface between two soils are still viewable even though
the main mesh details are not. This is intentional so that you can easily see and
graph data at nodes that are used for automatic tracking of interlayer fluxes in the
VADOSE/W model.
Boundary flux modeling with rainfall infiltration, runoff, snow melt etc. can be
very numerically demanding from a convergence perspective. Potential problems
can be made worse if the shape of the surface mesh is not realistic. Consider the
two meshes illustrated in Figure 4-21 and Figure 4-22. In the first figure, the
ground profile has rounded corners which are much more natural and much more
numerically friendly. In the second figure, changes in slope angle are represented
by a sharp break. This sharp break is not only un-natural, but the shape of the
individual elements right at the transition points creates numerical problems if
Page 88
SIGMA/W
Chapter 4: Meshing
there are large changes in boundary condition type at different nodes within the
same element. This would be the case when the corner node at the bottom of the
slope becomes a seepage face point while the next node up slope is still an
infiltration node. Basically, it is better to build the mesh to look somewhat natural.
g
Figure 4-21 GeoStudio region mesh showing rounded surface slope breaks
Page 89
Chapter 4: Meshing
SIGMA/W
and crest of the slope. Also notice that three region points are used on the bottom
of the mesh beneath the toe and crest location. This is a useful tip to remember.
When you want to have more control over the trans-finite element mesh you
should add region points on opposite sides of the mesh from where you need the
detail. As a final note, adding region points can be done at any time even after
the surface mesh is created. When the region beneath a surface mesh is changed,
the surface mesh above it will be automatically regenerated to ensure mesh
compatibility with the region below.
Figure 4-23 Region mesh with region corner points viewed and surface
details not viewed
Joining regions
Compatibility must be maintained between regions to ensure the regions are
connected. Regions must be joined at the region Points and Points must be
common to adjoining regions for the regions to be properly connected. GeoStudio
has a number of features to assist in achieving region compatibility.
The following are some of the main characteristics:
Page 90
SIGMA/W
Chapter 4: Meshing
Consider the diagram in Figure 4-24. Region 1 is drawn first and Region 2 can be
drawn by clicking on Points 7, 3, 9 and 8. Points 4, 5 and 6 are automatically
added to Region 2.
9
2
7
3
5
6
4
1
7
5
6
4
1
1
Page 91
Chapter 4: Meshing
4.5
SIGMA/W
Page 92
SIGMA/W
Chapter 4: Meshing
Page 93
Chapter 4: Meshing
SIGMA/W
along a vertical line. In this case it is necessary to find all the elements intersected
by a cut-line and then to project the results onto the cut-line from the nearest
nodes. This can be done, but this means that interpolated values are displayed, as
opposed to looking at the exact values computed at the nodes.
Vertical profiles
It is easier to spot check the computed results against hand-calculations when using
quadrilateral elements in a structured mesh than it is with triangular elements at
some random orientation. Take for example the case of determining in-situ
stresses. In a structured mesh it is fairly straightforward to spot check the stress in
an element. The vertical stress is directly related to the vertical distance from the
surface and since the element is aligned with the x and y directions, the hand
calculation is trivial. It is not as straightforward for triangular elements, which are
all at different orientations.
Unstructured meshing techniques are clearly better for highly irregular geometric
shapes. Care is required in geotechnical numerical modeling to not make the
geometry unnecessarily complicated as discussed elsewhere in this chapter and in
the previous chapters. Just because it is easy to mesh complicated shapes is not
sufficient reason to make the geometry unnecessarily complicated.
Another advantageous feature of unstructured meshing is that the method has no
restrictions on the number of element divisions along region edges. In transfinite
mapping, as discussed above, it is necessary to define corners and opposing sides
between the corners must have the same number of divisions. Unstructured
Page 94
SIGMA/W
Chapter 4: Meshing
meshing has no such restrictions. This makes it easy to transition from a fine mesh
to a course mesh.
Unstructured meshing also opens the door to adaptive meshing in the future. Mesh
refinement and adaptive meshing is possible with a structured mesh pattern, but it
is more restrictive than with triangular unstructured meshing. Adaptive meshing
during a solve process does not necessarily address all numerical problems related
to finite element analysis. For example, adaptive meshing can create many small
elements in areas of high gradients, but solving the equations in these areas
requires smaller time steps in a transient analysis. This is fine in principle, except
that the time steps are common to the entire geometry and can therefore become
too small for other regions that have larger element sizes.
Generally, some extra effort and care up front in creating a mesh pays dividends
when it comes to solving, interpreting, visualizing, discussing and presenting the
results. Often there is a trade off in effort between defining the problem and
dealing with the results. Extra effort is either required at the start or at the end of
the analysis. Our experience is that it is better to spend the extra effort defining the
problem and in the mesh creation, than struggle to interpret and justify results at
the end of the analysis.
The ability to construct either structured or unstructured meshes in GeoStudio
offers the best of both approaches. Users can select and use whatever they wish
based on personal preference or as dictated by problem requirements.
4.6
Page 95
Chapter 4: Meshing
SIGMA/W
In a simulation of consolidation, the time step size for the first time step needs to
be sufficiently large so that the element next to the drainage face consolidates by at
least 50 percent. Achieving this is related to the element size; the larger the
element the greater the required initial time step. If the time step size is too small,
the computed pore-water pressures may be unrealistic.
The important point in this section on meshing is to realize that meshing, more
particularly element sizes, comes into play in a transient analysis. Rules and
guidelines for selecting appropriate time stepping are discussed elsewhere with
reference to particular types of analysis.
4.7
Interface elements
In a finite element formulation, generally all the elements are connected at the
nodes and remain connected at the nodes. Sometimes it is desirable, however, to
allow elements to move relative to each other, such as along a joint between two
rock blocks or along the contact surface between soil and structure. Another
example might be the potential slippage that may exist between a geo-synthetic
and surrounding soil. In a seepage analysis, a wick drain will open up the flow
capacity in the axial direction of the drain. Flow can in essence follow the contact
between elements. In a thermal analysis, an interface element could be used to
simulate a thin layer of insulation for the case where actually drawing the thin
layer is hard to do because of the thickness relative to the scale of the entire model.
Features and behavior like this can be simulated with what is known as interface or
contact elements.
In the current version of GeoStudio, only SIGMA/W has a form of an interface
element which can be used to simulate the slippage between neighboring elements.
They are called slip elements. The slip elements are fairly primitive and in need
of enhancement.
A more general interface formulation with a wider range of applications has been
formulated for GeoStudio and will be available in a future release.
In GeoStudio it is important to recognize that if the interface element is present in
one analysis, it will be present in all the other products within a particular data file.
Page 96
SIGMA/W
4.8
Chapter 4: Meshing
Infinite elements
X1
Pole
When infinite elements are used, the x-coordinates of Nodes 6 and 8 are modified
so that the distance between Node 1 and 8 is the same as between Node 1 and the
pole. That is, the x-coordinates of Nodes 6 and 8 are modified so that X2 in Figure
4-29 is equal to X1. Next the decay function is created using the x-coordinate of
the pole, the x-coordinate of Node 1 (or 2 or 5) and the modified x-coordinate of
Page 97
Chapter 4: Meshing
SIGMA/W
Node 8 (or 6). Nodes 3, 4 and 7 are considered to be at infinity and are not used in
creating the shape functions.
Infinite elements need to be 8-noded quadrilaterals. The secondary nodes (6 and 8)
are required to form the decay function. Triangular elements cannot be used for
this purpose.
Infinite elements are a convenient way of extending the far field of a problem, but
they need to be used with some caution. The infinite elements work best if the
specified boundary conditions at the far field are the primary unknowns in the
analysis; that is, head, displacement, temperature, and so forth. Specifying
gradients (for example, unit rates of flow or stress) along the element edge
representing the far field will work, but the effect is less certain than specifying
primary field variables.
Boundary conditions should not under any conditions be specified on the element
edges that are parallel to the infinity direction. In the element in Figure 4-29,
boundary conditions should not be applied along edge 1-8-4 and edge 2-6-3. You
cannot apply a boundary condition that is a function of edge length when the edge
length is considered to be infinite.
The distance between the pole and the inner edge of the infinite element obviously
influences the coefficients of the decay function and the modification of the
secondary node x-coordinates. The pole position consequently does have some
influence on the behavior of the elements. Fortunately, the results are not all that
sensitive to the pole position. Generally, the objective is to make the distance
between the pole and the inner edge of the element as large as conveniently
possible, but not beyond the extents of the overall mesh.
We recommend that you always complete an analysis without infinite elements
first. Then after you have obtained an acceptable solution, save the problem to a
new file and then add the infinite elements to refine your analysis. Having
solutions with and without infinite elements is a useful way of understanding the
effect of extending the far field boundary and understanding the computed
response of these elements.
Infinite elements in SIGMA/W
Infinite elements can be drawn in both the x-direction and y-direction in
SIGMA/W so that the mesh will be compatible with other GeoStudio models.
However, caution should be used when specifying infinite elements in the y-
Page 98
SIGMA/W
Chapter 4: Meshing
4.9
All elements should be visible to the naked eye when the mesh is
printed at a zoom factor of 100 % and when the horizontal and vertical
scales are the same. The exception to this guideline is a surface mesh.
Page 99
Chapter 4: Meshing
SIGMA/W
Page 100
SIGMA/W
Chapter 4: Meshing
In summary, higher order elements are essential for the all nonlinear soil models
(except elastic-plastic), particularly in regions where the stress and strain changes
are high. In regions where the stress and strain changes are small, or the material is
assigned a linear-elastic model, the ordinary 4-noded and 3-noded elements may be
adequate.
Number of elements
Based on many years of responding to GEO-SLOPE user support questions, most
users start with a mesh that is too complex, containing too many elements for the
objective of the analysis. The thinking when users first start doing finite element
analyses seems to be the more elements, the better; that a large number of elements
will somehow improve the accuracy of the solution. This is not necessarily true. If
the mesh is too large, the time required to obtain a solution can become
unattainable. Sometimes it also becomes very difficult to interpret the results,
particularly if the solutions appear to be unreasonable. The effort required to
determine the reason for an unreasonable solution increases dramatically with
mesh size.
We highly recommend that you try and create a mesh with less than 1000
elements, particularly at the start of an analysis. Our experience is that most
geotechnical problems can be modeled with 1000 elements or less. Obviously there
are exceptions, but this number is a good goal to strive for. Later, once you have a
good first understanding of the critical mechanisms in your problem, you can
increase the mesh density to refine the analysis.
Effect of drawing scale
Another good guideline is that all elements should be visible to the naked eye
when the mesh is printed or viewed at a 100% zoom factor. Groups of elements
that appear as a solid or nearly solid black smudge on the drawing are too small.
This means a suitable element size is related to the drawing scale. A drawing at a
scale of 1:100 can have much smaller elements than a drawing at a scale of 1:2000.
In other words, if it is necessary to zoom in on an area of the drawing to
distinguish the elements, the elements may be unnecessarily small.
All elements should be readily distinguishable when a drawing is viewed when the
vertical scale is equal to the horizontal scale. It is possible to draw a nice looking
mesh at a vertical exaggerated scale, but when viewed at a true vertical scale the
mesh appears as a wide black line. This effect is illustrated in Figure 4-30. The top
part of the figure shows a nice mesh at 10V:100H, a 10 times vertical
Page 101
Chapter 4: Meshing
SIGMA/W
Figure 4-30 Mesh at an exaggerated scale (upper) and at a true scale (lower)
Mesh purpose
The How To modeling chapter notes that in good numerical modeling practice it
is important to form a mental imagine of what the solution may possibly look like
and to clearly define the purpose of the model before trying to create a model.
Meshing is closely tied to this guideline. The mesh should be designed to answer
specific questions. Trying to include all possible details in a mesh makes meshing
unnecessarily time consuming and can sometimes make it difficult to interpret the
results.
Page 102
SIGMA/W
Chapter 4: Meshing
Let us assume that we are interested in estimating the seepage though the clay core
of a zoned dam with rock shells. Figure 4-31 shows a typical case. The rock shells
are considered to be many orders of magnitude more permeable than the core. In
addition, the granular drain filter layers between the clay and the rock are clean
and can easily handle any seepage though the core without impeding the drainage.
In other words, the granular filter layers and rock shells make no contribution to
dissipating the hydraulic head on the upstream side of the core. If this
consideration is true, then there is nothing to be gained by including the highly
permeable materials in the analysis. A mesh such as in Figure 4-31 is adequate to
analyze the seepage though the core.
1
Figure 4-32 shows the total head contours (equipotential lines) in the core. From
this the seepage quantities through the core can be computed.
Sometimes a mesh may be required to include the shells in the analysis for other
reasons, such as a stress-deformation analysis. In such a case, the mesh can exist,
but does not need to be included in the analysis. This is accomplished using null
elements as shown in Figure 4-33. Elements in GeoStudio can be null (not active)
by leaving a key material property undefined. In SEEP/W the elements are null if
there is no specified conductivity function for the material. In the example in
Figure 4-33 the rock shells have no conductivity function assigned to the material.
Page 103
Chapter 4: Meshing
SIGMA/W
One of the attractions inherent to numerical modeling is that the geometry and
finite element mesh do not necessarily have to conform strictly to the physical
conditions. As in Figure 4-31, the core can be analyzed in isolation. This would not
be possible in physical modeling. The dam with a toe drain in Figure 4-34 is
another good example. The toe drain does not have to be included in the numerical
analysis. This, of course, would not be possible in a physical model.
8
10
11
Simplified geometry
A numerical model needs to be a simplified abstraction of the actual field
conditions. This is particularly true when it comes to the geometry. Including all
surface irregularities is unnecessary in most situations. Geometric irregularities can
cause numerical irregularities in the results, which distract from the main overall
solution. The main message can be lost in the numerical noise.
Simplifying the geometry as much as possible is particularly important at the start
of an analysis. Later, once the main processes involved are clear, the geometry can
be altered to determine if the geometric details are important to the main
conclusions.
The situation is different if the main objective of the analysis is to study the effects
of surface irregularities. Then the irregularities of course need to be included. So,
Page 104
SIGMA/W
Chapter 4: Meshing
once again, the degree of geometric complexity depends on the objectives of the
analysis.
Also, the level of geometric detail that needs to be included in the problem must be
evaluated in light of the certainty with which other factors such as the boundary
conditions and material properties are known. There is little to be gained by
defining a very detailed geometry if the material properties are just a rough
estimate. A simplified geometry is more than adequate if the material properties
are rough estimates. There needs to be a balance in complexity between all the
aspects of a finite element analysis, including the geometry.
Over-complicating the geometry is a tendency when users first get into numerical
modeling. Then, as modelers gain more experience, they tend to use more simple
geometries. This obviously comes from understanding how the mesh can influence
the results and what level of complexity is required. The situation should be the
reverse. It is the modelers with limited experience who should use simplified
geometries.
The main message to remember when starting to model is to keep the problem as
simple as possible until the main engineering issues are well understood.
Page 105
Chapter 4: Meshing
SIGMA/W
Page 106
SIGMA/W
Material Properties
Page 107
5.1
SIGMA/W
Linear-elastic
Anisotropic Linear-elastic
Page 108
SIGMA/W
Strain-Softening
Page 109
SIGMA/W
5.2
Linear-elastic model
The simplest SIGMA/W soil model is the linear elastic model for which stresses
are directly proportional to the strains. The proportionality constants are Young's
Modulus, E, and Poisson's Ratio, . The stresses and strain are related by the
equation:
Page 110
SIGMA/W
x
E
y
=
z (1 + )(1 2 )
xy
0
0
0
0
x
0
y
0
z
1 2
xy
2
For computational purposes, can never be 0.5. Even values greater than 0.49 can
cause numerical problems. Consequently, SIGMA/W limits the maximum value
for Poissons ratio, , to 0.49.
Table 5-1 Linear-elastic properties
Edit Box Label
E Modulus
Property
Young's Modulus
Poissons Ratio
constant value
Cohesion
constant value
Friction angle
degrees
Data for the linear-elastic model includes the cohesion and friction angle. This
information is not used in the solution, but it is used in the Contour program to
help illustrate regions of the soil where the computed stresses have exceeded the
yield strength. Recall that for a linear-elastic model, there is no yield value defined
and computed strains may be very unrealistic. Using cohesion and friction angle
along with the Mohr-Coulomb failure criterion will enable the computed shear
stress to be compared visually with the theoretical yield stresses.
Page 111
5.3
SIGMA/W
Natural ground deposits are often stratified and inclined. Therefore, it is desirable
to consider the possibility of having different stiffness values in two orthogonal
directions. Consider the case illustrated in Figure 5-1. The soil strata are crossanisotropic in the local orthogonal directions x and y with x, making an angle
with the global x-axis. The sign convention for can be described as follows:
when counter-clockwise from the x-axis, is positive. The anisotropic elastic
parameters in the local directions are defined by the following parameters:
The parameter vxy ' is the Poissons Ratio of horizontal (x) strain to vertical (y)
strain caused by a stress change in the y-direction. These parameters must satisfy
the following restrictions (Pickering, 1970):
and,
1 x > 2 ( Ex E y ) yx
The constitutive matrix [C] for anisotropic conditions is symmetric and can be
defined as (Britto and Gunn (1987):
c11 c12
c
c
[C ] = c0 c21 c22
31
32
0
0
c13
c23
c33
0
0
0
0
c44
where:
Page 112
SIGMA/W
c0 =
Ey
(1 + x ) (1 x 2 ( Ex
E y ) yx2
c11 = ( Ex E y ) 1 ( Ex E y ) 2yx
c12 = ( Ex E y ) yx (1 + x )
c13 = ( Ex E y ) x + ( Ex E y ) 2yx
c21 = c12
c22 = (1 2x )
c23 = c12
c31 = c13
c32 = c23
c33 = c11
c44 = Gxy c0
Page 113
SIGMA/W
When the local and global axes do not coincide (when is non-zero), it is
necessary to transform the [C] matrix from the local x-y coordinate system to the
global x-y coordinate system. The transformation of [C] to [C] is given by the
following equation (Zienkiewicz and Taylor, 1989):
[C ] = [T ][C ][T ]
where:
cos 2
sin 2
=
T
[ ]
0
sin cos
sin 2
cos
2
0
sin cos
2sin cos
0 2sin cos
1
0
0 cos 2 sin 2
0
Property
E Modulus (1)
P. Ratio (1)
E Modulus (2)
P. Ratio (2)
G Modulus (2)
Angle
5.4
Page 114
SIGMA/W
Page 115
SIGMA/W
Initial modulus
When a soil is subjected to zero shear stress (i.e. when ( 1 3 ) = 0), its stressstrain behavior is modeled using the initial modulus, Ei . This initial tangent
modulus is controlled by the confining stress, 3 and is calculated as follows:
Equation 5-1
Ei = K L Pa 3
Pa
where:
Ei
KL
pa
confining stress
When the exponent n is equal to 1.0, the initial tangent modulus, Ei , is directly
proportional to the confining stress. When n is equal to 0, Ei is independent of the
confining stress.
Two different methods are used in SIGMA/W to calculate the initial soil modulus.
When K L is equal to zero (i.e., not defined by the user), the initial tangent
modulus, Ei , is set to the user-specified value for E and remains constant
throughout an analysis. If K L is nonzero (i.e., defined by the user), Ei is computed
using the equation above.
When the confining stress is zero or negative (i.e., the soil is in a state of tension),
it is possible for the initial modulus, Ei , to become zero or negative. In order to
Page 116
SIGMA/W
Tangent modulus
A soil is said to be following a loading path when it is subjected to a shear stress
higher than it has previously experienced, for example, from point O to point A in
Figure 5-2. Along this loading path, its constitutive behavior is governed by the
tangent modulus, Et . This tangent modulus is defined in the Duncan and Chang
model as a function of soil properties, triaxial deviatoric stress, ( 1 3 ) , and
R ( 1 3 )(1 sin )
Et = 1 f
Ei
c
+
2
cos
2
sin
(
)
3
where:
Ei
Et
tangent modulus,
Rf
ratio between the asymptote to the hyperbolic curve and the maximum
shear strength. This value is usually between 0. 75 and 1. 0,
Et can be limited to a minimum value. The default is equal to Pa . This userspecified lower limit is called Emin in the material properties dialog box. If Et
becomes too small it causes convergence difficulties.
Page 117
SIGMA/W
Unloading-reloading modulus
When a soil is unloaded from a higher shear stress state, (for example, going from
point B to point C in Figure 5-2), the non-linear model uses the unload-reloading
modulus, Eur .
The unload-reloading modulus, Eur , is computed in a manner similar to the
computation of the initial modulus, Ei except that the unloading-reloading
modulus number, K ur replaces K L in Equation 5-1. Thus, the unloading-reloading
modulus is calculated as:
Eur = K ur Pa 3
Pa
Bm = K b Pa 3
Pa
where:
Bm
bulk modulus,
Kb
modulus number,
Page 118
SIGMA/W
Pa
The relationship of the bulk modulus to the Poisson's ratio can be defined as
follows from theory of elasticity.
Bm =
E
3 (1 2 )
Letting the Poissons ratio, v, be equal to zero in the above equation gives a bulk
modulus, Bm , equal to E/3. When letting the Poissons ratio, v, be equal to 0.49,
Bm has a value of 17E. SIGMA/W imposes both upper and lower limits on the
computed value of Poissons ratio. The upper limit is 0.49. The lower limit is
either the user-defined constant value or 0.01.
If K b is not defined (i. e. zero), SIGMA/W assigns the user input constant value to
the Poissons ratio. Otherwise it is calculated using Equation 5-2.
Negative pore-water pressure
The cohesive component of a soil can consist of two components; namely,
effective cohesion and cohesion due to matric suction (Fredlund and Rahardjo,
1993):
c = c + (ua uw ) tan b
where:
c
(ua - uw) =
Page 119
SIGMA/W
1 3 1 + 3
sin R f c cos
In Duncan and Changs formulation of the hyperbolic model, Rf, the failure ratio,
is used in the following manner:
( 1 3 ) f
= R f ( 1 3 )ult
The ultimate strength term, (1 -3)ult, represents the asymptote which the
hyperbolic stress-strain curve will approach at large strains. The (1 -3)f is the
deviatoric stress at failure. From a Mohrs diagram, it can be seen that:
( 1 3 )ult ( 1 + 3 )ult
2
sin c cos
When multiplied by Rf and substituting in the second equation above, this equation
can be written as follows:
( 1 3 ) f ( 1 + 3 ) f
2
sin R f c cos
Comparing this equation with Equation 5-3, it is seen that the inequality given in
Equation 5-3 provides an indicator as to how close the stress state is to the failure
state.
Page 120
SIGMA/W
Bm =
( 1 + 2 + 3 ) / 3
where:
Page 121
SIGMA/W
Further details of the hyperbolic model can be found in the report by Duncan et al.
(1980). This report also contains a list of hyperbolic parameters for some 135
different soils.
Table 5-3 Non linear elastic (hyperbolic) properties
Edit Box Label
E Modulus
Property
Initial soil stiffness, it is used when K(load) is zero.
Poissons Ratio
K(load)
n(exponent)
K(ur)
K(bulk)
m(exponent)
Min. P. Ratio*
Pa
Rf
Cohesion
Phi
Phi B
A value
5.5
Elastic-plastic model
Page 122
SIGMA/W
Plastic matrix
In SIGMA/W, soil plasticity is formulated using the theory of incremental
plasticity (Hill, 1950). Once an elastic-plastic material begins to yield, an
incremental strain can be divided into an elastic and a plastic component.
{d } = {d e } + {d p } ,
{d e } = {d } {d p }
or
Only elastic strain increments, de, will cause stress changes. As a result, stress
increments can be written as follows.
Equation 5-4
{d } = [Ce ]{d e } ,
or
{d } = [Ce ] ({d } {d p })
A function which describes the locus of the yield point is called the yield function
and is defined using the symbol, F. In the Elastic-Plastic model of SIGMA/W, the
Page 123
SIGMA/W
yield point depends only on the stress state. Consequently, the yield function can
be written as follows in equation form.
F = F ( x , y , z , xy )
An incremental change in the yield function is given by:
dF =
F
F
F
F
d x +
d y +
d z +
d
x
y
z
xy xy
dF =
{d }
The theory of incremental plasticity dictates that the yield function, F < 0, and,
when the stress state is on the yield surface, dF is zero. This latter condition is
termed the neutral loading condition, and, can be written mathematically as:
dF =
{d } = 0
{d } = G
p
where:
G
Substituting the plastic strain from Equation 5-5 into the incremental stress
equation (Equation 5-4) gives:
Page 124
SIGMA/W
Substituting the stress vector, {d}, into the neutral loading condition, the
following expression for the plastic scaling factor, , can be derived.
[Ce ]
G
[Ce ]
{d }
From the previous two equations, a relationship between stress increments and
strain increments can be obtained.
{d } = ([Ce ] C p ) {d }
where:
G F
[Ce ]
G
F
[Ce ]
[Ce ]
C p =
To evaluate the plastic matrix, [Cp], the yield function, F , and the plastic potential
function, G , need to be specified.
Yield criterion
SIGMA/W uses the Mohr-Coulomb yield criterion as the yield function for the
Elastic-Plastic model. The following equation provides a common form of the
Mohr-Coulomb criterion expressed in terms of principal stresses.
J
I
The Mohr-Coulomb criterion can also be written in terms of the stress invariants
I1 , I 2 and . The yield function, F, can then be written as follows (Chen and
Zhang, 1991).
Page 125
SIGMA/W
J
I
where:
J2 =
2
2
1
2
x y ) + ( y z ) + ( z x ) + xy2
(
the second
3 3 J3
32
2 J2
1
3
= cos 1
J 3 = xd yd zd zd xy2
I1 = x + y + z
id = i
I1
3
where i = x, y or z.
When the angel of internal friction, , is equal to zero, the Mohr-Coulomb yield
criterion becomes the Tresca criterion (Smith and Griffiths, 1988):
F = J 2 sin + c
3
The plastic potential function, G, used in SIGMA/W has the same form as the yield
function, F (i.e. G = F) except the internal friction angle, , is replaced by the
dilation angle, . Thus, the potential function is given by:
Page 126
SIGMA/W
J
I
The derivatives of the yield function in terms of the stress invariants are computed
using the chain rule of differentiation.
F
F I1
F J2
F
=
+
+
I1
J 2
Derivatives of the Mohr-Coulomb yield function, with respect to the stress
invariants, can be written as follows:
F
sin
=
3
I1
F
1
1
=
sin cos +
sin + +
J2 2 J2
3
3
3
Equation 5-6
J
F
3
3
3
The derivatives of the stress invariants with respect to the stresses are:
I1
= 1 1 1 0
J2
= xd yd zd
2 xy
J3
3J J 2
3
=
3
2 J 2
2 J 2 sin 3
J3
J
J
J
= yd zd + 2 xd zd + 2 xd yd + 2 xy2
3
3
3
3
2 zd xy
Page 127
SIGMA/W
The material properties required for this model are given in Table 5-4.
Table 5-4 Elastic-plastic material properties
Edit Box Text
Property
E Modulus
Poissons Ratio
Constant value
Cohesion
Phi
Dilation Angle
in degrees
(0 ) .
A value
5.6
Strain-softening model
Page 128
SIGMA/W
Yield criterion
The strain-softening model in SIGMA/W is an elastic-softening plastic model. Its
yield criterion is a function of the stress state and the equivalent plastic strain. The
yield function, F, for the strain-softening model can be written in terms of shear
stress, q, and shear strength, cu.
F = F ( , p )
= q 3 cu
Under plane strain condition, it can be shown that at failure:
2 =
1 + 3
2
q = 3J 2
Page 129
SIGMA/W
The shear strength, cu, is dependent on the total equivalent plastic strain, p, as
given by the following equation:
Cu
cu = Cu R p
C
r
when p = 0
when 0 < p pr
when p pr
where:
( 1 3 )
Cu
Cr
pr
plastic strain where the softening line intersects the residual line, and
rate of softening.
2 at peak conditions,
p is the softening parameter for this model and is a measure of the total
(accumulated) deviatoric plastic strain given by:
p = d p
d p =
2
2
2
2
2 p
4
dex deyp ) + ( deyp dezp ) + ( dezp dexp ) + ( dexy )
(
3
9
where:
1
deip = d ip d vp , i = x , y , or z component
3
1
dexyp = d xyp
2
p
d v = ( d xp + d yp + d zp )
Page 130
SIGMA/W
Plastic matrix
An incremental change in the yield function, F, is given by the following total
differential:
dF =
{d } +
F cu
d
cu p p
When the stress state is on the yield surface, the incremental change in yield
function, dF, is zero.
dF = 0 =
Equation 5-7
{d } +
F cu
d
cu p p
The incremental plastic strain vector {dp} is related to the derivative of the plastic
potential function G through the plastic scaling factor, .:
{d } = G
p
Since only elastic strains can cause stress changes, the constitutive relationship can
be written as:
{d } = [Ce ] ({d } {d p })
Equation 5-8
G
= [Ce ] {d }
Substituting the above expression for incremental stresses, {d}, into Equation 5-7
gives:
[Ce ]{d } =
F F cu
d p
cu p
[Ce ]
The associated flow rule states that plastic potential function is the same as the
yield function (i.e. G =F). Applying the associated flow rule and using the
definition of incremental plastic strain, the incremental equivalent plastic strain can
be expressed as follows.
Page 131
SIGMA/W
2
2
2
2 q q q q q q 4 q
d p =
+
+
+
9 x y y z z x 3 xy
From these two equations, an expression for the plastic scaling factor, , can be
derived:
F
[Ce ]
=
{d }
F
F F cu
d p
[Ce ]
cu p
Finally, the incremental stress corresponding to a given incremental strain is
obtained by substituting into Equation 5-8.
{d } = ([Ce ] C p ) {d }
The plastic matrix [Cp] is calculated as follows:
F F
[Ce ]
C p =
F
F
F cu
d p
[Ce ]
cu p
[Ce ]
The plastic matrix [Cp] is the same as that used for the elastic-plastic model except
for the additional second term in the denominator.
Evaluation of the plastic matrix
To evaluate the plastic matrix, [Cp], it is necessary to compute the derivatives of
F cu
F
,
, and the
, and other derivatives terms
cu p
incremental equivalent plastic strain, d p .
the yield function,
Page 132
SIGMA/W
The yield function for the strain-softening model can be written as:
F = q 3 cu
Since the shear strength, cu, is independent of stresses, the derivatives of the yield
function, F, with respect to stresses is given by:
F q
=
Considering that the shear stress, q , is equal to q = 3 J 2 , this equation can be
rewritten as follows.
q 1
=
2
3 J2
J 2
F
= 3
cu
cu
=0
p
These derivatives are used to evaluate [Cp], as given previously.
Von Mises criterion
When the rate of softening, R, is zero and the residual strength, Cr, is the same as
the peak strength, Cu, the stress-strain curve takes the form of an elastic-rigid
plastic curve. Under these conditions, the yield criterion for the strain-softening
model becomes the von Mises criterion and is given by the following equation:
F = q 3 Cu
Page 133
SIGMA/W
where:
Cu
5.7
Property
E Modulus
Poissons Ratio
Constant value
C(peak)
C(residual)
Softening Rate
Cam-clay model
Page 134
SIGMA/W
Soil parameters
The Cam-clay model is an effective stress model which requires the following soil
properties:
Specific volume at the critical state when p ' is 1.0 (or, ln ( p ' ) is
0)
v Specific volume
These five parameters are illustrated in Figure 5-7. The critical state line shown in
the p ' q plane is the locus of critical states projected onto that plane. The critical
state line has a slope of, , which is related to the angle of internal friction of the
soil. For the case of triaxial compression, can be expressed as:
Page 135
SIGMA/W
6sin '
3 sin '
By definition, the specific volume, v , is related to the void ratio, e , through the
following expression:
v = 1+ e
Yield function
The yield function for Cam-clay, as illustrated in Figure 5-8, can be expressed in
terms of stress invariants p' and q as follows (Atkinson and Bransby, 1978):
F=
p
q
+ ln 1
Mp
px
where:
p'x
the peak mean stress and is the value of p at the critical state line or t.
The peak mean stress, p'x, is related to the pre-consolidation pressure, p'c, by:
Page 136
SIGMA/W
ln px = ln pc 1
or
px = p 2.71828
The stress invariants p' and q are alternative forms of the first stress invariant, I1',
and the second deviatoric stress invariant, J2, respectively.
p =
I1'
3
and:
q = 3J 2
By definition, the first stress invariant, I1', can be expressed in terms of effective
stresses as:
I1' =
1
( x + y + z )
3
Page 137
SIGMA/W
and the second deviatoric stress invariant, J2, can be written as:
J2 =
2
2
1
2
+ ( z x ) + xy2
(
)
(
)
x
y
y
z
The second deviatoric stress invariant, J2, is the same for either total or effective
stresses.
Equations for the over-consolidation line and the critical state line can be used to
calculate the peak mean stress, p'x. As illustrated inFigure 5-8, the specific volume
at critical state, vx, for a particular over-consolidation line can be written as:
vx = v + ln p ln px
From the critical state line, the same specific volume, vx, can also be calculated
using:
vx = ln px
Eliminating the specific volume, vx, from these two equations gives the following
expression for the peak mean stress, p'x.
Equation 5-9
v ln p
px = exp
Page 138
SIGMA/W
Plastic matrix
The plastic matrix [Cp] is developed using the yield function, F, in a manner
similar to the development used for the Elastic-Plastic and Strain-Softening
models. In the following discussion, subscripts e and p denote elastic and plastic
state respectively.
As shown in Equation 5-9, the peak mean stress, p'x, is a function of the specific
volume, v. If it is postulated that a change in the peak mean stress, p'x, can only be
caused by a plastic change in the specific volume, an incremental change in the
yield function, F, is given by:
Equation 5-10
dF =
{d } +
F px p
dv = 0
px v p
Page 139
SIGMA/W
dv p = dv dv e = ( )
dp
p
dv p = v0 d vp
where:
dvp
v0
dv
Since only elastic strains can cause stresses to change, the incremental effective
stress constitutive relationship can be written as follows:
{d } = [Ce ] ({d } {d p })
G
= [Ce ] {d } p
where:
[C'e]
Substituting {d'}, from the previous equation into Equation 5-10 gives:
Equation 5-11
F
F
G F px p
Ce ]{d } =
Ce ] p
dv
[
[
p
px v
Page 140
SIGMA/W
The plastic increment in specific volume, dvp, can be calculated from the original
specific volume and the incremental plastic volumetric strain, d vp .
dv p = vo d vp = vo p (
G G G
+
+
)
d x d y d z
Applying the associated flow rule (i.e. the plastic potential function, G, is the same
as the yield function, F), the following expression is obtained for the plastic scaling
factor, p.
F
[Ce ]
p =
{d }
F
F F px
[Ce ] p v0 Fm
px v
where:
Fm =
G G G F F F
+
+
=
+
+
d x d y d z d x d y d z
{d } = ([Ce ] C p ) {d }
where:
F F
[Ce ]
C p =
F
F F px
Ce ]
vF
[
p 0 m
px v
[Ce ]
Page 141
SIGMA/W
p
F px
F
,
, v0 , and Fm , px and Fm , and the
dv
dv
= p
d ln p
dp
d v =
dV
de
dv
=
=
1+ e
V
v
where:
total volume
Combining the previous two equations, the following expression for the
incremental volumetric strain is obtained.
d v = (
vp
)dp
Page 142
SIGMA/W
K =
dp
d v
K =
vp
(1 + e ) p
In turn, the following expression is obtained for the effective stress modulus E ' .
E = 3K (1 2 ) =
3 p
(1 + e )(1 2 )
F
F
=
x
F
y
F
z
F
xy
F F p F q
=
+
i p i q i
F I1
F J2
=
+ 3
I1 i
q i
where:
i
Page 143
SIGMA/W
F
q
1
=
+
2
p
p
p
1
F
=
q p
The remaining terms in the plastic matrix C p ,
F px
,
, v , and Fm , and
px v p 0
F
1
=
px
px
px
p
= x
p
v
Fm =
F F F
+
+
x y z
F p p p F q q q
+
+
+
+
p x y z q x y z
F
F
1) +
(0)
=
(
p
q
=
1
q
+
2
p
p
Initial condition
The initial soil conditions are defined by the initial specific volume, v0, together
with the initial stresses. In an analysis using the Cam-clay model, the initial
stresses must be defined. It is necessary to know the initial pre-consolidation
pressure, p'c, in order to compute v0. The relationship between the initial void ratio,
v0, and the pre-consolidation pressure, p'c, is illustrated in Figure 5-9.
A user can either specify the pre-consolidation pressure, p'c, directly when defining
the model parameters or allow SIGMA/W to approximate the pre-consolidation
Page 144
SIGMA/W
pressure, p'c, from the in-situ stresses, the over-consolidation ratio, OCR, and the
coefficient of earth pressure at rest, K0.
If the pre-consolidation pressure is not specified when defining a Cam-clay soil,
SIGMA/W estimates p'c in the following manner. In an analysis using the Camclay model, the initial stresses represent the present in-situ stress state. In equation
form, the present in-situ stress state is
x 0
{ 0 } = y 0
z0
xy 0
The maximum vertical stress that the soil has experienced in the past is usually
determined from an oedometer test. The ratio of the maximum vertical stress in the
past, 'vmax, to that at present, 'v, is known as the over-consolidation ratio, OCR.
In SIGMA/W, OCR is a user-specified parameter for each soil type. The
relationship between the maximum vertical stress, 'vm, and maximum horizontal
stress, 'hm, is approximated as follows:
K 0 = 1 sin
= K 0 vmax
hmax
Assuming that the shear stress is zero, the maximum stress vector, {'m}, can be
written as:
' y 0 K 0 OCR
xmax
' OCR
y0
{ m } ymax
=
0
The historic past maximum mean stress, p'm, and maximum shear stress, qm, are
respectively:
pm =
1
( xm + ym + zm )
3
Page 145
=
qmax
1
2
xmax
SIGMA/W
) + ( zmax
xmax
)
ymax ) + ( ymax zmax
2
The pre-consolidation pressure, p'c, occurs in the p'-q space where the shear stress,
q, is zero. Therefore, from the yield function of Cam clay, the following expression
can be obtained:
p
ln c = 1 , or, ln px = ln pc 1
px
Substituting the natural logarithm of the pre-consolidation pressure, ln p'x, the yield
function for Cam-clay can be rewritten as:
q
+ ln p ln pc = 0
p
Therefore, the pre-consolidation pressure, p'c, is given by:
exp max
pc = pmax
M pmax
Once the pre-consolidation pressure, p'c, is obtained either as a user-specified value
or from the above equation, SIGMA/W proceeds to calculate the initial specific
volume, v0. First, N, the intersection of the normal consolidation line and the
vertical axis (i.e., when ln p = 0), is calculated from the following expression:
= + ( )
From the normal consolidation line passing through the pre-consolidation pressure,
p'c, the specific volume at the critical state is given by:
vc = N ln pc
Also, from the over-consolidation line passing through the pre-consolidation
pressure, p'c, the initial specific volume, v0, can be expressed as:
p
v0 = vc + ln c
p0
Page 146
SIGMA/W
In this expression, the initial mean stress, p'0 , can be calculated from the initial
stresses using the equation:
p0 =
1
( x 0 + y 0 + z0 )
3
M=
6sin
3 sin
Cc
2.303
C
= s
2.303
Alternatively, and can be obtained from a plot of void ratio versus ln(p)
instead of void ratio versus log10(p).
The specific volume v is equal to 1 plus the void ratio (e).
Page 147
SIGMA/W
= N +
for Cam-Clay
= N ( + ) ln 2
The parameter is defined as the intersection point of the critical state line with
the vertical p = 1.0 line.
Another vitally important item of data required for a Cam-clay analysis is the
present (initial) in-situ stress state and the initial yield surface. Establishing the
present in-situ stress state is relatively simple and is the same procedure as for any
other SIGMA/W nonlinear analysis. Establishing the initial yield surface can be
done by specifying the pre-consolidation pressure (pc) as a material property, or by
computing pc from the initial in-situ stresses together with a specified overconsolidation ratio OCR. The SIGMA/W procedure for using this second
alternative is described in the chapter on Boundary and Initial conditions.
Table 5-6 summarizes the material properties required in the Cam-clay model.
Table 5-6 Cam clay / modified cam clay properties
Edit Box Label
O. C. Ratio
Property
Over-consolidation ratio
Poissons Ratio
Constant value
Lambda
Kappa
Gamma
Mu
Pc'
5.8
The modified Cam-clay model is similar to the Cam-clay model except that the
yield function is in the shape of an ellipse instead of a tear drop. The yield curve
for the modified Cam-clay model is illustrated in Figure 5-10.
Page 148
SIGMA/W
The yield function for the Modified Cam-clay model is given by the following
equation (Britto and Gunn, 1987):
q 2 = 2 ppc 2 p2
where:
p'c
pre-consolidation pressure.
The parameters used to define the Modified Cam-clay model are: M , , , and v .
These parameters are the same as those used for the Cam-clay model as discussed
in the previous section.
Similar to the Cam-clay model, SIGMA/W uses the peak mean stress, p 'x , to
determine the size of the yield locus for the Modified Cam-clay model. The peak
mean stress, p 'x , is the isotropic pressure when a soil reaches its critical state. At
the critical state, the shear stress, q, is given by the following equation:
Page 149
SIGMA/W
q = px
Substituting this value of shear stress, q, into the modified yield function equation
gives:
pc = 2 px
Therefore, the yield function, F, for the Modified Cam-clay model can be written
as:
F=
q2
+ 2 p 2 2 px
p
The derivatives of the yield function, F, with respect to stress invariants p ' and q
are respectively:
F
q2
2
= 2
p
p
F 2q
=
q p
F 2q
=
q p
px
q
are the same as those used for the
, [Ce ] , and
vp
Cam-clay model. The remaining variables in the plastic matrix, C ' p , for
p
,
The terms
q2
Fm = M 2
p
2
Page 150
SIGMA/W
F
= 2 M 2
px
Fm = M 2
q2
p 2
Except for the differences noted in the following discussion, the procedure to
define the initial condition of the Modified Cam-clay model is the same as that for
the Cam-clay model described in the previous section. If a pre-consolidation
pressure is not defined by the user, it is calculated using:
pc =
1
2
2
+ M 2 pmax
( qmax
)
M pmax
2
The interception of the isotropic normal consolidation line with the vertical axis
(i.e. ln p = 0 ), is given by:
= + ( ) ln 2
Using these values of the pre-consolidation pressure, pc , and the vertical
intercept, , the initial specific volume, v0 , can be computed using equations
outlined in the Cam-clay model section of this chapter.
5.9
Slip surfaces
SIGMA/W can simulate a slip surface using quadrilateral elements with Slip
Surface as its soil model. Using the Slip Surface material model, a user can specify
stiffness in two orthogonal directions. These directions are parallel and
perpendicular to the long side of the element. Figure 5-11 shows a typical slip
surface element with its long (x-) axis inclined at angle from the global x-axis
with nodes numbered as 1, 2, 3 and 4.
Page 151
SIGMA/W
SIGMA/W treats a slip surface element as a combination of four bar elements. For
example, for the element illustrated in the figure, there are two bar elements with
stiffness, Kn, in the local normal y direction, 1-4 and 2-3; and two bar elements in
the local tangential x-direction, 1-2 and 4-3, with stiffness Kt. Kt and Kn are the
user-specified tangential and normal stiffness for the element.
When formulating the element stiffness matrix [K], instead of evaluating the
integral
Kt
0
Kt
0
0
0
0
0
Kn
Kt
0
0
0
0
0
0
0
0
0
Kt
Kn
Kn
Kt
Kt
Kn
Kn
0
Kn
0
0
0
0
Kt
0
0
0
Kt
0
Page 152
0
K n
0
0
0
0
0
K n
SIGMA/W
[K'] is then rotated to the global coordinate system to obtain [K] using the
following transformation:
[ K ] = [T ][ K ][T ]
where:
cos 2
sin 2
[T ] =
0
sin cos
sin 2
cos
0
2
sin cos
2sin cos
0 2sin cos
1
0
0 cos 2 sin 2
0
When the normal stress on a slip is zero or tensile (negative), the stiffness values
are reduced by a factor of 0. 001.
Table 5-7 summarizes the slip surface element material properties.
Table 5-7 Slip surface properties
Edit Box Label
Property
K Normal
K Shear
In order to model a slip surface, a slip element should be rectangular and relatively
thin. However, it is highly recommended that you do not make the slip elements
too thin. The slip elements should have sufficient thickness such that they are
readily visible in the finite element mesh when the mesh is displayed at 100% and
the problem scale is the same in both horizontal and vertical directions. SIGMA/W
treats the longer side of a slip element as length and the shorter side as thickness.
Stresses and pore-water pressures in a slip element are obtained from the
neighboring elements above and below. In these neighboring elements, stresses are
first extrapolated from the Gauss points to the interface nodes. Then, from these
nodes, SIGMA/W interpolates for values at the Gauss point using shape functions.
The slip surface elements are treated as having drained behavior. Therefore, porewater pressures are not computed on the basis of stress changes in the element.
Page 153
SIGMA/W
A slip surface element is linear elastic and has no maximum shear resistance. If
you wish to model the slip material as having some maximum shear strength, it is
best to select a different soil model (e.g., the Elastic-plastic model).
Page 154
SIGMA/W
Boundary Conditions
This chapter discusses the various types of boundary conditions that can be applied
in SIGMA/W analyses. There are many ways to apply forces and displacements to
soil and structural elements in order to replicate real world loading conditions.
Before any loading is applied, it is important to know what the model will do with
the data you input. A discussion of the incremental finite element formulation has
been on-going throughout the book (see Chapter 3 and the Theory chapter). It is
critical to understand that a load should only be applied on one step in the model
and then removed. Sometimes you must remove a load manually (such as turning
off body loads after the in-situ conditions are solved) and sometimes SIGMA/W
will turn them off automatically (such as during fill placement when body loads
are only counted on the step the fill is activated).
This chapter shows what boundary condition options are available and how the
model uses the data. The following chapter on Analysis Types discusses the
differences between in-situ analysis and load-deformation analysis and when
various boundary conditions may be used. Therefore, the present discussion makes
limited reference to when different types of boundary conditions are applied.
6.1
Multiple boundary condition types are implemented to support virtually all loaddeformation modelling scenarios. In SIGMA/W, displacement, force or spring
boundary conditions may be applied to nodes, and stress and fluid pressure
boundary conditions may be applied to edges. Stress boundary conditions can be
applied to external element edges and may be specified as a normal and tangential
stress pair, or as an x-stress and y-stress pair.
An example of the unique way in which boundary conditions can be used is given
in the following figure, where struts used in a braced excavation can be modeled
by using a spring boundary condition at the location of the struts.
Page 155
SIGMA/W
6.2
Force or displacement
Fundamentally, there are only two types of boundary conditions that can be
applied in a stress-deformation model: force or displacement. How these
conditions are interpreted and applied is a different matter, because there are many
options depending on what your objectives are.
In all stress-deformation models it is critical to bound the problem. This means
that you must define some parts of the geometry as zero displacement boundary
conditions. If there is no bounding to the problem, how can there be any
reactionary force to the applied loading?
In general, it is common to bound the left and right sides of a problem along with
the bottom edge. A typical example of this is given in Figure 6-2, where the
bottom and side boundary nodes are specified as zero x- and zero y-displacement.
It is quite easy to see why there should be zero movement at the base of the model
in both directions, but there can be some question about when to also set the
vertical movement along the sides to zero. The answer to this latter question
depends on what you are trying to model. If you are modeling a larger domain
where you are saying the soil at the far sides is solidly attached to other soil even
farther away, then both x- and y-displacement should be zero. If you are modeling
a triaxial cell that is symmetric about the vertical axis, the axis of symmetry should
be fixed in the horizontal direction, and the bottom should be free to move in the x, but not y-direction as shown in Figure 6-3.
Page 156
SIGMA/W
+1
+2
+1
+6
+5
+6
+6
+5
+5
+6
+5
+4
+4
+4
+4
+5
+4
+3
+3
+3
+3
+3
+3
+4
+3
+2
+2
+2
+2
+2
+2
+2
+2
+3
+2
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+2
+1
+1
+3
+2
+4
+3
+5
+4
Figure 6-2 Bounding the model with zero displacement along edges
Page 157
6.3
SIGMA/W
Body loads
Body forces are included in an analysis by specifying a non-zero load (or weight)
per unit volume. For each element, SIGMA/W computes the volume of the
element, multiplies the volume by the specified unit load of the material, and
applies the total element body load as forces at the nodes of the elements.
Gravity body loads are included by setting the vertical body load to the unit weight
of the soil. Horizontal body loads can be specified to simulate pseudo-static
seismic forces. The direction of the body load is controlled by making the load
positive or negative, as described in Table 6-1.
Table 6-1 Body load sign convention
Force
Direction
Sign
Gravity
Vertical
Positive
Uplift
Vertical
Negative
Horizontal
Positive
Horizontal
Negative
Page 158
SIGMA/W
6.4
Page 159
SIGMA/W
Value
Symbol
X-Direction
Symbol
Y-Direction
Description
(none)
--------
--------
--------
none
Displacement
positive (+)
Displacement
negative (-)
Displacement
zero (0)
Force
positive (+)
Force
negative (-)
Force
zero (0)
Spring
not applicable
right
up
left
down
hollow red
arrow
hollow red
arrow
hollow red
triangle
solid red arrow
right
up
left
--------
down
--------
none
solid green
arrows
left or right
up or down
Spring
zero (0)
Rotation
--------
--------
none
zero (0)
--------
Moment
positive (+)
clockwise
rotation
hollow blue
circle
Moment
negative (-)
counterclockwise
hollow blue
circle
Boundary nodes have a negative x- or y-value point in the negative x- or y-direction, while
boundary nodes that have an x- or y-value that is positive or equal to zero point in the
positive x- or y-direction.
Page 160
SIGMA/W
6.5
Value
Symbol
Description
(none)
--------
none
--------
Normal Stress
positive (+)
Normal Stress
negative (+)
Normal Stress
zero (0)
Tangential Stress
positive (+)
Tangential Stress
negative (-)
Page 161
SIGMA/W
Tangential Stress
zero (0)
X-Stress
positive (+)
X-Stress
negative (-)
X-Stress
zero (0)
Y-Stress
positive (+)
Y-Stress
negative (-)
Y-Stress
zero (0)
Fluid Elevation
greater than
min. edge
y-coordinate
Fluid Elevation
NOTE: Edge boundary arrow symbols are only drawn for pressure boundary edges that are
on the perimeter of the mesh. Edge boundaries along internal element edges do not have
any arrow symbols shown, since the net resultant force on internal pressure boundaries is
always zero.
Page 162
SIGMA/W
As illustrated in Figure 6-5, the force computed from the specified pressure is
equally divided between the two corner nodes for a 4-noded element. For a higherorder 8-noded element, 1/6 of the force is assigned to each of the corner nodes and
4/6 is assigned to the middle intermediate node.
F/2
4/6
F/2
constant pressure
F/6
F/6
4-noded
8-noded
The nodal forces are actually computed numerically by integrating along the edge
of each element (see the Theory chapter). This generalizes the scheme and makes it
possible to consider a variable pressure distribution along the edge of the element.
It is useful to remember that specified boundary pressures cannot be used directly in a
finite element analysis - they must always be converted to nodal forces.
Page 163
SIGMA/W
Specified elevation
When you apply the fluid pressure type to an element edge, SIGMA/W internally
calculates equivalent force applied at each of the nodes defining the edge. The
nodal pressure is obtained by subtracting the nodal y-coordinate from the fluid
elevation and multiplying it by the fluid self-weight. SIGMA/W will only consider
positive fluid pressures. Therefore, for the fluid pressure boundary condition to
have an effect, the selected edge must have a y-coordinate less than the fluid
elevation.
NOTE: drawing a water table above the mesh does not generate a pressure
boundary condition on the mesh. It allows pore-water pressures in the soil to be
known, but it does not apply a total force (load) condition due to the weight of the
water on the ground surface. You must specify both water table and fluid pressure
if you need soil water pressures and surface water forces.
Defining a pressure boundary at internal element edges
By design, SIGMA/W allows you to define a pressure boundary only at external
element edges, which are the edges belonging to a single active element. If you
wish to define a pressure boundary at an internal element edge, you can use the
equivalent nodal force approach as described below.
Along an internal element edge, the equivalent nodal forces can be calculated by
multiplying the applied pressure with the corresponding contributing areas and
then applying those nodal forces as nodal boundary conditions. The contributing
areas are dependent on the analysis type (either two-dimensional or axisymmetric)
and on whether there is a mid-side node. Figure 6-7 and Figure 6-8 show the
contributing areas computed for corner nodes, a1 and a2 ; and the mid-side node,
Page 164
SIGMA/W
Figure 6-7 Contributing area for two dimensional elements of unit thickness
Page 165
SIGMA/W
Page 166
SIGMA/W
6.6
F = Fi Fi 1
where i is the time step number.
SIGMA/W obtains Fi and Fi-1 from the boundary function. The difference is
applied as the boundary condition for the i-th time step.
As a result, it is necessary to start defining the function at Time 0. The time
increment for the first step then is:
F1 = F1 F0
When you create a boundary function such as the one illustrated in Figure 6-9, you
should think of the Y-value as a cumulative value. If you want 10 lbs force applied
at load step 2, but at no other steps, then the function should look like the one
below. For all steps past step 2, there is no change in load so no load is applied. If
you had the function drop back to zero for all additional steps, then on the second
step, the program would use a change in load of 0 100 = -100 lbs and you would
undo what you tried to accomplish in the first step.
Page 167
SIGMA/W
10
Y-Force
0
0
Time
Figure 6-9 Step load of 10 on step 2 and zero on all other steps
Page 168
SIGMA/W
Analysis Types
7.1
Problem view
7.2
There are different ways to determine the in-situ stresses prior to a loaddeformation or consolidation analysis. SIGMA/W has two in-situ analysis types,
plus the load-deformation analysis can also be used.
In-situ 1 analysis
An In-situ 1 analysis is used to establish the initial in-situ stress conditions induced
by the self-weight of the soil. In this analysis type, you must specify the body load
(unit weight) and the relationship between horizontal and vertical effective stresses
(K0). This analysis only uses the linear-elastic soil model and it only applies to
horizontal ground surface where there is no shear stress in the ground.
Page 169
SIGMA/W
SIGMA/W computes the vertical stress y on the basis of the gravity load (unit
weight). The horizontal stress x is computed as y u multiplied by the userspecified K0 plus the pore-water pressure u. In equation form:
x = ( y u ) K 0 + u
The shear stress is arbitrarily set to zero.
This procedure is shown in Figure 7-1.
In-situ 2 analysis
An In-situ 2 analysis establishes the initial in-situ stress conditions using the
submerged weight of the soil. The initial pore-water pressure conditions are
obtained from the specified initial drawn water table or the selected external porewater pressure file. A discussion on drawing the water table is given later in this
chapter.
In an In-situ 2 analysis, SIGMA/W does the following: for any element that is
below the specified water table (or any positive pore-water pressure at a Gauss
point), SIGMA/W computes the submerged unit soil weight. The submerged unit
Page 170
SIGMA/W
weight is taken as the specified soil unit weight (specified as a vertical body load)
minus the water unit weight. With the modified soil unit weight(s), SIGMA/W
runs a gravity turn-on analysis (the applied load comes from the soil weight). This
step gives the in-situ effective stresses below the water table and the total stresses
above the water table. Next, below the water table the pore-water pressure is added
to the effective stresses to get the total stresses, and above the water table the
negative pore-water pressures are added to the total stress to get the effective
stress.
This procedure has the effect of applying K0 to the effective stresses below the
water table and applying K0 to the total stresses above the water table.
Load deformation analysis for in-situ stresses
A load/deformation analysis can also be used to establish the initial in-situ stress if
K 0 is less than 1.0. The reason for this limitation is that when there is no lateral
movement, K 0 is related to Poissons ratio by the equation:
K0 =
Page 171
SIGMA/W
K 0 = 1 sin
Two equations relating K nc , K 0 , OCR, and have been proposed by
Wroth (1975). The first equation is:
Page 172
SIGMA/W
K 0 = OCR K nc
v'
( OCR 1)
1 v '
3 (1 K
m
1 + 2K
) 3 (1 K 0 ) = ln OCR (1 + 2 K ) .
1 + 2K0
1 + 2K0
The value of m is an empirical constant that can be estimated from the plasticity
index (PI) using:
m
0. 022875 PI+1.22
For example, when ' = 30 , OCR =3, and v ' = 0.3 , K 0 must be 0.643.
The above equations are not used in SIGMA/W, but are presented here to help you
select an appropriate value for K 0 when you are establishing the current in-situ
stress state for a subsequent Cam-clay or Modified Cam-clay analysis.
In-situ stress example
Figure 7-3 shows a simple example we can use to illustrate how to spot check that
the correct initial in-situ stresses have been computed. The ground surface
elevation is 10 m. The surface of the free standing water is at EL. 12. The soil unit
weight is 20 kN/m3 and the water unit weight is 10 kN/m3 (rounded for
convenience here).
The effective stress at the ground surface should be zero. The total stress at the
ground surface should be equal to the weight of the free standing water (10 x 2 =
20 kPa). At the base of the problem, the effective stress v is (20 10 kN/m3) x 10
m = 100 kPa. Possions ratio is 0.334 and therefore Ko is 0.5
The effective horizontal stress is h = v Ko = 100 x 0.5 = 50 kPa
The total stresses are the effective stresses plus the pore-pressure which is 120 kPa
at the base of the problem: h = 50 + 120 = 170, v =100 + 120 = 220 kPa
The following graphs (Figure 7-4 and Figure 7-5) from SIGMA/W confirm and
correspond with these computed values. Note that the effective stresses at the
ground surface are zero and that the total stresses are 20 kPa.
Page 173
SIGMA/W
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
10
10
8
Vertical distance
Vertical distance
6
4
2
6
4
2
0
0
20
40
60
80
100
Y-Effective Stress
10
20
30
40
X-Effective Stress
Page 174
50
60
10
10
8
Vertical distance
Vertical distance
SIGMA/W
6
4
2
6
4
2
0
0
50
100
150
200
250
Y-Total Stress
50
100
150
200
X-Total Stress
It is good practice to always spot check your in-situ stresses using the Graphing
function in CONTOUR. For example, you can plot total and effective, vertical (y)
Page 175
SIGMA/W
and horizontal (x) stresses along a vertical profile through your section. It is
particularly important to make sure you are getting realistic horizontal effective
stresses. This was illustrated in the example above.
7.3
The Load/Deformation Analysis type is used whenever you want to apply loads
and find the resulting stress changes and displacements, including simulation of fill
placement and excavation construction procedures. In the case of a fill placement
analysis, the weight of the fill is added to your model on the first load step that
each fill layer is activated. In the case of an excavation analysis, SIGMA/W
calculates the resultant forces associated with removing the excavated elements
and applies these forces as negative values at the nodes along the excavation face.
The amount of the excavated forces are written to file so you can check them in the
Contour program.
If you check Adjust Fill to Design Elevation, then, after each time step, SIGMA/W
will adjust the top surface of any filled elements (that have settled) back to their
original design elevation. This simulates the continued placement of fill (while the
fill settles) until the fill reaches the prescribed elevation. This adjustment option is
available only for Load/Deformation analysis and is disabled for other analysis
types. By default, the fill adjustment option is not selected. More details of fill and
excavation are given in a separate chapter on the subject.
7.4
An initial water table is specified in two stages. First, the maximum negative
pressure head (i.e. the capillary rise) allowed in the initial condition is specified.
Secondly, the initial water table is drawn by clicking at each point in the initial
water table.
This feature is available for all analysis types in SIGMA/W. When you define an
initial water table, the initial pore-water pressure at each node is computed
proportionally to the vertical distance between the node and the defined water
table. The effect is that the pore-water pressure varies hydrostatically with distance
above and below the water table. Above the water table, the negative pore-water
pressure can be set to a limit to produce a pressure distribution such as shown in
Figure 7-6. The calculation of pressure at each point in the figure is given by:
Page 176
SIGMA/W
u a = ( y w - yr ) w
ub = ( yw - yb ) w
uc = ( yw - yc ) w
The pore-pressure at A is the specified maximum negative pressure head times the
unit weight of water.
The pore-pressure a B is the elevation of the water table minus the elevation of
Point B times the unit weight of water. This will be a negative number since
elevation B is higher than elevation of the water table.
The pore-pressure a C is the elevation of the water table minus the elevation of
Point C times the unit weight of water.
A
Max negative
pressure head
7.5
A series of initial conditions can be defined at the start of an analysis or after restarting the current analysis. SIGMA/W will add the computed incremental
changes to the user-specified initial condition if any are specified. Consequently,
the output files always contain the sum of the initial conditions and the incremental
changes.
The initial conditions that can be specified are:
Page 177
SIGMA/W
Nodal deformation
Strain
The type and objective of a particular analysis will determine which initial
conditions need to be defined. All or none of the initial conditions can be specified.
The initial conditions files must have the same number of nodes and elements as
the DEFINE data file, as well as the same number of integration points. In other
words, the initial conditions files must be compatible with the DEFINE data file
and previously computed results must be obtained from a compatible mesh.
For a linear-elastic analysis where the primary interest is settlement and no
excavation is involved, there is no need to open any initial conditions file. For
analyses using Cam-clay or Modified Cam-clay materials, it is necessary to define
the initial in-situ stress conditions, since the soil properties depend on the stress
state. For nonlinear elastic and the elastic-plastic materials, an initial stress
condition file is desirable, but not mandatory.
When the initial stress conditions have been established with an in-situ analysis,
the deformations and strains from such an analysis have no meaning.
Consequently, the initial deformation and strain should be zero, which is
accomplished by not opening an initial displacement or strain file.
Your can define the initial pore-water pressure conditions in two different ways:
Page 178
SIGMA/W
Structural Elements
8.1
Introduction
This chapter presents the formulation used in SIGMA/W for structural elements. In
SIGMA/W, a structural element can be either a bar element, which is capable of
resisting axial force only, or a beam element, which is capable of resisting both
bending and axial force. Structural elements can be used only in a two-dimensional
plane strain load-deformation analysis.
8.2
Beam elements
da
da
w = N1a1 + N1 + N 2 a2 + N 2
dr 1
dr 2
where:
Ni , Ni
The n -th derivatives in the local r and global x-coordinate systems are related
by:
Page 179
SIGMA/W
d nw l d nw
= n
n
dr 2 dx
Equation 8-1
where:
N1 = (2 + r )(1 r ) 2 4
N1 = (1 + r )(1 r ) 2 4
N 2 = (2 r )(1 + r ) 2 4
N 2 = (1 r )(1 + r ) 2 4
Stiffness matrix for a beam element
For a beam of flexural rigidity EI and length l , the strain energy b due to
bending is given by:
l
b =
0
EI d 2 w
dx
2 dx2
where:
Page 180
SIGMA/W
A beam element in SIGMA/W can be either 3-noded or 2-noded. However, a 3noded beam element is treated as a combination of two 2-noded elements. A 2noded element is formulated using cubic Hermitian interpolation functions. For
this element the lateral displacement can be expressed in terms of the Hermitian
interpolating functions as follows:
e
dw
dw
e
w = N1w + N1
+ N 2 w2 + N 2
dr 1
dr 2
e
e
1
l e dw
l e dw
e
w = N1w + N1
+ N 2 w2 + N 2
2 dx 1
2 dx 2
e
e
1
where:
le
Page 181
SIGMA/W
Again using Equation 8-1 and neglecting axial displacements, the curvature at any
point within the beam element can be expressed as:
d 2w
4
2 = e 2 0
dx l
d 2 N1
dr 2
d 2 N1 l e
dr 2 2
d 2 N2
dr 2
= [ B ]{a}
u1e
e
w1
d 2 N 2 l e ( dw dx )1
u2e
dr 2 2
w2e
e
( dw dx )2
where:
[B]
{a}
local coordinate ( 1 r 1) .
The strain energy due to bending in a beam element now can be written as:
le
= {a} [ B ] C [ B ]{a} dx =
T
e
b
{a} [ B ]
T
C [ B ]{a}
le
dr
2
where:
C
[K ]
le
2
[ B ] C [ B ] dr
T
Page 182
SIGMA/W
[ K ]g = [T ] [ K ] [T ]
e
For a beam element inclining at angle to the global x-axis, the transformation
matrix is:
cos
sin
0
[T ] =
0
0
sin
cos
0
0
0
0
0
0
0
0
1
0
0 cos
0 sin
0
0
0
0
0
sin
cos
0
0
0
0
0
0
8.3
Beam examples
Figure 8-3 shows two examples of the beam formulation applied in SIGMA/W.
The upper beam is fixed at the left end and on a roller at the right. A unit load has
been applied all along the top of the beam and the computed deflection is
illustrated. A closed form solution for this case was used to verify that the
SIGMA/W results are exact.
The lower beam in the figure is a cantilever beam with an applied moment
boundary condition on the right side. Again, the vertical displacement matches the
closed form solution.
In Version 6.0 of SIMGA/W it is necessary to define beams along the edges of
existing, non null, elements. In these examples, a single row of elements were
Page 183
SIGMA/W
specified and given almost zero strength. This way, the strength of the beam is
taking all the loading. In a future version of the software, it will be possible to
draw free beams that are not attached to two dimensional elements. This is possible
because the cross-sectional area of the beam is entered as a beam property and
does not need to be defined true to scale in the mesh.
Figure 8-3 Beam bending under constant load (upper), cantilever moment
load (lower)
Shear Force
0.0010
0.0005
0.0000
-0.0005
-0.0010
-0.0015
-0.0020
0
10
15
20
Distance
Page 184
SIGMA/W
Figure 8-4 shows a SIGMA/W generated shear force diagram for the cantilever
beam example. In addition to shear force, it is possible to generate graphs of axial
force, moment, and rotation in any structural member.
Another example of an application of a structural beam is given in Figure 8-5
where a series of beams are tied back during excavation. Full details of how this
analysis is set up and carried out are given in the Illustrative Examples chapter. For
now, it is useful to see how a beam can be used.
Figure 8-6 shows the bending moment diagram for the beam in the tie-back
example. It is clear from this figure that the additional anchoring of the beam by
bar elements is helping minimize the magnitude of the bending moment within the
beam.
20
18
Elevation - metres
16
14
12
10
8
6
4
2
0
0
10
15
20
25
30
35
Distance - metres
Page 185
40
SIGMA/W
Moment
100
50
-50
0
10
15
Distance
8.4
Bar elements
Page 186
SIGMA/W
1
N1 = (1 r )
2
1
(1 r 2 )
2
1
(1 + r )
2
1
(1 r 2 )
2
N2 =
N3 = (1 r 2 )
___
a =
0
EA du
dx
2 dx
where:
u
u e = N i uie
i =1
where:
Page 187
SIGMA/W
n
dN i e
du
=
ui
dx
i =1 dx
dN1
dx
dN1
dx
dN 2 u1e
, for a 2-noded element
dx u2e
u1e
dN 2 dN 3 e
u2 , for a 3-noded element
dx dx e
u3
where:
{n}
dN i dr dN i
=
dx dx dr
1 dN i
=J
dr
where:
r
local coordinate ( 1 r 1)
Jacobian operator = dr dx
Page 188
SIGMA/W
du
1 dN1
= J
dr
dx
= J 1
dN1
dr
e
dN1 u1
, for a 2-noded element
dr u2e
dN 2
dr
dN 3
dr
= J 1 [ B ]{a}
u1e
e
u2 , for a 3-noded element
u e
3
where:
[B]
{a}
The following expression for the strain energy due to axial deformation for a bar
element is obtained:
le
= {a} [ B ] C [ B ]{a} dx =
T
e
a
{a} [ B ] C [ B ]{a} J
T
dr
where:
C
[K ] =
e
[ B] C [ B] J
T
dr
Page 189
SIGMA/W
Bar elements can be set to become active at different load steps. If the active step
is defined as zero, the bar element is considered to be an active element from step
1. If the active step is nonzero, the bar element is considered to be an active
element from that active step.
Enter tension axial force in a bar element as a negative value and compression
force as a positive value. The reaction forces at the two end nodes will be shown as
inward arrows for tension axial force and outward arrows for compression axial
force those are the nodal forces that will be actually applied on the rest of the mesh
at the step defined. A pre-axial force can be applied before activating a bar
element.
8.5
Bar example
Figure 8-7 shows a structure composed of bar elements prior to and after loading.
Notice that the bars do not have to be drawn relative to any existing soil elements.
By definition, a bar can only support an axial load, so it can be drawn anywhere in
the domain as long as it is connected to valid nodes.
Another example of an application of a bar was given previously in the tie back
example (see Figure 8-5 above). In that example, the bar was use to tie back a
structural beam into the soil to help minimize bending moments within the beam
during excavation.
In order to define a bar, it is necessary to know the E modulus and the crosssectional area as well as any applied pre-axial force. Finally, you can also specify
what loading step in the analysis the bar becomes active. This is useful in a braced
excavation analysis, for example, because the bar does not become installed until
after the excavation has moved below the intended bar support elevation.
Another example of the use of a bar is given in the braced excavation file
discussed in more detail in the Illustrative Examples chapter. Figure 8-8 shows the
application of a strut (bar member) during excavation. Notice that the entire
excavation is included in the model and not just the left or right half. This is
necessary because the at this time, it is not possible to draw the fixed boundary
condition on the mid point of the bar while it is adjacent to a null element (as
would be the case if only half the geometry were modeled).
Page 190
SIGMA/W
90 deg.
metres
20 kN
30 kN
-1
-1
10
distance - metres
Page 191
11
SIGMA/W
Page 192
SIGMA/W
9.1
During a sequential fill placement simulation, forces due to the self-weight of each
lift are applied to the finite element grid as the lift is being placed. The forces are
computed by knowing the unit weight and the volume of each fill element. This
total force is then divided up and applied as a negative Y-force at each element
node. The advantage of doing this is that the fill element not only applies a load to
the soil below, it has some self-compression as well.
When a fill element becomes active, it must have the body load for its material
specified. In other types of analysis it is important to ensure that the body load is
only applied on one load step and it is your responsibility to make sure you do not
apply it in both an in-situ analysis and any subsequent analyses. Because you will
generally do a fill analysis over several load steps, it is not convenient to stop the
analysis just to turn off the body load force. SIGMA/W will recognize the special
nature of fill elements and only apply the body load on the first step that the fill
element becomes active.
SIGMA/W can optionally compensate for the settlement caused by self
compression of newly added fill elements. For each lift, SIGMA/W adds the
current displacement to the element nodal coordinates to compute the element
weight and volume, which in turn are used to compute the gravitational nodal
forces to simulate the fill placement. Adding the displacements gives the element
Page 193
SIGMA/W
additional mass which represents the additional fill placed to compensate for
settlement during construction.
Consider the Embankment example in Figure 9-1 which shows the final deformed
mesh after placing six lifts. Note that the final crest is close to the design elevation
while there is considerable settlement at the foundation-embankment contact.
40
35
Height (metres)
30
25
20
15
10
5
0
0
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
Distance (metres)
This effect is further illustrated by Figure 9-2 which shows the vertical settlement
along the dam and foundation centerline. The maximum settlement is near the base
of the dam, as it should be.
The settlement at the crest is 0.0647m from the placement of the last lift. This can
be minimized by making the last lift small.
Figure 9-3 shows a one-step analysis with gravity turned on. The crest now has
settled significantly. By comparing the two kinds of analyses, you can get an
estimate of the amount of fill that was placed to compensate for the settlement.
This example illustrates another important point. If entire embankment is applied
as a single load, then the maximum displacement is at the top of the dam, which is
not correct. In reality, the maximum vertical displacement is at the base of the dam
which is modeled correctly if several lifts are used to simulate construction.
Page 194
SIGMA/W
Y-Displacement vs. Y
40
30
20
10
0
-0.30
-0.25
-0.20
-0.15
-0.10
-0.05
0.00
Y-Displacement
35
30
25
20
15
10
5
0
0
10
20
30
40
50
60
70
80
90
100
Figure 9-3 Deformed mesh after a single load step embankment addition
Page 195
9.2
SIGMA/W
Excavation elements
Excavation elements work in a similar way to fill elements in that they can be
turned off at various load steps to simulate the process of digging an excavation
from the top down (or in the case of a tunnel, from the center to the liner). When
an element is removed, a force that pulls outward on the remaining soil must be
applied. Unlike the fill element where the forces are known by the weight of the
soil being added, an excavation element uses the existing total stresses at the
element edges adjacent to the elements being removed. In this way, the actual total
stresses acting in the x- and y-directions can be used to arrive at a nodal force load
for the element edge nodes.
Some caution should be used when doing an excavation analysis where pore-water
pressures are included prior to excavating. In general, when an excavation is made
it is assumed that the any water in the excavation will be pumped out. This means
that on the soil side of the excavation the total stress at the excavation face is
comprised of the effective stress and the pore-water pressure. When an excavation
element is removed, the total stress existing prior to removal of the element is used
to determine the unloading forces. If the total stresses are correct prior to
excavation (such that they already account for water pressures), then the
excavation analysis should be carried out without defining a water table or
importing pore-water pressure data from another source.
If the pre-excavation analysis did not account for any pore-water pressures, it
would still be wrong to then apply pore-water pressures simultaneously with
excavation. The reason for this is the total horizontal stress on the excavation face
will not be the same in both cases. Consider the following example where we wish
to know the excavation forces at a depth of 10m with the water table assumed to be
at surface, the unit weight of soil = 20, and K0 = 0.5.
For the following analysis we will use the total stress equation:
x = ( y u ) K 0 + u
where total stress equals effective stress plus pore-water pressure.
Case 1: pore-water pressure applied in pre-excavation analyses
In this case, the total stress is computed from the previous analysis step to
be:
Page 196
SIGMA/W
9.3
In general, this engineering book is not about how to use the software commands.
However, in this case, it is useful to see how fill and excavation elements are
designated and how they look once specified.
The Draw Fill/Excavation Elements command allows you to specify the time step
at which to add or remove selected elements to simulate fill placement and
excavation construction. For example, Figure 9-4 illustrates layers of elements
that are added incrementally at load times 1, 2, and 3 respectively:
+3
+3
+3
+3
+3
+3
+3
+3
+3
+2
+2
+2
+2
+2
+2
+2
+2
+2
+2
+2
+2
+2
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+3
+3
+2
+2
+1
+1
+1
Figure 9-5 shows part of an excavation problem in which elements are excavated
(removed) incrementally at load times 1 and 2. Elements with time 0 are
permanent elements which are present throughout an analysis.
Page 197
SIGMA/W
0
0 -1 -1
0 -1 -1
-1
-1 -10 0
-1 -1 -10
-1 -1
-1
-1
-1 -1
-2 -2
-2
-2
-2 -2
-2
0 -2 -2
0 -2 -2
0
-2
-2 -20
-2 -20 0
0
Fill elements are added to the problem at the selected time step, and do not exist
before this time step; excavation elements are removed from the problem at the
selected time step, and do not exist after this time step; permanent elements exist
throughout the entire problem, and therefore do not require a selected time step.
NOTE: An initial stress file is required when performing an excavation analysis. Without
any prior stress conditions, it would not be possible to compute the unloading forces caused
by removal of stress. In addition, excavation analysis should use correctly computed total
stress conditions which means do not apply the pore-water pressure load at the same time
as the excavation loads.
9.4
Braced excavations
Page 198
SIGMA/W
Ks =
F EA
=
l
L
where:
F
change in length,
Young's Modulus,
Some displacement is usually required before a strut is fully seated and acts like a
pure spring. You can simulate this by defining a spring boundary function using
the Boundary command in DEFINE. The spring function must be specified in
terms of the potential movement.
NOTE: The actual force acting at the node is given in the displacement file and can
be displayed in CONTOUR using the View Node Information command and
checking the Boundary Forces. You can use this force to check or design the size
of the strut. Furthermore, the force presented in the file is per unit width of wall (or
per unit radian in the case of an axisymmetric analysis).
A better alternative to using the spring node boundary condition is to use a bar
element. The bar has the advantage in that its stiffness is a function of its length
and the length is included as an actual value because it is drawn as part of the
problem geometry. In addition, the bar can be set to become active at different time
steps so you can define all the bars ahead of time and they will only become active
as the excavation passes their installation elevation.
The chapter on Illustrative Examples includes a model that uses structural supports
in an excavation analysis.
Page 199
SIGMA/W
Page 200
SIGMA/W
10
Consolidation
10.1
Page 201
x
1
x
x 1
xy
E 0
yz
0
zx
0
SIGMA/W
0
0
0
1
1
1
1
+
0
H
( x ua )
u
0
0
0
( y a )
( z ua )
1
0
0
0
0
2(1 + )
0
0 xy
0
0
2(1 + )
0 yz
0
0
0
2(1 + )
zx
( ua uw )
( u u )
a
w
( ua uw )
( ua uw )
( ua uw )
0
0
( ua uw )
where:
normal strain,
normal stress,
shear stress,
ua
pore-air pressure,
uw
pore-water pressure,
unsaturated soil modulus for soil structure with respect to matrix suction
(ua - uw), and
Poissons ratio.
Page 202
SIGMA/W
This relationship can be re-written in the following stress-strain form for a twodimensional space.
1 0 0
( x ua )
1 0
E (1 )
( y u a )
=
( z ua ) (1 + )(1 2 )
xy
u u
x a w
H
0
0
ua u w
y
H
0
1 2
u u
z a w
2 (1 + )
H
xy
[ ]{ } [ D ]{mH }( ua uw ) + {ua }
Equation 10-1 { } = D
where:
[D]
{mH }
1 1 1
0
H H H
If it can be further assumed that air pressure remains atmospheric at all times,
Equation 10-1 becomes:
{ } = [ D ]{ } + [ D ]{mH } uw
On the other hand, for a soil element which is fully saturated, the total stress on the
soil structure is given by:
{ } = {D}{ } + {m} uw
where:
{m}
Page 203
SIGMA/W
Comparing these last two equations, it can be seen that, when the soil is fully
saturated (i.e. S = 100%):
[ D ]{mH } = {m}
For a linearly elastic material, this condition is satisfied when:
E
H =
1 2v
10.2
k x 2uw k y 2uw w
+
+
=0
w x2 w y2 t
where:
kx, ky
uw
seepage velocity,
time.
The volumetric water content for an elastic material is given by the following
expressions (Dakshanamurthy et al. , 1984):
Equation 10-2
w =
v uw
and:
Page 204
SIGMA/W
E
1
, and,
H (1 2 )
Equation 10-3
1
=
R H
where:
KB
bulk modulus,
10.3
{ } { } dV = { } {F } dV
*
Page 205
SIGMA/W
where:
{*}
virtual displacements,
{*}
{}
internal stresses.
Substituting Equation 10-4 into this equation and applying numerical integration, it
can be shown that the finite element equations that SIGMA/W solves are given by:
[ B ] [ D ][ B ]{ } + [ B ] [ D ]{m }
T
{uw } = F ,
[ K ] = [ B ] [ D ][ B ] ,
T
[ Ld ] = [ B ] [ D ]{mH }
T
{mH }
1
H
1
H
N , and
1
H
where:
[B]
[D]
[K]
stiffness matrix,
[Ld]
coupling matrix,
{}
uw
For a fully saturated soil, the coupling matrix, [Ld], can be written as:
[ Ld ] = [ B ] {m}
T
N ,
The flow equation can similarly be formulated for finite element analysis using the
principle of virtual work in terms of pore-water pressure and volumetric strains. If
virtual pore-water pressures, u*w, are applied to the Flow Equation and integrated
over the volume, the following virtual work equation can be obtained.
Page 206
SIGMA/W
*
k y 2uw w
* k x uw
u
+
w w x 2 w y 2 + t dV = 0
k uw* uw k y uw* uw
* w
dV = uw* vn dA
x
+
dV + uw
x
x
y
y
t
w
w
where:
Vn
boundary flux.
k x uw* uw k y uw* uw
* ( v u w )
dV = uw* vn dA
+
dV + uw
t
w x x w y y
Using finite element approximations, this equation can be written as:
[ B ] [ K w ][ B ]{uw } dV
T
( )
dV
t
{m} [ B ]
T
(uw )
N
+
t
= N vn dV
T
where:
K f = [ B ] [ K w ][ B ] dV ,
T
[M N ] =
L f = N
N , and
T
{m} [ B ] dV
T
with:
[B]
gradient matrix,
[Kw]
Page 207
SIGMA/W
[Kf]
<N>
[MN]
mass matrix,
[Lf]
{m}
nodal displacement.
K {u } dt
w f w
t +t
(uw )
( )
L f
dt
+
dt
t
t
t
t +t
[M N ]
t
t +t
vn dAdt
Applying the time differencing technique using as the time stepping factor to this
equation, the following finite element equation is obtained.
( K
{uw }
L f { }
t +t
t +t
t
+ (1 ) K f {uw }
= t N
( v
n t +t
) [M
] (uw ) tt +t +
+ (1 ) vn
) dA
When the backward (fully implicit) time-stepping scheme is used (by setting = 1)
and assuming that and remain constant within a time increment, this equation
becomes:
K {u }
w f w
t +t
[ M N ]{uw } + L f { } = t {Q}
where:
{Q}
Page 208
t +t
SIGMA/W
K f + [ M N ] {uw } = t {Q}
w
L f { }
t +t
K f {uw } t
w
[ K ]{ } + [ Ld ]{uw } = {F } , and
t
K f + [ M N ] {uw } = t {Q}
w
L f { }
where:
[ K ] = [ B ] [ D ][ B ]
T
[ Ld ] = [ B ] [ D ]{mH }
T
{mH } =
1
H
1
H
1
H
N
0
K f = [ B ] [ K w ][ B ]
T
[M N ] =
L f = N
{m} [ B ]
Page 209
t +t
K f {uw } t
w
E
1
3K B
=
H (1 2v )
H
1 3
R H
SIGMA/W
In order for these equations to model the fully saturated case, the following
conditions must be satisfied:
1,
0, and
[Lf]
[Ld]T
10.4
dV = dVs + dVv
where:
dVs
dVv
If the volume change of the soil particles, dVs, is small and thus neglected, the
volumetric strain can be approximated as follows.
v =
dV dVv
V
V
Page 210
SIGMA/W
From the definition of void ratio, e, a change in void ratio, de, is given by:
V dV
dVv
d v
=
de = d v = v =
(1 n )V (1 n )
Vs Vs
where:
n
The slope of a void ratio versus matrix suction curve can be written as:
Equation 10-5
d v
de
=
d ( ua uw ) (1 n ) d ( ua uw )
In an unsaturated soil element, when only a change in matric suction occurs, the
incremental volumetric strain, dv, can be written as:
d v = d x + d y + d z =
3d ( ua uw )
H
or:
d v
3
=
d ( ua u w ) H
After substituting this into Equation 10-5, it can be seen that the slope of a void
ratio versus matric suction curve is:
3
(1 n ) H
10.5
Page 211
SIGMA/W
Page 212
SIGMA/W
10.6
The H is the unsaturated modulus that relates the volumetric strain of the soil to a
change in negative pore-water pressure or change in suction. The H-modulus may
be defined as a function of negative pore-water pressure. At saturation, H is related
to the elastic constants E and v by the equation:
E
H =
1 2v
Therefore, H must be set to E / (1 2v ) at zero pore-water pressure when defining
an H-Modulus vs. pore-water pressure function. As a soil dries and the pore-water
pressure becomes highly negative, the soil becomes very stiff. This increase in
stiffness can be represented by an increase in H. Figure 10-1 illustrates a potential
increase in H as a function of the negative pore-water pressure.
Page 213
SIGMA/W
Page 214
SIGMA/W
10.7
The formulation for saturated / unsaturated fully coupled consolidation has been
applied to a one dimensional model where results matched with expectations. This
model is discussed towards the end of this chapter. The application of fully
coupled saturated/unsaturated analysis in two dimensions is questionable because
of our inability to fully understand and define the relationship between the slope of
the soil water characteristic curve (Mv) and the soil stiffness (E) during variably
saturated conditions.
There are times, however, where you can successfully use a fully coupled analysis
in two dimensions. SIGMA/W has a powerful feature to only apply the fully
coupled formulation to the saturated zone within a model geometry that contains
both saturated and unsaturated zones. The assumption inherent in consolidating
within the saturated zones only is that there is sufficient water storage capacity
within the surrounding unsaturated zones to take in any dissipated water from
consolidating saturated zones without undergoing volume change. This is a
reasonable assumption in many cases.
If you choose the saturated zone consolidation analysis, SIGMA/W will solve the
fully coupled equations for all nodes in the mesh, however, it will fix the porewater pressures in the unsaturated zones to be equal to their value at the start of the
analysis. In other words, there will be no incremental change in pressure
throughout the analysis. The pore-water pressure in the saturated zones will be
dissipated according to the permeability defined in the seepage portion of the
analysis. This means that the rate of dissipation is likely controlled by the
permeability value associated with the unsaturated soils surrounding any saturated
zones experiencing consolidation. The unsaturated zones likely have a lower K
value than the saturated conductivity so they will control dissipation of pressures.
When this type of analysis is chosen, it is important to define the volumetric water
content function in the SEEP/W program. At a minimum, it is necessary to define
the Mv value of the function. This is necessary because SIGMA/W will set the
Poissons ratio to 1/3 and then compute the elastic modulus, E, from the Mv value
as follows:
1 3
E
H =
for a saturated soil (from Equation 10-2),
combined with =
R H
1 2v
results in
Page 215
E=R=
SIGMA/W
1
Mv
This is a valid approach because for a saturated soil, the change in water content is
equal to the change in volumetric strain. The Poissons ratio is set to be 1/3 which
validates the above assumptions for a linear-elastically behaved soil. This
combination of parameters will ensure that there is compatibility between
hydraulic and deformation moduli in the saturated only analysis; it conforms to the
theoretical coupled formulation in this zone; and it means the analysis can be
carried out without a full definition of the volumetric water content function (e.g.,
only the Mv is required).
10.8
tmin
l2
l 2 w (1 + )(1 2 )
=
12cv 12 k
E (1 )
where:
l
You can use this formula at a couple of typical locations in your problem to help
you select appropriate time step sizes. In most cases, the best approach is to apply
the formula to areas where you expect the largest hydraulic gradients.
Another method of computing time step sizes is to do a simple, hand-calculated 1D consolidation analysis using text-book, closed form solutions. A quick, rough
estimate of the time required to reach, for example, 10% and 50% consolidation
can help a lot in selecting a time-increment sequence for a SIGMA/W SEEP/W
consolidation analysis.
Page 216
SIGMA/W
Time step sizes can be too small or too large. If the time step sizes are too small,
your solution will be mostly determined by numerical noise, making the results
look totally unrealistic. If the time step sizes are too large, the computed
deformation can be in considerable error.
Time step sizes can vary over a fairly wide range, but they must be reasonable.
Unfortunately, an appropriate time-step sequence is problem dependent;
consequently, several trial runs must be made in the early stages of analyzing a
new problem.
10.9
Page 217
SIGMA/W
Page 218
SIGMA/W
Figure 10-2 shows a simple 1D mesh that can be used to illustrate several
components of coupled and un-coupled consolidation. In this 1m tall saturated
column problem, a 1000 kPa load has been applied to the surface. There is zero
drainage on the sides and base of the sample so the entire 1000 kPa total stress load
is transferred to the pore-water and the pressure instantly jumps up to 1000 kPa in
the entire sample. Over time, drainage out the top of the column permits this
pressure to dissipate which increases the effective stress and results in shrinkage.
Page 219
SIGMA/W
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0
100
200
300
400
500
600
(x 0.001)
The problem of dissipating the excess pore-water pressure can be modeled fully
coupled using SEEP/W and SIGMA/W formulations simultaneously. Figure 10-3
shows the dissipation of excess pore-water pressure over time as computed using
the coupled formulation. As expected, the pressures want to revert back to the preload condition. Figure 10-4 shows the surface node displacement over time during
consolidation. As shown, the maximum displacement is about -0.8m over the time
modeled.
You may question how a 1m tall column can displace by 0.8m. In reality it cannot,
but numerically it can. In this analysis we have used a linear-elastic soil model
which has no yield point. What this means is that it can displace as much as it
needs to in order to take up the stress. This is a good example to illustrate that it is
important to know what the different soil models mean. In this example, the
computed stresses are correct, but the computed strains given by the linear model
are not realistic.
Page 220
SIGMA/W
0.8
Elevation
2.0000e+002
3.0000e+002
0.6
4.0000e+002
6.0000e+002
0.4
8.0000e+002
1.0000e+003
0.2
1.2000e+003
1.4000e+003
1.6000e+003
0.0
0
200
400
600
800
1000
Pore-Water Pressure
Y-Displacement
-0.2
-0.4
-0.6
-0.8
0
500
1000
1500
2000
Time
Page 221
SIGMA/W
The output from the coupled solution can be put back into SIGMA/W to test the
uncoupled formulation. In other words, we can take the coupled pore-water
pressures from the end of the first load step in the coupled solution and use it as the
starting condition for an uncoupled solution. The ending condition for the
uncoupled solution can be set as the initial condition from the coupled solution. So,
we can simply test the un-coupled formulation by solving the coupled solution
backwards. Figure 10-5 shows the two pore-water pressure conditions applied in
the uncoupled analysis. These two conditions are directly obtained from the
previously solved SIGMA/W files. These are the two pressure conditions that will
give SIGMA/W the u at every node so that it can solve the consolidation
equations independently of a seepage solution. Figure 10-6 shows the computed ydisplacement at the end of the single load step un-coupled consolidation. The
maximum displacement at the surface is about -0.7 m which is not quite the same
as the -0.8 m computed in the fully coupled formulation.
The difference between the two displacements is explainable. In the coupled
formulation, the computed pressure written to file at the end of the first load step is
not quite equal to the total applied stress of 1000 kPa. In other words, during this
first load step, the load is applied and some of it has consolidated. So, when we
take the end of the first load step in the coupled formulation and use it as the start
of the un-coupled consolidation, we are missing a small amount of pressure
dissipation and this is giving some small difference in computed displacement.
Page 222
SIGMA/W
Elevation
0.8
0.0000e+000
0.6
0.4
0.2
1.0000e+000
0.0
0
200
400
600
800
1000
Pore-Water Pressure
This example has shown how output pressures from a previous SIGMA/W analysis
can be used to solve an un-coupled consolidation analysis without using the
SEEP/W program. In fact, SIGMA/W can use pore-water pressures generated by
itself, SEEP/W, QUAKE/W or VADOSE/W in its un-coupled formulation. There
is no need to specify hydraulic properties because the change in pore-water
pressures were previously solved in the external programs. SIGMA/W simply uses
these pressures to obtain the change in effective stress in order to determine the
volume change.
Page 223
SIGMA/W
Y-Displacement vs. Y
1.0
Elevation
0.8
0.0000e+000
0.6
0.4
0.2
0.0
-0.8
1.0000e+000
-0.6
-0.4
-0.2
0.0
Y-Displacement
Mendel-Cryer effect
The Mendel-Cryer effect involves pore-pressures that respond to loading such that
the pressure peaks at a higher value than the applied total stress. This can happen
when a total stress is applied to the surface of a sphere as illustrated in Figure 10-7.
In this case, a total stress of 100 kPa is applied on the first step and the pressures
are then allowed to dissipate with time. The seepage boundary condition is
specified in the SEEP/W program and it allows drainage out of the surface of the
sphere. Figure 10-8 shows the Mendel-Cryer effect where the pressure at the center
of the sphere slowly increases as the external pore-water pressure impact reaches
the center. After the peak, the excess pressures are allowed to dissipate back to the
pre-load conditions. Along with this increase in effective stress (e.g., dissipation of
excess pressure) comes shrinkage of the sphere as shown magnified 10x in Figure
10-7.
Page 224
SIGMA/W
Figure 10-7 Mendel-Cryer sphere and deformed sphere (symmetric about left
axis)
Pore-Water Pressure vs. Time
Pore-Water Pressure
150
100
50
0
0
10
Time
Page 225
SIGMA/W
The same analysis can be carried out in an uncoupled formulation. The excess
pressures of 100 kPa can be input into SEEP/W and then allowed to dissipate over
the same time period. When this happens, the Mendle-Cyer effect of the pressure
increase in the center of the sphere is not reproduced and the change in volume of
the sphere will be slightly different.
Two dimensional uncoupled consolidation example
40
35
25
20
15
160
100
80
10
120
160
Height (metres)
30
5
0
-5
-5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
Distance (metres)
Page 226
95
100
SIGMA/W
Figure 10-10 shows the deformed mesh generated in a single load step using the
change in pore-water pressure between the post construction condition (previous
figure) and pre-construction condition. The maximum displacement is computed to
be about 8cm at the base of the fill in the center of the embankment.
40
35
Height (metres)
30
25
20
15
10
5
0
-5
-5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
Distance (metres)
The most useful aspect to the un-coupled formulation is that you can use two
pressure conditions from any other GeoStudio analysis as start and end conditions
for a volume change (consolidation) analysis in SIGMA/W; and you obtain very
reasonable results without excessive computational effort or convergence issues.
NOTE: if you are dissipating an excessive pressure, and you have a non-linear soil
model, it may be useful to use the un-coupled option that applies the formulation
over successive time steps instead of a single load step. SIGMA/W can read
multiple pore-water pressure conditions and create the necessary loading sequence
to dissipate the pressures in smaller steps.
Page 227
SIGMA/W
Page 228
SIGMA/W
11
Numerical Issues
This chapter discusses several numerical issues related to solving the partial
differential equation for stress-deformation. It deals with non-linearity of soil
constitutive models and how they are dealt with. The chapter also discusses some
features in the SIGMA/W program that can help you assess convergence of your
solution such as graphing convergence data during the solve process or stopping
and restarting an analysis in order to make adjustments without having to re-start
an entire solve process.
11.1
Nonlinear analysis
{ } = { R ( a )} { F } = 0
t
where:
nodal displacements.
It is assumed that the solution will start from an equilibrium (or, at least, near
equilibrium) state such that (t-1) is approximately zero. The vector
Page 229
{R} = [ B ]{ } dv
SIGMA/W
Equation 11-2
i +1
} = { } + a { a }
i
where i is the iteration counter and the superscript t has been omitted for clarity. To
start the iteration process for time step t, displacements at the end of the previous
time step (i. e. , time t-1) are used as the initial estimate:
t
a1 = t 1a
R
a = a = [ KT ]
where:
[KT]
The iterative correction applied to the nodal displacements can be calculated as:
KT i { a i } = { i }
or:
{ a } = K
i
i
T
{ }
i
Page 230
SIGMA/W
a i +1} = { t 1a} + { a i}
k =1
In the Newton Raphson method, the stiffness matrix is tangential to the loaddisplacement curve and is updated every iteration. This process is illustrated in
Figure 11-1(a). For an analysis involving strain-softening materials, this process is
modified such that the stiffness matrix is not updated, but the stiffness matrix used
in the first iteration is retained throughout the time step, as illustrated in Figure
11-1(b). In order to achieve stability in SIGMA/W, the initial modulus is always
calculated based on elastic behavior.
It should be noted that the load term is the externally applied load for the first
iteration in a load step; and it is the unbalanced load for subsequent iterations.
Incremental displacements resulting from increment load {tF} at time t are
obtained upon solving Equation 11-1. These incremental displacements are added
to the displacements at the beginning of the time increment. In the displacement
output files, SIGMA/W reports the total displacements at each node since the
beginning of an analysis at Time Step 1.
Page 231
SIGMA/W
Page 232
SIGMA/W
11.2
Convergence
Page 233
SIGMA/W
Both criteria are expressed as the Euclidean norm of a vector quantity. The
(Euclidean) norm of a vector is a measure of its magnitude and is defined as:
|| D || = d 2
i =1
where:
|| D || =
( a
a =
n =1
2
xn
+ a yn 2 )
where:
N
axn
a yn
ai
i 1
100%
k
k =1
where:
Page 234
SIGMA/W
, and
k =1
The unbalanced load convergence criterion compares the norm of unbalanced load
in an iteration to the norm of the externally applied load in a time step. This
criterion is also expressed as a ratio in percentages, for the i-th iteration, it can be
expressed as:
i
F
100% f
where:
For an analysis without any externally applied load (i. e. when only displacement
boundary conditions are applied), SIGMA/W ignores the unbalanced load
criterion. For an analysis involving only linear elastic materials (the Linear-elastic
model and slip elements), SIGMA/W skips the iteration process because the
elasticity is not a function of confining stress and is therefore constant for all
elements at all times.
NOTE: You may select one or both convergence criterion; if you have selected
both criteria, iteration will only be stopped after both criteria are satisfied or when
the maximum number of iterations has been exceeded. As a general rule, the
unbalanced load convergence criterion is more useful when you are approaching a
limit load; otherwise, the displacement convergence criterion is usually sufficient.
Page 235
SIGMA/W
Graphing convergence
Displacement Norm
0.0025
0.0020
0.0015
0.0010
0.0005
0.0000
0
10
15
20
25
Iteration #
Page 236
SIGMA/W
Plotting displacement norms, the graph in Figure 11-2 seems to indicate that the
solution is converging quite nicely. Figure 11-3 shows unbalanced load norms for
the same problem. It shows significant fluctuations in unbalanced load indicating
that the analysis may be experiencing convergence difficulties.
Step Number 24
1
0
10
15
20
25
Iteration #
The convergence graph for the following step (Step 25) shows an unacceptable
increase in the unbalanced load norm (See Figure 11-4). The analysis needs to be
restarted at Step 25 after reducing the load step size in SIGMA/W DEFINE. A
graph of unbalanced load norms may provide a better indicator for convergence
and it is especially useful when performing a non-linear analysis.
Page 237
SIGMA/W
Step Number 25
150
100
50
0
0
20
40
60
80
100
Iteration #
When a problem involving sequential fill placement with fill height adjustment is
analyzed, the convergence record is reset to Iteration 1 after the fill height is
adjusted. The convergence record for the entire time step can be viewed once the
fill height adjustment process has been completed.
A trick that is sometimes useful in judging convergence is to set the maximum
number of iterations to a very high number and also set the convergence tolerance
to a very low number. This will force the solution far past an acceptable
convergence, but it can give you confidence that you have indeed reached an
acceptable, stable solution. You can use this knowledge to select an acceptable
criterion. In some cases, you may be able to relax the criterion while in other cases,
you may have to tighten the criterion.
11.3
The SIGMA/W finite element equations are a set of equilibrium equations. The
driving (applied) forces are in equilibrium with the resisting forces that arise from
the soil strength. The load that can be applied is therefore limited by the ultimate
resistance of the system. In the terminology of SLOPE/W (another GEO-SLOPE
product), the limiting load is the load that results in a factor of safety equal to 1. 0.
Page 238
SIGMA/W
11.4
Tension zones
The effect of tension zones can be handled indirectly with the Nonlinear Elastic
(Hyperbolic) soil model. Consider the case of a steep slope with tension in the
crest area. If the material has tensile strength, energy will be stored in the system,
preventing the soil from deforming to the point where cracks develop. However, if
the soil has no tensile capacity, cracks will form and the equivalent stored energy
will contribute to additional deformation. To prevent the soil from storing energy
representative of the tensile strength, the soil can be made very soft by assigning it
a low E modulus.
Page 239
SIGMA/W
In the Nonlinear Elastic (Hyperbolic) model, the soil stiffness can be a function of
the confining stress ( 3 ) . In a tension zone, the minor principal stress will tend
towards zero or even become negative. In response to this low confining stress, the
soil is assigned a very low E modulus value.
You can specify both Displacement Norm Tolerance and Unbalanced Load Norm
Tolerance. In this case, iterations will continue until both criteria are satisfied or
until the maximum number of iterations is exceeded.
In most modelling problems, only the displacement norm criterion is required. The
unbalanced load norm criterion is especially useful in an elasto-plastic analysis
when approaching the collapse load.
In summary, the Nonlinear Elastic model can be used to make the soil soft in a
zone of tension, essentially eliminating the energy representative of tensile
strength.
11.5
Stop-restart
11.6
Halt iteration
The iteration process for the current time step can be halted by clicking the Halt
Iteration, button. SOLVE uses the results from the last iteration to create the output
files (if necessary) for the current time step, and then continues analyzing the next
time step.
To halt the analysis at the current time step, you should press the Halt Iteration,
button and wait until the next time step is being analyzed; then, press the Stop,
button. This allows SOLVE to create the necessary output files for the current time
Page 240
SIGMA/W
step. If you press the Stop, button before the next time step is processed, the
analysis may be stopped before SOLVE can create the output files.
If you stop an analysis and wish to restart at the next time step, you must select the
latest displacement, stress, strain, pore-water pressure and property files as your
initial condition files as described in the Stop-Restart section.
The iteration number is reset to 1 when the next time step is started.
11.7
The solver program reads all input geometry, material properties, initial and
boundary conditions and then computes results for all user specified output load
steps. All computed solver output is written to text files that are then available to
the Contour program for visualization. The various output files are discussed
below. The letter designation associated with each file is the first letter in the
output file extension and is followed by an integer numeral that designates which
load step the file is associated with. Therefore myfile.d01 is the filename of the
displacement file generated for load step 1.
The displacement (D) files contain the displacement, the applied and computed
reaction forces at the nodes, pressure boundary conditions applied at the edges, and
axial and shear forces at structural element edges. The extension of each
displacement file name begins with D. The displacement file information can be
contoured, plotted, and viewed at each node.
The stress (S) files contain the x-stress, y-stress, z-stress, and x-y shear stress at
each element integration point. The extension of each stress file name begins with
S. The stress file information can be contoured, plotted, displayed as a Mohr
Circle, and viewed at each element Gauss region or approximated at each node.
The strain (E) files contain the x-strain, y-strain, z-strain, and x-y shear strain at
each element integration point. The extension of each strain file name begins with
E. The strain file information can be contoured, plotted, displayed as a Mohr
Circle, and viewed at each element Gauss region or approximated at each node.
The pore-water pressure (U) files contain the pore-water pressure at each element
integration point. The extension of each pore-water pressure file name begins with
U. The file information can be contoured, plotted, and viewed at each element
Gauss region or approximated at each node.
Page 241
SIGMA/W
The material property (L) files contain the yield conditions, deviator stress, and
material properties at each element integration point. The extension of each
material property file name begins with L. The various material properties can be
contoured, plotted, and viewed at each element Gauss region or approximated at
each node.
While it is useful to know the file formats, it is not necessary to ever access the
data using a text editor. It is possible to have the Contour program present all data
in graphical or tabular format. This is discussed in detail in the following chapter
on Visualization of Results.
Page 242
SIGMA/W
12
Visualization of Results
12.1
Introduction
An attractive feature of SIGMA/W is the varied yet flexible ways that the results
that can be viewed, evaluated and visualized. This is important for any finite
element analysis due to the large amount of data involved. Much of the data can
only be interpreted through graphical visualization using contours and graphs. And
yet at the same time it is often necessary to look at details by examining the digital
data at a particular node or at a Gauss integration point in an element.
This chapter describes and summaries the capabilities available in SIGMA/W for
viewing, evaluating and visualizing the results.
12.2
Load steps
Often it is of interest to look at the difference between two time steps. Say for
example that we want to examine only the excess generated pore-pressures. This
can be done by subtracting the initial static conditions from the results for any time
step or all the time steps.
There is an option to identify and select the load step that should be used for
comparison with all other steps. The results data conditions in this identified file
are subtracted from all the other time increments selected for viewing. If this file
Page 243
SIGMA/W
represents the initial static conditions, then the initial static conditions will be
subtracted from the other time increments selected for viewing. If for example, you
are looking at pore-pressures, the pore-pressures will be the pore-pressures at each
selected time increment less the initial static conditions.
12.3
(F )
xy
Equation
( )
xy
xy = x2 + y2
Fxy = Fx2 + Fy2
Page 244
SIGMA/W
( max )
Maximum Strain
Minimum Strain
Equation
( min )
Volumetric Strain
( v )
Deviatoric Strain
( q )
( max )
max
(
=
min
(
=
+ x )
2
y + x )
( y x ) xy 2
+
+
2
2
( y x ) xy 2
+
2
2
( y x ) xy 2
2
2
max =
v = x + y + z
y ) + ( y z ) + ( z x )
2
1
q =
2 + 3 2
xy
2
x
Page 245
SIGMA/W
( xy )
XY-Force
(F )
xy
Equation
xy = x2 + y2
Fxy = Fx2 + Fy2
Maximum Total
Stress
( max )
max
(
=
min
(
=
( min )
p=
Normal Total Stress
( n )
Tangential Total
Stress
( t )
n =
t =
+x )
2
+x )
2
( y x )
+ xy2
2
x + y
2
( x y )
2
Y-Effective Stress
y' = ( y u )
( x y )
Z-Effective Stress
z' = ( z u )
Maximum Effective
'
max
= max u
Stress
( 'max )
Minimum Effective
Stress
( 'min )
cos 2 + xy sin 2
sin 2 xy cos 2
( ' z )
x' = ( x u )
( ' )
+ y +z )
X-Effective Stress
( 't )
( y x )
+
+ xy2
2
'
min
= min u
Page 246
SIGMA/W
(p )
p' =
'
Normal Effective
Stress
( 'n )
Tangential Effective
Stress
=
'
n
( 't )
t' =
xy max
12.4
+ y +z )
x' + y'
2
( x' y' )
Maximum Shear
Stress
( x' y' )
2
cos 2 + xy sin 2
sin 2 xy cos 2
( y x )
+ xy2 = 1 3
2
xy max =
1 ( x y ) + ( y z ) + ( z x )
q=
2 +6 xy2
2
The Contour viewing node and edge information differs from the similar command
in Define. In Contour, this command displays the boundary pressure applied by
SOLVE along a particular element edge or at a node, while in Define, this
command displays all element edges along which the user has specified a
boundary pressure. In particular, if you have specified a pressure boundary
condition using a function, Contour will show you the pressure applied by SOLVE
at the time step you are viewing, while SOLVE will only show you the function
number.
Figure 12-1 and Figure 12-2 show the process of viewing node and element
information for a node and one of its adjacent Gauss points at the center and base
of a dam after consolidation of the excess pore-water pressure caused by placement
of the fill. You can see that there is a similarity between Gauss values and nodal
values, but that they are not always the same. The reason for this is given in the
next section.
In general, SIGMA/W has many ways to access and view results data; at the core
of these options is the ability to see what is happening at each point in the define
mesh.
Page 247
SIGMA/W
Page 248
SIGMA/W
12.5
x= N
{X }
where:
x
projected value outside the Gauss points at a local coordinate greater than
1. 0
<N>
{X}
The local coordinates at the element nodes are the reciprocal of the Gauss point
local coordinates when forming the element characteristic matrix. Figure 12-3 is an
example of the local coordinates at the element corner nodes when projecting
outwards from the four Gauss points in the element. The value of 1.7320 is the
reciprocal of the Gauss point coordinate 0.57735.
Page 249
SIGMA/W
Figure 12-3 Local coordinates at the corner nodes of an element with four
integration points
This projection technique can result in some over-shoot at the corner nodes when
variation in the Gauss point values is large. For example, consider that we wish to
contour stress in a highly deformed element which, consequently, has large
variations in stresses at the Gauss points. Projecting such large variation in stresses
can result in unrealistic stress at the nodes.
Extreme changes in the parameter values at the Gauss points within an element
often indicate numerical difficulties (the over-shoot at the nodes being just a
symptom of the problem). This over-shoot can potentially be reduced by a finer
mesh discretization. Smaller elements within the same region will result in a
smaller variation of parameter values within each element, therefore lowering the
potential for encountering unrealistic projections.
12.6
Contouring data
There is a long list of computed values that can be contoured, and it generally
includes any of the information that can be viewed at a node as illustrated above.
Figure 12-4, for example, is a contour plot of excess (final minus initial) porepressures in a selected zone in a dam.
Page 250
SIGMA/W
0
1 50
0
50
The contour dialog box, illustrated in Figure 12-5, shows the data range for the
selected parameter. In this example, the pore-pressure range is -418 to 4214 psf.
From this, SIGMA/W computes a starting contour value, a contour interval and the
number of contours. These are default values. Seldom do these default contour
parameters give a nice picture. The problem usually is some numeric noise in the
data at the extremities of the data range. Usually the contour parameters need some
adjustment to produce meaningful and publishable contour plots. The default
values are good for a quick glance at the results, but not adequate for presentation
purposes.
For the illustration in Figure 12-5, the range starts at -418, but the starting value
has been modified to be zero. The number of contours was adjusted to be 10 so the
last contour is at 4500, just greater than the maximum in the data range
It usually takes several iterations to obtain a contour plot that is meaningful and
presents the intended message. The contour coloring and shading scheme is
described in the software on-line help.
Page 251
SIGMA/W
Page 252
SIGMA/W
25
20
0
18
80
60
80
10
160
15
120
Height (metres)
30
180
5
0
-5
-5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
Distance (metres)
12.7
Graphing data
The graphing options in SIGMA/W are very powerful. In most cases, any type of
data you need to access and graph can be done right inside the program. In rare
cases, you can get tabular access to the raw graph data and paste it directly into an
external program such as Excel for additional graphing or manipulation.
At selected nodes, the Draw Graph command allows you to plot a graph containing
any of the computed nodal parameter values listed in Table 12-4.
Page 253
SIGMA/W
X-Boundary Force
X-Displacement
Y-Displacement
Y-Boundary Force
XY-Boundary Force
X-Total Stress
Y-Total Stress
Z-Total Stress
X-Effective Stress
Y-Effective Stress
Z-Effective Stress
Pore-Water Pressure
X-Strain
Y-Strain
Z-Strain
Maximum Strain
Minimum Strain
Volumetric Strain
Deviatoric Strain
Poisson's Ratio
Void Ratio
The parameters in this table are the dependent variables of the graph. Any of the
dependent variables can be plotted versus the following independent variables:
nodal x-coordinates, nodal y-coordinates, or distance between nodes (starting at the
first selected node).
The following independent variables are also available if more than one time
increment is being viewed: time, X-Displacement, Y-Displacement, X-Strain, YStrain, X-Y Shear Strain.
The independent graph variable that you choose affects how the selected nodes and
time steps are used in the graph:
Page 254
SIGMA/W
A new graphing option that is quite powerful in some cases is the option to graph
the sum of all selected nodal dependent variables versus the average of all selected
independent variables. Consider for example that you want to create a loaddeformation graph for designing a footing. In the past, you would have extract
force and displacement data for each node at the base of the footing and then take
the data into Excel to sum the forces and average the displacement. This can be
done automatically now as illustrated in Figure 12-7. In this figure, all the nodes
beneath the footing are selected and then the graphing routine sums the forces and
plots them against the average displacement.
This graphing option can be used in many scenarios throughout the GeoStudio
analysis tools.
Page 255
SIGMA/W
12.8
Contour draws the displacement at each node in the mesh. Most displacements
cannot be drawn un-scaled, since they would be too small to be visible. Specifying
a magnification value allows you to control the scale at which the displacement is
drawn. When you type a value in the magnification edit box, the maximum length
edit box is updated to display the length at which the maximum displacement will
Page 256
SIGMA/W
Magnification=
12.9
A complete picture of the stress state at a point can be viewed with a Mohr circle
diagram such as in Figure 12-8.
Effective Stress at Node 655
40
85.488
30
-24.559
36.421
sx
20
Shear
10
0
95.668
-10
-20
26.241
sy
-30
-40
20
30
40
50
60
70
80
90
100
Normal
Mohr circles can be viewed at nodes or Gauss regions in an element. Mohr circles
can also be generated and viewed in terms of total stresses or effective stresses. In
addition, strain Mohr circles can be viewed as well as stress Mohr circles.
Page 257
SIGMA/W
Page 258
SIGMA/W
13
Illustrative Examples
This chapter presents the analyses of some common stress and deformation
example problems. These examples are presented in order to:
The GeoStudio project files for each problem are included with the SIGMA/W
software. The files can be used to re-run each analysis and to verify that you can
obtain the same results as presented in this chapter.
13.1
Included files
Strip footing.gsz.
Contours of the vertical stress beneath a footing form what is commonly known as
a pressure bulb. The shape of the pressure bulb in a homogeneous linear elastic soil
can be used to verify the SIGMA/W formulation.
In this example, SIGMA/W is used to analyze a strip footing using plane strain
analysis. Drained (effective stress) material parameters are used. They are: E' =
100 kPa and v' = 0.45. In a homogenous elastic material, these parameters will
affect only the displacements, but not the stress distribution. The footing load is
applied as a Y pressure boundary with a pressure of unity (1.0) so that the resulting
contours can be presented as a ratio of the applied pressure.
Figure 13-1 shows the finite element mesh. Note the combination of element sizes.
Near the footing, smaller 8-noded elements are used while further away, the
elements increase in size. The boundary conditions on the bottom and right
indicate zero x- and y-displacement which assumes there is no influence from the
footing at these locations. The left side boundary condition is zero x-displacement
Page 259
SIGMA/W
which will allow vertical deformation only. Figure 13-2 presents the resulting
vertical stress contours.
B/2
1B
Strip Footing
2B
3B
Strip Footing
0.9
2B
0.3
0.
0.
0.7
1B
0.2
3B
Page 260
SIGMA/W
The contours have the correct general shape and are essentially the same as the
values predicted by closed form solution. Table 13-1 shows a comparison between
the position of the vertical stress contours along the center-line beneath the footing
obtained from a closed form solution and from SIGMA/W using the CONTOUR
function. The closed-form solution is obtained from the pressure bulb charts
presented in Lambe and Whitman (1979). The location of the two sets of contours
is similar with the exception of the stress near the bottom boundary of the problem.
Table 13-1 Comparison of SIGMA/W and closed form results for strip footing
13.2
SIGMA/W
0. 8
0. 52 B
0. 56 B
0. 6
0. 88 B
0. 89 B
0. 5
1. 10 B
1. 11 B
0. 4
1. 51 B
1. 48 B
0. 3
2. 25 B
2. 04 B
Included files
Round footing.gsz
In this example, the vertical stress distribution under a circular footing is modeled
using SIGMA/W. The vertical stress contours beneath a circular footing have the
same general shape as for a strip footing. However, the size of the pressure bulb is
much smaller.
Reflecting the circular geometry of the footing, this problem is modeled using an
axisymmetric view. The foundation material is considered to be homogenous
linear elastic with E' = 100 kPa and ' = 0. 45. In this material, the choice of these
parameters will affect only the displacement results, but not the stress distribution.
The footing load is applied as a y-pressure boundary with a downward pressure of
1.0 kPa. The details of the problem may be obtained by viewing the input file
using the DEFINE function.
Figure 13-3 shows the SIGMA/W computed vertical stresses beneath a circular
(round) footing.
Page 261
SIGMA/W
B/2
0.9
1B
2
0.
1
0.
2B
Figure 13-3 Vertical stress contours beneath a round footing
13.3
SIGMA/W
0. 9
0. 28 B
0. 26 B
0. 7
0. 43 B
0. 45 B
0. 5
0. 63 B
0. 64 B
0. 3
0. 94 B
0. 97 B
0. 2
1. 20 B
1. 26 B
Included files
Hyperbolic_i.gsz
Page 262
SIGMA/W
Hyperbolic.gsz
Value
cohesive strength, c
1000 psf
Page 263
SIGMA/W
47
loading exponent, n
47
failure ratio, Rf
0.9
Poissons ratio,
0.48
Minimum stress, S3
200
For an explanation of these parameters, see the non-linear elastic section of chapter
on Material Properties. In this example, the stiffness of the clay is assumed to be
independent of the confining stress (i.e., depth). This is simulated by setting the
hyperbolic exponent, n, to zero. The Poisson's ratio is assumed to remain constant
as the footing load is applied. The clay has an un-drained strength of 1000 psf; that
is, c equals 1000, and equals zero.
The initial tangent modulus applied in the analysis is 1x105 psf, and the
atmospheric pressure is set to 2116 psf. In this example, the initial soil stiffness is
assumed to be independent of the confining stress 3; that is, the exponent n is 0.
Consequently, Ei is equal to KL multiplied by 2116 and the modulus number KL is
47. In equation form:
A minimum allowable stress value, S3, of 200 has been applied to the model so
that there is some stiffness to the soil when the confining stresses approach zero.
This reduces convergence problems because it prevents the soil stiffness from
behaving like fluid clay which we do not anticipate in this case.
Before proceeding with the actual settlement analysis, the in-situ stress state needs
to be defined. The in-situ stresses are necessary when the soil stiffness under
loading is a function of the confining stress present in the ground prior to loading.
This is the case for the Hyperbolic soil model.
Page 264
SIGMA/W
In-situ stresses
A SIGMA/W in-situ analysis can be used to establish the initial stress conditions.
The analysis type is set to In-situ 1 because we have a horizontal ground surface
and we are not using submerged unit weight. It is assumed that the clay is uniform,
its bulk unit weight is 110 pcf, and the at rest earth pressure coefficient, K0, is 0.
92.
The self-weight of the footing is ignored in the in-situ analysis because it does not
exist at the time the stresses are developed in the in-situ soil. This is achieved by
removing the footing elements from the finite element grid either by setting the
footing material to be a null element, or by assigning its self-weight to zero.
Examination of the computed results confirms that, as expected, the vertical total
stress at a point is equal to the self-weight times the depth. In equation form:
where:
d
After the initial stresses are established and verified using hand calculations if
possible, a settlement analysis can be carried out using the non-linear elastic
model.
When developing the finite element model for this settlement analysis, the footpound (force) unit system is used. Finite elements representing the rigid concrete
footing are simulated as linear elastic materials with a very stiff modulus. The
Page 265
SIGMA/W
0
-1
Y-Stress (x 1000)
-2
-3
-4
-5
-6
-7
-8
-9
0
10
15
20
Time
When solving SIGMA/W, we set the output stress file from the in-situ analysis as
the initial condition file for stresses.
After solving, SIGMA/W Contour is used to process the results of the analysis. At
any given time step, nodal displacements at individual nodes can be obtained using
the View Node Information command. A nodal displacements graph as a function
of time can also be generated. In this case, it is useful create a total load versus ydisplacement graph for all nodes at the base of the footing. The graph Sum versus
Average option is selected so that an average footing displacement versus total
loading function can be obtained. In previous versions of SIGMA/W, this data
would have to be gathered into Excel and plotted there.
Page 266
SIGMA/W
Figure 13-6 shows a plot of the total applied load versus the average vertical
displacement. The shape of this load-displacement curve clearly shows the nonlinear nature of the problem.
Y-Boundary Force
-20000
-40000
-60000
-80000
-2.5
-2.0
-1.5
-1.0
-0.5
0.0
Y-Displacement
The ultimate footing pressure can be evaluated in terms of the Terzaghi bearing
capacity formula for a circular footing (Terzaghi, 1943). For a round footing with a
rough base, the ultimate bearing capacity is given as:
This approximate value is somewhat higher than the 6000 psf "ultimate" load
applied using the boundary stress function in the SIGMA/W analysis.
Page 267
SIGMA/W
Figure 13-7 shows the tangent modulii at the end of the loading. The initial
modulus E is about 100,000 psf. Note how under the footing the tangent modulus
has decreased to below 20,000.
80000
60
00
40
00
0
00
20
When modeled as a non-linear elastic material, a soil is considered past the yield
point when the stress state is within Rf of the soil strength. This consideration
provides an arbitrary means of showing zones that are stressed close to the soil
strength. Figure 13-8 shows the "yield zone" at Load Step 11. For a rigid footing,
this analysis indicates that the soil has not "yielded" below the center of the
footing. This is consistent with the known stress distribution beneath a rigid
footing.
This figure also shows contours of maximum shear stress. In this example, the
shear strength of the soil is 1000 psf, and Rf is 0.9, making the maximum 900 psf.
As shown in the figure, the 900 psf contour is at the edges of the yellow area which
is the area where the strength has reached 900 psf.
Page 268
SIGMA/W
900
70
0
0
90
30
0
0
50
Figure 13-8 Maximum shear stress contour values and yield zone (step 11)
It should also be noted that the shear stress contours are drawn only for the
foundation material. In this example, the footing material is much stiffer than the
foundation material. Consequently, the stress distribution in the two zones are also
very different. To present a true picture of the stress distribution, only the finite
elements in the foundation are selected when plotting the shear stress contours by
using the View Element Regions command in CONTOUR.
13.4
Elastic-plastic material
Included files
EP undrained.gsz
Page 269
SIGMA/W
For this analysis, only the displacement convergence criterion is specified and is
set to 1%. The footing load is applied as normal pressure along element edges
using a boundary function. Computation is continued until Step 35 when it is no
longer possible to meet the convergence criterion. It is important to watch the
convergence information during solving of non-linear soil models because it is a
sign that the soil is approaching a failure condition. Once past this point, the
solutions are not reliable. The reasons for this are discussed in the section of
Chapter 3 called Stability and Serviceability.
Page 270
SIGMA/W
Figure 13-10 shows vertical displacement under center of the footing as a function
of the time step number. The geometry of the example is similar to the previous
footing examples. The maximum displacement at this point is 0.091370m or
91.37mm. This being a flexible footing, the displacement, of course, varies under
the footing. At the outer edge the displacement is only 25.729 mm. With the
application of the last load, the displacement is beginning to increase rapidly
indicating the load is near the ultimate collapse load.
Y-Displacement
-0.02
-0.04
-0.06
-0.08
-0.10
0
10
20
30
40
Time
Figure 13-10 Settlement (meters) under center of footing
Page 271
SIGMA/W
Y-Boundary Force
-500
-1000
-1500
-0.08
-0.06
-0.04
-0.02
0.00
Y-Displacement
Figure 13-11 is the load versus deformation plot for this footing. In this example,
the maximum load at the end is about 1050 kN.
For a footing with a smooth base, which is the case in this problem, the Terzaghi
(1943) formula gives an ultimate bearing capacity of 5.14c (=514 kPa in this
problem). The footing is 2m wide so the ultimate load would be 1020 kN. This is
nearly identical to the SIGMA/W value of 1050 kN obtained for the last converged
solution.
Figure 13-12 shows the contours of deviatoric stress q and the yield zone at Load
Step 21. The deviatoric stress at yield for the Tresca yield criterion is given by:
As one would expect, the yield zone in this figure is approximately bounded by the
170 kPa contour.
Page 272
SIGMA/W
This example illustrates that the elastic-plastic model gives reasonable results for a
total stress, un-drained analysis (e.g., when the friction angle phi is zero). The next
example looks at the case of a frictional material.
Bearing capacity of a purely frictional material
Included files
EP Frictional.gsz
EP Frictional_i.gsz
This example presents an analysis of a strip footing on a purely frictional soil (e.g.,
cohesion is zero and phi is some value greater than zero). This could be a sand soil
type, for example, and it is modeled in this case as an elastic-plastic material. The
Page 273
SIGMA/W
objective here is to show the nonlinearity of a load deformation curve and discuss
some considerations unique to this type of analysis. A secondary objective is to
compare the SIGMA/W results with a closed-form bearing capacity solution.
In this problem the soil friction angle (phi) is set to 30 degrees and the cohesion is
set to zero. Two different cases regarding the soil dilation angle are considered.
One is with a dilation angle of 30 degrees (the same as the friction angle) and the
other is with a dilation angle of zero. When the dilation angle is equal to the
friction angle, an associated flow rule is used in the analysis. When these two
parameters are not equal, the flow rule is said to be non-associated.
For all non-linear constitutive soil models for frictional soils, the soil properties
depend on the stress state in the ground. In general, higher confining stresses mean
the soil is stronger and stiffer. For an analysis such as this, it is therefore necessary
to establish the in-situ stress conditions before applying the footing load. This
needs to be done as a separate preliminary step. The results from this in-situ
analysis will become the initial conditions for the subsequent loading part of the
analysis.
In-situ analysis
The in-situ stresses are established by first running an in-situ analysis. An In-situ 1
type of analysis is suitable here since the ground surface is horizontal. The unit
self-weight is 20 kN/m3, and a coefficient of soil pressure at rest, Ko, of 0.67 is
assumed (this is analogous to a Poissons ratio of 0.4).
It is always good practice to verify that the correct starting stresses have been
established by inspecting the stresses along a vertical profile through the section
with the Draw Graph command. The most crucial variable to check is that the
correct vertical and horizontal effective stresses have been established.
The finite element mesh for these analyses is shown in Figure 13-13. The mesh is
made up of six soil regions and each region is automatically meshed with
unstructured elements. It is worth noting the types of elements used. Experience
has shown that triangular elements with constant stress gradients give the most
reasonable results from a computation perspective when the frictional elasticplastic model is used. Using the automatic meshing feature with full control of
mesh density allows you to concentrate the density of elements where the stresses
are most likely to reach peak values.
Page 274
SIGMA/W
10
11
12
Figure 13-13 Finite element mesh for footing analysis on a purely frictional
material
After the initial stresses are established, the problem can be saved as a new name
and the loading analysis can be carried out.
For the loading part of the analysis the body load needs to be set to zero, the
analysis type needs to be set to a Load Deformation Analysis, and the footing
pressures need to be defined and applied. The details can be viewed by opening the
respective files to see the parameters used in the analysis.
In this case, the load is applied in increments of 10 kPa.. Since SIGMA/W is based
on an incremental formulation, the fixed action value of 10 kPa will be applied
each time step and accumulated during the loading sequence. There will be 88 load
steps which will result in a final applied pressure of 880 kPa. The results are saved
every fifth step plus the 88th step.
A structural element has been added across footing. This structural element is
stiffer than the soil (E = 2.4e8 kPa, I = 3.5e-5 m4, cross sectional area = 2e-2 m2 )
Page 275
SIGMA/W
and helps reduce excessive uneven footing deformations by pushing the footing
nodes down more evenly.
Figure 13-14 shows the extent of the yield zone at the end of the loading. The yield
zone under the footing is reasonable as one would expect from pressure bulb
shapes below a footing load and from bearing capacity theory. While not
illustrated in this figure, the yield zone just below the ground surface and well to
the right of the footing is likely not consistent with reality. The reason for the
indicated yield in this area is the low confining stress. The soil has no cohesion,
and consequently as the confining stress approaches zero, the elastic-plastic
relationship indicates yielding even under small changes in shear stress or if there
is a tendency for some tensile stresses. This side effect can often be reduced by
giving the soil a small cohesion which is likely more representative of actual field
conditions.
Figure 13-15 shows the load versus deformation curve generated in the Contour
program. Here we see a maximum load force on the entire footing of about 1700
Page 276
SIGMA/W
kN. If we divide this value by the 2m square surface area of the footing, we have a
bearing capacity on the footing of about 850 kPa. From closed form solutions the
ultimate bearing capacity for a four-meter wide strip footing is 800 kPa (based on a
bearing capacity coefficient of 20). The SIGMA/W computed results have
converged up to and slightly past this theoretical maximum bearing capacity.
Y-Boundary Force
-500
-1000
-1500
-2000
-0.10
-0.08
-0.06
-0.04
-0.02
0.00
Y-Displacement
Figure 13-15 Load versus displacement curve
One thing to note is the curvature of the load-deformation curves. They start out on
a straight line representing linear-elastic conditions, but then begin to bend over
and deviate from the linear elastic condition as they should. These data indicate
that the SIGMA/W formulation provides acceptable results. The trends are
reasonable and logically consistent with a traditional text-book understanding of
soil behavior, and the ultimate loads available are consistent with expectations
from bearing capacity formula.
Page 277
13.5
SIGMA/W
Strain-softening analysis
Included files
StrainSoft peak.gsz
StrainSoft residual.gsz
StrainSoft.gsz
Three different analyses are carried out. First, the residual strength is set equal to
the peak strength (100 kPa in this example) to obtain the load/displacement curve
for the case with no softening. This in effect makes the analysis similar to an
elastic-plastic analysis. Second, the analysis is repeated with the peak equal to the
Page 278
SIGMA/W
residual shear strength (50 kPa in this example). Again, the analysis will behave
elastic-plastic, but in this way, it is possible to bracket the range of possible
solutions for the actual strain-softening simulation. The analysis is then repeated a
third time using the strain-softening parameters.
Results of these analyses are presented in Figure 13-17. The upper and lower
curves are essentially elastic-plastic analyses using the von Mises failure criterion.
The center curve is the result for the strain-softening analysis case. It is noted that
the maximum load is between the peak and the residual. The footing pressure is
calculated by summing the Y-boundary force at the footing nodes and then
dividing this sum by the width of the modeled footing (i.e. 2 m). Both Y
displacement and Y boundary force data is extracted by CONTOUR using the
Draw Graph command. The data is then charted using Excel and the graph is
imported into this document.
The pressure versus displacement results obtained are similar to results obtained by
Chan (1986), who showed that the load-displacement curve for a footing on a
Page 279
SIGMA/W
strain-softening clay reaches a maximum between the peak and residual values and
remains fairly constant until the displacement becomes very large. The pressuredisplacement curve eventually bends down towards the residual line, but only after
significant footing displacement. In this example, the displacement is
approximately 0.8 m for a 4.0 m wide footing.
According to the closed-form solution by Prandtl (Smith & Griffiths, 1988), the
ultimate bearing capacity is given by:
The ultimate footing pressures of 257 and 514 kPa are calculated for cohesive
strength values of 50 and 100 kPa respectively. When peak and residual values are
used without invoking softening, SIGMA/W computes ultimate values which are 8
to 10% higher than the closed-form solution.
Another point worth noting is that it takes a very large number of small
displacements to achieve the curve for the strain-softening material. Each solid
black square on the curve represents a displacement increment. In addition, the
stress distributions under the footing are quite irregular. However, when we take
the nodal boundary forces along the footing and sum them for the entire footing
load, the results are reasonably good as in the above figure. Stated another way,
overall the results are reasonable, but locally at individual points they can be fairly
erratic.
13.6
Included Files
CamClay ini.gsz
CamClay OCR1.gsz
CamClay OCR5.gsz
This example simulates a triaxial test on Cam-clay soil. The main purpose here is
to verify the SIGMA/W formulation.
The analysis simulates a consolidated-undrained test; that is, the sample is
consolidated by applying a confining pressure and then loaded without any pore-
Page 280
SIGMA/W
The first step is to establish the consolidated stress state. This is done by applying
a normal pressure of 100 kPa on the top and right side of the mesh as shown in
Figure 13-18 (arrows towards the mesh). The left side of the mesh is the symmetric
axis.
Note the boundary conditions on the left and bottom which represent rollers.
Page 281
SIGMA/W
1.0
0.8
metres
0.6
0.4
0.2
0.0
0
200
400
600
metres (x 0.001)
Figure 13-18 Triaxial test mesh
The mesh consists of only four elements. They are higher order 8-noded elements
with 9-point integration. Four elements are sufficient for a test like this since the
stress is uniform throughout.
At the end of consolidation there is in essence no excess pore-pressure and there is
no significant hydrostatic pore-pressure. For this numerical analysis the initial
pore-pressures therefore can be zero.
This part of the analysis induces a stress of 100 kPa in each element and becomes
the initial conditions for the loading phase.
Hydraulic conditions
Page 282
SIGMA/W
An important issue in an analysis using Cam-clay is the initial yield surface. The
existing or pre-load yield surface must be defined. SIGMA/W has two methods of
doing this. The Over Consolidation Ratio (OCR) option is used here.
The initial yield surface is computed from the following stress states:
y max = y OCR
x max =
y max
(1 sin )
The results for a test with OCR equal to 1.0 (normally consolidated) are
summarized in Figure 13-19. As it properly should be, the total stress path is a
straight line at a slope of 3V:1H. The effective stress path is vertical until it reaches
Page 283
SIGMA/W
the initial yield locus. The effective stress then gradually moves up and to the left
until it arrives at the critical state line. The value of deviatoric stress parameter (q)
at the critical state is computed by SIGMA/W to be 54.1 kPa. The corresponding
hand calculated value is 54.6 kPa. Figure 13-19 was plotted using Excel using data
extracted using the Graph in CONTOUR. The initial yield locus was calculated
using Excel.
The effective stress path can be viewed directly in SIGMA/W as shown in Figure
13-20. The critical state line in this case is a 45-degree line which passes through
the end-point of the stress path.
Page 284
SIGMA/W
110
100
90
80
70
60
50
40
30
20
10
0
0
10
20
30
40
50
60
70
80
90
100
110
Two other plots of interest are presented in Figure 13-21 and Figure 13-22. Both
the deviatoric stress q and the pore-water pressure u reach a maximum and then
remain essentially constant as they should for a normally consolidated soil.
60
50
40
30
20
10
0
0.00
0.02
0.04
0.06
0.08
Y-Strain
Page 285
SIGMA/W
Pore-Water Pressure
80
60
40
20
0
0.00
0.02
0.04
0.06
0.08
Y-Strain
Figure 13-22 Pore-pressure versus vertical strain
Over-consolidated test
Page 286
SIGMA/W
The shear stress q reaches a maximum of 209.7 kPa and then drops slightly. The
pore-water pressure increases until the effective stress reaches the initial yield
locus. Then it decreases to zero and becomes negative as the effective stress path
moves towards the critical state line. All of these responses are in accordance with
the actual behavior of over-consolidated clay.
250
200
150
100
50
0
0
50
100
150
200
250
Page 287
SIGMA/W
250
200
150
100
50
0
0.00
0.05
0.10
0.15
Y-Strain
Pore-Water Pressure
50
40
30
20
10
0
-10
-20
-30
-40
0.00
0.05
0.10
0.15
Y-Strain
Page 288
SIGMA/W
13.7
Modified Cam-clay
Included files
MCamClay ini.gsz
MCamClay OCR5.gsz
This example is essentially a repeat of the previous one except that the triaxial tests
simulate a Modified Cam-clay model. The two models are of course very similar
except for the shape of the yield surface. The initial yield locus for the Modified
Cam-clay is elliptical.
Much of the preliminary discussion on setting up initial conditions and so forth are
the same and are consequently not repeated here.
Modified Cam-clay results
A summary of the results for this simulation of a triaxial test based on the
Modified Cam-clay model are presented in the following four figures. The results
have the shape and form as they should.
Also, the point at which the effective stress path meets the critical state line is in
good agreement with the theoretical values as summarized in Table 13-4.
Table 13-4 Intersection of effective stress path and critical state line
Test
Theoretical
(kPa)
SIGMA/W
(kPa)
OCR 1
65.9
66.7
OCR 5
209.1
206.1
Page 289
SIGMA/W
Page 290
SIGMA/W
250
80
70
60
50
40
30
20
200
150
100
50
10
0
0
10
20
30
40
50
60
70
80
90
100
110
50
100
150
200
250
0.20
0.25
60
80
50
60
Pore-Water Pressure
Pore-Water Pressure
40
30
20
10
40
20
0
-20
0
0.00
0.05
0.10
0.15
0.20
-40
0.00
Y-Strain
0.05
0.10
0.15
Y-Strain
13.8
Included files
Anisotropic Tunnel.gsz
Anisotropic Tunnel_i.gsz
Page 291
SIGMA/W
the elements to be excavated at specific load steps are identified using the Draw
Fill/Excavation Elements command in SIGMA/W Define. The SIGMA/W solver
then calculates the nodal force induced by the stresses inside the elements to be
excavated and applies an equal, but opposite nodal force along the excavation face.
An initial stress state should be specified if elements are to be excavated at the
starting step of an analysis. It is very important to make sure that the initial stresses
are correct. In particular, make sure the initial total stress has any added pore-water
pressure effects prior to excavating. A discussion of the reasons for this is given in
the chapter on Fill and Excavation.
This example shows a tunnel excavation in stratified rock. The rock strata is
assumed to be inclined at 30 to the horizontal. The longitudinal modulus is set to
60 GPa and the transverse modulus is set to 20 GPa. The stratified rock is modeled
using the anisotropic elastic material model.
The first step in an excavation analysis is to establish the in-situ stress conditions.
For the in-situ analysis, the self-weight of the rock is assumed to be 25 kN/m3 and
the coefficient of earth pressure at rest, Ko, is 1. Therefore, the initial stress state is
isotropic. The in-situ stress analysis output file is used as the initial stress file for
the excavation analysis.
In this example, the excavation is carried out in four steps. The number of time
steps is specified using the KeyIn Time Increments command in SIGMA/W.
Elements to be excavated are identified using the Draw Fill/Excavation Elements
command and they appear on screen with an integer number in each element that
shows the load step the element is to be activated (fill) or removed (excavated).
After choosing element type as excavation and entering the time step to excavate,
the desired elements can be selected by clicking and dragging the mouse. By
default, all elements are given a step number of 0, indicating that they will always
be included in an analysis.
The View Preferences command can be used to show the excavation sequence at
each time step. Elements to be excavated are identified with a negative step
number. Figure 13-31 shows the tunnel part the finite element grid used in the
excavation example where the excavation is carried out in four steps. The back
(top portion) of the tunnel is excavated in the first load step, followed by the
remainder of the tunnel in subsequent steps.
Page 292
SIGMA/W
0
0 0
0
0
0
0
0
0
0
0
-3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3
0
0
0
0
0
0
0
0
0 0
0
0
0
0
0
-3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3
0
-3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3
0
0
-3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3
-3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3
0
0
0
0
-4-4
-4-4
-4-4
-4-4
0
0
-4 -4
-4 -4
0
0
00
-4 -4
-4 -4
00
-4
-4
-4
-4
0
0
0
0
0
0
0
0
0
0
0 0
0
0
0
0 0
0
0
0 0
0
0
0
0 0
0
0
0
0
0
0
0 0
0
0
0
0
0 0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
-1 -1 -1 -1 -1
0
-1 -1
-1 -1
0
-1 -1
-1-1
0
-1-1
-1-1
-1-1
-1
0
-2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2
-2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2
0
-2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2
0
-2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2
0
-2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 -2
0
0
0
0
0
0
0
0
The deformed mesh after complete excavation is shown in Figure 13-32. After
excavation, the circular opening is deformed into an ellipse with its axis coinciding
with the orientation of the anisotropic strata.
Page 293
13.9
SIGMA/W
Included files
Embankment construction.gsz
Embankment consolidation.gsz
Page 294
SIGMA/W
40
Embankment Construction
35
+6
+5
+6
+6
+5
+5
+6
+5
+4
+4
+4
+4
+5
+4
+3
+3
+3
+3
+3
+3
+4
+3
+2
+2
+2
+2
+2
+2
+2
+2
+3
+2
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+2
+1
+1
Height (metres)
30
+3
+2
25
20
15
+1
+2
+1
+4
+3
+5
+4
10
5
0
-5
-5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
Distance (metres)
SIGMA/W can optionally adjust the weight of each lift to compensate for
settlement from the placement of previous lifts. You can select or disable this
option using the analysis settings command.
40
1.0000e+000
30
2.0000e+000
3.0000e+000
20
4.0000e+000
10
5.0000e+000
0
-0.30
6.0000e+000
-0.25
-0.20
-0.15
-0.10
-0.05
0.00
Y-Displacement
Figure 13-34 Vertical displacement vesus elevation for each load step
In this example, the material properties have been defined using total stress
parameters. If A and / or B pore-water pressure parameters are also specified, it is
Page 295
SIGMA/W
The previous analysis outlines the construction phase of the embankment and the
build up of excess pore-water pressure in the foundation soil. It is now useful to
predict the amount of volume change that will occur during consolidation (as
excess pore-pressures are dissipated).
The consolidation volume change can be estimated using the un-coupled
consolidation analysis option in SIGMA/W where the start and ending pore-water
pressure conditions are specified. No estimate of the time to reach a fully
consolidated state is made using this option. If that information is necessary, a
coupled consolidation analysis could be carried out using SEEP/W. The ability to
estimate the volume change can be done with pore-pressure information generated
solely by SIGMA/W.
Since we are interested in estimating the volume change due to dissipation of the
excess pore-water pressure, we are in effect saying that our ending pore-water
Page 296
SIGMA/W
pressure condition is the same as the condition prior to placing any fill. We also
know that the starting pressure condition is that which exists at the end of the
placement of the last lift. We can obtain both of these conditions directly from the
SIGMA/W project file.
After completing the embankment construction analysis, the file was saved as a
new name, the analysis type changed to un-coupled consolidation, the body load
was turned off, all fill elements were set to permanent, the infinite elements were
set to regular elements, and the load steps set to one. When selecting the pressure
conditions, the ending conditions as PWP=0 and the starting condition as the load
step 6 pressure condition from the construction project file. The PWP=0 option is
valid because we are only interested in volume change due to excess pressure
dissipation and we did not establish any pore-water pressure condition in the
ground prior to placing any fill. So, the u values needed by the solver will be the
excess pressure minus 0.
Figure 13-36 shows the volume change computed by the un-coupled consolidation
analysis where the initial and final pore-water pressure data was obtained from a
previously solved SIGMA/W analysis.
Embankment Consolidation
Vertical settlement: 6.8 cm
40
35
Height (metres)
30
25
20
15
10
5
0
-5
-5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
Distance (metres)
Page 297
SIGMA/W
This simple example illustrates using a truss element to simulate a crossexcavation brace or strut.
The first step is to do an analysis to establish the in-situ stresses. In a finite element
analysis, an excavation is simulated by applying forces on the excavation face
which represent the stresses that were in the ground before the excavation is made.
To know what forces to apply on the excavation face, we need to know the total
stresses in the ground prior to the excavation. This is the reason for first
establishing the in-situ stresses. This is done as a separate analysis, and the results
become the initial conditions for the excavation part of the problem.
26
24
22
20
18
16
14
12
10
8
6
4
2
0
-2
-5
10
15
20
25
30
35
40
45
50
55
60
65
70
Figure 13-37 Total in-situ Y stress contours and mesh used in strut example
Figure 13-37 shows the computed total vertical stress for the in-situ conditions.
Also shown is the finite element mesh used in the analysis. It is important to note
the use of both structured and unstructured finite elements. The region to be
excavated is made of a structured mesh which is very advantageous because we
want to excavate in even lifts. The remaining mesh can be structured or
unstructured. Also important to note is that even while the problem appears
symmetrical, the entire region is modeled. This is necessary because once the strut
becomes active it must be attached at both sides. If only half the problem was
modeled, the mid point of the strut would not have a valid boundary condition
attached to it because it would be adjacent to Null element nodes.
For this simple example, we'll select some arbitrary values for the brace properties.
E is 1 x 108 kPa and the cross-sectional area is 1 x 10-3 m2. The brace is 10 m long
and there is no pre-force in the member.
Page 298
SIGMA/W
The problem is set up to excavate the trench in five steps with the strut to become
active on the second step. Figure 13-38 shows the deformed mesh at the end of the
excavation with the strut in place. The deformed mesh is shown at 5x
magnification. It is clear that the strut is holding back deformation into the
opening.
Excavation with a Horizontal Strut
Strut
26
24
22
20
18
16
14
12
10
8
6
4
2
0
-2
-5
10
15
20
25
30
35
40
45
50
55
60
65
70
metres
Figure 13-39 shows the horizontal displacement of the left side of the excavation
during the sequence of excavating. The strut becomes active at step two and
virtually limits all further lateral displacement from this time onwards. Figure
13-40 shows the x boundary force at the end of the strut and it confirms a force
was added at step two. The force is not present at step 3 or 4 etc. because it was
already added as a boundary force once and should not be added again.
Page 299
SIGMA/W
25
0.0000e+000
1.0000e+000
20
Y
2.0000e+000
3.0000e+000
15
4.0000e+000
5.0000e+000
10
-0.02
0.00
0.02
0.04
0.06
0.08
0.10
X-Displacement
X-Boundary Force
25
20
15
10
0
0
Time
Page 300
SIGMA/W
Cryers Ball.gsz
One of the classic examples for the consolidation process is Cryers problem in
which a sphere of saturated clay is subjected to a sudden increase in all-round
confining pressure. Drainage is permitted on the surface of the sphere. Gibson et
al., (1963) have graphed the analytical solution to this problem showing the
variation of the pore-water pressure at the center of the sphere with time for
various Poissons ratios.
A fully coupled consolidation analysis of Cryers problem can be done with
SIGMA/W and the results can be compared with the published information.
Figure 13-41 shows the problem as defined for the SIGMA/W analysis. It is an
axisymmetric analysis and therefore represents a sphere. The sphere is 1 m in
diameter.
The coefficient of volume compressibility mv is 1 x 10-4 kPa-1. The problem is
analyzed as saturated only. This means that E is computed from mv.
E = 1/mv = 1 / 1 x 10-4 = 10,000 kPa.
Poissons ratio v is set to 0.334.
A normal pressure of 100 kPa is applied to the surface of the sphere. The soil is
free to move in the vertical direction along the central vertical axis and is free to
move horizontally along the horizontal central axis.
The hydraulic conductivity is 4.91 x 10-5 m/sec.
The mesh is an unstructured mesh with 6-noded triangular elements. The mesh has
two regions. The outer curved edge of the region has 23 Points. The x-y
coordinates of these points where computed in EXCEL and then pasted into the
Point dialog box in SIGMA/W.
Page 301
SIGMA/W
Figure 13-42 shows the pore-pressure response and dissipation at the center of the
sphere. The point of interest is that the pore-pressure rises above the applied
surface pressure and then dissipates. The peak is 118 kPa.
The effect is less pronounced towards the outer surface, but nonetheless exists as
shown in Figure 13-43. The peak is 106, but disappears very quickly.
Page 302
SIGMA/W
Pore-Water Pressure
150
100
50
0
0
10
Time
Pore-Water Pressure
150
100
50
0
0.0
0.1
0.2
0.3
0.4
Time
Page 303
SIGMA/W
In Figure 13-44 the pore-pressure is normalized to the initial applied pressure and
T is the usual consolidation related time factor defined as:
T=
cv t
kE (1 )
t
=
2
w (1 + )(1 2 ) r 2
r
This example helps to confirm that this aspect of SIGMA/W has been formulated
and coded correctly.
Consolidation 1D.gsz
Page 304
SIGMA/W
1.0
0.9
0.8
0.7
Metres
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
A 1000 kPa load is applied in Step 1 and then the excess pore-pressure is allowed
to dissipate. The load is applied as a boundary function (Figure 13-46). Recall that
SIGMA/W is an incremental formulation and loads are therefore applied as
increments. The change in the function between Time 0.0 and 1.0 is 1000. For all
other time steps the difference is zero and therefore no further load is applied.
Page 305
SIGMA/W
0.0
Y-Stress (x 1000)
-0.2
-0.4
-0.6
-0.8
-1.0
0
Time
E=
D (1 + )(1 2 )
(1 )
2
D
3
t =
w mv A
4 K sat
A is the area. The gradient will be the highest during the first time step at the first
element at the surface. Substituting into this equation gives:
Page 306
SIGMA/W
t =
w mv A 10 (1 e 3) 0.05 0.5
4 K sat
4 (5 e 6)
= 12.5
So a starting time step around 10 minutes will give a reasonable solution. Through
several runs it was learned that an initial time step of 5 minutes actually gives nice
results. This estimation formula is only approximate, but very useful to get within
the right order of magnitude. Time steps that are too small can lead to jagged
unrealistic pore-pressure dissipation curves near the surface. Often several trial
runs are required to establish an appropriate time stepping sequence.
The time steps increase exponentially up to a total elapsed time of 5115 minutes
(about 85 hours).
The dissipation of the excess pore-pressure is shown in Figure 13-47. The porepressure increase is 1000 kPa which is equal to the applied load. There is small
dissipation in the upper 0.2 m (20 cm) during the first load step. The dissipation is
also reflected in the vertical strain as illustrated in Figure 13-48.
1.0
0.0000e+000
5.0000e+000
0.8
1.5000e+001
3.5000e+001
0.6
7.5000e+001
1.5500e+002
0.4
3.1500e+002
6.3500e+002
0.2
1.2750e+003
2.5550e+003
5.1150e+003
0.0
0
200
400
600
800
1000
1200
Pore-Water Pressure
Page 307
SIGMA/W
1.0
0.8
0.6
0.4
0.2
0.0
0.0
0.2
0.4
0.6
0.8
1.0
Y-Strain
The same results can be obtained from a SEEP/W analysis if the only interest is to
dissipate the excess pore-pressure. Say that some kind of an analysis has resulted
in an induced excess pore-pressure of 1000 kPa. The excess pressure can be
dissipated in SEEP/W, and all else being the same, the time to dissipate the excess
pressure is the same as from a coupled analysis. The graph Figure 13-49 is from a
SEEP/W analysis and is identical to the coupled results in Figure 13-47. In the
SEEP/W analysis the excess pore-pressure comes from another analysis while in
the coupled analysis the excess pore-pressure comes from the applied load.
Page 308
SIGMA/W
1.0
0.0000e+000
5.0000e+000
0.8
1.5000e+001
3.5000e+001
0.6
Y
7.5000e+001
1.5500e+002
0.4
3.1500e+002
6.3500e+002
0.2
1.2750e+003
2.5550e+003
5.1150e+003
0.0
0
200
400
600
800
1000
1200
Pressure
Page 309
SIGMA/W
100
90
80
70
60
50
40
30
1.0
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2.0
E/R ratio
An R/E ratio (reciprocal of E/R) is somewhat analogous to the Skempton porepressure parameter B. An R/E of 0.9 is roughly equivalent to a B value of 0.9.
Controlling the stiffness of the soil grain structure (E) relative to the water stiffness
(mv) makes it possible to compute a pore-pressure response relative to the loading.
Sat-Unsat Consolidation.gsz
The consolidation component in SIGMA/W is formulated for general saturatedunsaturated consolidation. This example looks at a simple 1D problem that has
both a saturated and an unsaturated zone to illustrate the issues and concepts.
SIGMA/W is based on the formulation published by Wong, Fredlund and Krahn
(1998).
A key relationship in the saturated-unsaturated formulation is:
w = v ( ua uw )
Page 310
SIGMA/W
where w is the volumetric water content, v is the volumetric strain and (ua-uw) is
the soil suction or negative pore-pressure. The volumetric water content w can
change as a result of strain and or a change in suction. The constants and are
related to the soil stiffness and the degree of saturation.
If the soil is saturated, is zero and is 1.0. The change in volumetric water
content then is equal to the volumetric strain. In equation form:
w = v
Saturated consolidation analyses can consequently be performed through the
definition and control of and .
Problem description
This example consists of a 2-m high column (Figure 13-51). The water table is at
mid-height and the initial pore-pressure is disturbed hydrostatically both above and
below water table.
The column is free to move up and down, but is restrained laterally. This is like a
conventional consolidation test in a steel ring.
A 10 kPa load is applied to the surface. This is done through a boundary function
so that the load is applied only in Step 1.
This is like an undrained test and the hydraulic boundary condition around the
perimeter is consequently a no flow boundary (nodal flux is zero; default value).
Page 311
SIGMA/W
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0.0
0.2
0.4
0.6
0.8
1.0
Material properties
E=
D (1 + )(1 2 )
(1 )
2
D
3
Page 312
SIGMA/W
200
180
160
140
120
100
-40
-30
-20
-10
Pressure
Figure 13-53 shows the pore-pressure immediately after applying the surface
pressure of 10 kPa. Below the water table in the saturated zone the response is 100
percent; that is the pore-pressure increase is equal 10 kPa. In the unsaturated zone
above the water table the response is much less; towards the top there is almost no
response.
One way of looking at this is that below the water table the load is taken up by the
water since water is near incompressible; or the water is much stiffer than the soil
grain structure. At the top the soil grain structure absorbs the load since the voids
are not full of water. Just above the water table there is a smooth transition from
the water absorbing the load to the soil grain structure absorbing the load.
Page 313
SIGMA/W
2.0
1.5
1.0
0.5
0.0
-15
-10
-5
10
15
20
1.5
3.0000e+000
7.0000e+000
1.5000e+001
1.0
3.1000e+001
6.3000e+001
0.5
1.2700e+002
2.5500e+002
5.1100e+002
0.0
-15 -10 -5
5 10 15 20
1.0230e+003
Page 314
SIGMA/W
The picture is clear by looking at the pore-pressure at the end relative to the porepressure immediately after the load is applied as in Figure 13-55. The porepressure distribution again becomes hydrostatic, but the position where the new
hydrostatic line crosses the zero pressure line is now higher. This means the water
table now is higher.
2.0
0.0000e+000
1.5
1.0
1.0000e+000
0.5
1.0230e+003
0.0
-15 -10 -5
10 15 20
Intuitively this behavior is entirely consistent with what one might expect.
Commentary
Page 315
SIGMA/W
practice. The feature at this stage is useful for understanding the mechanism, but
no so much for making actual predictions of saturated-unsaturated consolidation
magnitudes.
Moreover, the formulation as shown here gives logical and reasonable results for a
1D analysis. The same is not true for general 2D analyses. More research and
development is required before this feature can be applied to 2D problems.
Uncoupled consolidation analyses should be used when both saturated and
unsaturated zones are present in a 2D field problem.
Truss.gsz
SIGMA/W has what are called truss elements. They are basically 1D elements
with loads or displacements at the ends. Loads or displacements cannot be
specified along the edge of the truss elements. Physically, these elements can only
have axial loads.
A positive force in a truss element means compression; a negative force means
tension. This is consistent with the sign convention in SIGMA/W for stresses in
ordinary elements.
Two of the main applications of truss elements are cross-excavation braces or the
free length of a tie-back anchor. One of the key features of truss elements is that
they can span across other ordinary elements.
This example is provided to demonstrate that the truss elements function properly.
This can be done by analyzing a simple truss. SIGMA/W is not intended as a tool
to analyze trusses, but such a structure can be analyzed to verify the software. The
following diagram shows a simple truss. This example is taken from the book by J.
L. Meriam, entitled Engineering Mechanics - Statics and Dynamics (SI Version)
published by John Wiley & Sons. The example is on page 134. The forces in each
member of the truss can actually be hand computed by the method of joints
(summation of moments and forces about certain joints).
For the SIGMA/W analysis, the members are all given the same E modulus and
cross-sectional area. The actual magnitude of these values is arbitrary, as long as
they are the same for all the members. The length of each truss member is 5 m.
Page 316
SIGMA/W
As just one example of a hand calculation, let us compute the force in 2-3. Taking
moments about 1,
5F - (-20*5) - (-30*10) = 0 (The applied forces are negative since they
point in the negative y-direction).
F = (-100 - 300) / 5 = -80 kN (negative means tension).
The SIGMA/W result can be viewed with the View Truss Element Information
command in CONTOUR. The SIGMA/W computed value is -80 kN. The closedform and SIGMA/W forces for all the members are given in Table 13-5 below.
Table 13-5 Truss verification data
Member
Closed-form
SIGMA/W
2-3
-80.00
-80.000
5-6
-34.64
-34.641
4-6
-17.32
-17.318
4-5
+34.64
+34.641
3-5
-34.64
-34.644
1-4
+63.51
+63.509
1-3
+11.55
+11.543
The SIGMA/W values are the same as the closed-form hand calculations
indicating SIGMA/W is computing these values correctly.
Page 317
SIGMA/W
Tie Back.gsz
Tie Back_ini.gsz
The objective here is to analyze the stability of a temporary shoring tied-back wall
for the proposed design shown in the following diagram. The excavation depth is 9
m and the shoring system consists of sheet piling with two rows of pre-stressed
anchors. The sheet piling will be driven to 3 m below the design excavation base,
and the anchors will be at 3 m and 6 m from the top of the wall.
Page 318
SIGMA/W
Excavate down 4 m to El 15 after the sheet piling has been put in place
Install top row of anchors and stress to the design working load
Install second row of anchors and stress to the design working load
The sheet piling is 12 m long, driven into the ground 3 m below the base of the
excavation. The properties of the sheet piling are:
Anchors will be installed at 2 m spacing along the wall with a working design load
of 200 kN per anchor. Since this is a 2D analysis, the anchor loads must be
Page 319
SIGMA/W
specified per unit length of wall. This makes the anchor loads 100 kN per meter of
wall in the analysis.
The anchors will be installed at an inclination of approximately 17 degrees. The
top anchors have a bond length of 4.7 m and a free length of 3 m; 7 m for a total
length of 8.4 m. The bottom anchors have a bond length of 4.7 m and a free length
of 2.6 m for a total length of 7. 3 m.
The anchors will consist of 40 mm diameter steel bars, pressure grouted into
150 mm diameter boreholes. The other anchor properties are:
The soil will be modeled as a linear elastic material with E = 5000 kPa and
Poissons ratio equal to 0.334. In a linear-elastic 2D analysis, a of 0.334
equates to a Ko of 0. 5. Recall that:
Ko = / (1- ) = 0.334 /(1 0.334) = 0.5
This value is important when we establish the initial in-situ stress state.
The unit weight of the soil is 20 kN/m3. This is specified in SIGMA/W as a body
load for computing the in-situ stresses. The friction angle of the soil is 32 degrees,
and for design purposes the cohesion will be considered to be zero even though the
soil likely has some cohesive strength component. (The soil strength actually does
not come into play in this linear-elastic analysis; it is used only when the
SIGMA/W stresses are used to look at stability).
Page 320
SIGMA/W
The first step is to establish the stress state in the ground before the sheet piling is
installed and before excavation starts. In this case, it is best done with an Insitu
type of analysis. The original ground surface is level so we can use the Insitu1
option. The soil is assigned a Body Load of 20 kN/m3 , and a Ko of 0. 5. The
boundary along the base is fixed (zero x and y displacements). Along both ends,
vertical movement is allowed, but lateral movement is not allowed. The ycondition is specified as none which means the node is free to move up or down.
Eight-noded elements are used throughout, since they give better stress
distributions within the elements than 4-noded elements. Four-point Gauss
numerical integration is adequate in this case, as opposed to the nine-point
integration.
The following figure shows the vertical stresses for a portion of the section.
Elevation 3 m is 16 m below the ground surface. The y-stress is 320 kPa which
correctly matches H which is 16 m x 20 kN/m3 = 320 kPa.
You can use the Draw Mohr Circle feature in CONTOUR and click on any node to
spot check the results. You will see that the horizontal stresses are half the vertical
Page 321
SIGMA/W
stresses as they should be for Ko equal 0. 5. The x-y shear stresses are essentially
zero making the vertical and horizontal stresses the principal stresses as they
correctly should be. It is vitally important that you ensure that the horizontal
stresses are reasonable, since these stresses come into play on the wall when the
ground is excavated.
Simulation of construction procedure
The construction will be simulated in five steps. This means we need to define a
time sequence of 5 steps.
The next task is to flag the elements that will be removed to simulate the
excavation stages. This is done with the Draw Fill/Excavation Elements command.
Looking at the problem definition, when you open the file, you will notice some
elements marked with a 0, -1, -3, and -5. The 0 elements are permanent. The
elements with -1 are removed in Step 1, the ones with -3 are removed in Step 3 and
the elements with -5 are removed in Step 5.
The grouted portions of the anchors are simulated as beam elements. The grouted
portion of the upper anchor will become active in Step 2.
In finite element modelling terms, the pre-stressing of the anchor is achieved by
applying boundary forces at the ends of the free lengths of the anchors. SIGMA/W
applies these forces based on the pre-stress force specified as part of the bar
element definition to simulate the anchor free (unbonded) length. A very important
point is that the pre-stress force must not be applied at the same time as the bar
element becomes active. The pre-stress forces cannot act directly on the bar
element. One way to think of this is that the pre-stress forces are applied first and
then they are locked into the bar element in the next step. In this example, the
upper pre-stress force is applied in Step 2 and the bar element representing the free
length becomes active in Step 3. With this background explanation, the simulation
activities in each of the steps are:
Step 2 activate upper grouted bond length and apply upper anchor
pre-force
Step 3 activate upper free length (bar element) and excavate next
3m
Page 322
SIGMA/W
Step 4 activate lower grouted bond length and apply lower anchor
pre-force
Step 5 activate lower free length (bar element) and excavate final
2m
You can check the specified values and activation times with the View Edge
Information command to look at the grouted (beam) lengths and the View Bar
Element Information to look at the bar elements.
Results
There are many, many ways of looking at the results. Only some are presented here
to illustrate what can be done and to provide a reference for checking that the
software is working properly.
Figure 13-59 shows the force in the free length of the upper anchor. The force is
zero until Step 2 when the 100 kN pre-force is applied. Then the force increases to
150 kN as the second lift is excavated; then the force decreases when the lower
anchor is pre-stressed and finally the force again increases when the last lift is
excavated.
-50
-100
-150
0
Time
Page 323
SIGMA/W
The following diagram (Figure 13-60) shows the lateral displacement on a vertical
profile through the wall. The displacement is in meters.
Notice how the wall gets pulled back with the application of the upper pre-stress
load (Step 2) and then moves back out with the next excavation stage (Step 3).
Applying the lower pre-stress has the same effect, but not nearly as significant.
20
1.0000e+000
15
2.0000e+000
10
3.0000e+000
4.0000e+000
0
0.00 0.02 0.04 0.06 0.08 0.10
5.0000e+000
X-Displacement
Figure 13-60 Lateral wall movement
The next diagram (Figure 13-61) shows the moment distribution in the sheet piling
at the five different construction stages. The maximum moment is about
125 kN-m.
Page 324
SIGMA/W
15
Distance
1.0000e+000
10
2.0000e+000
3.0000e+000
5
4.0000e+000
0
-50
5.0000e+000
50
100
150
Moment
Figure 13-61 Moment distributions in the sheet piling
The final axial force in the grouted portion of the upper anchor is shown in Figure
13-62. Worth noting is that the maximum axial load (about 120 kN) is less than the
140 kN in the free length portion. The reason for this is the interaction between the
grouted length and the soil. The soil is absorbing some of the energy.
The axial force is shown as leveling off at the ends. This is due to a numeric
boundary effect in the beam formulation. The formulation is based on a constant
strain in each element of the beam. For the inner structural elements is not
noticeable due to averaging to the nodes. It is noticeable only for the end elements.
It is the same as four 4-noded elements when insitu stresses are computed. The
first row at the surface and the last row at the bottom have constant stresses, which
cause a stress profile to deflect at the top and at the bottom. Without these end
effects the maximum axial load is 140 kN as in the free length.
Page 325
SIGMA/W
-20
Axial Force
-40
-60
-80
-100
-120
0
Distance
Figure 13-62 Axial force in the upper grout length of the anchor
The final diagram (Figure 13-63) here shows the lateral (x) stress distribution at
the end of construction. Notice how the stress contours are affected by the presence
of the anchors.
40
60
80
40
60
80
Page 326
SIGMA/W
Soft Ground_ini.gsz
Soft Ground.gsz
Problem description
The subsurface clay stratum is 9 m deep and the water table is 1 m below the
ground surface. The 1 m layer above the water table is highly weathered,
desiccated and fissured making its behavior similar to a fine granular soil. The
embankment to be analyzed is shown in the following figure and is constructed of
relatively sandy soil. The height of the embankment is 5 m with 3h:1v side slopes
and a 10 m crest width. Due to symmetry about the center, only half of the cross
section is used in the analysis.
15
14
13
Elevation - metres
12
11
10
9
8
7
6
5
4
3
2
1
0
0
10
15
20
25
30
35
40
metres
The objective of this example is to outline the correct modeling techniques for
determining the embankment stability and settlement.
The construction of an embankment can be simulated in a manner similar to
constructing an excavation. The difference is that elements are added (activated)
Page 327
SIGMA/W
instead of removed (nullified). The time step number assigned to added elements is
positive instead of negative.
Clay properties
The clay is modeled as a Modified Cam-clay (MCC) soil. It is highly plastic with a
liquid limit (LL) of 61%. The compression index Cc can be estimated from the well
known equation:
Cc = 0. 009 (L.L. - 10 percent)
= 0. 009 (61-10)
= 0. 46
The slope of the normal consolidation line can be computed from Cc / 2. 303);
this makes equal to 0.20.
Page 328
SIGMA/W
Value
Cc (compression index)
0. 46
0. 20
0. 04
3 x 10-4 (1/kPa)
k ( hydraulic conductivity)
5 x 10-3 m/day
26 degrees
1. 0
0. 56
( Poissons Ratio)
0. 36
1. 2
(specific volume)
2. 2
Sand properties
The embankment and upper meter of subsoil will be modeled as highly permeable
linear-elastic material, which is considered adequate since the major settlement
issue arises in the underlying compressible clay. The properties adopted for the
sand are:
E = 2000 kPa
= 0.36
k = 1.0 m/day
mv = 3 x 10-4 kPa-1
The k and mv values are actually arbitrary for the sand, since the pore-water
pressure conditions will be specified, meaning there is no change with time. In a
transient (consolidation) analysis, the software requires these values be specified,
but the actual values do not affect the results.
In-situ stresses
It is essential when using the Modified Cam-Clay (MCC) soil model to establish
the initial yield surfaces. Any changes due to the embankment loading are relative
Page 329
SIGMA/W
to the yield surfaces that exist prior to the loading. The initial yield surface is
related to the initial insitu stress and the over-consolidation ratio (OCR). The OCR
is specified as a soil property, but the insitu stress must be computed in a separate
step.
The finite element mesh must be complete at the start, but the elements
representing the embankment are ignored during the insitu stress analysis by
giving the embankment elements a soil model set to "none".
The two soils representing the subsurface are given a body load equal to the unit
weight of 20 kN/m3, and a Ko equal to 0.56. The water table is specified as a
horizontal line 1 m below the ground surface.
The unit weight of water has been set to 10 kN/m3. This makes it convenient to
check that the correct total and effective stresses and pore-water pressures have
been computed correctly. In real field problems, the unit weight of water would of
course be 9.81 kN/m3.
The bottom of the problem is fixed, while both vertical ends are allowed to move
vertically, but not horizontally.
When establishing the insitu stresses, it is important to select "Insitu" as the
analysis type. Selecting a Load/Deformation analysis type would result in wrong
horizontal effective stresses. The following figures show the total and effective
horizontal stress profiles in the subsoil before starting the embankment fill
placement.
The effective horizontal stress should be:
h
56.0 kPa
Page 330
SIGMA/W
6
Y
10
10
0
0
50
100
150
X-Total Stress
10
20
30
40
50
60
X-Effective Stress
Note that in the Modified Cam-Clay definition dialog box that pc is not defined.
When this is value undefined, SIGMA/W uses the OCR together with the insitu
stresses to compute pc at each Gauss point in each element. The initial yield
surface is consequently defined at each Gauss stress Loading Sequence.
Loading sequence
Fill Lift
1st
2nd
13
3rd
19
4th
25
5th
The analysis will be run for 40 days. Figure 13-66 figure shows the embankment
with the numbers in each row of elements indicating at what time (day) the
elements become active. The clay will consolidate during the time between each
lift placement and for 15 days after the last lift is placed.
Page 331
SIGMA/W
t =
w mv A
4 K sat
where:
A
During the initial time step, the size of the element area (A) where the
consolidation will be the greatest (just below the water table) is roughly one square
meter. Correspondingly, the initial time step should be approximately 1 to 2 days.
A value of one day is used as the initial time step for this problem.
A useful thing to remember is that a reasonable starting time step is roughly mv divided by
Ksat.
Page 332
SIGMA/W
The computed results can be viewed in several different ways. Only a few
approaches are presented here to illustrate some of the highlights. Many additional
graphs and contour plots can be viewed by using the CONTOUR option.
Figure 13-68 shows the pore-water pressure changes with time at a point one meter
to the right of the centerline and two meters below the water table. The pore-water
Page 333
SIGMA/W
pressure immediately rises after the placement of a lift and then falls until the next
lift is placed. The pore-water pressure drops from a high of 90 kPa down to about
55 kPa after the final lift has been placed. Significant excess pore-water, however,
remains at the end of 40 days indicating that additional long-term settlement will
occur.
Figure 13-69 shows the settlement that occurs at the center-line at the water table
elevation. Note how some settlement occurs immediately as the lifts are placed and
additional settlement occurs gradually during the consolidation phases. To be
precise, there is also some settlement due to consolidation during the one-day
interval that the lift is placed, but most of it is due to the additional load.
Pore-Water Pressure
100
80
60
40
20
0
10
20
30
Time
Figure 13-68 Change in pore-pressure with time
Page 334
40
SIGMA/W
Y-Displacement
0.0
-0.2
-0.4
-0.6
-0.8
0
10
20
30
40
Time
Figure 13-69 Immediate and consolidation settlement with time
Settlement profiles along the original ground surface on Days 0, 7, 13, 19, and 25
are shown in Figure 13-70. These are the days that the fill was placed. Of
particular interest is the location of maximum settlement. Note that the maximum
does not occur along the centerline, but is located more toward the outer edge of
the fill in the early stages of loading. At the end of 40 days the maximum is under
the center of the embankment, but not during the construction.
The fact that the maximum settlement is not under the center of the embankment
during the early stages of construction is due to the two-dimensional dissipation of
the excess pore-pressure. There is some lateral flow towards the toe area which
does not exist under the center. The excess pore-pressure relative to the load is also
the highest in the maximum settlement area as is evident from the low effective
horizontal stresses depicted in Figure 13-71. They are the lowest in this area.
In addition, the settlement towards the toe area is aided by the uplift outside the
toe.
Page 335
SIGMA/W
0.1
Y-Displacement
0.0
-0.1
-0.2
-0.3
-0.4
-0.5
-0.6
0
10
15
20
25
X distance
Figure 13-70 Settlement profiles along the original ground surface
25
40
50
Page 336
SIGMA/W
The results clearly show the advantages of the staged construction. Without the
staged construction, the excess pore-pressure would be much higher and could
potentially lead to failure during construction. This could be easily demonstrated
by rerunning the included files after all the fill elements have been assigned a
Step 1 activation flag.
Page 337
SIGMA/W
Page 338
SIGMA/W
14
Product Integration
14.1
The objective of this illustration is to observe how to use finite element pore-water
pressure results in a stability analysis. Including or deliberately ignoring negative
pore-water pressures can be critical to understanding and interpreting a slope
stability analysis.
In particular, the objectives of this illustration are to:
Repeat the analysis, but remove the advanced parameters from the soil
property information. Graph the pore-water pressure and strength
along the slip surface and note how the negative pore-water pressures
have been ignored).
The seepage portion of the analysis is illustrated in Figure 14-1. In general, the
slope is comprised of multiple layers with a finer, lower permeability layer located
half way up the slope face. Note that more coarse soil soils in region 1 and 3 have
the same hydraulic properties (Ksat 1x10-3 m/day). In addition, a steady-state
infiltration rate of q = 5.0x10-5 m/day is applied along the top and the slope with a
pressure equals zero condition on the downstream surface. A potential seepage
review has been applied along the face of the slope.
Page 339
SIGMA/W
26
24
22
20
18
Seepage Face
16
14
12
Pressure = zero
10
8
6
4
0
10
20
30
40
50
60
The soil property information for the SLOPE/W portion of the analysis is given in
Table 14-1.
Table 14-1 Slope soil information
Layer
Unit Weight
Phi
Cohesion
Unit Weight
above WT
Phi B
18
25
18
15
19
20
19
15
20
30
10
20
15
Figure 14-2 shows the location of grid and radius points applied in the stability
analysis, as well as the SEEP/W computed perched water table and the computed
factor of safety. SLOPE/W was able to read the seepage results directly from
SEEP/W in order to compute the actual pore-water pressures at the base of each
slice. Figure 14-3 shows the actual pore-water pressures applied on each slice.
Notice how the pore pressures on the slices change from negative to positive to
negative and back to positive as the slice number increases from left to right. This
type of pore-water pressure condition would not have been possible to accurately
establish without the use of a rigorous saturate-unsaturated seepage flow model.
Note in the SLOPE/W results that the finite element mesh used in the seepage
analysis is superimposed on the solution. The finite element mesh is not actually
used by SLOPE/W, but the data from the mesh at all locations can be used to
determine pore pressures on the base of any slice geometry.
Page 340
SIGMA/W
44
42
11
40
38
36
34
32
1.324
30
2
28
12
13
26
24
3
22
1
20
Infiltration
18
8
16
14
10
12
3
10
14
8
6
4
0
10
20
30
40
50
60
Figure 14-2 SLOPE/W solution showing perched water table from SEEP/W
Pore-Water Pressure vs. Distance
30
25
20
Pore-Water Pressure
15
10
5
0
-5
-10
-15
-20
0
10
20
30
40
Distance
14.2
The objective of this illustration is to observe how to use climate coupled finite
element pore-water pressures results in a stability analysis. You will see how
important including negative pore pressure can be to a slopes stability.
Page 341
SIGMA/W
The VADOSE/W analysis mesh as well as the SLOPE/W grid and radius are
illustrated in Figure 14-1. In general, a large rainfall event was applied over the
undulating ground profile on the first day of a 30 day long transient climate
coupled VADOSE/W analysis. This heavy rainfall created ponded water
conditions in the low point near the center of the mesh. The rainfall day was
followed by 29 days of infiltration from the pond and evaporation through the
climate-ground boundary. Stability results are compared for the day just after the
rainfall (e.g., day 2 from VADOSE/W) as well as after a long drying period (e.g.
day 30). A comparison is also made between using and not using soil strength
parameters that account for increased strength with increased negative water
pressure (e.g. Phi-B).
The soil property information for the SLOPE/W portion of the analysis is given in
Table 14-1. Two stability simulations were carried out: one with a Phi B parameter
defined to account for the effect of soil water suction on strength and one without
Phi B.
Table 14-2 Slope soil information
Layer
Unit Weight
Phi
Cohesion
Unit Weight
above WT
Phi B
18
25
18
15
Page 342
SIGMA/W
Figure 14-2 shows a minimum factor of safety of 1.16 in the stability analysis
based on advanced pore-water pressure parameters (e.g., soil suction strength
effects) as well as the VADOSE/W computed water. SLOPE/W was able to read
the seepage results directly from VADOSE/W in order to compute the actual porewater pressures at the base of each slice.
Figure 14-6 shows the computed factor of safety after the heavy rainfall day based
on soil property strength data that does not include the effect of suction on the
strength of the soil. In this analysis the factor of safety is reduced to less than 1.0.
One final analysis was carried out using advanced strength parameters to
determine the factor of safety after an extended drying period. Figure 14-7 shows a
factor of 1.74 after 29 days of infiltration from the pond (e.g. no rainfall) and
surface evaporation on the ground profile. It is clear that from this example that the
factor of safety is very dependent on ground-climate influences and on the
application of advanced soil strength parameters that take into account increased
strength due to negative water pressure in the soil (e.g. soil water suction).
Finally, Figure 14-8 shows the actual pore-water pressures applied on each slice in
the stability analysis for day 30 of the simulation. Notice how the pore pressures
on the slices change from positive to negative as the slice number increases from
left to right. This type of pore-water pressure condition would not have been
possible to accurately establish without the use of a rigorous saturate-unsaturated
climate coupled seepage flow model.
Note in the SLOPE/W results that the finite element mesh used in the seepage
analysis is superimposed on the solution. The finite element mesh is not actually
used by SLOPE/W, but the data from the mesh at all locations can be used to
determine pore-water pressures on the base of any slice geometry.
Page 343
SIGMA/W
1.159
0.999
Page 344
SIGMA/W
1.739
Figure 14-7 Factor of safety after long drying period (e.g. day 30 pore-water
pressure data from VADOSE/W)
10
0
-10
-20
-30
-40
Distance
Page 345
14.3
SIGMA/W
(x 1000)
1.11
1.08
1.05
1.02
0.99
0.96
0.93
0.90
-500
-400
-300
-200
-100
100
200
300
400
Distance - feet
Figure 14-9 Dam pore-water pressure prior to earthquake
Page 346
500
600
SIGMA/W
Acceleration ( g )
0.3
0.2
0.1
0.0
-0.1
-0.2
-0.3
-0.4
-0.5
-0.6
0
10
20
30
40
50
Time (sec)
Page 347
SIGMA/W
Pore-Water Pressure
9000
8000
7000
6000
5000
0
10
15
Time
1.14
(x 1000)
1.11
1.08
Head BC = 1110 ft
1.05
P=0
1.02
0.99
0.96
0.93
0.90
-500
-400
-300
-200
-100
100
200
300
400
500
600
Distance - feet
Figure 14-13 shows the model set up for the SEEP/W analysis that will compute
the dissipation of the excess pore-water pressure generated during the earth quake.
The boundary conditions for the seepage analysis are illustrated in the figure. In
Page 348
SIGMA/W
this analysis it is assumed that the reservoir remains full with a total head of 1110
ft (or 102 ft pressure head). The seepage analysis was carried out to simulate 5
days and the change in pore-water pressure at point A over time is illustrated in
Figure 14-14 below.
Pressure vs. Time
9000
Pressure
8000
7000
6000
5000
0
100000
200000
300000
400000
500000
Time
14.4
Seepage data from a steady state or transient SEEP/W analysis can be used in the
GeoStudio CTRAN/W module to predict contaminant transport with or without
adsorption, diffusion and decay. This is a one-dimensional contaminant transport
problem verification example analysis with a free exit boundary. The CTRAN/W
results are compared with closed form analytical and published solutions.
To generate a one-dimensional steady-state flow system, a one-row finite element
mesh is created using SEEP/W DEFINE. The finite element mesh is 1 m high and
40 m long and consists of a total of 30 elements and 62 nodes. The SEEP/W files
containing this example are named EXIT.GSZ.
Page 349
SIGMA/W
The head differential and hydraulic conductivity are selected to produce a constant
seepage velocity U of 0.05 m/s in the positive x-direction. The volumetric water
content is defined as a constant value of 0.5. The average linear velocity is:
=U /
= 0.05 / 0.5
= 0.1 m/s
The dispersivity L is set to 4 m, and the molecular diffusion coefficient D* is set
to zero. The resulting hydrodynamic dispersion coefficient is:
D = Lv + D *
= 4 0.1 + 0.0
= 0.4 m 2 /s
The time step sequence consists of 48 steps. Results are presented for Time Steps
8, 16, 24, 32, 40 and 48 with total elapsed times of 80 s, 160 s, 240 s, 320 s, 400 s
and 480 s respectively. Computed results for these six time steps are included with
the CTRAN/W software.
The boundary condition at the left end of the problem is assumed to be a source
boundary with concentration of the source Cs specified as 1.0 unit/m3. The
boundary condition at the right end is specified as a free exit boundary (Qd > 0).
The initial concentration of the flow system is set to 0.0.
Frind, (1988) has presented an analytical solution to the transport equation with a
free exit boundary. The solution is approximated by the analytical solution for
transport in a semi-infinite medium. The concentration C as a function of x and t is
expressed as:
C =
Cs
2
x - vt
erfc
2 Dt
x - vt v( x +vt )
vx
exp D erfc
1 + D
2 Dt
( x vt ) 2
v t
+
exp
4 Dt
D
where:
Page 350
SIGMA/W
concentration,
Cs
elapsed time,
erfc
Page 351
SIGMA/W
1.0
8.0000e+001
0.8
1.6000e+002
0.6
2.4000e+002
C
0.4
3.2000e+002
0.2
4.0000e+002
0.0
4.8000e+002
10
20
30
40
Distance
Page 352
SIGMA/W
1.0
CTRAN/W Solution
(Exit Qd > 0)
0.9
0.8
Analytical Solution
Concentration
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0
10
20
30
40
Distance
Figure 14-16 Comparison with analytical solutions for one exit boundary
condition
1.0
8.0000e+001
0.8
1.6000e+002
0.6
2.4000e+002
C
0.4
3.2000e+002
0.2
4.0000e+002
0.0
4.8000e+002
10
20
30
40
Distance
Figure 14-17 CTRAN/W solution with two zero dispersive mass flux exit
boundary conditions
Page 353
SIGMA/W
1.0
CTRAN/W Solution
(Exit Qd > 0)
0.9
0.8
CTRAN/W Solution
(Exit Qd = 0)
Concentration
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0
10
20
30
40
Distance
Figure 14-18 Comparison with analytical solution for two exit boundary
conditions
14.5
Seepage velocity and water content data from a steady state or transient
VADOSE/W analysis can be used in CTRAN/W to predict contaminant transport
with or without adsorption, diffusion and decay. The purpose of this example is to
show the influence of including climate (or vegetation) effects on the movement of
contaminants in soils.
In this example a simple advection dispersion analysis is carried out in
CTRAN/W after solving a transient VADOSE/W analysis. The CTRAN/W
program reads the transient seepage velocity and water content data at different
time steps and uses it in the solution of the advection dispersion equation.
The problem is kept quite simple for illustration purposes. In the seepage analysis,
a pressure equal to zero boundary condition is applied at the base of the low point
near the center of the mesh. This pressure boundary condition is saying there is a
small source of water at this location just enough to keep the ground saturated. It
Page 354
SIGMA/W
is assumed that a source of contaminant is also present with the water at the ground
surface.
Figure 14-19 shows the position of the water table and the concentration contours
for the case where the model only allowed infiltration at the source of contaminant.
For this case, there is no evaporation demand along the rest of the ground surface
and therefore the contaminant is limited in its migration to a somewhat radial
pattern beneath the source.
In Figure 14-20, the same concentration is applied in the pond location along with
the same source of water, however, the climate boundary condition is active at all
other ground surface nodes. The climate boundary in this case is removing about
5 mm per day of evaporation.
It is clear for the case with evaporation, that there is a lot more spreading of the
contaminant. The results show that the contaminant is being pulled up towards the
drying ground surface where it is either exiting with the vapor flow or being
deposited or both depending on the exit boundary condition specified in the
contaminant analysis.
Page 355
SIGMA/W
14.6
Henry (1964), developed an analytic solution for a simplified sea water intrusion
problem. The Henry problem has since become a benchmark verification
example for many numerical models of density-dependent flow. However,
Croucher and OSullivan, (1995), noted that none of the published numerical
model comparisons with Henrys solution that they examined were able to match
Henrys solution to a great extent. In addition to outlining the possible reasons for
the discrepancies, Croucher and OSullivan, presented a new, highly accurate
numerical solution to the problem. Their numerical solution is used here for
comparison with the results from CTRAN/W.
The system being modeled is shown in Figure 14-21. It consists of a 2.0m long
section of a 1.0m thick aquifer where the right boundary is in direct contact with
sea water and the left boundary has a constant influx of freshwater. The sea water
has a relative density, (specific gravity), of 1.025 at a reference concentration of
1.0. The concentration of sea water is fixed at 1.0 along the sea water boundary
and a fixed freshwater inflow rate of 6.6x10-5 m3/s is specified along the freshwater
boundary. The top and bottom boundaries are both impermeable. The aquifer is
homogeneous and isotropic and has a saturated hydraulic conductivity Ksat =
1x10-2 m/s, a porosity n = 0.35 and a velocity independent dispersion coefficient of
D = 1.89x10-5 m2/s. The aquifer is discretized using 0.05m square elements and the
solution is sought at steady state. The SEEP/W example file containing this
problem is called HENRY.GSZ, and the full definition of the problem may be
viewed using SEEP/W and CTRAN/W module DEFINE views.
Page 356
SIGMA/W
Left Boundary
Freshwater
Qf = 6.6E-5 m2/s
C=0.0
Right Boundary
Seawater
Hs = 1.0m
C = 1.0
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2.0
Aquifer Properties
Ks = 1E-2 m/s
n = 0.35
D = 1.89E-5 m2/s
Seawater Density
SG=1.025 @ Cref=1.0
The above Henrys problem has been analyzed with CTRAN/W, and a steady state
solution is obtained at an elapsed time of more than 11,000 seconds (at time step
35). Figure 14-22 shows the computed seawater concentration contours at steady
state along with the water flow velocity vectors. At steady state, seawater enters
the aquifer across the lower portion of the sea water boundary via density induced
gradients and mixes with freshwater flowing in the opposite direction. The
constant influx of freshwater from the left freshwater boundary causes the diluted
seawater to exit the system across the upper portion of the sea water boundary. In
this way, a seawater flow cell is established in which the seawater toe migration
towards the freshwater boundary is controlled by the rate of freshwater flow, the
density of the seawater, and the degree of mixing between the seawater and
freshwater. The degree of mixing is controlled by the dispersion coefficient used in
the modelling, which in this case is velocity independent.
It should be noted that in Figure 14-22 near the upper left of the aquifer,
CTRAN/W computed a few small negative concentration values. This slight
numerical oscillation is a direct result of the Peclet and Courant numbers being
exceeded in these areas because of the relatively high water velocity and relatively
coarse mesh and time step discretization. It is possible to eliminate the negative
concentration by reducing both the mesh size and the time step size, however,
since we are more interested in the solution in the lower portion of the flow system
Page 357
SIGMA/W
and we only use the 0.5 concentration contours in the comparison, refinement to
the finite element mesh and time steps were deemed unnecessary in this case.
Freshwater
C=0
Seawater
C=1
1.00
0.75
0.50
0.
8
0.
6
0.
4
2
0.
0.25
0.00
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
2.00
Page 358
SIGMA/W
1
0.9
0.8
CTRAN/W
0.7
Elevation (m)
0.6
0.5
0.4
0.3
0.2
0.1
0
1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
X-Coordinate (m)
14.7
The GeoStudio module TEMP/W can be used to model natural and artificial
ground freezing without or with the addition of heat added by flowing water.
Consider the frozen ground that forms around a buried pipe with a pipe wall
boundary condition of -2 degrees Celsius. Figure 14-24 shows the ground
temperature profile after 2 years of freezing. It is clear from this figure that a
region of frozen ground has formed around the pipe and that the ground freezes to
a deeper depth beneath the pipe where there is less influence from the warmer
yearly average ground surface temperature. Details of this analysis and comparison
of the results for this case to other published data can be found in the TEMP/W
engineering book. The SEEP/W example file containing this problem is called
PIPE WITH FLOWING WATER.GSZ, and the full definition of the problem
may be viewed using SEEP/W and TEMP/W module DEFINE views.
Page 359
SIGMA/W
1.6
2.5
1.4
Pipe
1.2
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
Figure 14-25 shows the temperature profile for same analysis except in this case
the influence of flowing water is considered. The SEEP/W module was set up such
that a hydraulic gradient of 0.27% was established across the region from right to
left with a saturated hydraulic conductivity of 0.1 m/day. It is clear from this figure
that the extents of the frozen ground are far less than for the static water condition
and that the shape of the frozen ground is more strongly influenced by the flowing
water than the warmer air temperature.
Figure 14-26 shows the actual computed water velocity vectors and total head
contours across the region after two years of analysis. This figure illustrates clearly
the diversion of water around the frozen ground region. Careful examination of the
figure shows that the velocity of water increases as it moves around the frozen
ground. This is because the cross sectional area available for flow is reduced due to
ice formation. For cases where several freezing pipes are installed to create a
frozen barrier wall, the increase in velocity of water between adjacent freeze pipes
can result in a situation where more heat is added by the water than can be
removed by the freeze pipes. When this occurs, closure of the frozen wall (due to
adjacent frozen regions growing together) will not happen. This effect can be seen
in Figure 14-27 for a typical shaft sinking artificial ground freezing project.
Page 360
SIGMA/W
1.6
2.5
1.4
1.5
1.2
Pipe
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
Figure 14-25 Temperature contours around freezing pipe with flowing water
When carrying out this type of analysis, it is very important to have well-defined
material properties and to use a fine mesh discretization and small time steps. The
seepage velocity generated by SEEP/W and added into the TEMP/W finite element
partial differential equation is a linear term, and as such, the computed results are
quite sensitive to numerically appropriate element size and time steps. A first guess
at time step sizes can be made by computing the Courant number and ensuring that
it is less than a value of 1. The Courant number is given by:
C=
vt
x
where:
Page 361
0.4
0.6
1.6045
0.2
1.6035
1.6025
5
1.601
1.6005
1.6
1.5
1.4
1.3
1.2
1.1 Pipe
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
-0.2
-0.2 0.0
SIGMA/W
0.8
1.0
1.2
1.4
1.6
Figure 14-26 Water flow vectors and head contours around frozen pipe
It is suggested when modeling convective heat transfer to use the adaptive time
stepping routine and set the maximum time step by first setting the Courant
number to 1 in the above equation and computing the permissible time step to
ensure numerical stability.
Page 362
SIGMA/W
Finally, while this example shows how water can influence freezing in ground, the
same combination of TEMP/W and SEEP/W can be used to study the influence of
freezing ground on the transient seepage of water when the parts of the ground are
subjected to cold temperatures. This might be the case for dam performance in
climates with both summer and winter seasons.
In both cases, you must set up a complete SEEP/W and TEMP/W analysis and
then start the solver process from the TEMP/W program. TEMP/W will launch
SEEP/W, pass it the ground temperatures and wait for SEEP/W to pass back water
content and computed seepage velocities. In the event that SEEP/W is passed a
ground temperature that is below the freezing point of water, it will compute a
reduced hydraulic conductivity corresponding to the magnitude of the temperature
below the phase change point. The calculation of the frozen ground conductivity is
discussed in the Material Properties chapter of this book.
14.8
Page 363
SIGMA/W
15
14
13
12
Elevation - metres
11
10
9
8
7
6
5
4
3
2
1
0
0
10
15
20
25
30
35
40
metres
The clay will be modeled as a Modified Cam-clay (MCC) soil. It is highly plastic
with a liquid limit (LL) of 61%. A summary of all clay properties is given in Table
14-3. The embankment and upper meter of subsoil will be modeled as highly
permeable linear-elastic material, which is considered adequate since the major
settlement issue arises in the underlying compressible clay. The properties adopted
for the sand are:
E
2000 kPa
0.36
1.0 m/day
mv
3 x 10-4 (1/kPa)
The k and mv values are actually arbitrary for the sand, since steady-state porewater pressure conditions will be specified, meaning there is no change with time.
In a transient (consolidation) analysis, the software requires these values be
specified, but the actual values do not affect the results.
Page 364
SIGMA/W
Value
Cc (compression index)
0.46
0.20
0.04
3 x 10-4 (1/kPa)
k (hydraulic conductivity)
5 x 10-3 m/day
26 degrees
1.0
0.56
(Poissons Ratio)
0.36
1.2
(specific volume)
2.2
The analysis will be run for 40 days. Figure 14-29 shows the embankment with the
numbers in each row of elements indicating at what time (day) the elements
become active. The clay will consolidate during the time between each lift
placement and for 15 days after the last lift is placed.
+25 +25 +25 +25 +25 +25
+25 +25
+19 +19 +19 +19 +19 +19 +19 +19 +19
+19 +19
+13 +13 +13 +13 +13 +13 +13 +13 +13 +13 +13 +13
+13 +13
+7
+7
+7
+7
+7
+7
+7
+7
+7
+7
+7
+7
+7
+7
+7
+7 +7
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1 +1
Page 365
SIGMA/W
volumetric water content function can be defined as a straight line with two points
resulting in a line with a slope of 3 x 10-4 (1/kPa). The slope is (90 x 0.001)/300 =
0.0003.
The hydraulic conductivity of the clay will be constant and equal to the saturated
hydraulic conductivity, defined by a horizontal function at 5 x 10-3 m/day.
The pore-water pressure along the water table is zero and the total head is therefore
equal to the elevation. As a result, the boundary condition is defined as H (P=0).
Changes in pore-water pressure are not considered within the upper meter of clay
or the embankment fill. As a result, deformations in these zones due to changes in
pore-water pressure are not computed. It is important to note that pore-water
pressure changes within these zones are not allowed. Setting the boundary
conditions in these zones to be H (P=0) as illustrated in Figure 14-30, ensures this
to be the case. While the actual pore-water pressures may not be correct, the effect
has been removed, which is an objective of this analysis.
All the other boundary conditions can be left undefined, resulting in a zero flow
boundary at the vertical ends of the clay located below the water table and along
the bottom.
Figure 14-31 shows the pore-water pressure changes with time at a specified
location one meter to the right of the centerline and two meters below the water
table. The pore-water pressure immediately rises after the construction of a lift and
then falls until the next lift is placed. After the final lift has been placed, the porewater pressure at the specified location drops from a high of 90 kPa to about 55
kPa. Significant excess pore-water remains at the end of 40 days, indicating that
additional long-term settlement will occur.
Page 366
SIGMA/W
Figure 14-32 shows the settlement profiles along the original ground surface on
days 0, 7, 13, 19, and 25. Of particular interest is the location of maximum
settlement, which does not occur along the centerline, but is located more toward
the outer edge of the fill in the early stages of loading. This result is due to the
zone of lower effective horizontal stresses as shown in Figure 14-33. In addition,
the toe of the fill has been lifted up slightly.
Pore-Water Pressure
110
90
70
50
30
10
0
10
20
30
40
Time
Page 367
SIGMA/W
Y-Displacement vs. X
0.2
0.0000e+000
Y-Displacement
0.0
7.0000e+000
-0.2
1.3000e+001
-0.4
1.9000e+001
-0.6
2.5000e+001
-0.8
0
10
15
20
25
30
25
30
25
30
Page 368
SIGMA/W
15
Theory
15.1
Introduction
This chapter presents the methods, equations, procedures, and techniques used in
the formulation and development of the SIGMA/W SOLVE function. It is of value
to be familiar with this information when using the software. An understanding of
these concepts will be of great benefit in applying the software, resolving
difficulties, and judging the acceptability of the results.
The development of the finite element equations for stress/deformation analysis
using potential energy, weighted residuals, or variational methods is well
documented in standard textbooks, and consequently is not duplicated in this
Users Guide. (See Bathe, 1982, Smith and Griffiths, 1988, Segerlind, 1984 and
Zienkiewicz and Taylor, 1989 for further information on the development of finite
element equations).
SIGMA/W is formulated for either two-dimensional plane strain or axisymmetric
problems using small displacement, small strain theory. Conforming to
conventional geotechnical engineering practice, the "compression is positive" sign
convention is used. In this chapter, the bracket sets < >, { }, and [ ] are used to
denote a row vector, a column vector and a matrix, respectively.
15.2
The finite element equation used in the SIGMA/W formulation for a given time
increment is:
[ B ] [C ][ B ] dv {a}
T
where:
[ B]
strain-displacement matrix,
[C ]
constitutive matrix,
{a}
Page 369
SIGMA/W
< N> =
volume of an element,
{Fn }
Summation of this equation over all elements is implied. It should be noted the
SIGMA/W is formulated for incremental analysis. For each time step, incremental
displacements are calculated for the incremental applied load. These incremental
values are then added to the values from the previous time step. The accumulated
values are reported in the output files. Using this incremental approach, the unit
body force is only applied when an element is included for the first time during an
analysis.
For a two-dimensional plane strain analysis, SIGMA/W considers all elements to
be of unit thickness. For constant element thickness, t, the above equation can be
written as:
Equation 15-1
([ B ] [C ][ B ] R ) dA {a}
T
SIGMA/W
where:
[K]
([ B] [C ][ B]) dA
= ([ B ] [C ][ B ] R ) dA
= t
(for axisymmetric)
{a}
{F}
{Fb}
{Fs}
{Fn}
SIGMA/W solves this finite element equation for each time step to obtain
incremental displacements and calculates the resultant incremental stresses and
strains. It then sums all these increments since the first time step and reports the
summed values in the output files.
Strain-displacement matrix
x
{ } = y
z
xy
The field variable of a stress/deformation problem is displacement which is related
to the strain vector through:
Page 371
Equation 15-3
SIGMA/W
{ } = [ B ]
where:
[B]
strain matrix,
u, v
SIGMA/W is restricted to performing infinitesimal strain analyses. For a twodimensional plane strain problem, Z is zero and the strain matrix is defined as:
N1
x
0
[ B ] =
0
N1
x
...
N8
x
N1
...
y
0
N1
y
...
0
N8
...
x
N8
y
0
N8
y
u
x
x v
y y
=
{ } =
z u
xy
u + v
y x
The associated strain matrix [ B ] is then:
Page 372
SIGMA/W
N1
x
0
[ B] =
N1
R
N1
x
...
N1
...
y
0
N1
y
N8
x
0
N8
R
N8
...
x
...
N8
y
N8
y
{ } = [C ]{ }
where [C ] is the constitutive (element property) matrix and is given by:
0
1
1
0
E
[C ] =
1 v
0
(1 + )(1 2 )
1 2
0
0
0
2
where:
E
Youngs modulus
Poisson's ratio
The [C ] matrix is the same for both the two-dimensional plane strain and the
axisymmetric cases.
Page 373
SIGMA/W
Body forces
SIGMA/W can model body forces applied in both the vertical and the horizontal
directions. These forces are applied to all elements when they first become active.
The body force in the vertical direction, bv, is due to gravity acting on an element.
For a given material, the unit body force intensity in the vertical direction is given
by its unit weight, s, which is in turn related to its mass density, :
s = g
where g is the gravitational constant. When the unit weight s is non-zero,
SIGMA/W evaluates the integral s
(
v
The term p
Page 374
SIGMA/W
x
dr
r
y
dy =
dr.
r
dx =
Substituting these differentials into the incremental force equation and applying the
finite element approximation, the following equations can be derived for the
equivalent nodal forces at node i along an element edge:
x
y
pn
)dr
r
r
S
x
y
+ pt
)dr.
Pyi = N i t ( pn
r
r
S
Pxi = N i t ( pt
x
)dr
r
S
y
)dr.
Pyi = N i t ( p y
r
S
Pxi = N i t ( px
The fluid pressure boundary is as a particular case of normal and tangential case in
which no tangential pressure is applied. During numerical integration, fluid
pressure is evaluated at each integration point, say the i-th integration point, using:
pni = f ( y f yi ),
( y f yi ) > 0
Page 375
SIGMA/W
where:
yf
yi
The fluid pressure is only computed when the fluid elevation exceeds the ycoordinate of an integration point.
Nodal forces
Page 376
SIGMA/W
{F } = v [ B ] { } dv
T
where:
{}
[B]
dv
elemental volume.
The resultant nodal forces are accumulated at each node. To simulate the removal
of soils, as in an excavation, the signs on the nodal forces are reversed before these
forces are incorporated into the finite element equation.
In the displacement output files, SIGMA/W reports the nodal forces at all nodes
where displacement or spring boundary conditions are specified. These nodal
forces are calculated in the manner previously described. However, there is now no
change of signs. The nodal forces represent boundary forces which may be used,
for example, to estimate the loads supported by struts in a braced excavation.
15.3
Numerical integration
[ B ] [C ][ B ] dA
T
B
j =1
C j
B j det J j W1 jW1 j
where:
Page 377
SIGMA/W
integration point,
det J j
W1 j , W2 j
weighting factors
Table 15-1 to Table 15-4 show the numbering scheme and location of integration
points used in SIGMA/W for various element types.
Table 15-1 Sample point locations and weightings for four point quadrilateral
element
Point
+0.57735
-0.57735
-0.57735
+0.57735
w1
w2
+0.57735
1.0
1.0
+0.57735
1.0
1.0
-0.57735
1.0
1.0
-0.57735
1.0
1.0
Page 378
SIGMA/W
Table 15-2 Sample point locations and weightings for nine point quadrilateral
element
Point
w1
w2
+0.77459
+0.77459
5/9
5/9
-0.77459
+0.77459
5/9
5/9
-0.77459
-0.77459
5/9
5/9
+0.77459
-0.77459
5/9
5/9
0.00000
+0.77459
8/9
5/9
-0.77459
0.00000
5/9
8/9
0.00000
-0.77459
8/9
5/9
+0.77459
0.00000
5/9
8/9
0.00000
0.00000
8/9
8/9
Table 15-3 Sample point location and weighting for one point triangular
element
Point
w1
w2
0.33333
0.33333
1.0
0.5
Table 15-4 Sample point locations and weightings for three point triangular
element
Point
w1
w2
0.16666
0.16666
1/3
1/2
0.66666
0.16666
1/3
1/2
0.16666
0.66666
1/3
1/2
Page 379
SIGMA/W
Secondary Nodes
Integration Order
Quadrilateral
no
Quadrilateral
yes
Triangular
no
Triangular
yes
15.4
The element matrix for each element in the discretized finite element mesh can be
formed and assembled into a global system of simultaneous equations. The finite
element solution requires the solving of the system of simultaneous equations.
Page 380
SIGMA/W
15.5
Element stresses
SIGMA/W computes the stresses and strains at each integration point within each
element once the nodal displacements have been obtained. Strains are computed
from nodal displacements using Equation 15-3.
Page 381
SIGMA/W
Stresses are computed at each Gauss point using the constitutive matrix [C] in the
following manner:
x
x
y
y
= [C ]
z
z
xy
xy
15.6
Equation 15-4
+ 2 + 3
u = 1
+
3
( 1 2 ) + ( 2 3 ) + ( 3 1 )
2
u = B 3 + a ( 1 3 )
= ( B AB ) 3 + AB 1
After substituting the triaxial stress condition, 2 = 3, into Equation 15-4, it can be
written as:
Equation 15-6
u =
2 3 + + 2 1
3
Page 382
SIGMA/W
2
3
Equation 15-7
( B AB ) =
Equation 15-8
AB = + 2
3
Solving Equation 15-7 and Equation 15-8 simultaneously leads to the following
expressions:
=B
=
B
1
A
3
2
Pore pressure parameters A and B are commonly measured using triaxial tests in
the laboratory. Consequently, SIGMA/W uses A and B as input data in an analysis
and internally calculates and .
Equation 15-4 is presented in terms of principal stresses. For computational
purpose, it is more convenient to express the same equation in terms of general
stress components, x, y, y, and xy. When the pore-water pressure equation is
written in terms of these stress components, it can be expressed as follows.
u = u + u
Equation 15-9
where:
u =
u =
+ y + z )
y ) + ( y z ) + ( z x ) + 6 xy2
2
Page 383
SIGMA/W
Page 384
SIGMA/W
16
16.1
Coordinate systems
The global coordinate system used in the formulation of SIGMA/W is the first
quadrant of a conventional x y Cartesian system.
The local coordinate system used in the formulation of element matrices is
presented in Figure 16-1. Presented as well in Figure 16-1 is the local element
node numbering system. The local coordinates for each of the nodes are given in
Table 16-1.
Table 16-1 Local element node numbering system
Element Type
Quadrilateral
Triangular
Node
+1
+1
-1
+1
-1
-1
+1
-1
+1
-1
-1
SIGMA/W uses the fourth node to distinguish between triangular and quadrilateral
elements. If the fourth node number is zero, the element is triangular. If the fourth
node number is not zero, the element is quadrilateral.
Page 385
SIGMA/W
1 (1, 1)
s
5
2 (1, -1)
4 (-1, 1)
7
Local coordinates (r, s)
Global coordinates (x, y)
3 (-1, 1)
x
A) Quadrilateral Element
y
s
3 (0, 1)
6
2 (1, 0)
r
5
Local coordinates (r, s)
Global coordinates (x, y)
1 (0, 0)
x
B) Triangular Element
Page 386
SIGMA/W
SIGMA/W assumes that the displacement distribution within the element follows
the interpolating functions described previously in this chapter. This means that the
displacement distribution is linear when secondary nodes are not used, and the
displacement distribution is quadratic when the secondary nodes are used.
The displacement distribution model at any given location inside a finite element is
given by the following set of equations:
Equation 16-1 u = N
{U }
Equation 16-2 v = N
{V }
where:
u
{U }
{V }
16.2
= E
In a two-dimensional plane strain problem, there are three basic strain components:
longitudinal strain in the x-direction, x , longitudinal strain in the y-direction, y ,
and shear strain in the x-y plane, xy . SIGMA/W is formulated for small
displacement, small strain problems. The strain components are related to x- and y
displacements, u and v, as follows:
Page 387
SIGMA/W
u
x
v
y =
y
u v
xy =
+
y x
x =
At any point within a finite element, displacements u and v are related to the nodal
displacement vectors {U } and {V } by Equation 16-1 and Equation 16-2. Strains,
when expressed in terms of nodal displacements, can be written as follows:
x =
u
N
=
x
x
{U }
y =
v
N
=
y
y
{V }
xy =
u v
N
+
=
y x
y
{U } +
N
x
{V }
Equation 16-3
r
N
s
x
r
=
x
s
y
s
N
s
in which:
Page 388
SIGMA/W
x
r
Equation 16-4
x
s
y
r = J , the Jacobian matrix.
[ ]
y
s
Thus, the derivative of the interpolation functions, with respect to x and y can be
determined by inverting Equation 16-3:
1
= [J ]
N
where [ J ]
N
s
The Jacobian matrix can be obtained by substituting Equation 16-1 and Equation
16-2 into Equation 16-4.
N1
r
[J ] =
N1
s
N8
N2
K
X Y
r
r 1 1
X Y
N2
s
N8
K
s
2
M
X8
Y8
2
As can be seen from Equation 16-3 and Equation 16-4, derivatives of the
interpolating functions are also required for calculating strains and the Jacobian.
The derivatives of the interpolation functions with respect to r and s used by
SIGMA/W for quadrilateral and triangular elements are given in Table 16-2
and
Page 389
SIGMA/W
N i ,r =
Ni
r
N1,r
(1+s)
-(Nv,r)
-(N8,r)
N2,r
-(1+s)
-(N5,r)
-(N6,r)
N3,r
-(1-s)
-(N6,r)
-(N7,r)
N4,r
(1-s)
-(N7,r)
-(N8,r)
N5,r
-(2r+2sr)
N6,r
-(1-s2)
N7,r
-(2r-2sr)
N8,r
(1-s2)
N1,s
(1+r)
- (N5,s)
-(N8,s)
N2,s
(1-r)
-(N5,s)
-(N6,s)
N3,s
-(1-r)
-(N6,s)
-(N7,s)
N4,s
-(1+r)
-(N7,s)
-(N8,s)
N5,s
(1-r2)
N6,s
-(2s-2sr)
N7,s
-(1-r2)
N8,s
-(2s+2sr)
Page 390
SIGMA/W
N1,r
-1. 0
-(N5,r)
N2,r
1. 0
-(N5,r)
-(N6,r)
N3,r
0. 0
-(N6,r)
-(N7,r)
N5,r
(4-8r-4s)
N6,r
4s
N7,r
-4s
N1,s
-1. 0
-(N5,s)
N2,s
0. 0
-(N5,s)
-(N6,s)
N3,s
1. 0
-(N6,s)
-(N7,s)
N5,s
-4r
N6,s
4r
N7,s
(4-4r-8s)
Page 391
SIGMA/W
M1,r
M1,r
M2,r
(-2-s+s2)/(1-r)2
M2,r
M3,r
(-2+s+s2)/(1-r)2
M3,r
-4(2+s)/(1-r)2(1-s)
M4,r
M4,r
M5,r
(1+s)/(1-r)2
M5,r
M6,r
2(1-s2)/(1-r)
M6,r
2(1+s)/(1-r)2(1-s)
M7,r
(1-s)/ (1-r)2
M7,r
4/(1-r)2(1-s)
M8,r
M8,r
M1,s
M1,s
M2,s
(-r+2s)/(1-r)
M2,s
M3,s
(r+2s)/(1-r)
M3,s
-4(2+r)/(1-r) (1-s)2
M4,s
M4,s
M5,s
(1+r)/(1-r)
M5,s
M6,s
-4s/(1-r)
M6,s
4/(1-r) (1-s)2
M7,s
-(1+r)/(1-r)
M7,s
2(1+r)/(1-r)(1-s)2
M8,s
M8,s
Page 392
SIGMA/W
References
References
Atkinson, J.H. and Bransby, P.L., 1978. The Mechanics of Soils: An Introduction
to Critical State Soil Mechanics. McGraw-Hill.
Bathe, K-J., 1982. Finite Element Procedures in Engineering Analysis. PrenticeHall.
Bettess, Peter., 1992. Infinite Elements. Penshaw Press.
Biot, M.A. , 1941. General theory of three-dimensional consolidation. Journal of
Applied Physics, Volume 12, February, pp. 155-164.
Britto, A.M. and Gunn, M.J., 1987. Critical State Soil Mechanics via Finite
Elements. John Wiley & Sons, Inc.
Chan, D. H.-K., 1986. Finite Element Analysis of Strain Softening Material. Ph.D.
Thesis, Civil Engineering Department, University of Alberta, Edmonton,
Alberta, Canada.
Chen, W.F. and Zhang, H., 1991. Structural Plasticity: Theory, Problems, and
CAE Software. Springer-Verlag.
Dakshanamurthy, V., Fredlund, D.G, and Rahardjo, H., 1984. Coupled threedimensional consolidation theory of unsaturated porous media.
Proceedings of the Fifth International Conference on Expansive Soils,
Adelaide, Australia, pp. 99 -103.
Duncan, J.M. and Chang, C.Y., 1970. Nonlinear analysis of stress and strain in
soils. Journal of the Soil Mechanics and Foundations Division, ASCE, vol.
96, no. SM5, pp. 1629-1654.
Duncan, J.M., Byrne, P., Wong, K.S., and Mabry, P., 1980. Strength, Stress-Strain
and Bulk Modulus Parameters for Finite Element Analyses of Stresses and
Movements in Soil Masses. Report No. UBC/GT/80-01. Department of
Civil Engineering, University of California, Berkeley, California.
Page 393
References
SIGMA/W
Fredlund, D.G. and Morgenstern, N.R., 1976. Constitutive relations for volume
change in unsaturated soil. Canadian Geotechnical Journal, Vol. 13, No. 3,
pp.261 276
Fredlund, D.G. and Rahardjo, H., 1993. Soil Mechanics for Unsaturated Soils.
John Wiley & Sons, Inc.
Gibson, R.E., Knight, K. and Taylor, P.W., 1963. A critical experiment to examine
theories of three-dimensional consolidation. Proceedings from the
European Conference on Soil Mechanics and Foundation Engineering,
Weisbaden, Germany.
Henkel, D.J., 1960. The shear strength of saturated remoulded Clay. Proceedings
of the ASCE Research Conference on Shear Strength of Cohesive Soil,
Boulder, Colorado, USA, pp.533-544.
Hill, R., 1950. The Mathematical Theory of Plasticity. Oxford University Press.
Hinton, E. and Owen, D.R.J., 1979 An Introduction to Finite Element
Computations, Pineridge Press Ltd., U.K.
Lambe, T.W. and Whitman, R.V., 1969. Soil Mechanics. John Wiley and Sons,
Inc.
Lambe, T.W. and Whitman, R.V., 1979. Soil Mechanics, SI Version. John Wiley
and Sons, Inc.
Lancaster, P. and Salkauskas, K., 1986. Curve and Surface Fitting: An
Introduction. Academic Press. pp.101-106
Pickering, D.J., 1970. Anisotropic Elastic Properties for Soils. Geotechnique, Vol.
20, No.3, pp.271-276.
Salkauskas, K., 1974. C splines for interpolation of rapidly varying data. Rocky
Mountain Journal of Mathematics, Vol. 14, No. 1, pp.239-250.
Segerlind, L.J., 1984. Applied Finite Element Analysis, 2nd Ed. John Wiley and
Sons, Inc.
Page 394
SIGMA/W
References
Page 395
References
SIGMA/W
Page 396
SIGMA/W
Index
Index
Anisotropic elastic model ............112
Cam-clay
soil parameters..........................135
File
open initial conditions ..............177
Fill placement
surface settlements ...................193
Contouring
Gauss point values....................249
problems when contouring Gauss
point values ...........................250
Convergence
Do numerical experiments.............30
Draw
Interface elements..........................96
Joining regions...............................90
Element nodes................................70
Page 397
Index
SIGMA/W
Meshing .........................................69
Surface mesh..................................86
Tension zones ..............................239
Transfinite meshing .......................82
Triangular regions..........................81
understand physical process ..........21
Units
consistent set ..............................67
Stop-Restart .................................240
Why model?...................................15
Page 398
Units
Metric
Imperial
Geometry
meters
feet
F/L
kN/m
pcf
F/L3
kN/m3
pcf
Cohesion
F/L
kPa
pcf
Pressure
F/L2
kPa
psf
Force
kN
lbs
kPa
psf
E (modulus)
F/L
Units
Metric
Imperial
Geometry
meters
feet
F/L
kN/m
pcf
F/L3
kN/m3
pcf
Cohesion
F/L
kPa
pcf
Pressure
F/L2
kPa
psf
Force
kN
lbs
E (modulus)
F/L2
kPa
psf