Professional Documents
Culture Documents
Declaration statement
The author certifies that all material contained within this report is his own work except
where it is clearly referenced to others.
Signed:
Date: 06/05/2006
Acknowledgements
Mike Drew of Cofra UK Limited, for his guidelines on the costing of prefabricated vertical
drains and their installation in practice.
Abstract
This project is concerned with the design of a program to calculate the optimal solution for
geotechnical problems involving the consolidation of soft ground by the use of prefabricated
vertical drains. The effects being taken into account include smear, well resistance, ramped
loading and multiple layers. The program has allowed the author to complete a series of
parametric studies into the effects of the factors which contribute to the rate of consolidation
using vertical drains. The final solution is a distributable program that uses an intuitive
graphical user interface. This allows the user to input soil parameters and assumptions and
then run the program to find the optimal spacing of drains to achieve a given consolidation in
a given time. An element of probabilistic analysis has also been incorporated into the program
to allow the creation of risk to cost curves for any parameters. This allows the user to make an
educated decision based on the allowable cost and the degree of certainty in the soil
parameters.
ii
CONTENTS
Contents.....................................................................................................................................iii
List of Figures ............................................................................................................................ v
List of Tables............................................................................................................................. vi
List of Symbols ........................................................................................................................vii
1
Introduction .................................................................................................................... 1
2.1.1
2.1.2
2.1.3
Smear.................................................................................................................. 2
2.1.4
2.1.5
2.2
2.3
2.4
3.2
3.2.1
3.2.2
3.2.3
3.2.4
3.3
3.4
3.4.1
3.4.2
Probabilistic analysis........................................................................................ 18
6.1
6.2
6.3
Installation of Verticalc............................................................................................ 32
8.2
8.3
8.4
8.5
8.6
8.7
8.8
Troubleshooting ....................................................................................................... 39
9.2
Equations.................................................................................................................. 41
9.2.1
Hansbo (1981):................................................................................................. 41
9.2.2
Olson (1977)..................................................................................................... 41
9.2.3
Carrillo (1942).................................................................................................. 42
iv
List of Figures
Figure 1: Consolidation curves as described by Hansbo (1981). n=25...................................... 5
Figure 2: Plot of the effect of Smear for varying depth and time .............................................. 6
Figure 3: Plot of the effect of Well Resistance for varying depth and time............................... 6
Figure 4: Effect of Smear in relation to depth and spacing of drains......................................... 7
Figure 5: Plot of variation of U with depth - double vs. single drained conditions. .................. 7
Figure 6: Comparison of the effect of ramped loading on the consolidation process................ 8
Figure 7: Internal workings of the matlab ramped loading script .............................................. 9
Figure 8: Comparison of ramped loading methods.................................................................. 10
Figure 9: Input matrix for the matlab working script............................................................... 10
Figure 10: Matlab script modification for multiple layers ....................................................... 10
Figure 11: Variation of the effect of Well Resistance in a multilayered soil........................... 13
Figure 12: Contribution to the consolidation process by vertical drainage.............................. 14
Figure 13: Relative Effect of Vertical Drainage ...................................................................... 15
Figure 14: Graphical comparison between Leo (2004) and the present paper......................... 17
Figure 15: Example of a Beta Distribution .............................................................................. 19
Figure 16: The updated matlab script for variable input values............................................... 20
Figure 17: PERT analysis matlab script................................................................................... 21
Figure 18: Risk versus Cost curve from PERT analysis .......................................................... 22
Figure 19: The effects of well resistance on the consolidation process ................................... 23
Figure 20: The effects of smear on the consolidation process ................................................. 23
Figure 21: The effects of spacing on the consolidation process .............................................. 24
Figure 22: Principle of the Graphical User interface ............................................................... 25
Figure 23: GUI Initial Screen - Generic Version ..................................................................... 26
Figure 24: GUI Initial Screen - Cofra UK Version.................................................................. 27
Figure 25: GUI output showing Leo (2004) model.................................................................. 28
List of Tables
Table 1: Values for the initial analyses ...................................................................................... 5
Table 2: Comparison of the superposition method to Olson (1977) ........................................ 11
Table 3: Assumptions and Values used in the Leo (2004) analysis......................................... 16
Table 4: Comparison of the Leo (2004) results to those of the author..................................... 16
Table 5: Installation Costs........................................................................................................ 18
Table 6: Z - values for beta distribution................................................................................... 19
Table 7: Converted parameters from the Leo (2004) Analysis................................................ 27
vi
List of Symbols
Symbol
description
ch
Coefficient of consolidation
kc
mv
qw
Zone of smear = ds / d
Time (years)
Th
Degree of consolidation
vii
INTRODUCTION
The aim of this project is to provide a valuable tool for the design of vertical drains in
practice, with the emphasis being on the combination of an accurate prediction of
consolidation settlement and the ease and speed of use. In practice currently there are two
feasible approaches to the design of vertical drains. Firstly to use simple design tools and
methods based on basic assumptions (such as a single homogenous layer) to calculate the
consolidation then interpolating between different methods to gain a more informed estimate
of the settlements. The second option is reserved for projects that have little tolerance (such as
nuclear power plants), which involves using a finite element or difference program to
calculate the settlements. This second method is much more time consuming, and many more
of the soil parameters are required. In practice generally the first option is used. The aim of
this project is to fill the void between the two methods to give a quick estimate of the
settlements with the minimal amount of required information and time. The factors that are
going to be taken into account in this project are Ramped Loading, Smear, Well Resistance
and Multiple Layers.
2
LITERATURE REVIEW
sand and had large diameters, 0.4 0.6m, but as the technology developed methods of
decreasing the construction costs led to the development of the band drain that have a
comparatively small diameter. This led to the problem of well resistance, where by the drain
no longer has sufficient discharge capacity to cope with the volume of water drained from the
clay. The problem was first addressed from a design aspect by Yoshikuni & Nakanado (1974)
who arrived at a rigorous solution. Work since then has been carried out by Hansbo (1981)
who arrived at a simplified solution for the same case, which from testing provides a strong
correlation with the more rigorous Yoshikuni & Nakanado (1974) solution.
2.1.3 Smear
During the installation of vertical drains a mandrel is used to force the drain into the clay
layer. Although there are many ways of installing drains such as auger drilling, water-jetting,
the most common method is by the closed end mandrel. The mandrel causes disturbance to
the surrounding soil, and leads to a change in horizontal permeability in this area. The
disturbed area is known as the smear zone. Again Yoshikuni & Nakanado (1974) calculated a
rigorous solution and Hansbo (1981) the simplified version.
2.1.4 Ramped Loading
For all the solutions mentioned above the load is assumed to be placed instantaneously, in
practice this is impossible, and ramped loading needs to be accounted for. Olson (1977)
introduced a solution for a single ramped load, for both vertical and radial drainage separately
without account for the influence of Smear and Well Resistance. When used in conjunction
with Carrillo (1942) formula for combining vertical and radial flow, average flow for a whole
homogenous layer can be calculated. Through a process of superposition of the ramps,
variation in ramping can be taken into account using Olsons (1977) solution. Zhu & Yin
(2004) also developed a solution for ramped loading for combined radial and vertical flow
independent of Carrillo (1942). The authors also extended their solution to cover the effects of
smear as well Zhu & Yin (2004), although well resistance is still not accounted for.
2.1.5 Multi-Layered Systems
Most solutions for problems based on the simplification that the clay layer is homogenous. In
practice this is never the case as even homogenous soils have varying values with depth (such
as the coefficient of compressibility). Very little research has been done to create a single
formula to take into account these variations. The layering of the soil dramatically reduces the
vertical permeability of the soil and means that assumptions such as those of Zhu & Yin
(2001, 2004) and Olson (1977) that combine vertical and radial drainage are void. Zhu & Yin
(1999) also formulated a solution for double soil layers under ramped loading, but again the
2
solution disregards the effects of smear and well resistance. Onoue (1988) developed a finite
difference solution that was able to take into account the vertical and horizontal flow from
each layer independently. These were combined with the effects of smear and well resistance,
although the actual formulae are not present in the paper. Onoue (1988) also developed a
method by which the effects of layering can be taken into account for simple analyses; this
method is recommended in Moseley and Kirsch (2004) for the practical design of drains.
2.2 Assumptions and Values
From the review above of the main work on calculating the settlements it can be seen that
only a select few have combined all of the major contributing factors into one solution, and no
author has created a way of calculating the solution quickly. (Finite difference programs do
take time to set up.) The authors program will be based on the Hansbo (1981) method of
calculating consolidation settlements as this is a relatively quick and accurate method that
lends itself to manipulation for taking account of ramped loading. I will use Olson (1977) to
check my results for the case of ramped loading as this is an accepted standard for the
calculation of the ramped load.
Other than the actual formulae used for the calculations, thought also has to be given to the
values used within such formulae. Well resistance is dictated by the discharge capacity of the
drain, the length of the drain itself and whether the drain is single or double drained. The
drainage length is fixed for any given test as are the drainage conditions, but the discharge
capacity is a function of the lateral earth pressure at depth. The discharge capacity is a known
variable and the data is provided by the drain manufacturer following laboratory tests.
In the case of smear one can not be certain of the parameters for the diameter of the smear
zone and the reduction in permeability caused in such a zone, without experimental data to
back up such values. Hird & Moseley (2000) suggest values of s = 1.6 and kc kc' = 3 for
heavily stratified clay this is backed up by small scale ( = 254mm) test models and pore
water pressure measurements. Hansbo (1981) also suggested similar values of s = 1.5 and
kc kc' = 3 but without any experimental data to back up the values. Sharma & Xiao (2000)
conducted a series of tests on a large scale ( = 1m) in a single homogenous clay layer with
values of kc kc' = 1.3 and the zone of smear being about 4 times the size of the mandrel. Note
this is not the same as stating s = 4 , in reality s is now a function of the depth as the depth
increases so does the size of the mandrel required to penetrate to that depth. Indraratna &
Redana (2000) suggest a value of s = 3 4 , but give no numerical value to the reduction in
horizontal permeability apart from stating that the relationship is linear.
Although some authors Hansbo (2001), argue that the conditions for darcian flow are not
3
always valid, for the purposes of this project darcian flow is assumed as in Hansbo (1981) and
Yoshikuni & Nakanado (1974).
2.3 Factors affecting performance
As with all geotechnics there is an inherent uncertainty in the soil parameters used for design.
This includes such vital information as the horizontal permeability of the soil, to which the
whole radial consolidation process is linked. Chu et al. (2004) investigated the effects of
different factors on drainage, including guidance on the selection of PVDs and soil
parameters. Chai and Miura (1999) investigated the effects of the rectangular band drain
compared with circular wells in relation to the effects of smear, concluding that a circular
analysis agrees very closely with that of a rectangular analysis. The analytical model used has
an effect on the accuracy of the final solution Hawlader et al. (2002) compared Barron (1948)
with a finite difference analysis, showing a very high degree of agreement between the two.
Chu et al. (2004) also compared the Hansbo (1981) equation to a finite element analysis
concluding that for most cases Hansbo (1981) is good estimation for design purposes. There
is no real necessity to use a finite difference or element analysis to create an overly accurate
prediction based on uncertain parameters, when simple equations Hansbo (1981) can be used
to calculate equally valid predictions.
2.4 Practical Data
From contact with Cofra UK Ltd, the specifications for various drains were supplied giving
the analysis a realistic basis. The specifications for the Mebradrain series of drains are
contained within appendix 2. Chu et al. (2004) describe how the soil parameters can affect the
possible choice of drain by factors such as clogging and buckling of the drain.
3
PROGRAM DESIGN
kc kc' = 3 . For the purposes of this paper the values below have been used for all analyses
5.5
0.25
1.5
0.06557
0.5
m2/year
m2/MN
m
m
years
H
l
z
qw
20
10
10
1736
m
m
m
m3/year
0.01
0
0.1
10
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Simple
Well
From figure 1 the effect of smear and well resistance can be clearly seen as having a
considerable effect on the rate of consolidation of the soil.
From the equations a series of studies were completed into the effects of each of the factors
direct contribution to the retardation of the consolidation process with relation to depth. It can
be seen from the relative maximum magnitude of the graphs in figure 3 that well resistance
can have a very considerable effect on the process with up to a 0.37 reduction in average
5
0.08
0.07
5m
10m
15m
20m
25m
30m
35m
40m
0.06
0.05
0.04
0.03
0.02
0.01
0
0.01
0.1
10
100
Time (years)
Figure 3: Plot of the effect of Well Resistance for varying depth and time
0.4
0.35
0.3
5m
10m
15m
20m
25m
30m
35m
40m
0.25
0.2
0.15
0.1
0.05
0
0.01
0.1
1
Time (years)
10
100
A study was completed to see the effects of smear when varying the drain spacing. The results
prove interesting as the relative effect of smear varies only a little with increasing spacing, but
once again smear has the greatest effect in shallow soils, or more accurately well resistance
becomes the dominating factor in deeper soils.
Figure 4: Effect of Smear in relation to depth and spacing of drains
0.14
0.12
0.1
5m
10m
15m
20m
25m
30m
35m
40m
0.08
0.06
0.04
0.02
0
1
1.5
2.5
3.5
4.5
Spacing D (m)
One thing to note is that the values for the average degree of consolidation used so far are for
mid-depth in the soil body. As Hansbos
Degree of Consolidation U
0
0.2
0.4
0.6
0.8
6
Depth D (m)
14
depth
16
conditions.
18
20
U Smear + Well
U Smear + Well
double
vs.
single
drained
Simple Consolidation U
shown in blue is independent of depth. Smear in also relatively independent of depth showing
an almost direct relationship with well resistance. The figure also shows the massive effect of
the drainage condition on the drain itself, with the dashed lines denoting a single (top) drained
well. This has implications for the practical applications of vertical drains which will be
discussed later.
3.2.2 Ramped loading application
The next step was to introduce the idea of a ramped loading factor into the calculations.
Traditionally in practice this was done by applying the load half way through the construction
period, as this gives a reasonable degree of accuracy for long periods of time but in the short
term it is only a very rough estimate. To compare the effects of ramped loading the Olson
(1977) equation was plotted on the same axes as the Hansbo (1981) equations. From figure 6
it can be seen that the effects of ramped loading far outweigh those of well resistance and
smear combined. Also the method of placing the full load half way through the construction
period is shown to be inaccurate for short times (although it does provide a more informed
view than Hansbos (1981) simple equation alone.)
Figure 6: Comparison of the effect of ramped loading on the consolidation process.
0.01
0
0.1
Time (Years)
10
0.1
Average Consolidation
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Simple
Well
Well + Smear
Ramping
The author then used Hansbos (1981) simple consolidation equation to model the Olson
(1977) curve, thus creating a link between the ramped loading consolidation and that of smear
and well resistance. This allows the combined analysis of smear, well resistance and ramped
loading in one program without the need for finite difference or element formulations. The
8
ramped loading was accounted for by the superposition of many small steps of loading at
equal time intervals, the smaller the step the more accurate the solution.
Figure 7: Internal workings of the matlab ramped loading script
tim
consol.m
program call
Initial
Values
drain
tc
Construction Period
s
kckcd
tmax
Step 1:
Set Limits of
Accuracy
Based on tim
value
acc
L=H
layer
qw
Step 2:
Set Boundary
Conditions
If 'drain' =
single
Step 3:
Calculate
Consolidation
without Ramping
If 'drain' =
double
For All
Times
L=H/2
For All
Depths
CALCULATE
CONSOLIDATION
Hansbo Formula
Store U Values for
all calculated
Depths
Hansbo
Accounts for
Smear and Well
Resistance
Uavg is divided by
accuracy to
account for the
superposition
Step 4:
Uavg is
superimposed onto
its self to account
for ramped loading
Step 5:
Output Calculated
Values and Curves
Uavg
The degree of
consolidation associated
with the input 'tim' is found
within the Uavg matrix
Average U
Uavg
The program then was written as a matlab script, as represented in figure 7 to allow a more
flexible interface, smaller step sizes and the ability to easily change variables. Figure 7
represents the basic version of the script, taking into account well resistance, smear and
ramped loading.
By changing the version of the Hansbo (1981) formula used within step 3, the effects of
smear and well resistance can be removed to allow a direct comparison with the Olson (1977)
consolidation curve. In figure 8 the consolidations were plotted against time; it is obvious
from this comparison that the values are very similar, as little distinction can be made
visually. The values were analysed numerically to assess the accuracy of the superposition
method for changing step sizes.
Figure 8: Comparison of ramped loading methods
0.01
0
0.1
Time (years)
10
0.1
Average Consolidation
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Olson (1977)
Matlab Simple
It can be seen from table 2 that the superposition model agrees very closely with the results
directly given by Olson (1977), especially in the case of the 10,000 step model.
Even in the case of the 100 step model the prediction may start out with a large relative error
but in absolute terms the difference is insignificant. The superposition method allows the
effects of smear and well resistance to be accounted for at the same time as the ramped
loading. This can be seen in figure 8, showing how the effects of smear and well resistance
influence the emulation of the Olson (1977) curve. As the method is in script form it becomes
much simpler to perform parametric studies on the influence of individual factors.
10
Olson
0.000165
0.000654
0.001454
0.002558
0.003953
0.005631
0.007583
0.009799
0.012271
0.014990
0.017949
0.021139
0.024553
0.028183
0.032022
0.036064
0.040301
0.044728
0.049338
0.054126
0
6
12
20
0
5.5
5.0
2.0
0
0.20
0.22
0.25]
to, the 3rd the associated kc value (m2/year) and the 4th the mv value (m2/MN). This allows the
script to process the different values as it moves down through the depth of the soil.
In the later versions of the script, step 3 was modified to account for the changes in soil
parameters with depth; the details of this change in step 3 have been demonstrated in figure
10. This modification allows multiple layers to be taken into account quickly and easily with
just a simple adjustment to the layer matrix seen in figure 9. The program assumes that the
assumptions made in the Hansbo (1981) model for well resistance are still valid for a multiple
layer analysis. This was investigated and proven by Onoue (1988) with the testing of a
rigorous finite difference model for multiple layers when compared with Hansbo (1981).
11
layer=[0
1
2
3
0
6
12
20
0
5.5
5.0
2.0
Step 3:
Calculate
Consolidation
without Ramping
0
0.20
0.22
0.25]
For All
Times
Counts up to
the number of
layers in the
system
For All
Depths
Layer
Monitor
layer
Layer
Data
mv
kc
Changes soil
parameters dependant
on depth / layer in soil
Store average U
for all times in
matrix 'Uavg'
Uavg is divided
by accuracy to
account for the
superposition
Uavg
Hansbo
Accounts for
Smear and Well
Resistance
CALCULATE
CONSOLIDATION
Hansbo Formula
average U
Step 4
This assumption for well resistance holds well for small differences between the properties of
the layers (as can be seen in figure 11), but starts to become less accurate when the
differences become larger in the region kc layer 1 / kc layer 2 = 400. The inaccuracies
created by the well resistance assumption lead to a more conservative result than that given by
the Onoue (1988) model. Figure 11 shows the variation in the degree of consolidation
according to depth, comparing the Hansbo (1981) models for smear and well resistance and
smear alone. In reality the pore water pressures created within the drain would show a
continuous distribution throughout the length of the drain, rather than the stepped distribution
12
shown in figure 11. For the purposes of the program, the degree of consolidation is calculated
at all depths, accounting for the variation in soil parameters as shown in figure 11. This
distribution is then averaged and stored in the matrix Uavg with the corresponding time of
interest, to allow for the superposition influence of the ramped loading.
Figure 11: Variation of the effect of Well Resistance in a multilayered soil.
Degree of Consolidation U
0
0.05
0.1
0.15
0.2
0.25
0.3
Calculation assumptions:
Layer 1: Kc = 5.5 m2/yr
mv = 0.20 m2/MN
Layer 2: Kc = 4.5 m2/yr
mv = 0.22 m2/MN
Layer 3: Kc = 3.0 m2/yr
mv = 0.25 m2/MN
Time of interest 0.1 years
Depth (m)
10
12
14
16
18
20
Well Resistance
Without Well
Average
radial drainage distance is massively shorter than the vertical drainage distance for 99% of
PVD problems. Again this assumption leads to a conservative solution, as any vertical
drainage will only add to the safety factor of the design. This is not the case however for a
homogenous soil where due to the nature of the soil, the coefficient of vertical consolidation
can be obtained with relative accuracy. The contribution made by vertical drainage can
therefore be calculated much more accurately, and thus its effect on the overall consolidation
can be taken into account with the aid of the Carrillo (1942) formula.
As there are no effects of smear and well resistance in the case of vertical drainage there is no
need to apply a version of the superposition method to this case. Instead, Olsons (1977)
formula for ramped vertical drainage can be used to calculate directly the contribution to the
overall consolidation. Figure 12 shows an example of the contributions of both radial and
vertical drainage to the consolidation process.
Figure 12: Contribution to the consolidation process by vertical drainage.
Average Consolidation Curve
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Radial Drainage
Vertical Drainage
Combined Drainage
0.9
1
0.1
0.2
0.3
0.4
0.5
0.6
Time (Years)
0.7
0.8
0.9
Figure 12 shows the contribution of vertical drainage when the spacing of the wells is much
shorter than the vertical drainage distance. (Spacing (D) =1.5m, Depth (H) =10m). If the
spacing becomes greater then the effect of vertical drainage is much more pronounced, to the
point where the vertical drainage can have more of an effect than the radial drainage. In
reality this would never occur as there would be no realistic benefit from installing the drains
at such great centres. Figure 13 shows the effects of increasing the spacing of the drains and
14
the effect this has on the relationship between the vertical and horizontal drainage. So long as
the Uv/Uh value is less than 1 the effect of Radial Consolidation outweighs that of Vertical
Consolidation, and it is worth while implementing a drainage scheme. This equates to a
maximum viable spacing for a 10m deep soil of 5.2m and 6.4m in the case of the 20m deep
soil. Bearing in mind that these spacings correspond to n values of 79 and 98 respectively
they are way outside the normal constructional limits (for Cofra, the maximum viable
installation n value is 38). In the lower regions of figure 13 there is little difference between
the different depth plots; this adds weight to the argument that if the spacing can be
minimised the effects of vertical drainage can also be minimised.
Figure 13: Relative Effect of Vertical Drainage
4.5
4
3.5
Depth = 20m
Depth = 10m
2.5
2
1.5
1
0.5
0
4
5
6
Drain Spacing - D (m)
10
finished model was accurate. Unfortunately there are few authors who have published work in
the same area; that of writing programs to model all the associated factors along with multiple
layers. Onoue (1988) published graphical outputs from a finite difference program, modelling
all the factors and more (such as vertical drainage in the individual layers), but these are
impossible to emulate as the input parameters and assumptions are unknown. Leo (2004)
wrote a spreadsheet program to model all the factors except for the consideration of multiple
layers. The input values and assumptions for the results calculated were given in the paper
and this allowed the direct comparison of the Leo (2004) model to that of the author.
Table 3: Assumptions and Values used in the Leo (2004) analysis
ch = 2.25 m2/year
kh / ks = 3
ch / cv = 3
rw = 0.033 m
h = 4.5 m
re = 0.4725 m
rs = 0.075 m
s = 1.6
With the values in table 3 the author used the matlab script created for a single layer analysis
to model the results generated by Leo (2004). One thing to note is that as the soil is relatively
shallow (4.5m) Leo (2004) has not included the effects of well resistance in the analysis. To
this effect the author has used the ideal drain model (infinite drain permeability) to more
accurately compare the models.
Table 4: Comparison of the Leo (2004) results to those of the author
Ramp
Leo (2004)
0.389
0.632
0.893
0.975
0.994
Ramp No Well
Present
Relative Error
0.394
1.269%
0.634
0.284%
0.894
0.078%
0.976
0.062%
0.994
0.040%
0.083
0.292
0.542
0.792
tc = .083
0.236
0.783
0.950
0.989
0.237
0.784
0.951
0.989
0.590%
0.166%
0.095%
-0.030%
0.232
0.775
0.946
0.987
-2.239%
-1.265%
-0.518%
-0.182%
0.083
0.292
0.542
tc = .208
0.448
0.876
0.972
0.454
0.878
0.972
1.300%
0.182%
-0.010%
0.445
0.870
0.969
-1.977%
-0.920%
-0.341%
Averages
0.335%
Time (yrs)
0.167
0.250
0.458
0.708
0.958
tc = 0.167
-1.048%
Table 4 demonstrates the accuracy of the program, with the maximum relative error being
16
1.3% and the average error for all results being only 0.335%. The graphical check in figure 14
shows without a doubt the high degree of correlation between the models. Figure 14 also
demonstrates the validity of Leo (2001) in disregarding the effect of well resistance for the
analysis as it had a maximum influence of 2.1% and an average influence of only 1%. If the
soil in the analysis was deeper or the resistance of the drain very low then well resistance
would have a much more influential part to play.
Figure 14: Graphical comparison between Leo (2004) and the present paper
Time (years)
0.15
0.3
0.25
0.35
0.45
0.55
0.65
0.75
0.85
0.95
0.4
0.5
0.6
0.7
0.8
0.9
1
Leo 2004
Present Paper
From the program checks done above it can be concluded that the single homogenous layer
soil analysis, taking into account well resistance, smear, ramped loading and vertical drainage,
is 99% accurate in testing. The expansion of the model for multiple layers is valid (Onoue
1988) and although it cannot claim the same degree of accuracy due to inherent errors in the
assumptions and the loss of the vertical drainage, it can claim to always give a conservative
result for the cases analysed. There are no precedents to test the multiple layer model against
to ascertain its exact accuracy.
3.4
Program Optimisation
allowable for the required degree of consolidation. The values for the installation costs in
table 5 were provided by Mike Drew of Cofra UK Ltd.
Table 5: Installation Costs
N value
Material Cost
Installation Cost
17 20
0.20 / m
0.34 / m
20 - 30
0.20 / m
0.38 / m
30 - 38
0.20 / m
0.48 / m
Deep Layers of soil that require pre-drilling 0.95 / m
Mobilisation of Equipment to site 8000
Knowing the area of the installation, the depth of the soil, the time allowed and the degree of
consolidation required, it is possible to calculate the spacing required to achieve the
associated degree of consolidation. From the spacing the number of drains required in the
installation can be calculated and using the installation costs in table 5 this can be turned into
a cost for the project. This approach enables the program to find the optimal spacing to
achieve the requirements provided, and thus the lowest cost solution to the problem. The only
problem with this approach is the variance in the soil parameters, as a small difference from
the calculated solution to those in practice would throw out the answers and not necessarily
lead to the project completing on time. For example a small decrease in Kc due to variation
across the site would lead to the required degree of consolidation not being reached in the
allotted time. The choice of the values for the soil parameters is left at the discretion of the
Geotechnical Engineer based upon the results of a ground investigation. Using a method of
probabilistic analysis the variation in soil parameter could be automatically accounted for,
thus giving the Engineer a method of backing up his instincts.
3.4.2 Probabilistic analysis
What was needed was a method of taking into account the variation in the values of the soil
parameters quickly and easily without needing any more data about the soil conditions
surrounding the installation. As the aim is to provide a design tool it was deemed by the
author unrealistic to expect the user to have to provide any amount of input data extra to that
required by the usual methods of ground investigation, as this would make the program more
complicated and require additional cost to implement at a design stage.
The Programme Evaluation and Review Technique or PERT is one such method. PERT was
originally devised to provide a time estimate based on best, expected and worst case scenarios
with no other inputs into the method, where a = minimum possible time, m = expected time
and b = maximum time. The PERT analysis assumes that the distribution of variables
corresponds to a beta distribution.
18
a + 4m + b
6
(b a)
6
te =
Variance,
v = te2
TS TE
te
Z
0.0
0.3
0.6
0.9
1.3
1.5
2.0
2.5
3.0
P
0.50
0.62
0.73
0.82
0.90
0.93
0.98
0.99
1.00
element, as no direct account of the spread of the results is taken into account.
To allow the PERT analysis to be performed three estimates for the possible duration need to
be calculated - the best, expected and worst case scenarios. These can be calculated by
allowing the user to enter a spread of values for the soil parameter Kc - a maximum,
minimum and average value. When the consolidations are calculated using the different
values of Kc it will give a spread of U values corresponding to the best, expected and worst
case scenarios.
19
Figure 16: The updated matlab script for variable input values
tim
Initial
Values
supplied by
risk.m
consolsimple_fast.m
program call
layer
drain
tc
Construction Period
kckcd
tmax
Step 1:
Set Limits of
Accuracy
Based on tim
value
acc
L=H
qw
Step 2:
Set Boundary
Conditions
If 'drain' =
single
Step 3:
Set upper, lower
and average kc
values
Step 4:
Calculate
Consolidation
without Ramping
layer
If 'drain' =
double
L=H/2
For All
Times
consolmulti_fast.m
includes an extra loop to
account for the changing
kc an mv values with
depth. Steps 6 and 7 are
also removed.
Hansbo
Accounts for
Smear and Well
Resistance
Uavg is divided by
accuracy to
account for the
superposition
Step 5:
Uavg is
superimposed onto
its self to account
for ramped loading
Step 6:
Olson's ramped
equation for vertical
drainage is used for
single layer
Uavg
The degrees of
consolidation associated
with the input 'tim' are
found within the Uavg
matrix for each case.
Step 7:
Carrillo's formula is
applied to combine the
effects for vertical and
radial drainage
avgUmin
a
m
b
20
averageU
Ouput returned to
risk.m to allow
PERT analysis of
soil data
avgUmax
This was written as an extension to the matlab script to allow for the variation in the soil
parameters, which can be seen in figure 16. These values can then be entered into the PERT
analysis to give a degree of confidence in the end consolidation completing on time- fig 17.
Figure 17: PERT analysis matlab script
Arbitrary D
value
tim
risk.m
program call
layer
Initial
Values
drain
tc
'layers'==1
SINGLE LAYER
Construction Period
'layers'==2+
MULTI LAYER
kckcd
qw
consolsingle_fast.m
program call
target U
consolmulti_fast.m
program call
Step1:
Run scripts with
Arbitrary D value to
establish base points
avgUmin
averageU
New D value
avgUmax
No
Step 2:
Use PERT analysis
to calculate the
expected U
Check:
expected U =
target U
expected U
Warning
Message
Displayed
Not
Possible
Yes
D
Last D value
calculated frovides
excellent start point
for further analysis
Step4:
Run scripts with
Arbitrary D value to
establish base points
For All P
Values
Step 3:
Use PERT P and Z
table combined with
formula to change
target U to account
for variable data
probable.m
program call
target U
Warning
Message
Displayed
No
'layers'==1
SINGLE LAYER
'layers'==2+
MULTI LAYER
consolsingle_fast.m
program call
consolmulti_fast.m
program call
Degree stores
average U, avgUmax
and avgUmin
Degree
Not possible
Step 5:
Use PERT analysis
to calculate
associated U
Check:
associated U =
target U
Yes
Step 6:
U value is stored
along with spacing in
a matrix
Step 7:
Cost analysis is
run using Spacing
21
The higher the probability of a timely completion, the closer the spacing needs to be and the
closer the spacing, the more drains are needed to cover the installation area. As the number of
drains goes up so does the cost associated. This link allows the plot of a risk (probability)
versus cost curve (figure 18) to demonstrate how much extra would need to be spent on a
project to greaten the chances of a timely completion.
Figure 18: Risk versus Cost curve from PERT
analysis
5
1.8
x 10
1.75
Target U = 0.95
Completion Time = 1 year
Area of Project = 30000m2
Kc (m2/yr) min= 4, avg= 5.5, max= 6
Calculated Necessary Spacing = 1.20 m
1.7
Cost ()
1.65
1.6
1.55
1.5
1.45
1.4
0.5
0.55
0.6
0.65
0.7
0.75
Probability
0.8
0.85
0.9
0.95
PARAMETRIC STUDIES
To some degree during the writing of the program some parametric studies have been
completed to assess the individual effects of factors affecting the consolidation process. The
full possibilities of using parametric studies could not be fully investigated until the matlab
script for the program was fully completed. Here the author has used parametric studies to
further explore the factors affecting the consolidation process.
Figure 19 shows the effects of well resistance with increasing depth by comparing the time to
reach 90% consolidation between an ideal drain (no resistance) and one with well resistance.
The figure demonstrates the massive effect that well resistance can have on the consolidation
process with a maximum of a 600% increase in the time required (this is the most extreme
case with the drain being 60m in length). What is interesting is the exponential behaviour the
curve demonstrates with the effects being much less noticeable with drains up to a length of
10m (effect at 10m = 15% increase).
22
n = 25
n = 50
While
3
the
influence
of
smear
is
qw/kh = 400 m2
1
of
consolidation.
visual
10
20
30
40
Drain Length (m)
50
60
2.2
Th90(smear) / Th90(no smear)
Kc/Kc' = 2
Kc/Kc' = 3
Kc/Kc' = 4
1.8
Hird & Moseley (2000)
1.6
1.4
1.2
Hansbo (1981)
1
1
1.2
1.4
1.6
1.8
2.2
23
2.4
2.6
2.8
Figure 20 demonstrates the point that a small change in the parameters can have a large effect
on the rate of consolidation, with a 7% difference between the Hansbo (1981) and Hird &
Moseley (2000) factors.
Vertical drains rely on the principle that if you decrease the spacing of the drains then the
consolidation process will accelerate. This said there is a limit to how many drains can be
installed into a soil before it becomes the horizontal permeability of the soil itself that
becomes the limiting factor and not the spacing of the drains. The author conducted a
parametric study into the spacing (n value) of drains and its effect on the rate of consolidation.
Figure 21: Effect of spacing on the consolidation process
20
18
16
14
12
10
8
6
4
2
0
0
10
20
30
40
50
60
70
80
90
100
n (d/D)
Figure 21 shows a flat section in the curve between n=1 and n=17. In this area any decrease in
the spacing has a minimal effect on the rate of consolidation, and thus is uneconomical to
install. Of course figure 21 is only valid for a specific set of initial conditions set in table 1,
although the general trend will always be the same. Although technically it would be possible
to do parametric studies into the effect of ramping and multiple layers the outcomes are much
more dependant on the input parameters provided. For example we know the effect of ramped
loading is directly proportional to the construction period as they are intrinsically linked. This
makes their value as an educational tool much less than the parametric studies in figures 1921.
24
PROGRAM DISTRIBUTION
Previously the matlab scripts that have been used to model the problem have been command
line based, that is there is no visualisation to the script. This approach is used for the
development stages of the program as it allows the author to rapidly change the scenarios and
calculation methods utilised within the scripts. Users other than the author may have problems
in understanding the notation and methodology of this approach, which is where a Graphical
User Interface (GUI) comes in. The GUI allows the input to have a visual element to it,
making the use of the program much simpler and more intuitive. As the task is to create a
design tool the GUI plays an important part in the ease and speed of the use of the program.
To accompany the program a manual would also be written to aid in the rapid use of the builtin features.
Figure 22: Principle of the Graphical User interface
verticalc.m
program call
Initial
Values
Values
input
Switches:
'Drainage'
'Spacing'
'Layers'
Graphical User
Interface
Assumptions
input
Optimum Solution
Creation
In depth Analysis
"CALCULATE"
button
"OPTIMISE"
button
'layers'==1
'layers'==2+
MULTI LAYER
SINGLE LAYER
Input Value
Check
risk.m
program call
DIFFERENT
consolsingle_fast.m
program call
consolmulti_fast.m
program call
SAME
'layers'==1
'layers'==2+
MULTI LAYER
SINGLE LAYER
'acc'=100
probable.m
program call
consolsingle.m
program call
consolmulti.m
program call
olsonvert.m
program call
OPTIMUM SPACING +
ASSOCIATED U VALUES FOR
PROBABILITIES
'spacing'=='tri'
olsonvert.m
program call
speedcalc.m
program call
'spacing'==0
costing.m
program call
SQUARE
TRI
Consolidation
Curve Plot
25
Figure 22 demonstrates the role that the GUI plays in the interaction with the programmed
matlab scripts. The two main functions can be seen clearly, with the optimise function on
the left which acts to calculate the most suitable calculation parameters for the in-depth
calculate function on the right of the diagram. The calculate function can also be used
separately to analyse predictions and other case histories. Figure 23 shows a screenshot of
how the GUI looks when run under Windows XP, with the main area for the generated plots
to the top right, and the area for input parameters around to the bottom and left of the screen.
Figure 23: GUI Initial Screen Generic Version
Two versions of the GUI were created. Figure 23 is the generic version, which has no
assumed values within its programming, so can be used to analyse any combination of
parameters. Figure 24 is designed for use by Cofra UK, with the values for the drain
parameters pre-programmed into the software, along with Cofras own costing data from table
5. From a design point of view customised versions of the software can be created specifically
for a company with their own values inserted into the program, as this makes the program
easier to use for employees as the parameters for the companies products do not need to be
looked up. The differences between figure 23 and 24 occur in the Drain Properties box with
the installation costs being removed from figure 23 and the drop down menu for drain choice
being inserted into figure 24.
26
For the purposes of this paper the author has decided to insert a worked example showing the
capabilities of the program at modelling case histories. In this case the author has decided to
verify the results for the Leo (2004) case, shown in table 4.
Table 6: Converted parameters from the Leo (2004) Analysis
Kc = 5.0 m2/year
Mv = 0.22652 m2/MN
ch / cv = 3
d = 0.066 m
t = 0.25 years
S = 0.9m
Drainage = Single
Depth = 4.5 m
D = 0.945 m
kh / ks = 3
s = 2.273
Tc = 0.167 years
Arrangement = Triangular
qw = inf
As the Leo (2004) analysis does not account for well resistance, putting in an infinite value
into the drain permeability (qw) accounts for this. To model this analysis the Calculate
function will have to be used, as the parameters are already defined and are in no need of
optimising.
27
The program returns an average consolidation of 0.634 which is the same as the value in table
4, column 3, row 4.
For the full details of how to use the program please consult the program users guide in the
appendix. For the purposes of program users guide, the working of the generic version will
be described as this involves more steps than the Cofra UK version.
6
CONCLUSIONS
The aim of this project was to create a design tool to speed up the calculation of vertical drain
installations in practice. In this requirement it was a complete success, as the program has
developed its own intuitive interface (the GUI) rather than using the usual confines of an
excel spreadsheet. While the theory behind the model is relatively simple, it uses Hansbo
(1981) and Olson (1977) methods, which are well known and well used benchmarks for the
industry. Cofras own in-house software for calculating vertical drains solely depends on
these same equations, but without any attempt to manipulate them to account for more than
one factor at any one time, i.e. it uses Olson (1977) for ramped loading and Hansbo (1981) for
28
The author has taken into account the main factors in the designing of a vertical drain
installation. This said there are many other factors that can affect the consolidation process,
such as creep and the inclusion of vertical drainage in multiple layers. With time these could
be included into the program to allow a more in-depth analysis of the soil, although these
were seen as being outside of the scope of the current project. Another possibility would be
for the program to predict the increase in strength in the soil from the associated
consolidation, although this would require many more input parameters to predict accurately.
To allow for a degree of confidence to be placed in the program, a series of tests could be run
on case histories to compare the results predicted by the program to those in practice.
6.3 Further Computing Developments
The author can see a near infinite amount of scope for enhancements to the user interface and
possibilities associated with it. For example a second GUI could be implemented to deal with
the data directly from the ground investigation positions of the samples across the site and
their associated values. This would allow the model to begin to appreciate the three
dimensional nature of the problem and be able to suggest different schemes for different areas
on the site where the parameters vastly differ. The ramped loading could also be expanded to
a number of smaller ramps, rather than a single one, with options for small steps or a linear
approach to the load increments.
29
REFERENCES
Almeida, M. S. S., Santa Maria, P. E. L., Martins, I. S. M., Spotti, A. P., Coelho, L. B. M.
(2000). Consolidation of a very soft clay with vertical drains. Geotechnique 50, No. 6,
633-643.
Barron, R. A. (1948). Consolidation of fine-grained soils by drain wells. Trans. ASCE, Vol.
113, Paper No. 2346.
31
This guide will demonstrate how to run the program under Windows XP, please contact the
author if you require a version of the program to run under any other operating system.
Step 1: Run MCRinstaller.exe This installs the matlab component runtimes
which allows you to run the Verticalc program.
Step 2: Run verticalc_genericm.exe This runs the Verticalc program on your
system. During future uses of the program only Step 2 needs to be
followed.
Once the install has completed you are now ready to use the Verticalc program.
Figure 8.1: Start-up Screen
Once the program has run successfully you should be looking at a screen that looks like that
in figure 8.1. This is the main screen from where all the facilities within the Verticalc program
are available.
32
The Graphical User Interface or GUI is the heart of the use of the program. This section of the
manual will hopefully explain the basic capabilities and how to go about using Verticalc.
Figure 8.2: The GUI breakdown
Radio Button
Chart
Plot Area
Input Boxes
Option
Panel
Drop Down
Menu
Execute
Buttons
Input Boxes allow the input of data through selecting the box and typing in the new
value.
Option Panel refers to the sections of the GUI, this allows for better referencing of the
individual components within the GUI.
Chart Plot Area is where the program generates its visual plots of the consolidation
curves and the cost versus risk curve, more on these later.
Drop Down Menus allow the choice of many different factors from within the
programmed choice, but only one may be selected at any one time.
Execute Buttons click on these to run sections of the analysis suite of the program.
The program itself has two main functions accessed by the Execute Buttons in the Output
Option Panel: Optimise and Calculate.
33
By selecting any white textbox with the mouse the initial values of the program can be
changed. All the parameters are self explanatory, although it should be noted that the reason
for the 3 values for Kc is for use in the probabilistic analysis. The maximum, minimum and
average values from the site investigation should be input.
Figure 8.3: Optimisation Output
For users who wish to disregard the probabilistic element of the program, input the same
value in all three boxes, and the probabilistic analysis will be disabled.
34
Once all the values are in place press the Optimise button to generate an output, as seen in
figure 8.3. By comparing the Spacing Option Panels of figures 8.2 and 8.3 you can see that
the optimise function has created a new set of values for the drain spacing, now based on the
soil parameters and the probabilistic analysis.
8.4 Using the Probabilistic Analysis Tools
The values created in the Spacing Option Panel are only valid for a specific Probability,
which is defined in the Risk Option Panel; the default setting is a 50% probability.
The plot of probability versus cost allows you to decide on the correct balance between
financial cost and the element of risk. To be 100% sure of the consolidation occurring in the
stated amount of time you will need to spend more capital than if you only needed to be 62%
sure. By changing the probability in the Risk Option Panel, the spacing details are updated
accordingly, as is the cost and degree of consolidation in the Output Option Panel.
Figure 8.4: Using the Probabilistic Analysis
Figure 8.4 shows the Probability options in the drop down menu, here we can see by selecting
the 93% probablilty the Spacing values have changed, as has the information in the Output
Option Panel. If any of the details are changed in the input parameters the Optimise function
needs to be re-run to take into account the changes.
35
Once the Probability has been chosen, the Calculate function can be run to give the
consolidation curve for given soil parameters and generated drain spacing.
8.5 Using the Calculate Function
36
A common mistake is to forget to select the drain arrangement; in this case the program will
pop up a warning like the one in figure 8.6 to ask you for the additional data.
Figure 8.6: Error Message
The program easily accounts for layered soils, by just inputting the values for the different
layers into the layers option panel, the program will automatically take the different values
into account. The resulting output is much the same as figure 8.5 but as the program no longer
accounts for vertical drainage there is only a single plot in the chart plot area.
Figure 8.7: Multiple Layer Output
As you can see in figure 8.7 the Layers Input Panel now has an extra set of data for the 2nd
soil layer, extending down from 10-15m depth. The Optimise function also works for multiple
layers using the same method as mentioned here.
37
The advanced features are contained within the Plot View input panel. The Cost and Consol
buttons allow you to switch between the consolidation plot and the cost versus probability
plot, this is very useful if you need to change the probability used once a calculation has been
run without having to re-run the Optimise function.
The other feature in the Plot View input panel is the Comparison drop down menu. This
allows you to compare the effects of changing the input parameters or the probability visually
in the chart plot area.
To compare two sets of Parameters:
Run the Calculate function on the first set of soil parameters, then change the
parameters and re-run the Calculate function taking care to make sure the drop
down menu is set to Comparison.
Figure 8.8: Comparison Feature
Figure 8.8 is comparing the effects of changing the drainage conditions from single drainage
conditions to double, this has the effect of speeding up the consolidation as the vertical
drainage has much more of an effect in the double drainage conditions.
38
The only real problems can arise in the Optimise function where there is no solution for the
set of parameters chosen. Or the wrong kind of inputs are used in the program. Examples of
the error messages are shown in figures 8.9 and 8.10.
Figure 8.9 Optimisation Error Message
If
you
accidentally
input
non-numeric
39
Standard
Unit
MD7007
Configuration
Channels
Material
Weight
Width
Thickness
Mechanical properties
Tensile Strength
Elongation
Elongation at 0.5 kN
Grab strength
Bursting Strength
Tear Strength
Hydraulic properties drain
In-plane flow cap. qp(10/1.0)
In-plane flow cap. qp(100/1.0)
In-plane flow cap. qp(350/1.0)
Discharge cap.
qw(300/0.1)
Discharge cap.
qw(500/0.1)
D.C. buckled
qwb(200/0.1)
Transmissivity
(10/0.1)
Transmissivity
(200/0.1)
Discharge cap.
qw(200/0.1)
Discharge cap.
qw(300/0.1)
Discharge cap.
qw(500/0.1)
D.C. buckled
Permeability
k
Pore Size
O95
Transport Details
Roll length
Roll diameter
Inside diameter
Weight Roll
40 ft container
g/m
mm
mm
38
PP
75
100
3.0
44
PP
70
100
3.0
44
PP
85
100
3.5
44
PP
110
100
5
EN 10319
EN 10319
EN 10319
ASTM D4632
ASTM D3785
ASTM D4533
kN
%
%
N
kPa
N
2.2
60
2
970
1000
270
1.8
40
2
580
900
180
2.2
60
2
970
1000
270
4.2
60
1.5
970
1000
270
EN 12958
EN 12958
EN 12958
EN 12958
EN 12958
EN 12958
ASTM D4716
l/m.s
l/m.s
l/m.s
10-6 m3/s
10-6 m3/s
10-6 m3/s
10-3 m2/s
1.1
0.75
0.59
49
1
60
1.2
50
14
55
0.6
0.8
0.68
0.57
70
20
82
0.94
2.7
2.5
1.8
155
25
130
2.5
ASTM D4716
ASTM D4716
ASTM D4716
ASTM D4716
ASTM D6918
10-3 m2/s
10-6 m3/s
10-6 m3/s
10-6 m3/s
%
0.7
87
55
14
37
0.5
2.2
55
25
22
0.9
94
76
42
18
175
80
28
EN 11058
ASTM D4491
ASTM D4491
ASTM D4751
mm/s
s-1
10-4 m/s
m
16
0.3
1.3
75
5
0.3
0.3
75
16
0.3
1.3
75
16
0.3
1.3
75
m
m
m
kg
km
300
1.1
0.15
22
160
300
1.1
0.15
22
160
250
1.2
0.23
25
130
200
1.2
0.23
25
80
40
9.2 Equations
8Th u
In which,
n=
kM
D
, ch = c
d
w
and Th =
ch t
D2
u r = u +
z (2l z )
2
n qw
k
n 2 n kc
3
s2
s 2 kc 1 s 4 1 2
1
ln
ln
1
s
s + 1 + z (2l z ) c 1 2
+
2
2
2
2
2
4 n 1 4n k ' c n 1 4n
qw n
n 1 s k 'c
Fn = ( N ) 2
, N = e and A =
2
N 1
4N
rw
Fn
2
For Tr Trc ,
Ur =
Tr
1
[1 exp( ATr )]
A
Trc
1
[exp( ATrc ) 1]exp( ATr )
Trc
T 2
1
2
1 4 1 exp( M T )
Tc T
M
41
2
Tc
1
M [exp(M
4
Tc ) 1 exp( M 2T )
42