Morgenstern etc.

, Spencer’s method is preferred since it satisfies all equilibrium

conditions (e.g. overall moment, individual slice moment, horizontal and vertical force
equilibriums) and it is suitable for slip surface of any shape (Pockoski and Duncan,
2000). Additionally, Mohr-Coulomb strength type is used in the analyses. Normal
distribution is used in the absence of statistical information about variables.

Despite having some differences, Slope/W 2012 is also working with the same
principle (e.g. LEM) as Slide software and they are practically the same. For examples,
Slope/W does not have LH sampling options and does analysis comparable to GM
type of analysis. However, as opposed to Slide, Slope/W has a slip surface search
option by defining the possible entry and exit locations of the surface. In some of the
cases, this may be useful.

3.2 Finite Element Analyses: PLAXIS 2D and Phase2

PLAXIS 2D 2015 is one of the 2D finite element geotechnical analysis tools. Besides
its plastic deformation analysis and other analyses options, it can also carry out slope
stability analysis with strength reduction method. It consists of mainly two parts. These
are input and output programs. Input program has 5 subsections, namely soils,
structures, mesh, flow conditions and staged construction. Units, dimensions of
working place, model type (plane strain or axisymmetric) and finite element type (6 or
15-noded) can be defined in the project properties part. Model geometry can be defined
by either inserting borehole or drawing soil polygons in soil or structures subsections,
respectively. Materials can be created and assigned into the soil regions by suing the
material sets options. In there, soil parameters, type of drainage, constitutive material
model, seepage and groundwater conditions can be defined. In the mesh section,
geometry is divided into finite elements (forming a mesh) with a desired level of
fineness. The software has 5 options for mesh dimensions. These are very fine, fine,
medium, coarse and very coarse. Additionally, local fineness can also be added into
the meshes. After that, ground water level (e.g. phreatic level) can be added in the flow
conditions section. After finalizing the input geometry, desired level of calculation
stages (e.g. phases) can be created in accordance with the purpose of the analysis. In
slope stability analysis, there should be at least two sages. First one creates the initial
field conditions which can be either K0 loading or gravity loading. For non-horizontal
57
layers, gravity loading should be used in order to create initial stresses. Second phase
performs the strength reduction analysis by reducing the cohesion and tangent of the
friction angle with the same number until failure occurs, giving the strength reduction
factor (SRF). If there are any loading or structural elements, they can be activated or
deactivated within these stages. There are also other advanced options available for
the stages. In the output, several analysis results, including deformations and stresses,
can be viewed for individual stages. After going to the strength reduction phase results,
one can obtain the failure surface zone from “Incremental Cartesian Strain” option
viewing the incremental shear strains, and reached SRF value can be seen from the
“Calculation Information”.

Although PLAXIS is a powerful and useful geotechnical tool, due to lack of

probabilistic built-in option, not every aspects of the problem can be considered. Since
2014, however, PLAXIS 2D user interface has changed with anniversary edition
version and the designers can interrupt the software with PLAXIS VIP license and
Python coding. This is explained in one of the later sections. In addition to that, latest
update of PLAXIS 2D 2015 is capable of doing a sensitivity analysis but it is very
limited.

Despite having lots of common properties, Phase2 v8 has several advantages compared
to PLAXIS 2D in terms of probability analysis. It can carry out probabilistic finite
element analysis with Rosenblueth’s Point Estimate Method. It can easily import from
or export into Slide software so both LEM and FEM analyses can be carried out, if
desired. After the probabilistic analysis, it gives failure zone, factor of safety and
reliability index information.

3.3 Random Finite Element Analyses: Rslope2D

As stated before, G.A. Fenton and D.V. Griffiths created a series of software in 1992.
Rslope2D is one of them. It is a 2D finite element slope stability analysis software that
considers the spatial variation and correlation of the soil properties. The software takes
into account of the mean, standard deviation and spatial correlation length of the input
parameters as well as local averaging of the properties over the finite elements (Fenton
and Vanmarcke 1990). The software applies the Monte-Carlo method for the
probabilistic analysis. It has a user interface in which input parameters and several
58
other options can be specified. Vertical and horizontal correlation distance, statistical
distribution of the random material and the number of analyses can be specified as
well. Additionally, soil material properties can also be correlated to one another. After
the analysis, the software provides a strength reduction value and deformed mesh view
which can show the failure surface location.

Despite its advantage of spatial correlation consideration, Rslope2D has some

limitations, therefore, disadvantages. One of them is that the software cannot
incorporate the existence of the ground water level, i.e. it can only deal with dry cases.
Another is that it can only work for simple slope geometry and single layered soils. In
other words, geometries having more than one inclined slope surface and layers cannot
be analyzed with this software. However, as stated before, it is an open-source coded
software so it can be improved if desired.

Rslope2D uses initial generator seed value for the random number generation process.
It can either be positive integer or zero. If desired, exact same result (e.g. same
sequence of random numbers) can be reproduced by starting the same seed number
and input parameters. This is especially useful when investigating the effects of the
input parameters on the results (e.g. parametric studies). If the initial seed value is
given as 0, the software starts the initial seed value from the computer. Depending on
the operating system, it uses either the process ID or system time. If more than one
analyses are carried out, second seed number is the previous seed plus one and this
goes like that. Therefore, if one wants to regenerate a certain analysis step (say nth
analysis), it can be done by giving the seed number of initial seed plus n-1.

3.4 Random Finite Element Analyses: PLAXIS Remote Scripting with

Python

As explained before, PLAXIS 2D 2014 or newer versions allows for interrupting the
flow of software with remote scripting via Python which is one of the available
programming languages. Both inputs as well as outputs can be created, called, stored
or altered. For this purpose, aside from the aforementioned software, PLAXIS VIP
license and internet connection are required. In the PLAXIS, there are several
commands for every specific action and they can be viewed from PLAXIS Command
Reference manual which comes with the installation of the software. Along with this,
59
required amount of Python syntax knowledge is necessary. Python 3 is utilized for the
scripting purpose.

Remote scripting can also be made partially for a specific purpose. For example, after
manually creating inputs and carrying out the analysis, certain output value can be
called and stored into an array object via scripting. Even this provides convenience in
terms of human effort and time. However, remote scripting is more efficient when it
is applied from the start till the end. For example, if a remote scripting file for a slope
stability analysis is created with a general, always valid expressions, it can be reused
in the future without any further effort. Remote scripting works can also be used for
optimization problems where the designer should change almost every input
parameters till the desired optimal result is reached. Since doing this procedure
manually is quite time consuming, scripting provides convenience for the user.

As explained before, Rslope2D has three main limitations. In order to resolve these
limitations, remote scripting is done for PLAXIS 2D that enables the software to carry
out multiple slope stability analyses with a randomly assigned material properties. This
way complex layered soil geometries with a ground water level can be analyzed with
the powerful computational capacity of PLAXIS. At the same time, multiple analyses
with a randomly generated and assigned material properties can be carried out with a
desired number of realization for example for Monte Carlo simulation. Scripting is
also done so that several soil regions are created with different soil strength properties.
This way heterogeneous and anisotropic soil media can be created and analyzed.

For the current study, coding is developed so that all analysis steps are automatized.
Created Python script is provided as an appendix at the end. General working
principles of the coding along with the procedure steps are as follows:

(1) Remote Scripting Server: Under the “expert” tab of PLAXIS 2D, the users can
configure and start the server (Figure 3.1). In here, availability of a valid license
(should be VIP), connection to remote service website and local connection is
shown. This is the first step of remote scripting and it is done before running
the Python coding file. As it is stated before, coding is done so that the user
does not have to do anything in the PLAXIS 2D environment after this step.
The scripting itself connects and carries every step in the PLAXIS 2D.
60
(2) Running the Python coding file created in this thesis study: First, script sets up
a connection between PLAXIS and Python. Then, it askes a number of input
parameters to the users via the Python interface.

Figure 3.1 Configuration of remote scripting server in PLAXIS 2D

(3) Inputs: The created Python code asks information about the external geometry,
number of layers and their boundaries, coordinates of the ground water level,
finite element mesh dimension and material properties (e.g. unit weight,
Young’s modulus, cohesion etc.) with their COV values. The script also asks
the horizontal and vertical spacing so that it can divide the geometry in order
to create soil regions for heterogeneous and anisotropic analysis. Finally,
before starting the analysis, it asks the number of analyses to be performed
(each with a different random value for the variables) and desired saving
location for the files.

(4) The script, first, sets up the working space in accordance with the dimension
of external geometry. Then, it makes PLAXIS to draw the external geometry,
layer boundaries and to create the sub regions. After this step, soil materials
are created from the given inputs. If chosen random, each material is created

61
 and assigned randomly different from its mean value. Finally, finite element
meshing, phreatic level and calculation stages are established.

(5) After strength reduction analysis is finished, PLAXIS file is saved and
calculated SRF value is written and stored on a file. Then, all inputs are cleared
out in order to be ready for the next analysis.

(6) The analysis is repeated specified run number times.

(7) After all of the analyses are finished, the script calls the file which contains all
SRF values and performs a number of statistical computation. In here,
probability of failure and reliability index are calculated. Then, they are also
written on the existing file.

(8) PLAXIS can sometimes give errors due to numerical calculation problems etc.
In order to compensate for this, one try-catch loop is inserted into the code so
that if a previously specified error occurs, PLAXIS can skip that analysis and
continue with the next one. This is also considered in PF and RI computations.

62