Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
SUBJECT TO WEATHERING:
AN ANALYTICAL AND NUMERICAL STUDY
Tesi presentata per il
conseguimento del titolo di Dottore di Ricerca
Politecnico di Milano
Dipartimento di Ingegneria Strutturale
Dottorato in Ingegneria Sismica, Geotecnica
e dellinterazione Ambiente-Struttura - XVI Ciclo
Stefano Utili
April 2004
EVOLUTION OF NATURAL SLOPES
SUBJECT TO WEATHERING:
AN ANALYTICAL AND NUMERICAL STUDY
Ph.D Candidate:
Eng. Stefano Utili
Supervisor:
Prof. Roberto Nova
April 2004
Dottorato in Ingegneria Sismica, Geotecnica
e dellInterazione Ambiente-Struttura del Politecnico di Milano
Scientific Committee:
Prof. Alberto Castellani (Coordinator)
Prof. Carlo Andrea Castiglioni
Prof. Claudio Chesi
Prof. Annamaria Cividini
Prof. Claudio di Prisco
Prof. Ezio Faccioli
Prof. Cristina Jommi
Prof. Sergio Lagomarsino
Eng. Paolo Negro
Prof. Roberto Nova
Prof. Roberto Paolucci
Prof. Maria Adelaide Parisi
Prof. Federico Perotti
Prof. Vincenzo Petrini
Prof. Giandomenico Toniolo
Prof. Carlo Urbano
(Picture on the front cover: air photo of mudslide corries at Beltinge (UK); after [Hutchinson, 1970])
One may define the human being, therefore, as the one who seeks the
truth.
[] The capacity to search for truth and to pose questions itself
implies the rudiments of a response. Human beings would not even begin
to search for something of which they knew nothing or for something
which they thought was wholly beyond them. Only the sense that they can
arrive at an answer leads them to take the first step. This is what
normally happens in scientific research. When scientists, following their
intuition, set out in search of the logical and verifiable explanation of a
phenomenon, they are confident from the first that they will find an
answer, and they do not give up in the face of setbacks. They do not judge
their original intuition useless simply because they have not reached
their goal; rightly enough they will say that they have not yet found a
satisfactory answer.
(Fides et Ratio)
Giovanni Paolo II
Acknowledgements
I wish thank my supervisor prof. Roberto Nova who has been
fundamental for my professional training. He taught me a method of
work, knowledge and critical engineering assessment. He also gave me a
very interesting problem to deal with.
Thanks to Riccardo Castellanza, not only for his sincere friendship but
also for his support. His presence and his hints have been fundamental
for me in many circumstances.
Then I would like to thank prof. Claudio di Prisco for all the times I went
to his office asking him something. He always found time to answer me.
Finally, I wish thank all my friends who accompanied me during these
three years with their friendship. In particular, I would like to thank
Dorotea and Antonio.
Index
i
Index
Introduction
Introduzione
1. Evolution of natural slopes subject to weathering:
an introduction.........................................................1
1.1. Introduction ................................................................1
1.2. Weathering phenomenon.............................................3
1.3. Slope weathering: models from the literature..............7
1.3.1. Geomorphologic and geological models ............................ 7
1.3.2. Engineering models.......................................................... 9
1.4. Objectives of the thesis ............................................. 10
1.5. Conceptual framework of reference ........................... 11
1.6. Choice of the methods used in the thesis .................. 13
Index
ii
1.7. Formulation of the problem...................................... 16
1.8. Development of the thesis ......................................... 18
1.9. References.................................................................. 19
Part 1: Analytical study
2. Limit equilibrium methods ..................................... 25
2.1. Introduction .............................................................. 25
2.2. General formulation .................................................. 28
2.3. Analysis of assumptions which make the problem
determinate ............................................................... 35
2.3.1. First group..................................................................... 35
2.3.2. Second group ................................................................. 39
2.3.3. Third group ................................................................... 40
2.3.4. An encompassing algorithm........................................... 44
2.4. Limit equilibrium solutions: physical admissibility and
optimum.................................................................... 45
2.5. Conclusions and some critical considerations ............ 50
2.6. References.................................................................. 53
3. Retrogressive failure analytical law........................ 57
3.1. Introduction .............................................................. 57
3.2. Mechanism of first failure: formulation based on the
limit equilibrium method........................................... 59
3.2.1. Introduction................................................................... 59
Index
iii
3.2.2. Determination of the analytical solution........................ 60
3.2.3. Case of inclined slope..................................................... 64
3.3. First failure mechanism, formulation by the limit
analysis upper bound method.................................... 66
3.3.1. The limit analysis upper bound theorem....................... 66
3.3.2. Determination of the analytical solution........................ 69
3.3.3. Charts relative to the first mechanism........................... 72
3.3.4. Numerical solutions by limit analysis ............................ 74
3.4. Determination of the second failure surface .............. 76
3.4.1. Introduction................................................................... 76
3.4.2. Determination of the analytical solution........................ 77
3.4.3. Charts relative to the second mechanism ...................... 84
3.4.4. Some aspects of the first two mechanisms ..................... 86
3.4.5. A different problem: which is the best slope profile? .... 90
3.5. Determination of the successive failure surfaces........ 93
3.5.1. Procedure for the determination of the successive failure
surfaces .......................................................................... 93
3.5.2. Discussion of the results ................................................ 95
3.5.3. Other failure mechanisms .............................................. 99
3.6. Undrained conditions = 0.................................... 100
3.6.1. First failure surface...................................................... 100
3.6.2. Second failure surface .................................................. 102
3.7. References................................................................ 103
Part 1 Conclusions...105
Index
iv
Part 2: Numerical study
4. Description of the distinct element method ..........109
4.1. Introduction ............................................................ 109
4.2. Description of PFC-2D............................................ 110
4.2.1. Law of motion.............................................................. 112
4.2.2. Force-displacement law................................................ 113
4.2.3. Calculation cycle.......................................................... 118
4.2.4. Stability of the numerical scheme................................ 120
4.2.5. Damping ...................................................................... 120
4.2.6. Other discrete element methods .................................. 121
4.2.7. PFC-2D parameters..................................................... 122
4.3. References................................................................ 124
5. Numerical simulations ...........................................127
5.1. Introduction ............................................................ 127
5.2. Calibration procedure.............................................. 128
5.2.1. Calculation of stresses and strains ............................... 129
5.2.2. Description of the biaxial test...................................... 130
5.2.3. Specimen generation procedure.................................... 131
5.2.4. Biaxial test execution .................................................. 134
5.3. Calibration of micromechanical parameters to
reproduce a frictional cohesionless material ............ 139
5.3.1. Influence of particle rotation on ................................ 139
5.3.2. Range of confining pressures to be investigated........... 143
Index
v
5.3.3. The role played by contact stiffness............................. 145
5.3.4. Dependence of on confining pressure ........................ 148
5.3.5. Choice of a suitable value of contact stiffness.............. 150
5.3.6. Determination of
- relationship ............................. 152
5.4. Calibration of micromechanical parameters to
reproduce a frictional-cohesive material .................. 154
5.4.1. Modelling of the contact behaviour ............................. 154
5.4.2. Calibration of the new micromechanical parameters ... 162
5.5. Simulations relative to weathering slope................. 167
5.5.1. Features of the slopes analysed.................................... 167
5.5.2. Particle radii and micromechanical properties
assigned ....................................................................... 168
5.5.3. Slope generation procedure .......................................... 168
5.5.4. Simulation of the retrogressive failure ......................... 171
5.5.5. First failure occurrence ................................................ 171
5.5.6. Case of strong erosion conditions................................. 174
5.5.7. Case of no erosion conditions....................................... 181
5.6. References................................................................ 183
Part 2 Conclusions.......187
6. Slope weathering: natural time scale.....................180
6.1. Introduction ............................................................ 180
6.2. Experimental validation of time-weathering laws ... 181
6.2.1. Case 1: Warden Point.................................................. 183
6.2.2. Case 2: Miramar .......................................................... 186
Index
vi
6.3. References................................................................ 189
Final conclusions..205
Appendix
A. Limit analysis results.....A.1
A.1. Tables of results....A.1
A.2. Linear interpolation of cohesion - crest retreat
relationships.....A.28
Introduction
I
Introduction
This thesis is aimed at building a model capable of making
quantitative predictions about the future evolution of natural slopes
which may be significant from an engineering viewpoint (decades).
Slopes subject to weathering undergo a series of landslides, occurring at
different times, which cause the progressive retrogression of the slope
front. In order to make predictions it is necessary to model the discrete
succession of the failures occurring within the slope. To this end,
methods typical of slope stability analysis and the distinct element
method have been used.
Weathering processes within slopes are not object of investigation.
Only the effects of weathering on the stability of slopes have been
studied. Therefore chemical and/or physical weathering processes
occurring in soil are completely disregarded.
The main work hypothesis assumed in the thesis concerns soil
strength: it has been assumed that it may be suitably expressed by two
parameters: c, (cohesion and internal friction angle) according to the
Mohr-Coulomb strength criterion. As regards weathering, it has been
assumed that it influences only cohesion causing its progressive decrease
in time. Moreover, it has been assumed that weathering is uniform within
the slope. Indeed, weathering is not uniform, but this assumption made
Introduction
II
the problem simpler and numerically treatable. However some
experimental data indicate that, at least in some cases, this assumption
leads to results that are not far from reality.
The first chapter has been conceived to give to the reader a
framework of the work done. In the chapter, the phenomenon studied is
described in detail; the assumptions made and the methods used to tackle
the problem are explained.
As regards the successive chapters, the thesis is divided into two
parts:
1. Analytical study: (Ch. 2 and 3) classical methods have been
used: limit equilibrium methods and limit analysis upper bound
method. An analytical law describing the discrete succession of
retrogressive failures of slopes subject to weathering has been
achieved by the limit analysis upper bound method (Ch. 3). This
law has been obtained as the solution to a mathematical problem.
The law relates the length of crest retreat and the succession of
profiles of the slope front to cohesion decrease. Since the
analytical law determined can be achieved also by limit
equilibrium methods according to a slightly different
formulation, a theoretical study of the limit equilibrium methods
(the most widespread and known methods in geotechnical
engineering) has been performed (Ch. 2).
2. Numerical study: (Ch. 4 and 5) the discrete element method has
been used. According to this method, soil is modelled as a
discrete assembly of particles (micromechanical approach). An
experimental numerical campaign has been performed in order to
determine the micromechanical parameters which must be
assigned to the system of particles so that it reproduces a c,
continuum. The campaign made possible to achieve simple
relationships between micro and macromechanical parameters.
This is an important result which opens interesting possibilities
concerning the use of the distinct element method in soil
mechanics. Simulations of the retrogressive failure of slopes
subject to weathering have also been run. The results achieved
have been compared with the predictions obtained by the
Introduction
III
analytical law determined by the limit analysis method (1
st
part).
The agreement found between the predictions made by the two
methods was good. This fact corroborates the validity of the
results found since the two methods used (limit analysis and
DEM) are completely different.
The last chapter (Ch. 6) is devoted to the determination of the time
scale relative to the evolution of natural slopes from experimental data.
Simple relationships between time and crest retreat of slopes have been
determined.
Introduzione
I
Introduzione
Questa tesi si propone di costruire un modello capace di elaborare
predizioni quantitative sullevoluzione di pendii naturali. Linteresse
della tesi rivolto a modellare levoluzione che i pendii subiscono in un
periodo di tempo significativo dal punto di vista ingegneristico (decadi). I
pendii soggetti a degradazione subiscono una serie di eventi franosi che
avvengono in tempi diversi e causano larretramento progressivo del
fronte dei pendii stessi. Per poter elaborare predizioni necessario
modellare una successione discreta di rotture allinterno dei pendii. A tal
fine sono stati usati sia metodi tipici dellanalisi di stabilit dei pendii che
il metodo degli elementi distinti.
Questa tesi non ha lo scopo di studiare i processi responsabili del
degrado del terreno costituente i pendii, ma gli effetti che tale degrado
induce sulla stabilit dei pendii; pertanto i processi chimici e/o fisici
responsabili del degrado del terreno non sono stati presi in alcun modo in
considerazione.
Lipotesi di lavoro principale riguarda la resistenza del suolo: si
ipotizzato che la resistenza del materiale costituente i pendii esaminati sia
caratterizzata da coesione e attrito (c, ) in accordo al criterio di rottura di
Mohr-Coulomb. Si assunto, inoltre che la degradazione influenzi solo
la coesione causando la sua progressiva diminuzione nel tempo. Si
Introduzione
II
assunto ancora che il degrado sia uniforme allinterno dei pendii.
Questultima ipotesi non certamente verificata nella realt, ma essa
permette di rendere il problema pi semplice da affrontare specialmente
dal punto di vista numerico. Tuttavia, alcuni dati sperimentali indicano
che, in certi casi, lassunzione fatta conduce a risultati non lontani dal
reale.
Il primo capitolo costituisce un inquadramento generale del lavoro
svolto. L sono descritti in dettaglio: il fenomeno studiato, le assunzioni
fatte ed i metodi usati.
La tesi composta da due parti:
1. Studio analitico: (Cap. 2 e 3) esso stato basato su metodi
classici dellanalisi di stabilit di pendii: i metodi
dellequilibrio limite e il metodo cinematico dellanalisi limite.
Questultimo ha permesso di determinare una legge analitica che
descrive la successione discreta delle rotture che avvengono nei
pendii soggetti a degradazione (Cap. 3). La legge analitica stata
ottenuta come soluzione ad un problema matematico. La legge
relaziona larretramento del fronte dei pendii (e la successione
dei diversi profili che i pendii via via assumono) alla diminuzione
di coesione. Dato che la legge analitica determinata pu essere
anche ricavata, con una formulazione leggermente differente,
mediante i metodi dellequilibrio limite, si svolto un
approfondito studio teorico di questi ultimi (nella pratica
dellingegneria geotecnica i pi conosciuti e diffusi tra i vari
metodi usati per lanalisi di stabilit di pendii).
2. Studio numerico: (Cap. 4 e 5) stato condotto con il metodo
degli elementi distinti. Coerentemente al metodo, il terreno
stato modellato come un insieme discreto di particelle (approccio
micromeccanico). Per determinare i parametri micromeccanici da
assegnare al sistema di particelle affinch esso riproduca un
continuo caratterizzato da coesione e attrito (c, ) si eseguita
una campagna numerica di prove sperimentali. La serie di prove
eseguite ha permesso di determinare semplici relazioni tra
parametri micro e macromeccanici. Questo risultato apre
interessanti possibilit di sviluppo per limpiego del metodo degli
Introduzione
III
elementi distinti nella meccanica delle terre. Infine sono state
condotte simulazioni della rottura retrogressiva di pendii soggetti
a degradazione. I risultati ottenuti sono stati confrontati con le
predizioni ottenute con la legge analitica determinata con
lanalisi limite (1
a
parte). Si cos potuto riscontrare un buon
accordo tra le predizioni elaborate con i due metodi. Questo fatto
corrobora in modo significativo la validit dei risultati trovati
dato che i due metodi usati (analisi limite e DEM) sono
completamente differenti.
Nellultimo capitolo (Cap. 6) si determinato il tempo scala relativo
allevoluzione di pendii naturali a partire da dati sperimentali. Si
sono determinate delle semplici relazioni tra tempo e arretramento
del fronte dei pendii.
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
1
1.
Chapter 1
1. Evolution of natural slopes subject to
weathering: an introduction
1.1. Introduction
This thesis is aimed at studying the evolution of natural slopes
subject to weathering processes. These processes cause a progressive
irreversible degradation of the slope material manifested as a reduction of
its mechanical properties. Therefore, as time goes on, landslides occur.
The time scale relative to the evolution of natural slopes ranges from
decades to thousands of years depending on soil type and weathering
processes acting on slopes. Weathering processes may have very
different velocities depending on the type of weathering. In the next
paragraph, an overview of such processes will be supplied.
Natural slopes are made of rock (granite, chalk, shale, marl,
limestone, etc.) or cohesive soil (clayey, silty, etc.). According to
[Terzaghi and Peck, 1967] soil means a natural block of grains
separable by a simple physical action like water agitation, whereas rock
defines a natural block of minerals jointed by strong and permanent
bonds. In reality, however, an intact granite and a mature quartzitic
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
2
sand deposit are only the extrema of a continuous transformation process
of the material composing the earth crust. There is no distinct threshold
beyond which a geomaterial ceases to be a rock and starts to be a soil
[Nova, 1997]. In fact the earth crust is subject to physical and chemical
processes which continuously transform rocks into soils and viceversa as
shown in Fig. 1.1. These processes can be grouped into four different
types: weathering, erosion, diagenesis and metamorphism. The latter two
are phenomena leading to rock formation, whereas the former two are the
opposite.
Fig. 1.1: schematic representation of the transformation processes among geomaterials
(after [Dobereiner and De Freitas, 1986]).
In literature, weathering is approached essentially in terms of its
description and classification. This phenomenon is very complex since it
depends on both the mineralogy of the intact rock or the decomposed soil
and environmental conditions. Usually different processes act together
and it is difficult to analyse them separately. For this reason, many and
different classification indices and scales have been defined. However, a
common feature of such processes is their capability to cause a
progressive irreversible degradation of the slope material manifested as a
reduction of its mechanical properties.
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
3
1.2. Weathering phenomenon
Weathering may be defined as the process of alteration and
breakdown of rock and soil materials at and near the Earthsurface by
chemical decomposition and physical disintegration [Geol. Soc. Eng.
Group Work. Report, 1995].
Weathering reduces hard rocks into soft rocks which maintain the
structure of the intact rocks, but are characterised by higher void ratios
and reduced bond strengths. Soft rocks are transformed into granular
soils generally called residual soils. Residual soils differ from their parent
rocks in mineralogical composition and structure.
Weathering processes belong to two categories depending on the
type of process: physical or chemical. Physical weathering causes the
mechanical destructuration of rocks without mineralogical change
whereas chemical weathering is due to chemical reactions leading to the
decomposition of the constituent minerals to stable or metastable
secondary mineral products.
Physical weathering is due to processes such as: freeze-thaw cycles,
temperature variations causing swelling-shrinkage cycles, wetting-drying
cycles, salt crystallisation, wind and rain action, living organism action
(see Fig. 1.2).
Former and surface
Erosion
New and surface
PROCESSES:
differential thermal expansion and insulation
wet-dry expansion
freeze-thaw action
rooth growth and burrows
wind & rain action
cristilization & expansion
EFFECTS:
Unloading (stress relief)
Joints formed
Incipient fractures opened
Intergranular and rock mass disintegration
PHYSICAL WEATHERING
Fig. 1.2: physical weathering processes (adapted from [Geol. Soc. Eng. Group Work.
Report, 1995]).
Freeze and thaw cycles cause rock fracturing (see Fig. 1.3). Water
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
4
thawing process needs a volume expansion (of the order of 10%) to occur
which if constrained by surrounding rock, leads to an increase of pressure
enlarging and developing cracks. Pressure exerted by ice may reach up to
200 MPa.
a) b)
Fig. 1.3: a) rock falls from natural slopes. b) Stages of freeze-thaw cycles (after [Martinati,
2003]).
Temperature variations cause variations of the state of stress within
rock mass leading to swelling-shrinkage cycles responsible of rock
fracturing. Temperature variations between night and day may reach up
to 50 Celsius degrees in desert or mountainous areas. Anyway, thermal
excursions may be caused by fires or volcanic eruptions as well.
A similar phenomenon is due to wetting-drying cycles which cause
variations of the state of stress within clayey soils leading to swelling-
shrinkage cycles responsible of cracking development.
Salt crystallisation occurs during rainy seasons and in marine coastal
areas where waters with high salt content go into rock cracks and pores.
Crystals formed within cracks and pores exert pressures which develop
fracturing.
Living organism actions such as tree root growth and burrows
contribute to cracking development as well.
Chemical weathering is mainly due to chemical reactions which
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
5
oxygen in the atmosphere, water and weak acids present into it are
involved in.
PROCESS:
solution
Weathered rock
(solid product)
Fresh rock
PROCESSES:
Chemical alteration
Volume change
Textural change
EFFECTS:
Alteration of minerals:
Feldspar + CO2+ H2O => Clays + Silicas + Cations
(Solid colloids) (Solutions)
Pyrite + O2+ H2O => Iron Oxide-Hydroxide
+ (Acid solutions)
Dissolution of limestone:
CaCO3 + H2O + CO2 => Ca(H2CO3)2 (solute)
products
solution
CHEMICAL WEATHERING
Fig. 1.4: chemical weathering processes (adapted from [Geol. Soc. Eng. Group Work.
Report, 1995]).
The reaction of oxygen with minerals gives rise to a process called
oxidation. The chemical element more subject to oxidation is iron. A
typical example is given by the pyrite oxidation:
4FeS
2
+ 10H
2
O + 15O
2
2Fe
2
O
3
.H
2
O + 8H
2
SO
4
Pyrite Limonite Sulfuric acid
Pyrite is a common rock-forming mineral in soft sedimentary rocks.
Examples of the effects of pyrite oxidation on slopes in Japan, have been
reported by [Chigira and Oyama, 1999].
Another typical example of weathering is due to the chemical
decomposition of granite. When water comes into contact with intact
granite, potassium feldspar and mica are transformed into kaolinite
(clay). This type of weathering is typical of granitic batholiths in the
Hong Kong region. Since long time ago, [Lumb, 1962] made a
systematic study attempting to characterise the mechanical behaviour of
the weathered granite. Such knowledge was requested to satisfy the
demand, particularly strong in that region, of engineering structures such
as high buildings, bridges, tunnels and roads.
Usually physical and chemical weathering occur together, in such a
way that one acts to accelerate the other. In fact, the progress of chemical
weathering relies on cracks opened or enlarged as a result of physical
weathering. In the same way, cracks may develop in response to changes
in volume and weakening induced by chemical weathering.
In the light of these observations, a mechanical approach which
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
6
neglects the specific weathering processes in order to model only the
weathering effects on the mechanical behaviour of the material subject to
degradation, appears reasonable.
In Fig. 1.5 and in Fig. 1.6 some case examples where it is evident the
need of an engineering study in order to predict the evolution of natural
slopes are shown. Many times civil structures (e. g. roads, railways,
buildings, waste deposits) have been located or designed to be located
nearby slopes subject to strong weathering conditions. In these cases, it is
of fundamental importance to suitably model the retrogression of the
slope front in order to establish if the structure will be reached by the
retrogressive slope front and if it will be reached during its design life.
a) b)
Fig. 1.5: photographs of steep slopes in weakly-cemented sand. a) Mogn, Gran Canaria;
urban development at the base (after [Delgado, 1991]. b) Daly City (California,
USA); urban development on the top (after [Clough et al., 1981]).
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
7
a) b)
Fig. 1.6: Esplanade Drive City of Pacifica (California, USA); a) cliff weathering affecting the
slope crest (dashed lines indicate the successive failure lines) b) damages caused by
slope retrogression. Many houses had to be demolished (after [Snell et al., 2000]).
1.3. Slope weathering: models from the literature
1.3.1. Geomorphologic and geological models
From an historical point of view, the first models aimed at describing
the evolution of natural slopes come from geomorphology and geology.
They have been conceived to describe the evolution of natural slopes
occurred in past eras on a geological time span. Therefore, all these
models are aimed at describing slope evolution in the long term (i.e.
thousands of years). They do not consider a succession of discrete events
(landslides), but a continuous process. In this process the sediment flux
due to landslides represents a long-term average of what is due to
individual events. The use of these average rates assumes that individual
slides are small enough not to change the slope profile significantly
[Mills H. and Mills R., 2001]. In the time scale of decades (i.e.
engineering time) this is not true at all. Therefore, the need of a model
predicting the evolution of a slope taking explicitly into account the
single landslides which modify its profile, is underlined.
Slope evolution models were born in 1866 thank to Fishers work
[Fisher, 1866] and successively they have been developed by [Bakker
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
8
and Le Heux, 1946, 1952], [Kirkby, 1973, 1984, 1987] and many others.
All these models approach the evolution of slope profiles in 2D and are
based on a mass balance or continuity equation relative to the slope mass:
z S
t x
=
(1.1)
where t is time, x is the Cartesian co-ordinate relative to the horizontal
direction, z the elevation of the slope profile and S the total downslope
flux of sediment. S is given by:
S C W L = + + (1.2)
where
1
C K z x = is called creep term, W is called wash term and L
represents the sediment flux due to landslides. Substituting Eq. (1.2) into
(1.1) a differential equation is obtained. This equation is non-linear
because of the complicated expressions relative to W and L.
The simplest models, for instance [Andrews and Hanks, 1985],
neglect the contribution given by W and L terms, so that the substitution
of Eq. (1.2) into (1.1) gives rise to:
2
2
z z z
K K
t x x x
= =
(1.3)
called linear diffusion equation. This equation is well known to
mathematicians like the Fourier equation (widely studied in literature).
Instead, if the wash and landslide contributions are taken into
account a non-linear differential equation is achieved and models are
known as non-linear diffusion models. Non-linear diffusion models have
been widely applied in North-America [Andrews and Hanks, 1987].
Recently, a computer code, based on [Kirkby, 1984] non-linear diffusion
model, the most encompassing model attempt according to [Mills and
Mills, 2001], has been implemented by [Kirkby et al., 1992].
In all these models, the parameters and variables introduced have
only an apparent physical meaning. They are aimed at taking into
account phenomena such as creep, wash, landsliding, but they are not at
all related to any mechanical quantity characterising the soil slope as well
as to any chemical quantity. The only variables with clear physical
meaning are the geometric ones.
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
9
Anyway, parameters and variables without physical meaning, may be
still used to describe a phenomenon if they are able to reproduce the
experimental data relative to it. As stated previously, these models were
born to describe slope evolution in the long-term; therefore they have
been applied to reproduce the evolution undergone by natural slopes in
thousands of years. In order to obtain the succession of slope profiles
formed in the past, the so-called time-location technique or space-time
substitution is used: some slope profiles surveyed in the same area are
taken as the temporal succession of the profile of a slope. [Kirkby, 1984]
took the profiles reported by [Savigear, 1952] (between Laugharne and
Pendine in Wales, UK) as a temporal sequence of profiles of a slope
which he used to validate his model. Assuming the time-location
technique reliable, Kirkby model matches well the experimental data.
In conclusion, geomorphologic and geological models are valid to
describe the slope evolution occurred in a long term period even if they
have not a solid physical and mechanical background, but they cannot be
used to predict the future evolution of slopes on an engineering time
span. In fact, these models take into account discrete events such as
landslides only by representing them with a variable that contributes
continuously to slope erosion. Moreover, the achieved profiles cannot be
in any way actual profiles but only evolution profiles averaged on long
time spans.
1.3.2. Engineering models
Another category of models, belonging to Soil and Rock Mechanics
disciplines, are featured by being related to the mechanical characteristics
of natural slopes. In these models, material characteristics such as unit
weight, strength and deformability, are taken into account. Weathering,
chemical or physical, is not taken into account from a phenomenological
point of view i.e. the specific weathering processes are disregarded.
Weathering is considered as a phenomenon causing variations of the state
variables used to describe soil/rock characteristics. This means that, for
instance in case of chemical weathering, kinetics of chemical reactions,
diffusivity and so on do not appear in the models.
If quantitative predictions about the future evolution of natural slopes
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
10
in terms of actual slope profiles and times at which landslides occur,
engineering models have to be used. One of them is that supplied by this
thesis.
1.4. Objectives of the thesis
As already stated in 1.1, a clear distinction between rocks and soils
does not exist in nature. In Geotechnical Engineering, aggregates of
particles (grains) whose chemical bonds are weak and may be broken by
load levels typical of Civil Engineering, are considered soils; whereas
materials characterised by strong chemical bonds whose mechanical
behaviour is governed by discontinuities (joints and faults) rather than
intact rock mechanical properties, are considered rocks. Many materials
are intermediate between rocks and soils, because they behave as rocks if
subject to low stress levels and as soils if subject to stress levels high
enough to break the chemical bonds among their grains [Nova, 2002].
This thesis is aimed at modelling the evolution of dry natural slopes
made of cohesive soils and soil-like materials i.e. materials whose
discontinuities play a non significant role in determining the conditions
leading to landslide occurrence. According to the definition of soil given
above, rock slopes whose joints play a negligible role into landslide
occurrence are included in this study. For instance, this is the case of
many coastal cliffs made of chalk (South England, Normandy, Greece,
Sardinia, Abruzzi, etc.).
Slopes are considered dry since this case is the simpler problem to be
tackled upon which extensions to more complex cases (seepage
conditions, earthquake occurrence, consolidation processes) may be dealt
with. This is the first necessary step towards the study of more complex
situations.
As time goes on, natural slopes are subject to a progressive reduction
of their mechanical properties leading to multiple landslides causing the
retrogression of the slope crest. This phenomenon occurs in decades for
coastal cliffs because of high chemical weathering caused by the sea
whereas it needs a longer period to occur (up to thousands of years) for
inland slopes.
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
11
This thesis is aimed at determining a law able to predict times at
which landslides occur. Moreover, the evolution of slope profiles with
particular interest to the retrogression of the slope crest, are object of
investigation (see Fig. 1.7).
t
1
t
3
t
2
t
4
t
5
t
6
Fig. 1.7: schematic representation of an initially vertical slope subject to weathering. Each
line represents a failure line at the onset of a landslide at the time t
i
.
It is pointed out that this law intends to describe the evolution of
natural slopes from a qualitative viewpoint and not from a quantitative
one. This is due to the simplifications introduced to model the problem
(see 1.7) and to the phenomena significant in triggering landslides that
have been neglected on purpose.
1.5. Conceptual framework of reference
Nowadays, the mechanical behaviour of soil-like materials
(according to an engineering classification) at macroscopic level may be
described by constitutive equations with a good level of accuracy. On the
contrary, at microscopic level they present strong heterogeneities as they
are composed of bonded particles with very different shapes and sizes,
bond strength between particles is inhomogeneous, etc.. Moreover, at this
scale, weathering processes are strongly inhomogeneous (see Fig. 1.8).
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
12
a) b) c)
Fig. 1.8: calcarenite (sedimentary soft rock formed by cementation of organogeus calcareous
grains) of Gravina (Puglia, Italy) SEM photos. a) Microfabric (magn. 30). b) Shells
grains (magn. 130). c) Shell surface (magn. 3500). (after [Castellanza, 2001]).
Therefore, the only available way to tackle the problem is by
considering the slope material at macroscopic level. Within a certain
volume (Representative Elementary Volume) soil is assumed
homogeneous. Hence, the conceptual framework of reference is that of
Continuum Mechanics. Within this framework, soil behaviour is
characterised by constitutive equations relating stresses to strains.
Weathering is modelled as a reduction of soil strength according to a
certain law. This law depends on both space and time. Its dependence on
space has been assumed, whereas its dependence on time has been
determined as result of the study (see Ch. 6).
In this work, weathering has been assumed uniform throughout the
whole slope. This hypothesis may appear very strong. In fact, considering
a coastal cliff subject to chemical weathering (oxidation, dissolution
reactions) and physical weathering (wind and rain actions) it is
reasonable to think that the deeper the soil is, the weaker the attack; or in
other words, that the most weathered soil lies into the most shallow
regions of the cliff. But experimental evidences support this hypothesis.
Yokota and Iwamatsu performed SPT tests into a slope onto Kyushu
Island (Japan) to test the variation of soil (soft pyroclastic rock) hardness
with depth [Yokota and Iwamatsu, 1999]. From Fig. 1.9 emerges that the
hardness is almost constant in the upper part of the slope whereas in the
lower part of the slope (toe) it can be considered constant from a small
depth inwards. According to these data, weathering can be, in first
approximation, considered uniform.
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
13
a)
b) c)
Fig. 1.9: Kyushu Island (Japan) hardness distribution in a steep slope: a) and b) data from
the upper part of the slope; c) data from the slope toe (after [Yokota and
Iwamatsu, 1999]).
1.6. Choice of the methods used in the thesis
The conditions, in terms of level of actual soil strength, at which a
stable slope becomes unstable need to be determined. For this reason,
some methods developed for slope stability problems have been used in
the thesis.
From an historical point of view, the first methods to analyse slope
stability are the limit equilibrium methods. Successively, limit analysis
methods have been applied to slope stability problems and later on the
finite element method too.
Limit equilibrium methods are the most widely used methods in
engineering practice for slope stability problems. Usually, they are used
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
14
to assess the stability of a slope in terms of a factor of safety. A feature of
these methods is represented by the few parameters needed to describe
the soil properties: unit weight, internal friction angle, cohesion. This is
very important since there is a great lack of data relative to natural
slopes. Moreover, it can be generally stated that if two models manage to
reproduce a phenomenon with the same degree of accuracy, the better is
that which needs the smaller number of parameters.
Limit analysis has been applied in many fields of Engineering. One
of them is represented by slope stability. The limit analysis methods,
lower and upper bound, are much more versatile than limit equilibrium
methods and above all, supply solutions which are rigorously lower and
upper bounds on the true collapse load.
The limit analysis methods need to assume an associated flow rule.
As it is well known in literature, soil behaviour does not obey to
associativeness (dilation angle is less than friction angle), but for the
studied problem the influence of dilation is small since soil is not
confined.
Therefore these methods have been chosen in order to derive an
analytical law describing the evolution of slopes subject to weathering
(see Ch. 3). Soil strength has been characterised by the Mohr-Coulomb
failure criterion (see Fig. 1.11). All the parameters needed by this law to
describe the soil properties are: unit weight, internal friction angle,
cohesion.
In the literature, lower bound solutions for slope stability problems
are a few ([Pastor, 1978]) as the determination of a static stress field is a
very difficult task for a complex slope profile. On the contrary, the
determination of a kinematic admissible velocity field is still an
affordable task even for complex slope profiles. For this reason, only the
upper bound method has been used to determine an analytical law
describing the evolution of slopes subject to weathering (see Ch. 3).
Recently, limit analysis has been incorporated into a numerical
formulation that uses finite elements for the discretisation of the soil
mass [Sloan, 1988, 1989], [Sloan and Kleeman, 1995], [Kim, 1998],
[Kim et al., 1999, 2002]. The lower and upper bound theorems have been
formulated by Sloan [Sloan, 1988, 1989] as linear problems to be solved
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
15
using linear programming technique. Based on finite element
discretisation of the slope the velocity field is optimised to find the
lowest upper bound and the stress field is optimised to obtain the highest
lower bound. Unlike traditional displacement based finite elements, each
node of the finite elements used in limit analysis is unique to a given
element. In order to use linear programming techniques, all the
conditions required for static and kinematic admissibility are formulated
as linear constraints on the nodal variables of the finite element
discretisation.
Some numerical lower and upper bound solutions obtained by
[Loukidis et al., 2003] have been used in 3.3.1 for comparison with the
results obtained by the classical upper bound method.
The finite element method (perhaps the most world-wide known
method in Engineering) has been used for slope stability analysis since
1975 [Zienkiewicz et al., 1975].
With this method, it is difficult to determine a failure line since only
a region where deformations localise (shear band) results from analyses.
If soil is modelled within the framework of the classical plasticity theory,
shear band width is mesh dependent. In order to avoid meaningless
results, advanced continuum models are needed: non-local models such
as those in the framework of gradient plasticity theory and viscoplasticity
theory. These continuum models need far more parameters to
characterise soil than the 3 parameters needed by limit analysis and limit
equilibrium methods. Apart the complication due to so many parameters,
often with an obscure physical meaning, their calibration is a very
difficult task.
The evolution of natural slopes is characterised by many landslides.
After each landslide a new slope profile is formed. Remeshing is needed
because of the new geometry. This becomes time-expensive if an
automatic remeshing in the finite element pre-processor is not available.
In conclusion, in comparison with the other methods used in this
thesis, the finite element method appears by far the less convenient.
The distinct element method is the youngest method, introduced by
[Cundall, 1971] and applied to soil mechanics by [Cundall and Strack,
1979]. This method is not based on continuum mechanics but on a
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
16
discrete micromechanical approach. In Ch. 4 a detailed description of the
principles of this method will be supplied.
In the numerical code PFC2D (Particle Flow Code 2 Dimensions),
soil is represented by an assembly of rigid disks upon which contact
forces act. Walls apply the boundary conditions relative to displacements.
Dynamic equilibrium equations are imposed for the system of disks.
Explicit schemes of direct integration in time are used to follow the
evolution of the system from a steady state to the successive one (for a
detailed description of the code see Ch. 4).
All the difficulties met by the finite element method, are not
encountered by this method. In fact, when a landslide develops, because
of decreasing of soil strength, a part of disks starts to move, giving rise to
a failure line easily detectable. For this reason, this method has been
chosen to numerically simulate the boundary value problem under study
as it will be formulated in the next paragraph. The second part of this
thesis is devoted to this study.
Unfortunately, a micromechanical description of natural slopes is not
available. The method needs parameters which, in general, are not known
for geomaterials. These parameters can be defined as micromechanical
since they rule the mechanical behaviour at a micromechanical scale. In
order to determine them, calibration is needed. Calibration consists in
determining the micromechanical parameter values such that if a volume
of particles (disks) large enough to be representative is considered, its
mechanical behaviour is identical to that of the continuum described by
known macromechanical constitutive laws and parameters whose
behaviour is intended to be simulated.
Therefore, thank to calibration process, the input parameters needed
to characterise the slope material are the parameters used in the
macromechanical description of it. For this study they are: unit weight,
internal friction angle and cohesion.
1.7. Formulation of the problem
In this paragraph the problem tackled by the analytical and numerical
study of this thesis is formulated.
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
17
Let us consider a plane uniform slope made of homogeneous
material, as shown in Fig. 1.10, whose height is H. No loads are present
on the slope. The known soil mechanical properties concern density and
resistance. The former is characterised by unit weight and the latter by
the Mohr-Coulomb failure criterion. According to this criterion, two
parameters are needed to describe the soil strength: , internal friction
angle and c, cohesion. In the literature, this ideal type of soil material is
also known as c- material.
In the limit analysis upper bound method (part 1), the stress-strain
behaviour assumed is rigid perfectly plastic, with associate flow rule, and
as a consequence dilation angle equal to internal friction angle . In the
distinct element method (part 2), the constitutive laws are imposed at
micromechanical level since force-displacement laws rule the interaction
between distinct rigid elements at contacts. The force-displacement laws
used are characterised by a micromechanical friction angle 0
and a
micromechanical dilation angle 0
= .
Chemical weathering affects soil mechanical strength causing its
progressive decrease in time. It is assumed that only cohesion decreases
whereas friction remains constant. Cohesion decrease is continuous and
uniform. In Fig. 1.11, the failure surface is represented in the Mohr plane.
According to the adopted assumptions, the failure surface evolves
remaining a straight line and lowering with constant inclination until soil
becomes uncohesive, c = 0.
H
Fig. 1.10: uniform slope.
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
18
fresh material
partially weathered
material
uncemented soil
time increasing
t
c
Fig. 1.11: failure criterion evolution due to weathering.
According to Fig. 1.11, the tension strength t is equal to cot c . This
assumption is an overestimation of the real material tension strength. But
the simplification is needed by the type of mechanisms assumed in the
limit analysis method (see Ch. 3). Assuming cot t c implies the
possibility of the development of tension cracks which cannot be taken
into account by the mechanisms used in the limit analysis method.
Anyway, it is assessed that this assumption should have a small effect on
the predicted slope profiles and the retrogression of the slope crest.
1.8. Development of the thesis
A whole chapter (2) is devoted to limit equilibrium methods showing
their common features and their differences. At this end, all the presented
methods are derived as particular cases of a general formulation. A
classification of the methods according to their features (assumptions and
solution schemes) has been attempted. The performance of these methods
depend on the type of slope and failure surface examined. Therefore an
assessment relative to the various methods for different types of failure
surfaces has been attempted in order to let the reader have an overview as
complete as possible of the capabilities and limitations of the methods.
In chap. 3, an analytical law describing the evolution of the slope
described in 1.7, has been achieved by use of both limit equilibrium
methods and upper bound limit analysis method. It will be shown that
limit equilibrium methods and limit analysis upper bound method lead to
the same equations. Nevertheless, it has been decided to use both
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
19
methods since the limit equilibrium formulation is simpler and more
known in engineering practice than the limit analysis formulation
whereas limit analysis upper bound method has a more solid theoretical
background.
In part 2, the evolution of the slope described in 1.7, is studied by
use of the distinct element method. First, an experimental campaign has
been run on numerical specimens in order to assign the micromechanical
properties to the set of rigid disks and contact bonds representing the
slope. Second, the evolution of the slope has been studied by simulating
the progressive weathering of the slope, decreasing the strength of
contact bonds between disks.
In Ch. 6, simple time-retrogression of the slope front laws have been
determined. At this end, the retrogressive failure analytical law
determined in Ch. 3 and the results obtained in part 2 have been used as
well as experimental data relative to some monitored natural slopes
subject to strong weathering.
1.9. References
Andrews D. J. and Bucknam R. C., 1987. Fitting degradation of
shoreline scarps by a nonlinear diffusion model. J. Geophys. Res., 92,
pp. 12857-12867.
Andrews D. J. and Hanks T. C., 1985. Scarp degraded by linear
diffusion: inverse solution for age. J. Geophys. Res., 90, pp. 10193-
10208.
Castellanza R., 2001. Weathering effects on the mechanical behaviour of
bonded geomaterials: an experimental, theoretical and numerical study.
PhD. thesis, Politecnico di Milano, Milan.
Cendrero A. and Dramis F., 1996. The contribution of landslides to
landscape evolution in Europe. Geomorphology, 15, pp. 191-211.
Chigira M., Oyama T., 1999. Mechanism and effect of chemical
weathering of sedimentary rocks. Engrg. Geol., 55, pp. 3-14.
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
20
Clough G. W., Sitar N., Bachus R. C., Rad N. S., 1981. Cemented sand
under static loading. J. Geotech. Engrg. Div., ASCE, 107(GT6), pp. 799-
817.
Cundall P.A., 1971. A computer model for simulating progressive, large-
scale movements in blocky rock systems. Proc. Symp. Int. Soc. Rock
Mech., Nancy, 2, art. 8.
Cundall P.A. and Strack O. D. L., 1979. A discrete numerical model for
granular assemblies. Gotechnique, 29, pp. 47-65.
Delgado E., 1991. A vista de parajo (Espaa a vuelo). RTVE
Publications, Estagraf.
Dobereiner L., de Freitas M. H., 1986. Geotechnical properties of weak
sandstones. Gotechnique 36(1), pp. 79-94.
Kirkby M. J., 1971. Hillslope process-response models based on the
continuity equation. In: Brunsden D. ed., Slopes: form and process. Inst.
Brit. Geogrs. Special Publication 3, pp. 15-30.
Kirkby M. J., 1973. Landslides and weathering rates. Geol. Appl. e
Idrogeol., Bari, 8, pp. 171-183.
Kirkby M. J., 1984. Modelling cliff development in South Wales:
Savigear re-viewed. Zeitschrift fur Geomorphologie., 28, pp. 405-426.
Kirkby M. J., 1987. General models of long-term slope evolution through
mass movement. In: M. G. Anderson and K. S. Richards eds., Slope
Stability. Wiley, pp. 359-379.
Kirkby M. J., Naden P. S., Burt T. P., Butcher D. P., 1992. Computer
simulation in physical geography. Wiley, pp. 85-90.
Martinati S., 2003. Modellazione degli effetti della degradazione chimica
di geomateriali cementati in opere sotterraneee. (in Italian) Degree
thesis, Politecnico di Milano, Milan.
Mills H. H., Mills R. T., 2001. Evolution of undercut slopes on
abandoned incised meanders in the Eastern Highland Rim of Tennessee,
Chapter 1 Evolution of natural slopes subject to weathering: an introduction
21
USA. Geomorphology, 38, pp. 317-336.
Nova R., 1997. On the modelling of the mechanical effects of diagenesis
and weathering. ISRM News Journal, 4(2), pp. 15-20.
Nova R., 2002. Fondamenti di meccanica delle terre. McGraw-Hill
edition.
Pastor J., 1978. Limit analysis: numerical determination of complete
statical solutions. Application to the vertical cut. J. Mecanique applique
2(2), pp. 167-196 (in French).
Savigear R. A. G., 1952. Some observations on slope development in
South Wales. Trans. Inst. Brit. Geogrs., 18, pp. 31-52.
Sloan S. W., 1988. Lower bound limit analysis using finite elements and
linear programming. Int. J. Numer. Anal. Methods Geomech., 12, pp. 61-
77.
Sloan S. W., 1989. Upper bound limit analysis using finite elements and
linear programming. Int. J. Numer. Anal. Methods Geomech., 13, pp.
263-282.
Snell C. B., Lajoie K. R., Medley E. W., 2000. Sea-cliff erosion at
Pacifica, California caused by 1997/98 El Nio storms. In D. V.
Griffiths, G. A. Fenton, T. R. Martin eds., Slope stability 2000. Proc.
Geo-Denver 2000, Denver, Colorado, USA.
Terzaghi K. and Peck R. B., 1967. Soils mechanics in engineering
practice. Wiley, 2
nd
edition.
Yokota S. and Iwamatsu A., 1999. Weathering distribution in a steep
slope of soft pyroclastic rocks as an indicator of slope instability. Engrg.
Geol., 55, pp. 57-68.
Zienkiewicz O. C., Humpheson C., Lewis R. W., 1975. Associated and
non-associated visco-plasticity and plasticity in soil mechanics.
Gotechnique, 25(4), pp. 671-689.
Part 1
Analytical study
Summary
This part is devoted to the theoretical study made relative to the
problem tackled: the evolution of natural slopes whose behaviour may be
suitably described by assuming that the slope soil is of c, type.
In chapter two a comprehensive review of limit equilibrium methods
is supplied. The chapter is aimed at showing the main features of such
methods which are the most used in the today geotechnical engineering
practise for the analysis of slope stability.
In chapter three the analytical law achieved by the author to describe
the evolution of the investigated slopes is illustrated. This law can be
derived by using two different conceptual frameworks based on two
different theories referring to limit equilibrium methods and to limit
analysis upper bound method respectively.
This part has been called analytical to stress the main result of this
theoretical study: an analytical law describing the retrogression of the
slope front caused by the progression in time of the slope weathering.
Chapter 2 Limit equilibrium methods
25
2.
Chapter 2
2. Limit equilibrium methods
2.1. Introduction
Limit equilibrium methods have been used in geotechnical
engineering for decades in order to assess the stability of slopes. The idea
of discretizing a potential sliding mass in vertical slices was introduced
early in the 20
th
century. In 1916, Petterson [Petterson, 1955] presented
the stability analysis of the Stigberg Quay in Gothenberg, Sweden where
the slip surface was taken circular. But the first method of slices is
associated to the Felleniuss name [Fellenius, 1927, 1936]. His method
also known as the Ordinary method, the Swedish circle method, the
conventional method and the US Bureau of reclamation method assumes
no interslice forces and the factor of safety is achieved by the overall
moment equilibrium around the centre of a circular slip surface.
In the mid-50s, Janbu [Janbu, 1954] and Bishop [Bishop, 1955] made
advances in the method. Janbu developed his method for generic slip
surfaces whereas Fellenius and Bishop developed their methods for
circular surfaces only (later Bishop extended his method to generic
surfaces).
In the 60s and 70s most methods were invented: some making the
Chapter 2 Limit equilibrium methods
26
limit equilibrium method a more powerful and refined tool of analysis of
slope stability (Spencer, Morgenstern & Price, Sarma methods) and other
making it more suitable for hand calculations (force equilibrium
methods). Many articles were published in these years on this topic: in
some of them a real contribution to the improvement of the method was
given whereas in others only slight modifications or different
formulations of earlier methods were given.
In the late 50s, these methods began to be implemented in computer
codes. Little and Price (1958) were the first who used a computer to run
stability analyses by the Bishop simplified method.
The advent of powerful desktop personal computers in the 1980s
made economically viable to develop commercial software based on limit
equilibrium methods.
Nowadays, these methods are routinely used for stability analyses in
geotechnical engineering practice and many programs are available
(examples will be supplied further on).
Methods of slices can be classified according to different criteria:
1. suitable only for circular failure surfaces or applicable to any
shape of surface.
2. rigorous and simplified: the former satisfy all equilibrium
equations whereas the latter satisfy only a part of equilibrium
equations. Some authors developed two versions (simplified and
rigorous) of the same method: Bishop, Sarma, Janbu, etc. Within
simplified methods, a large group is given by the methods of
forces (Lowe & Karafiath, Corps of engineers method, Seed &
Sultan).
3. depending on assumptions made to render the problem statically
determinate: 3 groups of methods can be recognised on the basis
of the hypotheses introduced about the interslice forces
[Espinoza et al., 1992a, 1994].
4. based on the parameter used to determine the critical surface: the
traditional factor of safety Fs or other parameters such as the
critical horizontal uniform acceleration [Sarma, 1973, 1979],
[Spencer, 1978].
All methods approximate the bottom boundary of slices with linear
Chapter 2 Limit equilibrium methods
27
bases. Formulations are based either on differential equations (e.g.
[Janbu, 1954], [Bishop, 1955], [Spencer, 1967]) or algebraic equations
making difficult to compare different methods to the inexperienced
reader. As the factor of safety is calculated by algebraic equations and
limit equilibrium methods are based on a slope division into a discrete
number of slices, here the latter formulation is preferred.
Equations Condition
A
B
C
2n
n
n
force equilibrium in two directions for each slice
moment equilibrium for each slice
Mohr-Coulomb failure criterion
4n total number of equations
Unknowns Description
D
E
F
G
H
I
L
1
n
n
n
n-1
n-1
n-1
factor of safety
normal force at the base of each slice, P
i
location of normal forces at the base of slices
shear force at the base of each slice, S
i
interslice horizontal force, E
i
interslice vertical force, T
i
location of interslice forces (line of thrust) h
i
6n-2 total number of unknowns
Table 2.1
In table 2.1 the number of equations and unknowns are summarised.
The difference between equations and unknowns gives the number of
assumptions to render the problem statically determinate:
6 2 4 2 2 n n n = . All methods make n assumptions on the locations of
P
i
. Most methods assume the normal force acting at the base centre
(uniform stress distribution), even if other methods assume P
i
acting at
the point of intersection between the resultant vertical force and the base
of the slice [Janbu, 1954] or according to a linear stress distribution
[Morgenstern & Price, 1965]. However, the influence of the location of
P
i
on the safety factor is negligible for a sufficiently large number of thin
slices. It may become an important factor in a wedge type analysis, when,
Chapter 2 Limit equilibrium methods
28
for instance, only two or three slices are used [Espinoza et al., 1994].
Concerning the remaining n-2 assumptions, methods differ greatly.
Usually, n-1 assumptions are made about interslice forces and an extra-
unknown to be determined together with the factor of safety is
introduced.
2.2. General formulation
Here, a general formulation shared by a large number of methods
(the rigorous ones) is given in order to show the main features common
to most methods and their differences. In the following, notation has been
taken according to Fig. 2.1.
Fig. 2.1: notation relative to the forces acting on a slice
In order to achieve the normal and shear base forces, the equilibrium
of forces along two perpendicular directions is written for all the n slices.
Directions may be either vertical and horizontal or parallel and normal to
the slice base. From imposing the equilibrium along the direction normal
to the slice base, P
i
is achieved:
( )cos sin sin
i i i i i i i i
P W T E Q = + (2.1)
where
1 i i i
E E E
+
= ,
1 i i i
T T T
+
= ,
i si i
W W V = + ,
si
W is the slice weight,
V
i
is the vertical applied load and Q
i
the horizontal one. Similarly, from
Chapter 2 Limit equilibrium methods
29
equilibrium along the direction parallel to the slice base, S
i
is achieved:
( )sin cos cos
i i i i i i i i
S W T E Q = + . (2.2)
Along the potential slip surface, shear stress is at the limiting of failure;
therefore:
tan
f
c = + (2.3)
according to the Mohr-Coulomb criterion.
The factor of safety F is defined as that value by which the available
shear strength parameters must be reduced in order to bring the soil mass
into a state of limiting equilibrium along a given slip surface. Hence, the
mobilised shear strength is defined by:
tan
m
c
F F
= + . (2.4)
Accordingly, the shear force acting along the base of each slice is given
by:
( )
1
tan
i i i
S c l P
F
= + (2.5)
where cos
i i i
l x = . From Eq. (2.5), S
i
is substituted into Eq. (2.2),
obtaining:
( ) ( )
1
sin cos cos tan
i i i i i i i i i
W T E Q c l P
F
+ = + (2.6)
From Eq. (2.1), P
i
is substituted into Eq. (2.6); rearranging:
( )
1
sin cos tan
i
i i i i i i
E W W c l m
F
(
= +
(
tan
cos sin
i
i i i i
T m Q
F
| |
+ +
|
\ .
(2.7)
where
tan
cos sin
i
i i
m
F
= + . This parameter was defined first by Janbu
[Janbu et al. 1956].
Chapter 2 Limit equilibrium methods
30
Summing the increments of horizontal interslice forces throughout the
soil mass gives:
( )
1 1
1 1 1
1
0 sin cos tan
i i
n n n
i n i i i i i
i i i
E E E W m W c l m
F
+
= = =
= = = +
1 1
tan
cos sin
i
n n
i i i i
i i
T m Q
F
= =
| |
+ +
|
\ .
(2.8)
In fact, according to the assumed notation,
1 1
0
n
E E
+
. The non null
interslice forces range from
2 2
, E T to ,
n n
E T . Rearranging Eq. (2.8), the
safety factor is obtained:
( )
1
1 1
cos tan
sin
i
i
n
i i i
i
ff
n n
i i i ff
i i
W c l m
F
W m Q I
=
= =
+
=
+ +
(2.9)
where
1
tan
cos sin
i
n
ff i i i
i ff
I T m
F
=
| |
= |
|
\ .
.
The subscript ff is used to indicate that the calculated factor of safety
has been achieved by the equilibrium of forces. As
ff
I and
i
m
depend
on F
ff
, thus iterative schemes must be used to determine F
ff
.
From the overall moment equilibrium equation, another factor of
safety F
mm
can be achieved, as well. First, two equations must be
imposed for each slice: the equilibrium of forces along the vertical
direction (this direction is chosen so that the normal interslice forces do
not appear) and the Mohr-Coulomb failure criterion. Second, if the
overall equilibrium of moments of all forces about an arbitrary point is
imposed, the factor of safety is obtained.
The choice of the pole for the moment equilibrium equation depends
on the slip surface assumed. In case of circular slip, the centre of the
circle is selected as calculations become enormously more simple. In
case of a non circular slip, any point may be chosen, but there are
convergence numerical reasons which determine a criterion of choice. In
fact, there are zones in which the location of the pole can result in
Chapter 2 Limit equilibrium methods
31
numerical instabilities due to computer round-off errors [Fredlund et al.
1992]. A good choice, may be the centre of the circle tangent to the
envelope of the slip surface.
In the following, the factor of safety of the moments is derived. First
the equilibrium equations along the vertical direction are imposed for
each slice:
sin
cos
i i i i
i
i
W T S
P
= (2.10)
then, from Eq. (2.10) P
i
is substituted into Eq. (2.5) obtaining S
i
:
( )
1
cos tan
i
i i i i i
S c l W T m
F
= + (
. (2.11)
Similarly, from Eq. (2.5) S
i
is substituted into Eq. (2.10) obtaining P
i
:
sin
i
i i
i i i
c l
P W T m
F
| |
=
|
\ .
. (2.12)
Imposing the overall equilibrium equation of moments leads to:
1 1 1 1
n n n n
i Wi i Qi i Si i Pi
i i i i
Wb Qb S b Pb
= = = =
+ = +
(2.13)
where b are the arms of the forces. The sign convention adopted refers to
a pole above the toe region of the slip surface (see Fig. 2.2). Substituting
the interslice forces from Eq. (2.11) and (2.12) into Eq. (2.13) leads to:
1 1 1
sin
i
n n n
i Wi i Qi i i i i Pi
i i i
c
Wb Qb W T l b m
F
= = =
(
+ = +
(
1
tan tan
cos
i
n
i i i i Si
i
c l W T b m
F F
=
(
+ +
(
(2.14)
and rearranging:
1 1 1 1
tan
i i
n n n n
i Wi i Qi i Pi i Pi Si
i i i i
Wb Qb W b m T b b m
F
= = = =
| |
+ + + =
|
\ .
Chapter 2 Limit equilibrium methods
32
( )
1
1
cos sin tan
i
n
i i Si i Pi i Si
i
c l b b W b m
F
=
(
= +
. (2.15)
From Eq. (2.15) the factor of safety is obtained:
( )
1
1 1 1
cos sin tan
i
i
n
i i Si i Pi i Si
i
mm
n n n
i Wi i Qi i Pi mm
i i i
c l b b W b m
F
Wb Qb W b m I
=
= = =
(
+
=
+ +
(2.16)
where
1
tan
i
n
mm i Pi Si
i mm
I T b b m
F
=
| |
= +
|
\ .
.
Substituting b
Pi
and b
Si
into the above Eq. (2.16), leads to:
( )
( )
1
1 1 1
tan cos sin
cos sin
i
i
n
i i i i i i i
i
mm
n n n
i i i Qi i i i i i mm
i i i
c l Y W Y X m
F
W X Qb W X Y m I
=
= = =
+ + (
=
+ +
(2.17)
with
i i O
X X X = and
i i O
Y Y Y = . Eq. (2.17) differs from Eq. (2.16)
because the co-ordinates of the centre of the slice bases (X
i
;Y
i
) appear.
This is the expression used by computers to perform calculations.
(X
i
;Y
i
)
(X
o
;Y
o
)
Fig. 2.2: slope slices and pole used to calculate the moment equilibrium for a generic slip
suface.
Chapter 2 Limit equilibrium methods
33
In case of circular slip surface only Eq. (2.11) is used since P
i
do not
appear in the moment equilibrium equation. Further, simplified
expressions relative to the force arms are achieved:
Si
b R = and
sin
Wi i
b R = . Therefore the moment equilibrium equation reduces to:
1 1 1
sin
n n n
i i i Qi i
i i i
R W Qb R S
= = =
+ =
(2.18)
Substituting S
i
from Eq. (2.11) into Eq. (2.18) allows to obtain the safety
factor:
| |
1
1 1
cos tan
sin
i
n
i i i
i
mm
n n
i i i Qi mm
i i
c l W m
F
W Qb I
=
= =
+
=
+ +
(2.19)
where
1
tan
i
n
mm i
i
I T m
F
=
=
.
If 0
i
Q = (horizontal external forces absent) and 0
i
T = , the safety
factor of the Bishop simplified method, also known as Bishops
modified method [Bishop, 1955], is achieved:
| |
1
1
cos tan
sin
i
n
i i i
i
Bishop
n
i i
i
c l W m
F
W
=
=
+
=
(2.20)
In Bishop simplified method, n-1 assumptions are made ( 0
i
T = ). As one
more assumption is made than required, one equilibrium condition
cannot be satisfied. Therefore, the horizontal equilibrium of one slice
cannot be satisfied with the computed safety factor.
If 0
i
E = as well, the vertical equilibrium equations for each slice
(2.10) are still the same and therefore Eq. (2.20) does not change.
For this case ( 0
i
Q = , 0
i
T = , 0
i
E = ) Fellenius
I
derived the factor of
I
In the original formulation, the resultant of interslice forces was assumed to act parallel
to slice base. This formulation leads to violate the principle of action and reaction. Here,
a more correct formulation is preferred.
Chapter 2 Limit equilibrium methods
34
safety of the moments in a different way. He imposed the equilibrium of
forces along the direction normal to the slice base for each slice,
obtaining:
cos
i i i
N W = . (2.21)
Substituting Eq. (2.21) into Eq. (2.18), a simpler factor of safety is
achieved:
| |
1
1
cos tan
sin
n
i i i
i
Fellenius
n
i i
i
c l W
F
W
=
=
+
=
. (2.22)
No iterations are needed to calculate F
Fellenius
. For this reason, Fellenius
method is also known, in literature, as direct method. Note that the
assumptions made are n+2(n-1): rows E, H, I in table 2.1. Therefore there
are n assumptions more than requested. In fact, this method does not
satisfy equilibrium in the direction parallel to the base of each slice, or
equivalently either horizontal or vertical force equilibrium is not
satisfied.
As in case of force equilibrium,
mm
I and
i
m
t
Fig. 2.5: forces and arms involved into slice moment equilibrium about O.
Neglecting the second order terms: 2
i
T x , 2tan
i i
E x ,
i i
E h
and after some rearrangements, the n equations become:
tan
i
i i
i i i i i Qi
Q
h E
T E E h h
x x x
= +
(2.27)
If the slice width x becomes infinitesimal, the equation reduces to:
( ) ( )
( )
tan
Q
dM x dQ
T x E x h
dx dx
= + (2.28)
where ( ) ( ) ( ) M x E x h x = . Eq. (2.28) is a differential equation, very
known in literature, valid independently of the assumptions made to
render the problem statically determinate. Its form suggests to use an
assumption on the line of thrust h(x) such as:
( ) ( ) ( )
3 3
h x f x H x = (2.29)
rather than the relationships given in Eq. (2.23), (2.24) and (2.25). In Eq.
(2.29)
3
is a positive dimensionless scaling parameter to be evaluated
with the factor of safety, f
3
(x) a given function such that ( )
3
0 1 f x and
H(x) the height of the slice. Eq. (2.29) must be used with equations
expressing the moment equilibrium for each slice. Eq. (2.28) is elegant,
but methods of slices are based on sets of algebraic equations, therefore
Eq. (2.29) must be substituted into Eq. (2.27) to calculate shear interslice
Chapter 2 Limit equilibrium methods
42
forces from a known set of normal interslice forces.
Janbu was the first who proposed assumptions concerning the line of
thrust [Janbu, 1954]. In its original formulation Eq. (2.27) was expressed
in a slightly different way:
tan
i i
i i T i Qi
E Q
T E h h
x x
= +
(2.30)
where tan tan
i
T i
h
x
=
r
y
r
x
v
B
D
C
L
Fig. 3.2: first failure mechanism.
According to the assumed failure mechanism, the logarithmic-shaped
region D-B-C rotates as a rigid body about the centre of rotation O, as yet
undefined, with the material below the logarithmic surface BC remaining
at rest (see Fig. 3.2). The assumed mechanism is completely defined by
two variables which can be chosen between the angles, x and y, relative
to the logarithmic spiral and the Cartesian coordinates, X and Y, of its
centre of rotation. For the sake of convenience, the angles are selected.
Chapter 3 Retrogressive failure analytical law
61
The equation of the logarithmic spiral surface is given by:
( ) exp tan
x
r r x = (
(3.1)
and therefore
( ) exp tan
y x
r r y x = (
(3.2)
where r is the radius of curvature of the spiral and r
x
and r
y
are the
minimum and maximum radii of curvature, respectively.
From geometrical considerations, the slope height H and the crest
retreat length L can be expressed as a function of spiral parameters:
( ) { }
exp tan sin sin
x
H r y x y x = (
(3.3)
( ) { }
exp tan cos cos
x
L r y x y x = ( +
(3.4)
In order to calculate the moment due to the soil weight, the area D-B-
C, area A, is expressed as the difference between the area O-B-C, area
A
1
, and the two triangles O-D-B and O-D-C, areas A
2
and A
3
respectively (see Fig. 3.2). In calculations, moment giving rise to
clockwise rotation has been assumed positive.
d
dA
O
v
2
cos
3
r
2
1
A
2
d r d =
Fig. 3.3: differential element of region A
1
.
Considering first the logarithmic spiral region O-B-C, a differential
element of the region is shown in Fig. 3.3. The moment due to the self
Chapter 3 Retrogressive failure analytical law
62
weight of this differential element is given by:
2
1
2 1
M cos
3 2
d r r d
| || |
=
| |
\ .\ .
(3.5)
and integrating over the entire area A
1
, it becomes:
( )
3 3
1
1 1
M cos exp 3tan cos
3 3
y y
x
x x
r d r x d = = (
. (3.6)
After integration of the obtained expression by parts, the moment results:
( )
3
1 1
M , ,
x
r f x y = (3.7)
where
( )
( ) ( )
( )
1
2
exp 3tan 3tan cos sin 3tan cos sin
, ,
3 1 9tan
y x y y x x
f x y
( +
=
+
(3.8)
The moment due to the weight of the triangular region O-D-B is:
( )
2
1 1
M 2 cos sin
3 2
x x
r x L Lr x
( | |
=
|
(
\ .
(3.9)
and after some manipulations, using the geometrical relationships (3.3)
and (3.4), it becomes:
( )
3
2 2
M , ,
x
r f x y = (3.10)
where
( ) ( ) ( )
2 2
2
1
, , sin cos exp 2tan cos
6
f x y x x y x y
(
=
(3.11)
In the same way, the moment for the triangular region O-D-C can be
calculated:
3
2 1
M cos cos
3 2
y y
r y Hr y
| || |
=
| |
\ .\ .
(3.12)
and after some manipulations using Eq. (3.3), it becomes:
( )
3
3 3
M , ,
x
r f x y = (3.13)
Chapter 3 Retrogressive failure analytical law
63
where
( ) ( ) ( ) ( ) ( )
2
3
1
, , exp 2tan cos exp tan sin sin
3
f x y y x y y x y x ( =
.(3.14)
In order to calculate the resistant moment, a differential element of
the logarithmic spiral region O-B-C needs to be considered again (see
Fig. 3.4). Along its arc dl, normal and shear stresses act. The moment due
to these stresses for this element is given by:
( )
r
M cos sin
cos
r d
d r
= + (3.15)
O
d
r
l
cos
r d
d
=
v
Fig. 3.4: stresses acting on the sliding soil mass.
Substituting the adopted failure criterion (Mohr-Coulomb, see Eq. (2.3))
into Eq. (3.15), the total resistant moment is obtained:
( )
2
r
M sin cos sin
cos
y y
x x
r
r c d r c d
= + =
(3.16)
and after integration:
( )
2
r
exp 2tan 1
M
2tan
x
y x
cr
(
= (3.17)
Chapter 3 Retrogressive failure analytical law
64
Imposing the equilibrium of moments about the centre of rotation O,
leads to:
1 2 3 r
M M M M 0 + = (3.18)
Since all terms in (3.18) contain
2
x
r , its simplification leads to a linear
equation in
x
r . Finally, by substituting
x
r with H by Eq. (3.3), the
following expression is obtained:
( )
1
=
, ,
c H
f x y
(3.19)
where
( )
( ) { } ( ) { }
( )
1 2 3
exp 2tan 1 exp tan sin sin
, ,
2tan
y x y x y x
f x y
f f f
( (
=
. (3.20)
In order to find the critical cohesion value, maximisation of the
function ( ) , , , , = c c x y H is needed. This will be explained in the next
paragraph 3.3 where the problem will be formulated again according to
the limit analysis upper bound method.
3.2.3. Case of inclined slope
Up to here it has been studied a vertical slope. The case of inclined
slope can not be neglected as it is much more general. In this case, it is
necessary to introduce a geometrical parameter ( the angle between the
slope and the horizon (see Fig. 3.5)) in order to describe the initial slope
profile. The geometrical relationship (3.4) has to be modified into:
( )
( ) ( ) sin sin
exp tan
sin sin
x
y x
L r y x
+ +
= ( +
`
)
. (3.21)
Chapter 3 Retrogressive failure analytical law
65
H
A
2
A
3
O
x
y
r
y
r
x
B
D
C
L
Fig. 3.5: inclined uniform slope.
Considering the expressions of the moments (M
1
, M
2
, M
3
, M
r
), only
the equations relative to M
2
and M
3
are different.
As regards the moment M
2
due to the weight of the triangular region
O-D-B, Eq. (3.9) is still valid. Substituting Eq. (3.21) instead of Eq. (3.4)
into Eq. (3.9) leads to a different expression from Eq. (3.10):
( )
3
2 2
M , , ,
x
r f x y = (3.22)
where
( ) ( ) ( ) {
2
2
2
1 sin
, , , exp 2tan sin
6 sin
x
f x y y x y
= ( + +
( ) ( ) 2exp tan sin sin cos y x x y + ( + +
( ) ( ) }
2
sin 2cos sin sin x x x + + + . (3.23)
As regards the moment M
3
given by the self-weight of the triangular
region O-D-C, Eq. (3.12) does not hold any more. The new expression is
Chapter 3 Retrogressive failure analytical law
66
given by:
( )
3
2 1
M cos cot cot sin cos
3 3 2
y y
H
r y Hr y y
| | (
= + +
|
(
\ .
(3.24)
and after some manipulations, it becomes:
( )
3
3 3
M , , ,
x
r f x y = (3.25)
where
( )
( )
( ) ( ) {
3 2
sin
1
, , , exp tan exp 2tan sin
6 sin
y
f x y y x y x y
+
= ( (
( ) ( ) ( ) }
2
sin sin cos 2exp tan sin sin cos sin y y y x x y x + + + + ( (
(3.26)
Imposing the equilibrium of moments Eq. (3.18) leads to:
( )
1
=
, , ,
c H
f x y
(3.27)
where
( )
( ) { } ( ) { }
( )
1 2 3
exp 2tan 1 exp tan sin sin
, , ,
2tan
y x y x y x
f x y
f f f
( (
=
. (3.28)
Note that the function obtained (Eq. (3.28)) depends on one
parameter more than the function obtained in case of vertical slope (Eq.
(3.20)).
3.3. First failure mechanism, formulation by the limit
analysis upper bound method
In this paragraph, all the calculations developed in the previous 3.2,
will be derived again on the basis of the upper bound limit analysis
theorem. Before doing this, the upper bound limit analysis theorem will
be recalled.
3.3.1. The limit analysis upper bound theorem
Chapter 3 Retrogressive failure analytical law
67
Considering a three-dimensional solid, a virtual rate of displacement
which satisfies the following relations:
ext
W 0
F
j j j j
V S
F u dV f u dS
= + >
(3.29)
( )
, ,
1
2
ij i j j i
u u = + (3.30)
ij
ij
g
and 0
(3.31)
with g vector of plastic modes making a convex domain in the stress
space, gives rise to a kinematically admissible act of motion. Assuming
such an act of motion, the upper bound limit analysis theorem states that:
the loads determined by equating the rate at which the external forces do
work:
ext
W
F
j j j j
V S
F u dV f u dS
= +
(3.32)
to the rate of internal dissipation:
V
W V =
d ij ij
d
(3.33)
will be either higher or equal to the actual limit load. Therefore, it can be
inferred that the lowest load among all the loads relative to admissible
failure mechanisms, determined by equating the rate of external work to
the rate of energy dissipation, is the best approximation to the limit load.
This load is an unsafe upper bound on the limit load.
As regards the studied case, the only force present is the weight force
(a body force) and no tractions are present on the solid boundary. Eq.
(3.31) are satisfied since a c- type of soil has been assumed. Further, the
studied problem is two dimensional: equations (3.32) and (3.33) become:
ext
W
j j j j
V A
F u dV F u dA =
(3.34)
V A
W V A
d ij ij ij ij
d d =
(3.35)
A further assumption about kinematics has been made: only rigid
Chapter 3 Retrogressive failure analytical law
68
body motions are considered. This means that strains develop only along
a narrow separation layer (discontinuity surface) between a sliding rigid
body and a fixed one (see Fig. 3.6). All energy dissipation occurs along
the separation layer. Discontinuity surfaces can be: plane, circular or
logarithmic spiral shaped. Any other shape of discontinuity surface needs
strains to develop somewhere within the sliding body, in order to not
violate compatibility.
Fig. 3.6: strains along the separation layer.
According to the assumptions made, the rate of energy dissipation is
given by:
( ) W l
d
d
= +
(3.36)
Stresses and strains are shown in Fig. 3.7.
,
,
Morh-Coulomb
failure surface
Fig. 3.7: stresses and strains along the separation layer.
As regards the slope self-weight, it is given by:
V A 1 F mg g g = = = . Since the area A is proportional to the slope
height H, the load is proportional to H as well.
Chapter 3 Retrogressive failure analytical law
69
Now let us consider the well known problem of determining the
critical height of a vertical cut in a c- soil. Assuming the hypothesis
made about kinematics valid, this problem differs from the problem
tackled here (to determine the c value at which a vertical slope collapses)
in the role played by H and c. In the case of the design of a vertical cut, c
is known and H must be determined; vice versa, in the case of a natural
slope subject to weathering, H is known and c is sought. Considering the
former case, the most critical mechanism is characterised by the lowest
height value which is given by the minimum of the function
H=H(considered mechanisms) obtained by the energy balance equation:
ext
W W
d
=
(3.37)
Note that the minimum of the function is higher than the true critical
height since it is an upper bound on it. Considering the latter case, the
most critical mechanism is characterised by the highest cohesion value
which is given by the maximum of the function c=c(considered
mechanisms) obtained by the same energy balance equation (3.37). Note
that the maximum of the function is lower than the true critical cohesion
value.
3.3.2. Determination of the analytical solution
In the following, first the rate of external work will be calculated and
second the rate of energy dissipation.
Considering a differential element of region A
1
, as shown in Fig. 3.8,
the rate of external work done by the soil weight is given by:
1 1 1
W u F =
i d d (3.38)
where
1
u is the displacement vector and
1 1
A d d = F k , with k vertical
unit vector. Calculations lead to:
( )
( )
( )
1
1 1 1 1 1 1 1 1
1 1
W cos A cos
cos
G O
X X
d d d d d
d
= = = u F u F u F
u F
2
2 1
cos
3 2
r r d
| || |
=
| |
\ .\ .
(3.39)
Chapter 3 Retrogressive failure analytical law
70
which is identical to (3.5) except for , rate of angular displacement of
the element. So, developing calculations in the same way as in 3.2.2,
the rate of work becomes:
( )
3
1 1
W , , =
x
r f x y (3.40)
dA
O
d
1
u
2
1
A
2
d r d =
A d k
G
Fig. 3.8: differential element of region A
1
.
As regards the area A
2
, the rate of external work is:
( )
( )
2
2 2 2 2 2 2
2 2
W A cos
cos
u F u F
u F
= = =
i
G O
X X
d
d
( )
1 1
2 cos sin
3 2
x x
r x L Lr x
(| |
=
|
(
\ .
(3.41)
and for the area A
3
:
( )
( )
3
3 3 3 3 3 3
3 3
W A cos
cos
u F u F
u F
= = =
i
G O
X X
d
d
2 1
cos cos
3 2
y y
r y Hr y
| || |
=
| |
\ .\ .
(3.42)
identical to (3.9) and (3.12) respectively, except for . Again developing
Chapter 3 Retrogressive failure analytical law
71
calculations in the same way as in 3.2.2, the rates of work become:
( )
3
2 2
W , , =
x
r f x y (3.43)
( )
3
3 3
W , , =
x
r f x y (3.44)
As regards the internal work, all soil energy dissipation occurs along
the logarithmic spiral surface B-C. The rate of energy dissipation along a
differential element of the spiral, is:
( )
d
W l d d = +
(3.45)
According to the normality plastic flow rule tan = , therefore
substituting this expression into Eq. (3.45) leads to:
( )
2
d
W tan tan cos
cos cos
r d r d
d c c r c r d
= + + ( = =
and the total rate of energy dissipation is:
2
W
y
d
x
c r d =
(3.46)
identical to Eq. (3.16), except for and sign. After integration, it
becomes:
( )
2
exp 2tan 1
W
2tan
d x
y x
cr
(
=
(3.47)
Equating the rate of external work to that of energy dissipation, leads
to:
1 2 3
W W W W
d
=
(3.48)
All terms contain . After its simplification, the same equation as
(3.18) is obtained, and after making the same manipulations as in 3.2,
Eq. (3.19) is obtained again.
In order to determine the most critical mechanism is necessary to
find the maximum of the function ( ) , , , , c c x y H = obtained by the
energy balance Eq. (3.48) respect to the two variables x and y. This
corresponds to find the minimum of the function ( ) , , f f x y = , as it can
be seen from Eq. (3.19).
Chapter 3 Retrogressive failure analytical law
72
Let us consider again the problem of finding the critical height of a
vertical cut; in order to determine the most critical mechanism is
necessary to find the minimum of the function ( ) , , , , H H x y c =
obtained from Eq. (3.48) as well. This leads to write Eq. (3.19) in a
different way:
( ) = , ,
c
H f x y
(3.49)
From Eq. (3.49) can be easily seen that the minimum of
( ) , , , , H H x y c = , is found by determining the minimum of the function
( ) , , f f x y = .
In conclusion, even if the two problems are different, their solution is
supplied by the same analytical function.
3.3.3. Charts relative to the first mechanism
The surface representing ( ) , , f f x y = in a Cartesian X,Y,Z space is
very irregular, with a lot of points of stationarity. Only in a small region
of the domain is bowl shaped (the bottom of the bowl is the relative
minimum solution of the problem). Any algorithm of any mathematical
code needs a starting value to find the minimum of a function. It has been
experienced that convergence to the solution is achieved only if a value
within the bowl region is supplied. The starting value must be varied
depending on and in case of inclined slope on as well.
In the following charts, the dimensionless stability number defined as
N H c = and the logarithmic spiral angles x and y are plotted against
values ranging from 0 to 42.5, in case of
o
90 = . Higher values are not
justified, in practice.
Chapter 3 Retrogressive failure analytical law
73
stability number vs.
0
1
2
3
4
5
6
7
8
9
10
0 5 10 15 20 25 30 35 40 45
[degrees]
N
=
H
/
c
Chart 3.1: values relative to the first failure occurring to a slope initially vertical ( = 90).
x vs.
0
10
20
30
40
50
60
0 5 10 15 20 25 30 35 40 45
[degrees]
x
[
d
e
g
r
e
e
s
]
Chart 3.2: values relative to the first failure occurring to a slope initially vertical ( = 90).
Chapter 3 Retrogressive failure analytical law
74
y vs.
0
10
20
30
40
50
60
70
80
0 5 10 15 20 25 30 35 40 45
[degrees]
y
[
d
e
g
r
e
e
s
]
Chart 3.3: values relative to the first failure occurring to a slope initially vertical ( = 90).
3.3.4. Numerical solutions by limit analysis
Loukidis studied the problem of a uniform inclined slope (see Fig.
3.5) subject to earthquake by limit analysis [Loukidis et al., 2003]. He
used the so called pseudo-static approach to take into account the
earthquake effects. According to this approach, the seism-induced
loading is represented by statically applied inertial forces. The stability of
the slope is expressed in terms of a single parameter: the critical
acceleration factor
C C
K a g = (see p. 39). The known data relative to the
slope are: , H, , , c; whereas the unknown to be determined is K
C
.
He determined the critical acceleration factor K
C
by using both
numerical lower and upper bound methods implemented in a finite
element limit analysis code [Sloan, 1988, 1989]. Moreover, he compared
the obtained results with those achieved by classical upper bound
method having assumed the same mechanism used here: a rotational
mechanism based on rigid body motion with logspiral failure surface
passing through the slope toe.
He found that the average difference of the numerical bounds, with
respect to their mean values (values obtained by averaging the results
Chapter 3 Retrogressive failure analytical law
75
achieved for different cohesion values), for three slope inclinations was
4 % , 6 % and 10 % for = 20, = 30 and = 45 respectively.
The classical upper bound method yielded values that were considerably
lower than those from the numerical upper bound limit analysis reducing
the difference between lower and upper bounds, with respect to their
mean, by approximately 50%. In fact, the difference between the
classical upper bound method and the numerical lower bound was,
2 % , 3 % and 3 % for = 20, = 30 and = 45 respectively.
Moreover, the performance of classical upper bound solutions is better at
low cohesion values (see table 3.1).
Data Method of analysis
[deg.] c [kPa] Num. l.b. Num. u.b. Clas. u.b.
20 20 0.107 0.133 0.114
20 40 0.271 0.304 0.287
20 60 0.399 0.431 0.420
30 20 0.291 0.331 0.302
30 40 0.464 0.504 0.477
30 60 0.593 0.631 0.615
Table 3.1: comparison of horizontal critical acceleration coefficient K
C
obtained for a
uniform slope with H = 20 m, = 20 kN/m
3
, = 30 by limit analysis methods
(after [Loukidis et al., 2003]). Note that classical upper bound solutions are
better at low cohesion values.
It is worth to remark that the difference between numerical lower
bound and classical upper bound solutions is very small. Therefore the
latter solutions are very close to the true ones which must lie between
lower and upper bounds by definition. Moreover it can be noted that
classical upper bounds are much lower than numerical upper bounds.
The analytical expression relative to the energy balance (equivalence
of the external work rate to the dissipated energy) was achieved by
Chang [Chang et al., 1984]. This energy balance has the same structure
as what has been achieved here (see Eq. (3.48)). But, the horizontal
inertial force gives rise to an additional term, included into the rate of
external work. This term is obtained in the same way as the gravity
term (
1 2 3
W W W + +
) has been calculated. If the energy balance equation
is manipulated in such a way that the function ( ) , , , , , K K x y c H = is
Chapter 3 Retrogressive failure analytical law
76
achieved, the function results different from ( ) , , , , c c x y H = because of
this term. Therefore, the values of the variables x and y corresponding to
which the two functions have their minimum and maximum respectively,
do not coincide. Hence, the results obtained for the case studied by
Loukidis cannot be extended to the case studied here (no earthquake
action but weathering action), in a rigorous way. Nevertheless, as the two
problems are similar, in the light of Loukidis results it can be reasonably
expected that numerical lower bound solutions are very close to the
solutions obtained by the upper bound method with logspiral mechanism.
This supports the opinion that the solutions here achieved are very close
to the true collapse solutions.
[Abdi et al., 1994] compared the classical upper bound solution (log-
spiral mechanism) to numerical lower and upper bound solutions
(determined by using a finite element limit analysis code) for the case of
a uniform reinforced vertical slope. They found out that the classical
upper bound solution is closer to the numerical lower bound than the
numerical upper bound for any value of . Therefore also in this case, the
classical upper bound solution resulted better than the numerical upper
bound solution since it was closer to the true collapse solution.
Anyway, it is useful to recall that a true collapse solution is true
within the conceptual framework of the constitutive model adopted
which is an approximation of the real soil behaviour and not the real one.
3.4. Determination of the second failure surface
3.4.1. Introduction
After the logarithmic spiral shaped region D-B-C is slipped away,
the slope is characterised by a logarithmic spiral boundary, as shown in
Fig. 3.9.
Note that the material accumulated on the slope toe cannot be taken
into account by the methods here used (limit equilibrium method and
limit analysis upper bound method), since they are not able to give any
information about the position reached by the landslided material.
Therefore it has been assumed that all material brought by landslides and
accumulated on the slope toe is removed by atmospheric agents and/or
Chapter 3 Retrogressive failure analytical law
77
marine erosion before a new landslide develops. This condition is known
as strong erosion in the literature (weathering-limited process in
geomorphological models).
In this case, no information was known by the author concerning
which is the most critical surface among the plane, circular and
logarithmic spiral as the demonstration developed by Chen concerns only
the case of a plane slope [Chen, 1975]. However, assuming a logarithmic
spiral shaped mechanism makes it possible to consider any mechanism
within the group of mechanisms based on rigid body motion. In fact a
circular shaped mechanism is obtained in case of = 0 and a plane
mechanism is a particular case of a circular one ( r = ).
H
Fig. 3.9: slope profile after the first failure.
3.4.2. Determination of the analytical solution
Let us consider the double logarithmic spiral shaped region B-C-E
which rotates as a rigid body about the centre of rotation O
N
, as yet
undefined, with the material below the logarithmic surface C-E
remaining at rest (see Fig. 3.10). For the sake of clarity, the same
symbols adopted in 3.3.2, are used to indicate the geometrical data
relative to the two logarithmic spirals. The symbols used for the spiral B-
C, named old, are upper lined, whereas the symbols used for the spiral E-
C, named new, are the same as those shown in Fig. 3.2.
Chapter 3 Retrogressive failure analytical law
78
O
O
O
N
N N N N O
1 2 3
A =A -A -A -A
O O O O
1 2 3
A =A -A -A
N
2
A
N
A
H
x
y
O
A
N
3
A
O
2
A
O
3
A
C
D B E
v
L
L
N:= new
O:= old
x
r
y
r
x
y
r
y
r
x
Fig. 3.10: second failure mechanism: grey lines are relative to the old spiral (B-C) whereas
black lines to the new one (E-C).
Since the slope profile is more complicated than the previous case of
plane slope, more geometrical data are needed to describe the slope
features. They are the angles x and y (see Fig. 3.10) relative to the
logarithmic spiral B-C. The geometrical relationships (3.3) and (3.4) are
still valid, keeping in mind that x and y are the angles relative to the
sliding spiral surface E-C. Two more relationships, concerning the spiral
B-C, have to be considered:
( ) { }
exp tan sin sin
x
H r y x y x = (
(3.50)
( ) { }
exp tan cos cos
x
L r y x y x = ( +
(3.51)
Chapter 3 Retrogressive failure analytical law
79
In order to determine the rate of external work done by the soil
weight, the same method as that used in 3.3.2, based on the difference
of areas, has been used. The rate of work done by the double logarithmic
spiral shaped region B-C-E, area A
N
, is obtained as the work done by the
region O
N
-C-E, area
N
1
A , minus the work done by the regions O
N
-D-E
and O
N
-D-C, areas
N
2
A and
N
3
A respectively, minus the work done by the
region B-C-D. The latest is expressed again as the difference between the
work done by region O-B-C, area
O
1
A , and the two triangular regions O-
D-B and O-D-C, areas
O
2
A and
O
3
A respectively. Therefore the rate of
external work done by the soil weight is given by:
( )
N N N O O O
ext 1 2 3 1 2 3
W W W W W W W =
(3.52)
As regards
N
1
W
,
N
2
W
and
N
3
W
,
2
W
and
3
W
= = u F u F
u F
i (3.53)
where
1
2
cos cos cos
3
N G y y
O
X X r r y r y = + .
Therefore the external work for the entire region becomes:
O 2 3
1
1 3 3
W cos cos cos
3 2 2
y y
y y
x x
r y r y r d r d
( | |
= + +
| (
\ .
(3.54)
and after integration by parts:
( ) ( )
O 2
1
exp 2tan 1
1 3 3
W cos cos
3 2 2 2tan
y y x
y x
r y r y r
| |
= + +
|
\ .
Chapter 3 Retrogressive failure analytical law
80
( ) ( )( ) ( )
3
2
exp 3tan sin 3tan cos sin 3tan cos
1 9tan
x
y x y y x x
r
( + +
+ (
+
(
. (3.55)
From Eq. (3.55), emerges that the final equilibrium equation will
assume a different structure from that relative to the first mechanism
obtained in the previous 3.3.2, since it will not contain any more
2
x
r in
all its terms. So in order to achieve the function which has to be
minimised, Eq. (3.55) must to be further manipulated.
Substituting Eq. (3.50) and (3.3) into Eq. (3.55), a more convenient
expression is obtained:
( ) ( )
( ) ( )
( ) ( )
( ) ( )
O 3
1
2
exp tan cos exp 2tan 1
W
exp tan sin sin
4tan exp tan sin sin
y x y y x
H
y x y x
y x y x
= +
( ) ( )
3
1
exp tan sin sin y x y x
+
(
( ) ( )( )
( )
2
exp 3tan 3tan cos sin 3tan cos sin
3 1 9tan
y x y y x x
( ) ( ) ( ) ( )
exp 2tan 1 exp tan cos
4tan
y x y x y
( (
(
`
(
)
. (3.56)
Considering now the region
O
2
A , whose gravity centre is G
2
, the rate
of external work is:
( )
( )
2 O O O O O O
2 2 2 2 2 2
O O
2 2
W A cos
cos
N G
O
X X
= = u F u F
u F
i (3.57)
where
( )
2
1
cos 2 cos cos
3
N G y x y
O
X X r y r x L r y = + + .
After some manipulations and substituting Eq. (3.50), (3.51) and (3.3)
into Eq. (3.57), the following expression is obtained:
Chapter 3 Retrogressive failure analytical law
81
( ) ( )
( ) ( )
( ) ( )
( ) ( )
O 3
2
2
exp tan cos cos sin exp tan cos
W
exp tan sin sin
2 exp tan sin sin
y x y x x y x y
H
y x y x
y x y x
( +
= +
( ) ( )
( ) ( )
( ) ( )
3
exp tan cos cos sin
cos 2exp tan cos
6 exp tan sin sin
y x y x x
x y x y
y x y x
( +
( +
`
(
)
. (3.58)
As regards the region
O
3
A , whose gravity centre is G
3
, the rate of
external work is:
( )
( )
3 O O O O O O
3 3 3 3 3 3
O O
3 3
W A cos
cos
N G
O
X X
= = u F u F
u F
i (3.59)
where
3
1
cos cos
3
N G y y
O
X X r y r y = .
After some manipulations and substituting Eq. (3.50) and (3.3) into Eq.
(3.59), the following expression is obtained:
( ) ( )
( ) ( )
( ) ( )
( ) ( )
O 3
3
exp tan cos exp tan cos
W
exp tan sin sin 2 exp tan sin sin
y x y y x y
H
y x y x y x y x
= +
( ) ( )
( ) ( )
2
exp tan cos
1
6 exp tan sin sin
y x y
y x y x
(
(
`
(
)
. (3.60)
As regards the internal work, all soil energy dissipation occurs along
the logarithmic spiral surface E-C. The rate of energy dissipation along
this spiral needs to be calculated in the same way as previously done for
the first failure mechanism in 3.3.2. Hence, it is given by Eq. (3.47):
( )
2
exp 2tan 1
W
2tan
d x
y x
cr
(
=
.
All terms contained in the energy balance equation (3.37) have been
calculated, but in order to obtain a function in the form
( ) , , , , , , c c x y x y H = it is necessary to develop further Eq. (3.40), (3.43),
(3.44) and (3.47) substituting H in them by Eq. (3.3). This leads to:
Chapter 3 Retrogressive failure analytical law
82
( )
( ) { }
1 N 3
1
3
, ,
W
exp tan sin sin
f x y
H
y x y x
=
(
(3.61)
( )
( ) { }
2 N 3
2
3
, ,
W
exp tan sin sin
f x y
H
y x y x
=
(
(3.62)
( )
( ) { }
3 N 3
3
3
, ,
W
exp tan sin sin
f x y
H
y x y x
=
(
(3.63)
( )
( ) { }
2
2
exp 2tan 1
W
2tan exp tan sin sin
d
y x
cH
y x y x
(
=
(
(3.64)
Equating the rate of external work (Eq. (3.52)) to that of energy
dissipation (Eq.(3.47)), leads to:
( )
N N N O O O
1 2 3 1 2 3
W W W W W W W
d
=
(3.65)
All terms contain and
2
H . After their simplification, and
rearranging Eq. (3.65) becomes:
1
( , , , , )
c H
g x y x y
= (3.66)
In the following, function ( ) , , , , g g x y x y = has been written trying to
make as few manipulations as possible from the previously achieved
analytical expressions referring to work rates.
( )
( ) ( )
( ) ( ) ( ) ( )
2 3
exp 2tan 1
1
, , , ,
2tan exp tan sin sin exp tan sin sin
y x
g x y x y
y x y x y x y x
( (
( ) ( )( )
( )
2
exp 3tan 3tan cos sin 3tan cos sin
3 1 9tan
y x y y x x
+
+
+
( ) ( ) ( )
2 2
sin exp 2tan cos cos
6
x y x y x +
+
Chapter 3 Retrogressive failure analytical law
83
( ) ( ) ( ) ( ) ( )
2
exp 2tan cos exp tan sin sin
3
y x y y x y x
(
(
+
(
( ) ( )
( ) ( )
( ) ( )
1
1 2
exp tan cos
, , , ,
exp tan sin sin
y x y
a x y a x y
y x y x
+ +
`
)
(3.67)
where ( )
( ) ( )
( ) ( )
1 2
exp 2tan 1
1
, ,
2tan
2 exp tan sin sin
y x
a x y
y x y x
= +
(
( ) ( )
sin exp tan cos cos x y x y x ( + + +
( ) ( ) ( ) ( ) }
exp tan cos exp tan sin sin y x y y x y x ( +
and ( )
( ) ( )
2 3
1
, ,
exp tan sin sin
a x y
y x y x
=
(
( ) ( )( )
( )
2
exp 3tan 3tan cos sin 3tan cos sin
3 1 9tan
y x y y x x
( ) ( ) ( ) ( )
exp tan cos exp 2tan 1
4tan
y x y y x
(
+ +
( ) ( ) ( ) ( )
2 2
sin 2exp 2tan cos 3exp tan cos cos cos
6
x y x y y x x y x ( +
+ +
( ) ( ) ( ) ( )
2
exp 2tan cos exp tan sin sin
6
y x y y x y x
(
)
.
Note that ( )
1
, , a x y and ( )
2
, , a x y depend only on parameters, therefore
they assume constant values when the minimum of ( ) , , , , g g x y x y = is
sought by Matlab algorithms.
In order to determine the failure mechanism is necessary to find the
maximum of ( ) , , , , , , c c x y x y H = obtained by the energy balance Eq.
Chapter 3 Retrogressive failure analytical law
84
(3.65) respect to the two variables x and y. The maximum of this function
is obtained by finding the minimum of ( ) , , , , g g x y x y = .
3.4.3. Charts relative to the second mechanism
The ( ) , , , , g g x y x y = function presents characteristics similar to
( ) , , = f f x y as it is very irregular and only in a small part of the domain
is bowl shaped (see 3.3.2).
In the following charts, the stability number and the logarithmic
spiral angles x and y are plotted against values. For friction angles
close to 0 ( 2 < ), failure surfaces pass below the toe. In order to take
into account such a failure, another variable to describe the mechanism
should be introduced, according to Chen. Since only the = 0 case
possesses an engineering interest, failure surfaces passing below the toe
have not studied in a general way whereas the = 0 case has been
tackled in 3.6. Values, plotted on charts, relative to very low friction
angles ( 2 < ) have been marked by hollow points to recall the reader
that slightly different values (lower stability numbers) are found if a more
critical mechanism is considered (failure surface passing below the toe).
stability number vs.
0
5
10
15
20
25
0 5 10 15 20 25 30 35 40 45
[degrees]
N
=
H
/
c
Chart 3.4: values relative to the second failure occurring to a slope initially vertical ( = 90).
Chapter 3 Retrogressive failure analytical law
85
x vs.
0
10
20
30
40
50
60
0 5 10 15 20 25 30 35 40 45
[degrees]
x
[
d
e
g
r
e
e
s
]
Chart 3.5: values relative to the second failure occurring to a slope initially vertical ( = 90).
y vs.
90
92
94
96
98
100
102
104
106
108
110
0 5 10 15 20 25 30 35 40 45
[degrees]
y
[
d
e
g
r
e
e
s
]
Chart 3.6: values relative to the second failure occurring to a slope initially vertical ( = 90).
Chapter 3 Retrogressive failure analytical law
86
3.4.4. Some aspects of the first two mechanisms
It seems interesting to show the difference between the spiral
surfaces relative to the first two failures. In the following charts, the co-
ordinates of the centres of rotation of the spirals are shown. The axis
origin has been set on the slope toe, as shown in Fig. 3.11. Moreover, co-
ordinates are normalised to the slope height H. This way of showing
results makes it possible to apply these charts to any real slope height.
H
Y
X
Fig. 3.11: axis origin.
X vs. , first spiral
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
0 5 10 15 20 25 30 35 40 45
[degrees]
X
/
H
Chart 3.7: values relative to a slope initially vertical ( = 90).
Chapter 3 Retrogressive failure analytical law
87
X vs. , second spiral
-0.5
-0.5
-0.4
-0.4
-0.3
-0.3
-0.2
-0.2
-0.1
-0.1
0.0
0 5 10 15 20 25 30 35 40 45
[degrees]
X
/
H
Chart 3.8: values relative to a slope initially vertical ( = 90).
It is worth to note that the X co-ordinate of centre of rotation of the
second spiral is negative for any value. This means that the centre is on
the right respect to the slope toe. Considering the tangent to the spiral at
the slope toe, its inclination is lower than the friction angle.
Y vs. , first spiral
1.2
1.4
1.6
1.8
2.0
2.2
2.4
2.6
2.8
0 5 10 15 20 25 30 35 40 45
[degrees]
Y
/
H
Chart 3.9: values relative to a slope initially vertical ( = 90).
Chapter 3 Retrogressive failure analytical law
88
Y vs. , second spiral
1.2
1.4
1.6
1.8
2.0
2.2
2.4
2.6
2.8
0 5 10 15 20 25 30 35 40 45
[degrees]
Y
/
H
Chart 3.10: values relative to a slope initially vertical ( = 90).
Up to here, it has been implicitly assumed that the logarithmic spiral
failure surface passes through the slope toe. As already stated, the failure
surface passes below the toe in case of very low friction angles; but it
could pass above the slope toe as well (see Fig. 3.12), since the
geometrical symmetry guaranteed by a plane slope profile is not valid
any more.
H
O
O
O
N
h
v
Fig. 3.12: failure surface passing above the slope toe.
Chapter 3 Retrogressive failure analytical law
89
In order to consider this possibility, the assumption of spiral failure
surface passing through the slope toe must be abandoned. Of course,
removing an hypothesis makes the problem more complicated. In fact,
now, the height of the spiral failure surface does not coincide any more
with the slope height and it is unknown.
To solve the problem, the spiral slope has been divided into a
discrete number of points and each point has been assumed as the toe of a
slope whose height is known. The maximum cohesion value and the
logarithmic spiral angles x and y corresponding to the most critical
mechanism, have been determined for all the slopes; each spiral having
height h
i
lower than the slope height H, as shown in Fig. 3.13. The most
critical mechanism is the mechanism with the highest cohesion value and
it has to be taken.
N
3
O N
2
O
N
1
O
H
c
3
h
2
c
2
c
1
O
x
y
2
O
Fig. 3.13: failure surfaces relative to the different mechanisms considered.
The results obtained in case of a slope initially vertical ( = 90)
have shown that the second failure mechanism passes through the slope
toe for any value of . This means that charts 2.4, 2.5, 2.6, 2.8 and 2.10
referred to a failure surface passing through the slope toe, describe the
actual failure mechanism.
Chapter 3 Retrogressive failure analytical law
90
3.4.5. A different problem: which is the best slope profile?
Let us consider a problem which is, for some aspects, dual to the
problem tackled up to here: assuming that the material strength is
suitably described by cohesion and friction (c, ), which is the best
profile joining point A to point B (see Fig. 3.14)? The two points are
fixed in space; therefore H and L are assigned. This problem is relevant
in some situations of geotechnical engineering (the construction of
artificial embankments and the stabilisation of natural slopes). In these
cases, the height of the embankment or slope is assigned and the
optimum profile is that which guarantees the largest safety at the lowest
cost. The length L is a geometrical constraint due to many reasons. The
extension of artificial embankments along the horizontal direction can be
limited typically by constructions (urban environment) or by agricultural
lands (rural environment). Works to stabilise natural slopes can be
subject to the same limitations. The natural way to formulate this
problem is based on variational calculus. But, up to now, no solution has
been found by variational calculus to the problem. Here, some results
have been achieved.
H
A
B
L
f
1
f
2
Fig. 3.14: problem under investigation: which is the best profile joining point A to point B?
Material strength (c, ) is uniform.
Let us consider profiles made by logarithmic spirals. They are
described by the following analytical expression:
( ) exp
x
r r a x = (
. (3.68)
Eq. (3.68) is relative to a family of logarithmic spirals. Eq. (3.68) is
similar to Eq. (3.1), but a general parameter a has been introduced
instead of tan. The family of spirals depends onto three parameters: r
x
,
Chapter 3 Retrogressive failure analytical law
91
a , x. Substituting Eq. (3.3) into Eq. (3.68), the parameters become: a , x,
y. Imposing that all the spirals pass through points A and B leads to an
analytical expression relative to the spirals where only one parameter
misses to be determined. This parameter rules the shape of the spirals
passing through A and B.
Let us assume tan a = . Once A and B are assigned, the logarithmic
spiral profile is fully determined. Let us compare this profile with the
plane one (see Fig. 3.15). The critical values of cohesion are determined
by finding the minimum of ( ) , , f f x y = (see Eq.(3.27)) in case of plane
profile and the minimum of ( ) , , , , g g x y x y = (see Eq. (3.66)) in case of
log-spiral profile. In order to compare the two cases, let us consider a
plane profile characterised by a factor of safety equal to unity, 1 Fs = ; i.e.
the slope height assumed, H, is critical for the profile considered. In case
of 63 = and 30 = , the critical value of cohesion results equal to
( ) 0.0702 c H = . Since it has been assumed that 1 Fs = for the plane
profile, ( ) 0.0702 c H = is the cohesion of the soil which is therefore
characterised by: 30 = and ( ) 0.0702 c H = . Now a logarithmic spiral
profile passing through the same points A and B is considered (see Fig.
3.15). The critical value of cohesion relative to this profile is equal to
( ) 0.0641 c H = . Since soil cohesion is equal to ( ) 0.0702 c H = , if this
profile is adopted the factor of safety results equal to 1.094 Fs = (the
factor of safety has been calculated as the reduction to assign to cohesion
in order to bring soil to failure). In the case analysed, adopting the
logarithmic spiral profile instead of the plane one leads to a factor of
safety 9.4% larger. This appears very interesting since the log-spiral
profile needs less material: an excavation is required to increase the
safety of the embankment analysed! Similar results are obtained for other
values of and .
Chapter 3 Retrogressive failure analytical law
92
H
A
B
L
Fig. 3.15: the two profiles analysed: plane and logarithmic spiral shaped.
The results shown above explain the choice operated by some
Japanese engineers who designed, several centuries ago, many castles
using logarithmic spiral profiles for their walls made of stones (see Fig.
3.16).
Fig. 3.16: different views of Kumamoto castle (Japan). The wall profiles are clearly log-
spiral shaped.
In all the calculations developed in previous sections, the parameter
a , one of the three parameters ruling the shape of logarithmic spirals (see
Eq. (3.68)), has been taken equal to tan . This assumption has been
made as logarithmic spirals have been introduced in order to describe the
failure lines developing within the slopes. These lines are log-spiral
shaped because mechanisms characterised by rigid motion of the
detaching soil mass have been assumed. a is equal to tan because the
slipping away soil mass moves making an angle equal to with the
failure lines (see Fig. 3.2 and Fig. 3.10). But in this section, logarithmic
spirals have been introduced in order to describe fronts of slopes. The
problem tackled consists into determining which is the best profile
joining two points fixed in space (i.e. H and L are assigned). To answer
the question, all the possible profiles should be considered. In this section
Chapter 3 Retrogressive failure analytical law
93
a particular family of profiles (logarithmic spirals) has been taken; the
more general the family is, the more exhaustive the answer will be. When
tan a is assumed, Eq. (3.66) does not hold any more: all calculations
developed in 3.4.2 must be derived again assuming Eq (3.68) instead of
Eq. (3.1) for describing the profile of the slope front. In order to
determine the best slope profile (among the log-spiral shaped ones) it is
necessary to derive a suitable analytical expression ( ) , , , , , f f x y x y a =
where two different spiral shapes appear: one relative to the log-spiral
slope front, with tan a , and the other one relative to the log-spiral
failure mechanism, with tan a = . The slope profiles considered (each
characterised by a different value of a ) will be characterised by a
different value of critical cohesion. Intuition (but it is only intuition!)
says that the best spiral (the one which maximises Fs) could be achieved
for tan a = (the case here treated).
In conclusion, it has been proved that the logarithmic spiral profile
improves significantly the performance of artificial embankments of c,
materials in terms of either factor of safety or maximum height that can
be reached. Finding the best embankment profile, even reducing the
research among the family of logarithmic spiral profiles, is still an open
problem which requires further research work to be solved.
3.5. Determination of the successive failure surfaces
3.5.1. Procedure for the determination of the successive
failure surfaces
After the second failure has occurred, the slope profile is
characterised again by a logarithmic spiral. In order to determine the
third failure surface the same procedure used to find the second one, has
been used.
The obtained results have shown that the third failure mechanism
passes above the slope toe independently of the initial inclination of the
slope (). After this failure, the slope geometry is characterised by a
boundary made of two logarithmic spirals (see Fig. 3.17). To determine
the fourth failure, a further difficulty is met since the procedure realised
Chapter 3 Retrogressive failure analytical law
94
up to here, is not applicable to such a profile. It has been assumed that the
fourth failure passes above the point of intersection of the two spirals
intending to verify this hypothesis a posteriori. Hence the procedure has
been applied on the upper spiral of the slope boundary (height h in Fig.
3.17).
h
H
Fig. 3.17: slope profile after the third failure.
The obtained results have shown that the fourth failure surface
passes above the point of intersection of the two spirals. This leads to
conclude that the hypothesis done was right. In fact, if the hypothesis
were false, a surface passing trough the point of intersection of the two
spirals should have been found.
After the fourth failure surface occurring, the slope boundary is
made by three spiral shaped parts. To determine the successive failure
surface, the same procedure has been applied on the upper spiral. A
failure surface passing above the intersection point of the upper spiral
with the middle one has been found.
Using this procedure repeatedly, as many failure surfaces as wanted
can be determined. The algorithm implemented in Matlab needs ,
(initial inclination of the slope) as input data, supplying cohesion, crest
retreat and all the geometrical variables describing the failure surfaces as
result. Some tables containing data relative to significant mechanisms,
(mechanisms describing the retrogression of the slope front up to it
becomes numerically negligible) for the cases of 90 = , 80 = ,
70 = , 60 = , 50 = and for some values of are shown in
appendix A.1. Depending on values, a few (in case of high values of )
or many (in case of low values of ) mechanisms are needed to be
Chapter 3 Retrogressive failure analytical law
95
determined to describe, in a complete way, the evolution of the slope.
3.5.2. Discussion of the results
In Fig. 3.18, a typical picture of the profiles assumed at different
times by an initially vertical slope is shown. From a glance at the figure,
it can be noted that all the failure surfaces are featured by similar shape
and shortening size. As time goes on, spiral centres of rotation become
closer to spirals (see Fig. 3.19). Moreover from a certain failure surface
on, depending on value, successive failure surfaces are featured by
exactly the same shape. They have only curvatures different. Their
stability number, M h c = , assumes a value which remains constant
from the fourth mechanism on, having considered negligible variations
less than 1%.
Fig. 3.18: slope evolution ( =90, =30).
Chapter 3 Retrogressive failure analytical law
96
Fig. 3.19: slope evolution ( =90, =30). Dots indicate the centres of rotation of spirals.
In Fig. 3.20, the slope evolution in terms of normalised cohesion
against crest retreat is shown. Cohesion normalisation has been done
dividing cohesion by soil unit weight and initial slope height. From the
figure, all the slope evolution from the beginning to the end of the
retrogressive phenomenon, is caught and well summarised. Cohesion
decrease is constant in time, but crest retreat occurs only when cohesion
reaches some critical values, therefore the retreat-cohesion relationship is
step like. Moreover, it is worth to note a region where dots lie along a
straight line (right side of Fig. 3.20, from the third dot on). These values
correspond to failure surfaces, previously described, featured by the same
shape and they have been interpolated so that linear crest retreat-cohesion
relationships have been achieved. In Fig. 3.21 these relationships are
shown for different friction angle values. Simple relationships between
and the straight line parameters (slope and intercept) may be achieved as
well (see appendix A.2).
Chapter 3 Retrogressive failure analytical law
97
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
L/H
c
/
(
*
H
)
linear interpolation
Fig. 3.20: dimensionless normalised cohesion vs. crest retreat ( =90, =30). Circles
indicate values obtained by limit analysis whereas lines indicate crest retreat and
cohesion evolution for increasing time.
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
L/H
c
/
(
*
H
)
= 18 = 20
= 15
= 45
= 5
Fig. 3.21: dimensionless normalised cohesion vs. crest retreat obtained by linear
interpolation for different values ( = 90). The first two failure mechanisms
have not been considered.
Note that the developed method does not take into account the
occurrence of secondary mechanisms. After some landslides have
Chapter 3 Retrogressive failure analytical law
98
happened, secondary failures start to occur. These failures develop
between two adjacent spiral pieces of the black profile. The first failure
which occurs lies between point A and point B (see Fig. 3.22). After
further cohesion lowering, successive secondary failure surfaces must
develop. In fact, according to the strong erosion condition (no permanent
material accumulation on the slope), at c = 0, the tangent to the slope
profile is not allowed to have an inclination higher than the friction angle.
Fig. 3.22: slope evolution, ( =90, =30). Gray lines indicate profiles at different times t
i
whereas black lines indicate the final profile.
As concerns inland cliffs (low or absent erosion conditions),
secondary mechanisms are unlikely to be activated since the part of slope
profile where they may occur is covered by residual material. These
mechanisms become significant and worth to be taken into account in
case of cliffs lying by the sea or, more in general, in any case of slope
subject to strong erosion.
In order to take into account such mechanisms it is necessary to
apply the method illustrated in 3.3.2 and 3.4.2 (calculation of external
work rate by subtracting different areas) to slope profiles made by two
spirals instead of one. This makes the use of the limit analysis method far
more complicated since more terms and parameters appear in the
equation of the external work.
Chapter 3 Retrogressive failure analytical law
99
Considering the final slope profile, obtained at c = 0 (see Fig. 3.22),
different results have been achieved in terms of average inclination of the
profile to the horizon. This may be defined by:
arctan
final
H
L
= .
The angle is higher than for friction values higher than a threshold
value ( 22 =
threshold
), vice versa for friction values lower than the
threshold value, is lower than .
Finally, it could have been expected that the slope assume a plane
profile making an angle with the horizon equal to the soil friction angle
(see Fig. 3.23). The results have shown that this conclusion is erroneous.
This can be explained by considering the inclination of the tangent to the
profile of a slope made of non cohesive material: it cannot assume a
value larger than the material repose angle, but it can assume a smaller
one. Therefore a profile made of spiral pieces or of whatever other shape
is physically admissible as long as the inclination of the tangent in any
point is lower than the soil friction angle. Only if all the failure surfaces
pass through the slope toe, the conclusion becomes true. But it has been
shown that this is not the case and therefore the final configuration of the
slope does not assume a plane profile.
failure surface
at c = 0
time increasing
Fig. 3.23: erroneous final slope profile.
3.5.3. Other failure mechanisms
In this chapter, mechanisms based uniquely on rigid body motion
have been assumed. Mechanisms based on straining body motion have
not been considered. Log-sandwich and two-triangle mechanisms (see
Fig. 3.24) are the easiest mechanisms to analyse. The former depends
Chapter 3 Retrogressive failure analytical law
100
onto two parameters, whereas the latter onto three. These mechanisms
have been used by Chen in calculating upper bound solutions for
retaining walls [Limit analysis and soil plasticity, ch. 8, 1975].
logspiral region
v
1
v
2
v
1
v
2
v
1
v
12
a) b)
Fig. 3.24: failure mechanisms based on straining body motion. a) Logspiral mechanism; it is
completely described by two variables (, ). b) Two-triangle mechanism; it is
completely described by three variables (, , ).
But once a failure has occurred, the slope profile becomes spiral
shaped. This boundary does not allow to consider two triangular wedges
any more. Therefore modified log-sandwich and two-wedge mechanisms
have to be used. They differ from the original ones because a triangular
wedge is replaced by a logspiral one.
Unfortunately, they need calculations far more complicated than
those tackled in previous paragraphs and improvement to solutions
achieved considering rigid body motions, is not guaranteed if any.
3.6. Undrained conditions = 0
3.6.1. First failure surface
Soil has to be analysed under undrained conditions, when the
consolidation phenomenon needs a long time interval to develop
compared with the weathering phenomenon. In practice, this can happen
in case of clayey soils characterised by very low permeability. This case
has been treated in a paper by [Michalowski, 2002] and it has been used
to achieve the results shown in the following.
Chapter 3 Retrogressive failure analytical law
101
When = 0 the logarithmic spiral surface becomes a circle. As
regards the first failure, in case of vertical slope, the stability number
assumes the well known value of 3.83, and therefore cohesion value at
failure is:
3.83
H
c
=
In case of inclined slope, cohesion value at failure is:
H
c
N
=
where 3.83 N .
The stability number becomes independent of the slope
inclination in case of less than about 50 (see Fig. 3.25). In fact, for
less than 50, the dimensions of the failure mechanism tend to infinity
(see Fig. 3.26 a). Thus the slope height H becomes negligible with
respect to the failure surface radius r. Consequently, the rate of energy
dissipation assumes a simple form:
( )
2
W 2
d
cr =
(3.69)
where is the angle drown in Fig. 3.26a. The centre of rotation O is
exactly above the midpoint of the slope. Hence the rate of external work
done by the slope weight, when r H , becomes:
2 2
ext
1
W cos
2
Hr =
(3.70)
Equating the rate of energy dissipation to the rate of external work
leads to:
( )
2
cos
2 2
c H
(3.71)
When = 23.2, the maximum cohesion value is found:
0.181 c H = (3.72)
and the stability number N assumes the value of 5.525.
Chapter 3 Retrogressive failure analytical law
102
Fig. 3.25: stability number versus with = 0 (after [Chen and Giger, 1971]).
a) b)
Fig. 3.26: large-size failure mechanism, = 0 analysis; a) in homogeneous soil, b) limited by
the presence of a bedrock (after [Michalowski, 2002]).
3.6.2. Second failure surface
As regards the second failure, in both cases of initially vertical
and inclined slope, the dimensions of the mechanism tend to infinity. In
fact, the circular slope profile, obtained after the first failure, can be
considered, in first approximation, equivalent to a plane profile such that
the soil volume under the two profiles is the same. The equivalent plane
inclination, which depends on the initial slope inclination, results less
Chapter 3 Retrogressive failure analytical law
103
than 50 in any case. Therefore the second failure surface is a circle with
r H (see Fig. 3.26a) and the cohesion value at the onset of the
mechanism is given by Eq. (3.72). These results are unrealistic and they
were already known to [Taylor, 1937, 1948]. A more rational stability
number may be obtained only by limiting the depth of the failure
mechanism to the depth of a bedrock (see Fig. 3.26b).
Finally, it can be concluded that the limit analysis and the limit
equilibrium methods are not suitable to study the evolution of slopes
under undrained conditions.
3.7. References
Abdi R., de Buhan P., Pastor J., 1994. Calculation of the critical height
of a homogenized reinforced soil wall: a numerical approach. Int. J.
Numer. Anal. Methods Geomech., 18, pp. 485-505.
Chang C. J., Chen W. F., Yao J. T. P., 1984. Seismic displacements in
slopes by limit analysis. J. Geotech. Engrg., ASCE, 110(7), pp. 860-874.
Chen W. F., Giger M. W., Fang H. Y., 1969. On the limit analysis of the
stability of slopes. Soils and Found., 9(4), pp. 23-32.
Chen W. F. and Giger M. W., 1971. Limit analysis of stability of slopes.
J. Soil Mech. Found. Engrg., ASCE, 97(1), pp. 19-26.
Chen W. F., 1975. Limit analysis and soil plasticity. Elsevier edition.
Loukidis D., Bandini P., Salgado R., 2003. Stability of seismically loaded
slopes using limit analysis. Gotechnique, 53(5), pp. 463-479.
Michalowski R., 2002. Stability charts for uniform slopes. J. Geotech. &
Geoenv. Engrg., ASCE, 128(4), pp. 351-355.
Sloan S. W., 1988. Lower bound limit analysis using finite elements and
linear programming. Int. J. Numer. Anal. Methods Geomech., 12, pp. 61-
77.
Sloan S. W., 1989. Upper bound limit analysis using finite elements and
Chapter 3 Retrogressive failure analytical law
104
linear programming. Int. J. Numer. Anal. Methods Geomech., 13, pp.
263-282.
Taylor D. W., 1937. Stability of earth slopes. J. Boston Soc. Civil
Engineers., 24(3), pp. 197-246. Reprinted in: Contributions to soil
mechanics 1925 to 1940. Boston Soc. Civil Engineers, pp. 337-386.
Taylor D. W., 1948. Fundamentals of soil mechanics. Wiley.
Part 1 Conclusions
105
Part 1 - Conclusions
A comprehensive review of limit equilibrium methods has been
supplied. The main features of such methods have been shown adopting a
unified formulation of the methods so that each method has been
illustrated as a particular case of a general theory. The physical
reasonableness of some hypotheses introduced in the methods and some
numerical aspects have been critically discussed. Comparisons among the
performances of different methods have also been made.
An analytical law describing the evolution undergone by natural
slopes subject to weathering has been achieved. The starting point of the
work done has been Chen study relative to the problem of the critical
height of a uniform plane slope. Here, a different problem has been
tackled. In order to determine the critical cohesion of a known slope, the
same function determined by Chen has been used but an original
formulation ( 3.2 and 3.3) of the problem has been supplied. According
to the formulation developed, the function ( ) , , f f x y = , already known
in the literature, has been derived consistently with the features of the
problem investigated. The function can be achieved by using both limit
analysis and limit equilibrium methods. Both formulations have been
illustrated in order to supply a large theoretical framework of reference to
the reader.
Part 1 Conclusions
106
An analytical function ( ) , , , , g g x y x y = which makes it possible to
determine the critical values of cohesion and the associated mechanisms
for all the successive retrogressive failures has been achieved. Thanks to
this function a complete description of the retrogressive failure
phenomenon has been obtained. According to the results obtained, the
retrogression of the slope front is linearly related to cohesion decrease
from a certain failure on (the third one in case of a slope initially
vertical). This is a significant indication relative to the phenomenon
studied since it emerges that the relationship between weathering (cause
of cohesion decrease) and retrogression of slopes will be linear from a
certain time on. When the first mechanism is considered, it has been
shown that the solution obtained (classical limit analysis upper bound)
is better than numerical upper bounds nowadays available; therefore it is
reasonable to suppose that the solutions relative to the successive
mechanisms may be still better than solutions nowadays achievable from
numerical limit analysis.
Finally, the function ( ) , , , , g g x y x y = made also possible to achieve
useful information about the choice of the best profile for the
construction of artificial embankments proving that the logarithmic spiral
profile has a better performance than the plane one.
Part 2
Numerical study
Summary
This part is devoted to the numerical study of the evolution of natural
slopes with the distinct element method.
After a brief description of the numerical method, calibration
procedures are illustrated. Calibration is the very crucial point of all the
work since the meaningfulness of the micromechanical parameters used
depends on it. The process of calibration used, allows to relate micro to
macro-mechanical parameters independently of the problem studied. Its
validity is general at all in order to reproduce a c- material type by the
distinct element method. For this reason, in the authors opinion this
represents the best result achieved by the numerical study performed.
Finally, simulations of the evolution of slopes subject to weathering are
shown. The solutions predicted by the distinct element method have been
compared with the solutions from limit analysis.
Chapter 4 Description of the distinct element method
109
1.
Chapter 4
4. Description of the distinct element method
4.1. Introduction
The distinct element method was introduced by [Cundall, 1971] for
the analysis of rock mechanics problems and then applied to soils by
[Cundall and Strack, 1979]. A thorough description of the method is
given in the two-part paper of [Cundall, 1988] and [Hart et al., 1988].
Moreover in [Introduction to mechanics of granular materials, Ch. 3,
1999] by Oda and Iwashita a clear description of some other DEM
approaches, different from the original method, developed by various
research groups in 80s and 90s, is supplied.
In the following, as the numerical code used to perform this
numerical study (PFC-2D) is based on the classical distinct element
method introduced by [Cundall and Strack, 1979], the main features of
the DEM referring to its classical version in the two dimensional case
will be described. A brief summary of the main differences of some other
distinct element methods with the classical DEM is supplied in 4.2.6.
Chapter 4 Description of the distinct element method
110
4.2. Description of PFC-2D
A distinct element code simulates the mechanical behaviour of a
system comprised of a collection of arbitrarily shaped bodies. The model
is composed of finite-size bodies or particles that displace independently
from one another and interact only at contacts or interfaces between the
particles. In the code used, PFC-2D, the particles are assumed to be rigid
and circular, and the behaviour of the contacts is characterised using the
so-called soft contact approach (see 4.2.2). The mechanical behaviour
of such a system is described in terms of the movement of each particle
and the interparticle forces acting at each contact point.
x
2
1
3
4
F
4
F
3
F
2
F
1
y
Fig. 4.1: representation of soil by the distinct element method (after [PFC-2D users manual,
1995]).
A discrete element code allows rotations and finite displacements of
discrete bodies, including complete detachment. The interaction of the
particles is treated as a dynamic process. Some codes (for instance
UDEC) consider deformable bodies of any shape. These codes have been
successively used in rock mechanics problems where joints and
discontinuities in the rock mass play the main role respect to the overall
behaviour. For instance [Allison and Kimber, 1998] have used UDEC to
simulate the possible failure mechanisms of a rocky cliff featured by two
sets of joints. These codes work efficiently only with a small number of
deformable bodies, therefore they are suitable for problems where a small
number of elements is enough to simulate the problem of interest. But in
Chapter 4 Description of the distinct element method
111
order to catch with an acceptable degree of accuracy the failure lines
relative to a succession of landslides developing within a slope a large
number of elements is needed. Therefore the code PFC-2D has been
chosen.
In the distinct element method, the contact forces and displacements
of a loaded assembly of particles are found by tracing the movements of
each particle. Movements may result from the propagation through the
particle system of disturbances caused by specified wall and particle
motion and/or body forces. The speed of propagation of the disturbance
depends on the physical properties of the discrete system.
The calculations performed in the distinct element code PFC-2D
alternate between the application of the Newton second law to the
particles and a force-displacement law at the contacts (see Fig. 4.2).
Newton second law is used to determine the motion of each particle
arising from the contact and body forces acting upon it, while the force-
displacement law is used to update the contact forces arising from the
relative motion at each contact.
Fig. 4.2: calculation cycle in PFC-2D (after [PFC-2D users manual, 1995]).
If the system is under static equilibrium conditions, a generic
disturbance (for instance the application of a body force) causes the onset
of a dynamic process leading to a new configuration of the system (a new
Chapter 4 Description of the distinct element method
112
fabric of the material under investigation) which ends when forces and
moments acting on each rigid body become zero again. This condition is
never reached rigorously, but it is reached numerically when the
inertial forces become negligible respect to the contact forces acting
among the bodies. Hence, when static equilibrium conditions or an
evolution process under quasi static conditions need to be reproduced, it
is necessary that the ratio between contact and unbalanced forces remains
small. When the ratio between the average contact to the unbalanced
force and the maximum contact to the maximum unbalanced force is
small (code default is 1%) the code considers the system of particles
under static equilibrium.
4.2.1. Law of motion
The motion of each rigid particle is described by the cardinal
equations of dynamics:
( ) ( ) ( ) t m t g = F x j (4.1)
2
1
( ) ( ) ( )
2
M t I t mR t = =
(4.2)
where x indicates the position of the gravity centre of the disk whereas
indicates its orientation respect to a Cartesian co-ordinate system (X,Y)
and j the vertical unit vector. F is the vector of the resultant of the forces
acting on the particle, due to the interaction with the surrounding
particles and M the resulting moment (see Fig. 4.1). R is the disk radius
while m and I are the mass and the moment of inertia of the particle,
respectively. g is the acceleration due to the body force present (in this
study the body force is given by the gravitational field and therefore it
acts only along the vertical direction). Since the mass is assumed to be
uniformly distributed, the centre of mass coincides with the centre of the
disk.
Note that since the code is two-dimensional, the out-of-plane force
component and the two-in-plane moment components are not considered
in any way in the equations of motion.
The equations of motion are integrated in time using an explicit
Chapter 4 Description of the distinct element method
113
centered finite-difference algorithm. This scheme corresponds to the
Newmarks procedure for the case of 0 = and 1 2 = .
The numerical scheme of integration starts from the calculation of
accelerations:
( )
1
( ) ( 2) ( 2) t t t t t
t
= +
x x x (4.3)
( )
1
( ) ( 2) ( 2) t t t t t
t
= +
(4.4)
From Eq. (4.3) and (4.4), knowing the force and moment acting on
each disk, the velocities are calculated at time / 2 t t + :
( )
( 2) ( 2)
t
t t t t g t
m
+ = + +
F
x x k (4.5)
( )
( 2) ( 2)
M t
t t t t t
I
+ = +
(4.6)
Finally, the velocities are used to update the position of the disk
centre:
( ) ( ) ( 2) t t t t t t + = + + x x x (4.7)
and disk orientation:
( ) ( ) ( 2) t t t t t t + = + +
(4.8)
The values of ( ) t t + F and ( ) M t t + to be used in the successive
timestep are obtained by application of the force-displacement law.
4.2.2. Force-displacement law
The force-displacement law is the relationship which models the
constitutive behaviour of the material. In fact, in the distinct element
method, the constitutive equations bind contact forces to relative
displacements between particles in a similar way as in continuum
mechanics constitutive equations bind stresses to strains.
The force-displacement law relates the relative displacement between
two entities (particle-particle or particle-wall) at a contact to the contact
Chapter 4 Description of the distinct element method
114
force acting on the entities. PFC logic is based on the soft contact
approach: that is overlap is allowed between particles. The force-
displacement law operates at a contact and can be described in terms of a
contact point x, lying on a contact plane that is defined by a unit normal
vector n (see Fig. 4.3). The contact point is within the overlapping
volume of the two entities (disks). The normal vector is directed along
the line defining the shortest distance between the disk centres or
between a wall and the disk centre.
a) b)
Fig. 4.3: contact point x and overlapping length U
n
in case of a) disk-disk contact b) wall-
disk contact. The contact plane passes through the contact point x and is normal to
the disk-disk or the wall-disk distance (after [PFC-2D users manual, 1995]).
This way of modelling contacts, known as soft contact approach, is
because the method was born for rock mechanics problems as stated in
4.1. The soft contact approach has been conceived to model rock joints
characterised by a finite thickness and a finite compliance. However, this
approach resulted good at modelling granular materials as the
geometrical irregularities of grains (grain shape is never circular but
elliptical or still more irregular) are taken into account by finite
overlappings at contact points.
The elastic compliance of contacts is modelled by a system of elastic
springs whose stiffnesses may be constant (linear elasticity) or function
of the contact forces (Herth-Mindlin elastic model). The Herth-Mindlin
model may be suitable for a detailed characterisation of nonlinear
Chapter 4 Description of the distinct element method
115
reversible behaviours but it needs more constitutive parameters than the
linear model. Therefore the linear elastic model has been adopted since it
is, by far, the simpler one.
According to this model, two uncoupled elastic springs relates the
two components of the contact force (the tangential and normal one) to
the corresponding components of the relative displacement between the
two contacting particles (see Fig. 4.4). In the calculation cycle, contact
forces are achieved from relative displacements:
( ) ( )
N N N
F t t U t t K + = + (4.9)
S S S
F U k = (4.10)
and ( ) ( )
S S S
F t t F t F + = + (4.11)
where F
N
and F
S
are the normal and tangential components,
respectively, of contact forces and U
N
and U
S
are the normal and
tangential components of relative displacements. Note that the normal
contact force is calculated on the basis of the current normal overlapping
length whereas the tangential component is calculated in an incremental
way
I
. Therefore K
N
represents an elastic secant modulus whereas k
S
an
elastic tangent modulus.
When a new contact is formed by two disks which get in contact, it is
unloaded: F
N
and F
S
are initialised to zero. During each subsequent
timestep F
N
is calculated directly from Eq. (4.9) whereas F
S
from Eq.
(4.10) and (4.11).
Tension is not allowed. This condition is enforced by checking the
overlapping expressed by U
N
. If 0
N
U (negative or zero overlapping),
the contact forces are set to zero 0
N S
F F = = .
Irreversible displacements are modelled by a frictional slider along
the tangential direction (see Fig. 4.4). Frictional sliding takes place when
the shear contact strength is reached. Each contact is checked for
I
Other distinct element codes (e.g. UDEC) compute also the normal force on the basis
of the incremental relative displacement. The computation of F
N
from the geometry
alone makes the code more stable and able to handle changes in disk radii after a
simulation has begun. This procedure has been used in all the simulations run (see
5.2.3 )
Chapter 4 Description of the distinct element method
116
slippage conditions by calculating the maximum allowable shear contact
force:
max ( ) tan ( )
S N
F t F t
= (4.12)
If max
S S
F F > then slip occurs and the new F
S
is given by:
( ) ( ) ( ) max ( )
S S S
F t sign F t F t = . (4.13)
a)
b)
K
N
k
S
jf
Fig. 4.4: PFC contact representation: a) normal interaction b) tangential interaction (after
[PFC-2D users manual, 1995]).
The force-displacement law is enforced by the code into two phases:
first an elastic trial is performed, second a plastic correction is added
(according to a nomenclature typical of finite element codes). First
contact forces are calculated according to the linear elastic law described
above (Eq. (4.9), (4.10) and (4.11)), second the failure criterion is
checked, third the plastic correction, if needed, is imposed assigning a
new value to F
S
(see Fig. 4.5).
In conclusion, the constitutive force-displacement law described is
featured by linear elasticity, purely frictional strength and no tension
condition. This is the constitutive law assigned to contacts by default in
PFC (called contact slip model). According to this law, the soil
micromechanical behaviour may be classified as elastic-perfectly
plastic. This consideration is clearly justified by considering the
constitutive law ruling the shear component of the contact forces (see
Fig. 4.6).
Chapter 4 Description of the distinct element method
117
Mohr-Coulomb
failure criterion
I
II
I
F
N
F
S
Fig. 4.5: force-displacement law according to the contact slip model. This law is enforced by
PFC into two phases. Phase I: linear elastic law (elastic trial); phase II (eventual):
frictional sliding (plastic correction).
max
S
F
U
S
F
S
1
k
S
Fig. 4.6: force-displacement law according to the contact slip model relative to the shear
component of contact force.
If the two components of the contact forces, F
N
and F
S
, are confused
with two stresses:
n
and respectively (for instance
n
and can be seen
as the stresses acting along the contact plane between two contacting
particles (see Fig. 4.3a)) the flow rule governing the constitutive law is
Chapter 4 Description of the distinct element method
118
non-associate and it corresponds to the case of zero dilation:
= 0, with
, ( ) t x , ( ) t , ( ) t F , ( ) M t , i.e. the
state of the system at the time t is wholly known, ( ) 2 t t + x and
( ) 2 t t +
( ), ( ) t t x
Compute ( ) t t + x , ( ) t t + from Eq. 4.7 and 4.8
( ), ( ) t M t F
Compute
S
U
Compute ( )
N
F t t + from Eq. 4.9
Compute ( )
N
U t t +
( ) 0
N
U t t + < ?
no
yes
0, 0
N S
F F = =
Compute ( )
S
F t t + from Eq. 4.10, 4.11
( ) ( ) max
S S
F t t F t t + + ?
Recompute ( )
S
F t t + from Eq. 4.13
Compute
( ), ( ) t t M t t + + F
READY FOR NEXT STEP
Fig. 4.7: algorithm relative to the calculation cycle run by PFC every timestep: black boxes
refer to calculations performed at disk level while grey boxes at contact level.
Chapter 4 Description of the distinct element method
120
4.2.4. Stability of the numerical scheme
The explicit centered finite-difference scheme of integration adopted
by PFC-2D is conditionally stable. The scheme is stable only if the
timestep chosen at the beginning of each calculation cycle is less than the
critical timesptep t
c
which is related to the minimum eigenperiod of the
system. In order to determine t
c
a global eigenvalue analysis is
prohibitive because of the large and constantly changing system of
distinct elements. Therefore a simplified procedure is used by PFC-2D to
estimate t
c
at the start of each cycle.
Considering the one dimensional mass-spring system described by a
point mass m and a spring stiffness k, the critical timestep is given by:
2
c
t m k = .
If an infinite series of point masses and springs are considered, the
smallest period of this system occurs when the masses move in
synchronised opposing motion. The critical timestep for this system is
given by:
trans
c
t m k = (4.14)
If rotational motion is considered, m must be replaced by I, the
moment of inertia of particles, k
trans
by k
rot
, the particle rotational
stiffness, and the critical timestep becomes:
rot
c
t I k = (4.15)
Thus, the critical timestep for a generalised multiple mass spring
system is given by the minimum between Eq. (4.14) and (4.15).
A critical timestep is found for each particle by applying Eq. (4.14)
and (4.15) separately to each degree of freedom and assuming that the
degrees of freedom are uncoupled. The critical timestep t
c
is taken to be
the minimum of all critical timesteps computed for all degrees of
freedom of all particles.
4.2.5. Damping
Note that the system of particles described up to here (rigid elements
connected by elastic springs), dissipate energy only by means of
Chapter 4 Description of the distinct element method
121
frictional slidings. In order to reduce the computational time needed by
the system of particles to reach its equilibrium state after a perturbation is
introduced or initial conditions are assigned, the code offers the
possibility of using a numerical damping. The default numerical damping
is non-viscous, but proportional to the particle accelerations. Introducing
the numerical damping, the equations of motion (4.1) and (4.2) become:
( ) ( ) 1 sgn ( ) ( ) x t m t g = F x j (4.16)
and
( )
2
1
1 sgn ( ) ( )
2
M t mR t =
(4.17)
respectively, where is a dimensionless constant. Particle accelerations
are decreased by damping whereas decelerations are increased. Note that
this type of damping does not act at contact level: that is no dashpots are
present in the contact laws, but only springs, sliders and no tension
constraints.
Indeed, this numerical damping affects the solution since the
equations of motion are altered. But in all the simulations run a very
small value has been used: 5%, whereas the default value suggested by
manual is 70%. Moreover the material is studied in quasi-static
conditions, therefore the influence of damping on the solution is
negligible. Instead, if dynamic conditions are studied damping plays a
key role [Tatarella, 1997].
4.2.6. Other discrete element methods
Successively to the first DEM codes (BALL, TRUBALL) developed
by Cundall et al., other codes which treat the soil as an assembly of rigid
bodies have been implemented: for instance the LMGC software
implemented by Jean within the framework of Contact Dynamic Method
developed by Moreau and Jean [Jean and Moreau, 1992], [Moreau, 1993,
1994]. The CD method is based on an implicit numerical integration of
the equations of motion (conversely PFC-2D is based on explicit
numerical integration). Since contacting bodies are considered perfectly
rigid, no stiffness is introduced in the contact law. A shock law, a
Chapter 4 Description of the distinct element method
122
unilateral condition and dry Coulomb friction are used. According to
[Lanier and Jean, 2000] the results obtained by LMGC code, performing
biaxial compression tests, agree very well with those achieved by PFC-
2D. This is an interesting result as the two codes treat the numerical
integration in time and the constitutive laws at contacts in different ways.
Another method GEM (granular element method) was proposed by
[Kishino, 1988]. GEM differs from classical DEM only for the
calculation algorithm which is based on the so-called stiffness method,
whereas contact laws are imposed in the same way. But this method up to
now has not been competitive compared to classical DEM because of the
high computational time requested by inversion of the stiffness matrices.
The MDEM (modified distinct element method) by [Iwashita and
Oda, 1998] is a variation respect to classical DEM since a moment-
rotation relationship is added to the force-displacement laws at contact
level. Therefore 3 types of interactions between particles are modelled:
normal and tangential contact forces and moment. Note that this is
possible even with PFC-2D using the so-called parallel bond model
II
.
This approach has not been chosen since more parameters need to be
introduced respect to the contact bond model
II
. This complication makes
much more difficult the calibration process with no significant
advantages.
4.2.7. PFC-2D parameters
This paragraph is intended to supply a summary of the parameters
used in the numerical simulations.
The unit weight does not influence the results achieved since the
simulations performed to calibrate the micromechanical parameters
(biaxial tests) have been run without gravity. Conversely, gravity is
present in the simulations relative to slope weathering, but the results
obtained have been compared to those obtained by limit analysis in terms
of normalised cohesion: c H . Therefore a value for the unit weight has
been chosen (
3
20 kN m = ), but even if another value were selected the
results would have been the same in terms of c H . Indeed, assigned a
II
About these models see 5.4.
Chapter 4 Description of the distinct element method
123
certain slope geometry, if a different unit weight is adopted both the
failure lines and the retrogression lengths of the slope change.
As regards the geometry of particles, three parameters are involved:
two relative to the particle sizes (R
max
and R
min
) and one to the porosity.
A uniform distribution of particle radii between R
max
and R
min
has been
assumed with R
max
= 3R
min
. The value of R
max
and R
min
used by [Calvetti,
1997] have been chosen. It can be noted that the range of radii is small if
compared to a real granular material. But according to Calvetti the ratio
adopted is large enough to reproduce the behaviour of granular materials.
Moreover lower ratios have been used by many researchers to perform
biaxial tests with PFC-2D: [Dubujet et al., 2003], [Hunt et al., 2003],
[Huang et al., 1999]. Other researchers have not assumed a uniform
distribution of radii, but only some discrete values (usually 3 or 5):
[Lanier and Jean, 2000], [Iwashita and Oda, 2000], [Liu et al., 2003],
[Kamp et al., 1999], [Cambou et al., 2000]. Nevertheless, a uniform
distribution of radii has been preferred because it is closer to the reality
as the variation of particle sizes is continuous in nature.
The porosity assumed ( 0.16 n = ) is a result of the calibration process
which will be illustrated in Ch. 5. This value may appear very small if
compared to real geomaterials. This is due to the two-dimensional feature
of the analyses performed. Disks are much less porous than three
dimensional grains.
As regards the micromechanical parameters, the normal contact
stiffness has been assumed equal to the tangential one and their values
have been determined by the calibration process; the intergranular
friction angle
III
and the tension bond force
t
III
have been studied varying their values within a range of physical
interest. Note that the last parameter is not free to vary since
tan t c
= (see Fig. 5.25 and 5.26).
III
For the definition of c
and t
see 5.4.
Chapter 4 Description of the distinct element method
124
PFC-2D parameters Values
s
= density of particles Assigned
s
= 2.38 t/m
3
n = porosity Determined
R
max
= maximum disk radius Assigned R
max
= 4.5 mm
R
min
= minimum disk radius Assigned R
min
= 1.5 mm
= numerical damping Assigned = 0.05
K
N
= normal contact stiffness Determined
k
S
= shear contact stiffness Assigned k
S
= K
N
0
k
0.17 0.16 0.14
Fig. 5.4: qualitative relationship between porosity and initial stress for various contact
stiffnesses. n=0.16 is the porosity value adopted for all the calibration simulations.
Finally, note that the limit value of porosity found 0.17 n = depends
on the chosen particle size distribution. In fact, if a smaller range of radii
were adopted, a higher limit value of porosity would have been obtained
and if the range reduces to a unique value (all particles with the same
radius) the highest limit value of porosity is achieved.
Chapter 5 Numerical simulations
134
5.2.4. Biaxial test execution
Biaxial tests have been performed onto the numerical specimens
described in 5.2. After specimens have been generated, they have been
brought to a prescribed confining pressure p. This has been done by
moving the boundary walls in stress control. In order to avoid
meaningless results, loading (in case of target pressure p higher than
0
)
or unloading (in case of target pressure p lower than
0
) has to be
imposed in quasi-static conditions as in laboratory tests on real soil
samples. In order to fulfil this requirement, only one wall per each
direction has been moved. In fact, doing so, the difference of forces
exerted by particles on opposing walls at any instant t during the test, is
entirely due to inertial forces induced by the movement of the walls. If
this difference is small in comparison with the forces exerted on walls,
the inertial forces can be considered negligible.
A numerical servo-control, created by the writer, prescribes wall
velocities depending on the difference between the required (final) state
of stress and the actual one. This difference has been multiplied by a
reduction factor which has been calibrated in order to obtain the
prescribed isotropic stress state maintaining the specimen in quasi-static
conditions. In Fig. 5.5a the stresses exerted on walls during isotropic
compression are shown. It can be seen that for the chosen value of the
reduction factor the differences between upper and lower and right and
left walls respectively are negligible. Moreover, the numerical servo-
control adopted makes it possible to reach the target confining pressure
smoothly without introducing oscillations of stresses around the target
value (see Fig. 5.5a).
After the completion of the isotropic compression phase, specimens
have been loaded in biaxial compression. In this case, a prescribed
velocity is applied to the upper boundary wall which moves downwards.
A numerical servo-control prescribes the motion of the right boundary
wall in such a way that the lateral stress remains constant during all the
test. The velocity of the upper wall has been calibrated in order to
maintain quasi-static conditions during the test (see Fig. 5.5b).
Chapter 5 Numerical simulations
135
0
50
100
150
200
250
0.0E+00 5.0E-05 1.0E-04 1.5E-04 2.0E-04 2.5E-04 3.0E-04
eps_y
s
t
r
e
s
s
e
s
o
n
w
a
l
l
s
[
k
P
a
]
left wall (fixed)
upper wall (servo-controlled velocity)
lower wall (fixed)
right wall (servo-controlled velocity)
a)
0
100
200
300
400
500
600
700
800
0.00 0.02 0.04 0.06 0.08 0.10 0.12
eps_y
s
t
r
e
s
s
e
s
o
n
w
a
l
l
s
[
k
P
a
]
left wall (fixed)
upper wall (constant velocity)
lower wall (fixed)
right wall (servo-controlled velocity)
b)
Fig. 5.5: stresses acting on the boundary walls a) during isotropic compression up to 200
kPa with the reduction factor =0.01 b) during the subsequent biaxial
compression at a loading rate v=0.01 m/s.
The biaxial tests have been run up to failure could be clearly
detectable. In all the tests, the peak stress is reached for low
d
values
Chapter 5 Numerical simulations
136
(see Fig. 5.6). Depending on the test conditions (confining pressure,
intergranular friction angle and contact stiffness) the value of
dev
at
failure varies from 4% to 10%. The features of the stress-strain
relationships achieved by the tests run will be illustrated, in a more
detailed and exhaustive way, in the next paragraphs ( 5.3 and 5.4).
The recorded fields of displacement can be used to verify the
reasonableness of the results achieved. In Fig. 5.7a a typical field of
displacement obtained at the end of the biaxial test is shown. This is in
good agreement with the results present in literature: see [Calvetti et al.,
1997] and [Lanier and Jean, 2000]. The latter ones performed biaxial
tests onto an assembly of rigid non-overlapping disks according to the
contact dynamics method using the LMGC software. In Fig. 5.7b is
shown the field of displacement obtained from their test (the only
difference with the present tests is due to the choice of the lateral wall
moving: left instead of right). Although this method is very different
(contact overlapping is not allowed, see 4.2.6) the field of displacement
is almost the same.
jf=0.2, k=5e4 [kN/m]
0
100
200
300
400
500
600
0.00 0.02 0.04 0.06 0.08 0.10 0.12
eps_dev
t
[
k
P
a
]
p=500 p=200 p=100 p=50 p=20
Fig. 5.6: stress-strain relationship achieved for various confining pressure values (k=5e4 and
tan
=0.2 (fixed
rotations).
5.3. Calibration of micromechanical parameters to
reproduce a frictional cohesionless material
In this paragraph, the calibration of micromechanical PFC-2D
parameters done in order to reproduce a frictional cohesionless
homogeneous material will be illustrated. The strength of such a material
is characterised only by one parameter: the internal friction angle. The
simulations run have been made possible to calibrate the
micromechanical parameters ruling particle interactions so that simple
relationships between one micromechanical strength parameter and one
macromechanical strength parameter,
and respectively, have been
obtained.
5.3.1. Influence of particle rotation on
The first aspect which needs to be put in evidence, is represented by
particle rotation. In fact, particle rotation plays a key role into
Chapter 5 Numerical simulations
140
determining the overall strength of the discrete assembly of particles. The
interested reader can make reference to [Skinner, 1969], [Konishi et al.,
1982, 1983], [Cundall et al., 1982], [Calvetti, 1997] for experimental and
numerical studies on this subject. Here the reasons of the choice operated
with reference to particle rotation will be explained.
PFC offers the possibility of inhibiting particle rotation. If rotations
are inhibited the global mechanical behaviour achieved is completely
different from the case of free rotations since an internal kinematic
constraint is added to the material. Assigned a value to the intergranular
friction
is assumed.
Depending on the grain size distribution adopted (different in the two
works), the maximum internal friction angle that can be achieved is no
more than 30. Moreover, the relationship achieved between
and is
non-linear.
In table 5.2 the internal friction angles achieved by [Panzeri and
Rinaldi, 1999] are reported. These values have been inferred by the thrust
exerted by an assembly of particles free to rotate, subject to their own
self-weight, on a vertical confining wall in conditions of active pressure
(see Fig. 5.9). The achieved internal friction angle values represent only
an average value, since the confining pressure varies along the vertical
wall and as a consequence the value of . The dependence of on the
confining pressure will be treated in 5.3.4. Anyway, the mean values
achieved are significant in order to show the impossibility of reproducing
middle-high values of internal friction angle adopting the kinematic
condition of free particle rotation.
tan jf
5.71 14.04 26.58 36.89 45.02
12.24 15.26 21.14 23.81 25.25
Table 5.2: values after [Panzeri and Rinaldi, 1999].
Chapter 5 Numerical simulations
141
Fig. 5.9: displacement field at failure (after [Panzeri and Rinaldi, 1999]).
Therefore in order to try to reproduce an internal friction angle in a
range large enough to describe a large class of c- material (from clays
featured by low friction angles to weakly cemented sands) it is necessary
to inhibit particle rotation.
In Fig. 5.10 the relationship obtained in case of inhibited rotations
between the intergranular friction angle
[
d
e
g
r
e
e
s
]
Fig. 5.10: relationship between
[
d
e
g
r
e
e
s
]
Fig. 5.13: contact stiffness vs. internal friction angle in semi-logarithmic scale (p=500 kPa
and tan
=0.2).
Now, it is useful to highlight that scaling the confining pressure is
equivalent to scaling the contact stiffness. In Fig. 5.14 and Fig. 5.15, the
relationships obtained by running two tests, one with confining pressure
and contact stiffness one order of magnitude lower than the other, are
shown. In Fig. 5.14 and in Fig. 5.15 the stress-strain and the
vol
-
dev
relationship respectively, are shown: in practise, they do not differ.
Therefore in order to choose a suitable value of contact stiffness, it is not
necessary to consider the whole range of pressures to be investigated, but
only tests run at a given pressure.
Chapter 5 Numerical simulations
146
jf=0.2
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18
eps_dev
t
/
s
k=5e5, p=500 k=5e4, p=50
Fig. 5.14: deviatoric stress normalised to the mean pressure vs. deviatoric strain (tan
=0.2).
jf=0.2
-0.04
-0.04
-0.03
-0.03
-0.02
-0.02
-0.01
-0.01
0.00
0.01
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18
eps_dev
e
p
s
_
v
o
l
k=5e5, p=500 k=5e4, p=50
a)
jf=0.2
-1.E-03
-1.E-03
-8.E-04
-6.E-04
-4.E-04
-2.E-04
0.E+00
2.E-04
4.E-04
0.E+00 1.E-03 2.E-03 3.E-03 4.E-03 5.E-03 6.E-03 7.E-03 8.E-03
eps_dev
e
p
s
_
v
o
l
k=5e5, p=500 k=5e4, p=50
b)
Fig. 5.15: volumetric strain vs. deviatoric strain (tan
[
d
e
g
r
e
e
s
]
k=5e5 [kN/m]
k=5e4 [kN/m]
Fig. 5.16: angle of internal friction vs. confining pressure for two values of contact stiffness
(tan
=0.2). Note that at very low confining pressure (p=20 kPa) there is a little
discrepancy respect to the general trend achieved at higher pressures.
Chapter 5 Numerical simulations
149
a) b)
Fig. 5.17: a) circular shaped numerical specimen after isotropic compression b) internal
friction angle vs. confining pressure referring to a middle-high range of confining
pressures (after [Sitharam, 1999]).
5.3.5. Choice of a suitable value of contact stiffness
Up to now, only the global strength behaviour has been considered
and no considerations have been made about compliance since the goal
of calibration simulations consists in reproducing a c- material type
regardless of its compliance. But if a physical sounder choice of contact
stiffness wants to be done, it is not possible to neglect such an aspect.
Therefore, the secant moduli of elasticity E
25
and E
50
have been
calculated (see table 5.5). Comparing the values obtained with Young
moduli relative to sand, k=5e4 kN/m appears the value which better
reproduces the experimental data. This value has been therefore adopted
in the successive calibration phase.
Chapter 5 Numerical simulations
150
p [kPa] E
25
[MPa] E
50
[MPa] k=K
N
=k
S
[kN/m]
500 299 215 5e4
500 3762 1976 5e5
200 303 197 5e4
200 2547 1508 5e5
100 323 188 5e4
100 1629 1068 5e5
50 339 170 5e4
50 1177 624 5e5
20 173 99 5e4
20 687 281 5e5
Table 5.5: secant Young moduli obtained at 25% and 50% of the material strength for
tan
=0.2.
In Fig. 5.18 the
vol
-
d
relationship achieved at different confining
pressures are shown. A physically sound trend has been achieved. The
compacting phase is larger for higher confining pressures (see Fig.
5.18a). As regards dilatancy, Fig. 5.18b shows the same final value
(slope of the curves in the right side of the graph) independently of the
confining pressure.
-5.0E-03
-4.0E-03
-3.0E-03
-2.0E-03
-1.0E-03
0.0E+00
1.0E-03
2.0E-03
0.0E+00 5.0E-03 1.0E-02 1.5E-02 2.0E-02 2.5E-02 3.0E-02
eps_dev
e
p
s
_
v
o
l
p=500 p=200 p=100 p=50 p=20
a)
Chapter 5 Numerical simulations
151
-4.0E-02
-3.5E-02
-3.0E-02
-2.5E-02
-2.0E-02
-1.5E-02
-1.0E-02
-5.0E-03
0.0E+00
0.0E+00 2.0E-02 4.0E-02 6.0E-02 8.0E-02 1.0E-01 1.2E-01 1.4E-01 1.6E-01
eps_dev
e
p
s
_
v
o
l
p=500 p=200 p=100 p=50 p=20
b)
Fig. 5.18:
vol
-
d
relationship at various confining pressures (k=5e4 kN/m and tan
= 0.2):
a) initial part of the test b) full test.
5.3.6. Determination of
- relationship
Up to here all PFC-2D parameters have been studied and suitable
values of porosity and contact stiffness have been selected. It can be
stated that the behaviour of a real granular material in terms of strength
and deformability may be satisfactorily reproduced.
What is still missing is the calibration of
. In order to determine
-
relationships, failure envelopes have been obtained by linearly
interpolating failure values in the s-t plane for different values of
(see
Fig. 5.19). The slope of the lines in the figure gives the internal friction
angle of the material.
In Fig. 5.20 the values of internal friction angle against intergranular
friction are shown. The relationship achieved by interpolating the values
of internal friction angle obtained in correspondence of different
is
linear in a wide range of such that all the values of engineering interest
(from 15 to 42.5) are covered. Therefore it is possible to conclude that
a simple relationship between micro and macromechanical strength
parameters has been obtained and a frictional cohesionless material has
Chapter 5 Numerical simulations
152
been satisfactorily reproduced.
Fig. 5.19: failure envelopes (for various tan
[
d
e
g
r
e
e
s
]
Fig. 5.20: vs.
(k=5e4 [kN/m]).
Chapter 5 Numerical simulations
153
Finally, the case of
and t
respectively (see Fig. 5.21). If these bonds are assigned to contacts, the
force-displacement laws along the tangential and normal directions are
elastic as far as contact forces remain lower than the bond values.
F
S
F
S
S
F c
<
N
F t
>
F
N
F
N
a) b)
Fig. 5.21: contact bonds according to the contact bond model implemented in PFC-2D. a)
Shear contact force. It remains in the elastic regime up to it equals the shear bond
strength c
I
, the contact is declared broken and the contact forces
are set to zero 0
N S
F F = = . If
S
F c
> instead of 0
N
U > (no-tension condition) and
S
F c
Strength of intact
contacts
B
1
A
1
B
2
B
3
A
2
A
3
Fig. 5.22: PFC contact bond model. A
1
and B
1
represent the forces relative to a compressive
and non-compressive contact respectively just before the imposition of the force-
displacement law. A
2
and B
2
represent the contact forces after the elastic trial. A
3
and B
3
represent the final contact forces after contact breakage.
This model is not suitable to simulate a c- material. It is very far
from the constitutive macromechanical behaviour which is intended to
model. In fact a c- material is characterised by two distinct strength
contributions: a frictional part and a cohesive one. Conversely in the PFC
contact bond model it is not possible to identify two distinct
contributions. Moreover, the bonded contacts are characterised by a shear
resistance independent of the amount of normal force whereas the broken
contacts are characterised only by frictional strength. Therefore bonded
contacts show a radically different constitutive law respect to non-bonded
I
Note that according to the convention adopted, t
becomes zero, the new contact model reduces to the slip contact
model whereas in the PFC contact bond model, if c
time increasing
partially weathered
material
fresh material
uncemented soil
Fig. 5.23: new model implemented in PFC: the contact strength is given by the summation
of two distinct contributions:
and c
Strength of intact
contacts
D
D
F
F
Fig. 5.24: implementation of the proposed model in PFC: when the contact strength is
exceeded two possibilities are given; fragile behaviour (F, grey): the contact
breaks and the failure surface reduces to the failure surface of non-bonded
contacts; ductile behaviour (D, black): the contact remains intact and the
strength unaltered.
As regards the calculation cycle, this model has been implemented
into the following way: after the calculation of contact forces from
relative displacements (see Eq. (4.9), (4.10) and (4.11)), the values of F
N
and F
S
are tested. The first check is on F
N
. In case of
N
F t
, if the
contact is fragile, it breaks (contact declared broken and contact forces
set to zero 0
N S
F F = = ); otherwise if ductile, it remains intact (
N
F t
=
and 0
S
F = ). The second check is on F
S
. In case of
S
F c
, if the contact
is fragile, it breaks (contact declared broken, F
N
remains unaltered
whereas tan
S N
F F
+ x , ( )
2
t
t
( ), ( ) t t x
Compute ( ) t t + x , ( ) t t + from Eq. 4.7 and 4.8
( ), ( ) t M t F
Compute
S
U
Compute
( )
N
F t t +
from Eq. 4.9
Compute ( )
N
U t t +
yes
0, 0
N S
F F = =
Compute ( )
S
F t t + from Eq. 4.10, 4.11
( ) ( ) max
S S
F t t F t t + + ?
Recompute ( )
S
F t t + from
tan
S N
F F c
= +
Compute
( ), ( ) t t M t t + + F
READY FOR NEXT STEP
N
F t
?
no
Fig. 5.25: algorithm relative to the calculation cycle run by PFC modified by the routine
implementing the DF contact model: the dashed rectangle encloses the new
instructions introduced by the routine. Black boxes refer to calculations
performed at disk level while grey boxes at contact level.
Chapter 5 Numerical simulations
159
Bonds have been assigned at the end of isotropic compression to
contacts bearing compression. In Fig. 5.26 the network of bonds is
shown. It can be seen that bonds are approximately uniformly distributed
within the specimen, although the network web is irregular. This is due to
the fact that the web of contacts in compression is irregular too. This
network appears suitable to reproduce real bonds in the soil, for instance
those relative to cemented sands. Moreover, an irregular network is a
necessary condition in order to reproduce an homogeneous cohesive
material in the same way as an irregular assembly of particles is
necessary to reproduce an homogeneous frictional material.
Fig. 5.26: network of bonds (p=500 kPa).
In Fig. 5.27 the stress-strain relationship achieved from all the three
models is shown. In order to make it possible a comparison among the
models the same contact bonds have been assigned in all the tests:
0.1 c
= kN, where c
together with
= 0.2).
In the light of the results achieved, the DF model has been adopted to
the end of reproducing a c- material.
Chapter 5 Numerical simulations
161
5.4.2. Calibration of the new micromechanical parameters
In this paragraph, the calibration of c
=0.1 kN; c
=0.05 kN; c
=0.025
kN) are shown. From the figure the ductile nature of the global behaviour
obtained by the DF model can be inferred. In order to evaluate the
influence of c
=0.2).
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.00 0.02 0.04 0.06 0.08 0.10 0.12
eps_dev
t
/s
c=0.1 kN
c=0 kN
c=0.025 kN
c=0.05 kN
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.00 0.02 0.04 0.06 0.08 0.10 0.12
eps_dev
t
/s
c=0.1 kN
c=0 kN
c=0.025 kN
c=0.05 kN
a) b)
Fig. 5.29: deviatoric stress normalised to mean pressure vs deviatoric strain a) confining
pressure p=500 kPa b) confining pressure p=50 kPa (k=5e4 kN/m, tan
=0.2).
In Fig. 5.30 the values of s and t at failure relative to tests run at
different confining pressures (see table 5.3) with various contact shear
strengths (c
=0.1 kN; c
=0.05 kN; c
parameter alone.
0
100
200
300
400
500
600
700
800
900
1000
1100
0 200 400 600 800 1000 1200 1400 1600 1800 2000
s [kPa]
t
[
k
P
a
]
c=0.1 kN
c=0 kN
c=0.025 kN
c=0.05 kN
Fig. 5.30: failure envelopes achieved by linear interpolation of failure values in the s-t plane
for various c
=0.2)
Note that the range of values in the s-t plane investigated lies in a
region at the right of the s = t line. This is a limit inherent to the type of
test run: biaxial compressive tests. In fact, in order to perform a test
whose stress path may lye on the left of the s = t line tension should be
exerted by boundary walls along one direction.
In Fig. 5.31 the values of s and t at failure relative to the same tests
(confining pressures in table 5.3 and c
=0.1 kN; c
=0.05 kN; c
=0.025
kN) but with a much lower intergranular friction angle, tan 0.04
=
instead of tan 0.2
=0.01 kN; c
=0.005 kN; c
=0.04)
The c - c
relationships.
In Fig. 5.32 is shown that cohesion, when lower than a limit value,
can be linearly related to the c
. In
fact, the slope of the c - c
relationships against the internal friction angle (see Fig. 5.33). The dots
corresponding to the different cases of
relationship can be
expressed by:
( )
1
c K c
= (5.1)
with ( )
2
1
tan
K
K
= and therefore:
2
tan
K
c c
= (5.2).
with K
2
a material constant parameter. From Eq. (5.2) the presence of a
Chapter 5 Numerical simulations
165
vertical asymptote is inferred. This asymptote should correspond to an
intergranular friction angle
becomes
infinite and therefore the specimen assumes infinite strength. But since
the case
=0 and 0 c
).
0.E+00
1.E+03
2.E+03
3.E+03
4.E+03
5.E+03
6.E+03
7.E+03
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
tan
s
l
o
p
e
o
f
c
-
c
c
u
r
v
e
Fig. 5.33: slope of the c-c
, c
and c
, c
=2.3 [deg.]
c
=3.5e-3 [kN]
=5.7 [deg.]
c
=6.6e-3 [kN]
=11 [deg.]
c
=9.7e-3 [kN]
=17 [deg.]
c
=12e-3 [kN]
50
=2.3 [deg.]
c
=8.7e-3 [kN]
=5.7 [deg.]
c
=17e-3 [kN]
=11 [deg.]
c
=24e-3 [kN]
=17 [deg.]
c
=30e-3 [kN]
100
=2.3 [deg.]
c
=17e-3 [kN]
=5.7 [deg.]
c
=33e-3 [kN]
=11 [deg.]
c
=49e-3 [kN]
=17 [deg.]
c
=60e-3 [kN]
200
=2.3 [deg.]
c
=35e-3 [kN]
=5.7 [deg.]
c
=66e-3 [kN]
=11 [deg.]
c
=97e-3 [kN]
=17 [deg.]
c
=0.12 [kN]
Table 5.7: determination of
, c
and c-c
and c-c
= 0.214 has
been assigned to particles and their rotations have been inhibited.
n r
min
[m] r
max
/r
min
disks k [kN/m] [deg.]
b. tests 0.16 1.5e-3 3 1216 1 5e4 (15-42.5)
Slope 1 0.16 0.1 3 74405 66.7 5e4 34
Slope 2 0.16 0.3 3 21385 200 5e4 34
Table 5.8 : values adopted for the numerical simulations run.
5.5.3. Slope generation procedure
Particles have been generated by the same procedure used in biaxial
tests (particle generation by radius expansion) with the only difference
that six instead of four rigid frictionless boundary walls have been used
to contain the assembly of particles. After the completion of the
generation procedure, gravity has been applied to particles. After inertial
Chapter 5 Numerical simulations
169
effects, due to gravity force application, disappeared, contact bonds, (DF
model) have been assigned to contact in compression. The network of
contact bonds obtained (see Fig. 5.35) is characterised by the same
irregular fabric as the networks relative to biaxial tests. An initial very
high value of c
*
H
)
slope 1
slope 2
Fig. 5.43: dimensionless crest retreat vs. normalised cohesion ( = 34). Circles represent
values achieved by limit analysis; triangles are values relative to slope 1; squares
are relative to slope 2. Lines indicate crest retreat and cohesion evolution for
increasing time.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0 0.2 0.4 0.6 0.8 1 1.2 1.4
L/H
c
/
(
*
H
)
Fig. 5.44 normalised cohesion vs. dimensionless crest retreat ( = 34). Circles represent
values achieved by limit analysis; triangles are values relative to slope 1; squares
are relative to slope 2. The line refers to a polynomial interpolating function.
Chapter 5 Numerical simulations
181
5.5.7. Case of no erosion conditions
In order to study more realistically the evolution of slopes, particles
have not been deleted. In this way it is possible to investigate the
evolution of slopes without neglecting the presence of the landslided
material on the base of the slope.
The differences with the previously studied case of strong erosion
conditions can be summarised as follows: once failure conditions have
been reached because of cohesion decrease, the wedge of soil involved in
the failure mechanism slips away and a soil flow process takes place. In
order to detect the end of such a process the kinetic energy of the system
of particles has been monitored. Successively to the end of the soil flow
process, cohesion has been further reduced up to the detection of the
subsequent landslide. In Fig. 5.45 the slope profile after the first landslide
occurrence can be seen. The presence of three distinct wedges of soil on
an inclined plane is due to the fact that wedges are formed by particles
still bonded and the inclined plane is formed by unbonded particles as it
is shown in Fig. 5.45.
Fig. 5.45: slope 1. Profile and network of bonds at the end of the soil flow after the
occurrence of the first landslide.
In Fig. 5.46 the field of velocities during the occurrence of the
Chapter 5 Numerical simulations
182
second failure is shown. Unlike the case of strong erosion (see Fig.
5.40a) only a part of the slope front is involved into the failure
mechanism. This is due to the stabilising effect exerted by the material
slipped away after the first failure and accumulated on the slope base.
This effect is also responsible of limit values of cohesion lower than
those recorded in the case of strong erosion (second failure occurs at
c
=0.016 kN instead of c
=0.022 kN). In Fig. 5.47 the slope profile
achieved after debris ended to flow is shown. The profile is plane
excepted for a small part in the upper region of the slope front. The
inclination of the plane is equal to the global friction angle of the material
= 34. In fact, all the material forming the plane is unbonded (see again
Fig. 5.47) and therefore it has been deposited on the underlying bonded
soil according to its angle of repose. At this point the simulation has
stopped since the number of particles involved in the successive failures
becomes too little in order that the analysis continues to be meaningful.
Fig. 5.46: slope 1. Field of velocities during the occurrence of the second landslide.
Chapter 5 Numerical simulations
183
Fig. 5.47: slope 1. Profile and network of bonds at the end of the soil flow after the
occurrence of the second landslide.
5.6. References
Bardet J.P., 1993. Numerical tests with discrete element method. Proc.
Modern approaches to plasticity. Kolymbas D. ed., Elsevier, pp. 179-197.
Bardet J.P., 1994. Numerical simulations of the incremental response of
idealized granular materials. Int. J. of Plasticity, 10(8), pp. 879-908.
Calvetti F., 1997. Micromeccanica dei materiali granulari: modellazione
discreta del comportamento meccanico. PhD. thesis (in Italian),
Politecnico di Milano, Milan.
Calvetti F., Combe G., Lanier J., 1997. Experimental micromechanical
analysis of a 2D granular material: relation between structure evolution
and loading path. Mech. of Cohesive-Frictional materials, 2, pp.121-163.
Cundall P.A., Drescher A., Strack O.D.L., 1982. Numerical experiments
on granular assemblies; measurement and observations. IUTAM Conf.
Deformation and failure of granular materials, Vermeer P.A. and Luger
Chapter 5 Numerical simulations
184
H.J. eds., Balkema, pp. 355-370.
Emeriault F. and Chang C. S., 1997. Interparticle forces and
displacements in granular materials. Computers and geotechnics, 20, pp.
223-244.
Hammad W.I., 1991. Modelisation non linaire et tude exprimentale
des bandes de cisaillement dans les sables. PhD. thesis (in French),
Universit J. Fourier, Grenoble, France.
Itasca Consulting Group, 1999. Particle flow code in 2-D (PFC-2D):
Users manual. Version 2. Augmented Fishtank.
Jiang M.J., Konrad J.M. and Leroueil S., 2003. An efficient technique for
generating homogeneous specimens for DEM studies. Computers and
Geotechnics, 30, pp. 579-597.
Konishi J., Oda M., Nemat-Nasser S., 1982. Inherent anisotropy and
shear strength of assembly of oval cross-sectional rods. Proc. IUTAM
Conf. Deformation and failure of granular materials, Vermeer P.A. and
Luger H.J. eds., Balkema, pp. 403-412.
Konishi J., Oda M., Nemat-Nasser S., 1983. Induced anisotropy in
assemblies of oval cross-sectional rods in biaxial compression. Proc.
Mech. of Granular Materials: New Models and Constitutive Relations.
Jenkins J.T. and Satake M. eds., Elsevier, pp. 31-39.
Kruyt N. P., 1993. Towards micro-mechanical constitutive relations for
granular materials. Proc. Modern approaches to Plasticity. Kolymbas D.
ed., Balkema, pp. 147-178.
Panzeri D. and Rinaldi O., 1999. Modellazione ad elementi distinti di un
versante in frana. Degree thesis (in Italian), Politecnico di Milano,
Milan.
Potyondy D.O. and Cundall P.A., 1998. Modeling notch formation
mechanisms in the URL mine by test tunnel using bonded assemblies of
circular particles. Int. J. Rock Mech. & Min. Sci., Special Issue (Proc.
3
rd
Am. Rock Mech. Symp. NARMS 98, Mexico), paper 67.
Chapter 5 Numerical simulations
185
Potyondy D.O., Cundall P.A. and Lee C., 1996. Modeling rock using
bonded assemblies of circular particles. Rock Mech. Tool and
Techniques (Proc. 2
nd
Am. Rock Mech. Symp. NARMS 96, Canada),
Balkema, pp. 1937-1944.
Potyondy D.O., Cundall P.A. and Sarracino R. S., 1996. Modeling of
shock and gas driven fractures induced by a blast using bonded
assemblies of spherical particles. Rock Fragmentation by Blasting (Proc.
5
th
Int. Symp. 96, Canada), Balkema, pp. 55-62.
Skinner A.E., 1969. A note on the influence of interparticle friction on
the shearing strength of a random assembly of spherical particles.
Gotechnique, 19, pp. 150-157.
Sitharam T.G., 1999. Micromechanical modeling of granular materials:
effect of confining pressure on mechanical behaviour. Mech. of
Materials, 31, pp. 653-665.
Part 2 - Conclusions
187
Part 2 - Conclusions
In order to reproduce a continuum by a synthetic assembly of rigid
particles with the distinct element method, an experimental campaign on
numerical specimens have been performed by running biaxial tests. The
porosity of specimens has been determined so that a uniform network of
contact forces within the specimens has been obtained and the magnitude
of these contact forces was small compared to the magnitude at failure. A
test procedure which guarantees the presence of quasi-static conditions
during the test execution has been followed. In order to reproduce a
material characterised by values of (global friction angle) of
engineering interest, all the tests have been run inhibiting particle
rotation. It has been shown that many factors affect the global strength of
samples which depends not only on micromechanical strength parameters
(intergranular friction angle), but also on the stiffness assigned to
contacts. As regards contact stiffness, it has been shown the existence of
a threshold value upon which it does not affect the global strength any
more. Moreover, the dependence of internal friction angle on confining
pressure has been investigated. A value of contact stiffness has been
assigned to contacts so that the obtained dependence of on confining
pressure was acceptable. In the choice of contact stiffness also global
stiffness has been taken into account: the contact stiffness has been
Part 2 Conclusions
188
determined so that the global secant moduli of elasticity obtained were
close to experimental data relative to biaxial tests run on sand. Finally, a
linear relationship between
). In the
proposed model, the case of pure frictional strength is a particular case of
the model (c
and c
adopted. In conclusion, thanks to the bond model proposed, simple
relationships between
and and c
H
)
Fig. 6.3: normalised cohesion vs. dimensionless crest retreat: the circles represent discrete
values from limit analysis and the line represents a fifth degree polynomial
interpolating function ( = 30).
In Fig. 6.4, the selected time-cohesion laws (Eq. (6.1)), calibrated to
fit the experimental data at best, are plotted. In Fig. 6.5, the crest retreat -
time curves obtained by substituting these laws into the cohesion crest
retreat relationship achieved by limit analysis (Fig. 6.3), are shown.
The hyperbolic law is the only one which makes the calculated
solution fitting well the experimental curve, whereas the exponential one
manages to catch the shape of the crest retreat - time curve from a
qualitative viewpoint. Instead, linear and parabolic laws are not suitable
to describe the cohesion decrease of soil in time.
Chapter 6 Slope weathering: natural time scale
196
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
0 100 200 300 400 500 600 700 800 900 1000
time [days]
c
/
(
H
)
linear hyperbolic parabolic exponential
Fig. 6.4: normalised cohesion vs. time laws calibrated to fit the experimental data at best.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 100 200 300 400 500 600 700 800 900 1000
time [days]
L
/
H
linear hyperbolic parabolic exponential experimental
Lfin
Fig. 6.5: evolution of the Warden Point slope; crest retreat vs. time curves obtained for
different time-cohesion laws ( = 30).
Chapter 6 Slope weathering: natural time scale
197
6.2.2. Case 2: Miramar
The second case analysed concerns the rear scarp of the Miramar
landslide, occurred in 1953, at Herne Bay (Kent, England) [Hutchinson,
1973] which underwent a strong degradation from an inclination of 73
in February 1953 to about 22 in March 1968 (see Fig. 6.6).
Fig. 6.6: a) Miramar rear scarp in 1953 two days after failure. b) Miramar rear scarp in
summer 1965 (after [Hutchinson, 1973]).
According to Hutchinson, conditions of no erosion were present. In
Fig. 6.7 the evolution undergone by the slope is schematically drawn.
From this figure, it can be inferred that material accumulated on the slope
toe plays a significant role into limiting the retrogression of the slope
front since failures occur only within the upper part of the slope where
the slope surface is not covered by landslided material.
Fig. 6.7: successive slope profiles generated by weathering action; no erosion conditions are
present (after [Hutchinson, 1973]).
The available observations concerning measures of crest recession at
different times are only three. At the time of the third observation, March
Chapter 6 Slope weathering: natural time scale
198
1968, a nearly plane surface 22 inclined was formed. Since slope profile
was plane, it can be assumed that the whole weathering process was
ended at that time. Therefore the third crest retreat measure is the final
one and the soil friction angle can be assumed equal to the angle of
repose =22.
From the three available observations, a relationship in terms of rate
of crest retreat (calculated considering finite time intervals) against time
was sought by [Hutchinson, 1973]. In the paper, the available
observations were plotted in a chart in bilogarithmic scale and two points
were obtained. A linear relationship was therefore inferred by
Hutchinson (see Fig. 6.8). In the writers opinion, data are too few to
conclude that the relationship is linear since given two points, any kind of
curve can easily fit them. The reported crest retreat rates are not realistic
(erroneous) since if crest retreat is calculated according to the
relationship shown in Fig. 6.8, the final crest retreat obtained (at t=15
years, time of the last survey) is L=2.96H. This retreat results larger than
the distance, D, between the upper and lower points of the final slope
profile: D=H/tan=2.48H (=22). Such a distance is given by the
summation of crest retreat, length of the slope base where fallen material
lies and the projection of the slope profile to the horizon (see Fig. 6.9).
Therefore these crest retreats are erroneous and the only reliable data
concern the initial and final slope profiles, documented also by
photographs (see Fig. 6.6).
Chapter 6 Slope weathering: natural time scale
199
Fig. 6.8: relationship inferred by Hutchinson in terms of time vs. finite rate of crest retreat
(after [Hutchinson, 1973]).
L
D
73
22
H
Fig. 6.9: initial (grey lines) and final (black lines) profiles of Miramar slope.
In order to model the evolution of such a slope, a numerical
simulation has been run with the distinct element method (see Ch. 4). In
this case, conditions of no erosion have to be simulated. Therefore the
material fallen after each failure has not been removed from the slope toe
(numerical simulation run according to 5.5.7). In Fig. 6.10 the
geometric domain adopted in the numerical analysis is shown. The
procedure used to generate the slope and to simulate the retrogressive
Chapter 6 Slope weathering: natural time scale
200
failure are the same as those described in 5.5.3 and 5.5.4 respectively.
About 75000 disks have been used.
3H
H
H
2H
73
Fig. 6.10: geometric domain adopted in the numerical analysis.
The analysis has been stopped after the second landslide occurrence
since the number of disks involved in successive failures become very
low. The trend of cohesion crest retreat relationship has been achieved
by polynomial interpolation of the recorded values at failure (see Fig.
6.11). In the figure the same relationship obtained by limit analysis is
also shown. Comparing the two solutions, the difference appears very
marked. This means that the stabilising effect exerted by the material
accumulated on the slope toe can not be neglected as it strongly affects
crest retreat. This result is in accordance with what has been achieved in
5.5.7 where a slope characterised by =34 and =90 has been
simulated. In the following, the numerical solution alone has been taken
since the slope is subject to no erosion as previously said.
Chapter 6 Slope weathering: natural time scale
201
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
0 0.5 1 1.5 2 2.5
L/H
c
/
(
H
)
fourth degree polynomial function
fifth degree polynomial function
Fig. 6.11: normalised cohesion vs. dimensionless crest retreat: the triangles represent
discrete values from numerical analysis whereas the circles represent discrete
values from limit analysis ( = 22). Lines represent polynomial interpolating
functions .
Since the scarcity of available data (the only reliable observations
concern the initial and final slope profiles) it is not possible to use the
case analysed first calibrating time - cohesion laws and then comparing
predictions to experimental data as it has been done in 6.2.1. Therefore
it has been decided to use the available information in order to check the
goodness of the time-cohesion laws already calibrated in 6.2.1. But the
cohesion at time 0 is different: c/(H)=0.429 instead of c/(H)=0.181 (see
Fig. 6.3); hence one parameter has been assigned again to the laws so that
the initial cohesion assumes the value determined by numerical analysis
c/(H)=0.429 whereas the other parameter has been taken equal to what
has been determined in 6.2.1. The linear and parabolic time - cohesion
laws have not been considered since they gave unsatisfactory fitting of
predictions to experimental data.
In Fig. 6.12, the crest retreat - time relationships obtained are shown.
The crest retreat at t=15 years (when slope profile has reached its final
configuration) should be equal to L
fin
if predictions were exact. More
precisely, according to the time cohesion laws adopted (hyperbolic and
exponential), L
fin
can be reached only asymptotically since cohesion
becomes zero at an infinite time (see Fig. 6.4). In case of hyperbolic law,
Chapter 6 Slope weathering: natural time scale
202
the predicted value of crest retreat at t=15 years is about half L
fin
whereas
crest retreat reaches a value close to L
fin
at t=50 years. In case of
exponential law, crest retreats remain much lower than L
fin
up to a very
large time value.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 5 10 15 20 25 30 35 40 45 50
time [years]
L
/
H
hyperbolic exponential
Lfin
Fig. 6.12: evolution of the Miramar slope; crest retreat vs. time curves obtained for the
hyperbolic and exponential time-cohesion laws ( = 22).
It can be concluded that assuming the proposed cohesion - time laws,
predicted times relative to the evolution of the slope are underestimated.
This means that the parameters ruling the selected cohesion - time laws
(Eq. (6.1)) are not valid independently of the slope studied. But if a
calibration of the considered cohesion - time laws is performed on the
basis of the monitored values of crest retreat relative to the studied slope
(as in 6.2.1), they could still fit the experimental data. This cannot be
done in the case here treated since the scarcity of data available.
6.3. References
Bromhead E. N., 1978. Large landslides in London Clay at Herne Bay,
Kent. Quarterly J. Engrg. Geol., 11, pp. 291-304.
Gostelow, T. P., 1974. Slope development in stiff overconsolidated clays.
PhD. thesis. Imperial College University of London, London, UK.
Chapter 6 Slope weathering: natural time scale
203
Hutchinson J. N., 1967. The free degradation of London Clay cliffs. Proc.
Geotech. Conf., Oslo, Norway, 1, pp. 113-118.
Hutchinson J. N., 1969. A reconsideration of the coastal landslides at
Folkestone Warren, Kent. Gotechnique, 19, pp. 6-38.
Hutchinson J. N., 1970. A coastal mudflow on the London Clay cliffs at
Beltinge, North Kent. Gotechnique, 20, pp. 412-438.
Hutchinson J. N., 1973. The response of London Clay cliffs to differing
rates of toe erosion. Geol. Appl. e Idrogeol., Bari, 7, pp. 221-239.
Hutchinson J. N., 1976. Coastal landslides in cliffs of Pleistocene
deposits between Cromer and Overstrand, Norfolk, England. In: Janbu
N., Jorstad F. and Kjaernsli B. eds., Laurits Bjerrum Memorial Volume,
Contributions to Soil Mech., Norwegian Geotechnical Institute, Oslo,
Norway, pp. 155-182.
Hutchinson J. N., 1998. A small-scale field check on the Fisher and
Bakker-Le Heux cliff degradation models. Earth Surface Processes and
Landforms, 23, pp. 913-926.
Hutchinson J. N., 2001. Reading the ground: Morphology and Geology
in Site Appraisal. Quarterly J. of Engrg. Geol. and Hydrogeol., 34, pp. 7-
50.
Hutchinson J. N., Bromhead E. N., Lupini J. F., 1980. Additional
observations on the landslides at Folkstone Warren. Quarterly J. of
Engrg. Geol., 13, pp. 1-31.
Wallace, R. E., 1980. Degradation on the Hegben Lake fault scarps of
1959. Geology, 8, pp. 225-229.
Final conclusions
205
Final conclusions
In this thesis work, an engineering model relative to the evolution of
natural slopes subject to weathering has been made up. The today
available models in the literature either do not take into account soil
mechanical properties or do take them into account by empirical
variables with absent or obscure meaning. The mechanical parameters
required by the engineering model made up as input data are: internal
friction angle and cohesion which are the very parameters usually
determined into the geotechnical engineering practice in order to
characterise soil mechanical strength.
In the first part of the thesis, an analytical law describing the discrete
succession of landslides occurring to slopes subject to weathering has
been obtained by the limit analysis method. This law makes it possible to
build tables where cohesion values responsible of failure occurrence for
different values of internal friction angle and initial slope inclination are
plotted (see appendix A.1). These values can be used in order to obtain a
first rough estimation of the time period involved into slope evolution.
The estimation obtained is rough since some assumptions introduced into
limit analysis in many cases cannot be considered realistic: homogeneous
slopes, uniform weathering throughout slopes, no influence of material
accumulated on the slope base on the successive failure mechanisms, no
Final conclusions
206
tension cracks.
In the second part of the thesis the discrete succession of failures
characterising slope evolution has been modelled by a numerical method
(DEM). This modelling is more powerful since the effect of soil
accumulated on the slope toe and the effect of dynamic impact of
detached soil on the slope base can be taken into account. Moreover, the
method can be used also to model inhomogeneous slopes subject to non-
uniform weathering. In fact, thanks to the calibration procedure of
micromechanical parameters made up and to the bond model introduced,
any value of c and of a c, soil type may be reproduced by a discrete
assembly of rigid particles. The disadvantage of the numerical method in
comparison with the analytical law, is represented by the fact that a non
negligible computational time is requested to perform such numerical
analyses.
In the last part of the thesis (Ch. 6) suitable relationships between
time and weathering, expressed in terms of cohesion decrease, relative to
the evolution of natural slopes have been sought. A two-parameter
relationship has been found out good at this end. Using this relationship,
predictions made by the engineering model made up, managed to
reproduce the few experimental data available from a qualitative
viewpoint.
Appendix A Limit analysis results
A.1
A.
Appendix A
A. Limit analysis results
A.1. Tables of results
In the following, some tables of values describing the evolution of (c,
) slopes are shown. They have been obtained by an algorithm
implemented by the author in Matlab. The algorithm is based on the
analytical study performed in Ch. 3. Different analytical functions are
minimised (first or subsequent failures) and a loop is performed to
determine as many failure mechanisms as wanted. The required input
data are: , . The data outputted in the tables are: the cohesion (c), the
total crest retreat (L), the height of each failure mechanism (h), the angles
(x, y) of each failure spiral and the co-ordinates (X
O
, Y
O
) of the centres
of rotation of each failure spiral. In Fig. A.1 the geometric data printed in
tables are illustrated. All the values are printed in normalised form so that
the tables can be applied to slopes of any height and unit weight. The
values of considered in the tables lye in a range of engineering interest
(from 16 to 45). The number of failure mechanisms calculated is larger
at low values of since the retrogression lengths associated to each
mechanism are smaller and therefore a larger number of mechanisms is
required to be taken into account to describe the phenomenon.
Appendix A Limit analysis results
A.2
H
O (X
O
; Y
O
)
h
Y
X
x
y
L
Fig. A.1: geometric data relative to a failure mechanism.
= 90, = 45
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.107138 0.356513 1 55.574 72.806 -0.85754 2.77132
0.041164 0.550762 1 51.762 91.365 0.03929 1.64908
0.022772 0.680122 0.667955 51.41 94.451 0.36723 1.39209
0.014147 0.760396 0.410637 51.328 94.341 0.56772 1.24074
0.008743 0.809943 0.25281 51.303 94.297 0.69123 1.1482
0.005392 0.840516 0.155899 51.302 94.297 0.76731 1.09138
0.003325 0.859366 0.096137 51.297 94.296 0.81422 1.05634
0.002051 0.870991 0.059283 51.297 94.293 0.84315 1.03474
0.001265 0.87816 0.036557 51.296 94.292 0.86099 1.02142
0.00078 0.882582 0.022543 51.299 94.292 0.872 1.01321
0.000481 0.885308 0.013901 51.297 94.293 0.87878 1.00815
0.000297 0.886989 0.008572 51.296 94.293 0.88296 1.00502
0.000183 0.888026 0.005286 51.3 94.293 0.88554 1.0031
0.000113 0.888665 0.00326 51.298 94.294 0.88713 1.00191
0.00007 0.889059 0.00201 51.296 94.293 0.88812 1.00118
= 90, = 40
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.120657 0.403116 1 52.131 70.917 -0.94628 2.73534
Appendix A Limit analysis results
A.3
0.047849 0.628619 1 47.649 91.873 0.05333 1.6311
0.027374 0.785007 0.690432 47.197 95.525 0.4203 1.39381
0.017665 0.885651 0.439733 47.085 95.358 0.65293 1.2503
0.011322 0.950178 0.281304 47.06 95.339 0.8013 1.15999
0.00725 0.991483 0.179938 47.054 95.323 0.89623 1.10235
0.00464 1.017909 0.115094 47.053 95.315 0.95697 1.06547
0.002968 1.034814 0.073617 47.052 95.312 0.99583 1.04188
0.001899 1.045627 0.047086 47.053 95.311 1.02069 1.02679
0.001215 1.052543 0.030117 47.053 95.31 1.0366 1.01713
0.000777 1.056967 0.019263 47.051 95.31 1.04677 1.01096
0.000497 1.059796 0.012321 47.056 95.307 1.05327 1.00701
0.000318 1.061606 0.007881 47.056 95.309 1.05743 1.00448
0.000203 1.062763 0.005041 47.054 95.309 1.06009 1.00287
0.00013 1.063504 0.003224 47.055 95.309 1.0618 1.00183
= 90, = 35
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.13476 0.452104 1 48.776 69.042 -1.03161 2.6934
0.055427 0.713365 1 43.565 92.528 0.07111 1.61086
0.032961 0.903068 0.716671 42.986 96.844 0.47881 1.39543
0.022229 1.03058 0.476218 42.855 96.62 0.74815 1.26204
0.014867 1.115831 0.317636 42.829 96.579 0.9274 1.17467
0.009928 1.172728 0.211826 42.822 96.556 1.04703 1.11649
0.006625 1.210719 0.141472 42.819 96.576 1.1268 1.07776
0.004423 1.236095 0.09449 42.824 96.581 1.18004 1.05195
0.002954 1.25302 0.063014 42.82 96.554 1.21563 1.03465
0.001971 1.264323 0.042085 42.824 96.573 1.23936 1.02314
0.001316 1.271872 0.028109 42.826 96.582 1.2552 1.01545
0.000879 1.276906 0.018745 42.82 96.554 1.26578 1.01031
0.000586 1.280269 0.012519 42.822 96.573 1.27284 1.00688
0.000391 1.282514 0.008362 42.819 96.583 1.27755 1.0046
0.000261 1.284012 0.005576 42.819 96.553 1.2807 1.00307
= 90, = 30
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.149546 0.503926 1 45.505 67.193 -1.11213 2.64479
0.064126 0.806833 1 39.499 93.374 0.09362 1.58791
Appendix A Limit analysis results
A.4
0.039884 1.038445 0.746051 38.774 98.436 0.54482 1.39652
0.028268 1.20211 0.52117 38.625 98.182 0.85694 1.27579
0.019863 1.317054 0.365188 38.594 98.133 1.07515 1.19307
0.013934 1.397622 0.255827 38.584 98.104 1.22812 1.13523
0.009767 1.454137 0.17945 38.584 98.123 1.33527 1.09484
0.006849 1.493777 0.125886 38.585 98.13 1.4104 1.06653
0.004804 1.521554 0.088185 38.582 98.099 1.46312 1.04661
0.003367 1.541036 0.061856 38.586 98.12 1.50006 1.03269
0.002361 1.554699 0.043392 38.584 98.13 1.52596 1.02293
0.001656 1.564276 0.030397 38.586 98.098 1.54413 1.01607
0.00116 1.570991 0.021322 38.588 98.121 1.55687 1.01127
0.000814 1.575701 0.014957 38.588 98.131 1.56579 1.00791
0.000571 1.579001 0.010478 38.584 98.1 1.57206 1.00554
= 90, = 25
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.165132 0.559084 1 42.317 65.378 -1.18683 2.58961
0.074261 0.911666 1 35.457 94.463 0.12195 1.5624
0.04865 1.197896 0.781096 34.553 100.445 0.62048 1.39764
0.036554 1.412154 0.577982 34.378 100.123 0.98484 1.29235
0.027194 1.571556 0.429089 34.349 100.084 1.25435 1.21678
0.020204 1.689924 0.318475 34.339 100.061 1.45448 1.16085
0.015001 1.777779 0.236343 34.341 100.046 1.60303 1.11939
0.011133 1.842971 0.175377 34.34 100.041 1.7133 1.08859
0.008262 1.891345 0.130133 34.338 100.039 1.79512 1.06573
0.006131 1.927241 0.096561 34.339 100.037 1.85584 1.04878
0.004549 1.953875 0.071649 34.337 100.037 1.9009 1.03619
0.003376 1.97364 0.053164 34.34 100.037 1.93433 1.02686
0.002505 1.988304 0.039448 34.339 100.037 1.95914 1.01993
0.001858 1.999185 0.029271 34.336 100.038 1.97754 1.01478
0.001379 2.007259 0.021719 34.338 100.036 1.9912 1.01097
= 90, = 20
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.181656 0.618101 1 39.202 63.606 -1.2541 2.52705
0.086269 1.031497 1 31.415 95.88 0.15793 1.53353
0.06006 1.391744 0.822782 30.304 103.024 0.70986 1.39852
Appendix A Limit analysis results
A.5
0.048353 1.680339 0.652079 30.104 102.615 1.14063 1.3129
0.038483 1.90992 0.517611 30.071 102.55 1.4816 1.248
0.030569 2.0922 0.410912 30.064 102.53 1.75218 1.19682
0.024271 2.236971 0.32635 30.064 102.54 1.96693 1.15631
0.019276 2.351912 0.259058 30.065 102.524 2.13754 1.12409
0.015302 2.443186 0.205737 30.066 102.538 2.27294 1.09855
0.012152 2.515645 0.163313 30.066 102.523 2.3805 1.07823
0.009646 2.573186 0.129699 30.064 102.54 2.46587 1.06212
0.007661 2.618861 0.102955 30.062 102.524 2.53367 1.04931
0.006082 2.655134 0.081765 30.063 102.537 2.58748 1.03916
0.004829 2.683928 0.064904 30.062 102.522 2.63022 1.03108
0.003834 2.706797 0.051545 30.064 102.538 2.66415 1.02469
0.003044 2.724949 0.040916 30.062 102.523 2.69109 1.0196
0.002417 2.739366 0.032494 30.066 102.537 2.71248 1.01557
0.001919 2.75081 0.025793 30.066 102.523 2.72947 1.01236
0.001524 2.759897 0.020484 30.065 102.538 2.74295 1.00981
0.00121 2.767112 0.01626 30.065 102.525 2.75366 1.00779
0.00096 2.772842 0.012914 30.063 102.541 2.76216 1.00618
0.000763 2.77739 0.010251 30.065 102.523 2.76891 1.00491
0.000605 2.781001 0.008141 30.066 102.538 2.77426 1.0039
0.000481 2.783868 0.006462 30.063 102.523 2.77852 1.0031
0.000382 2.786145 0.005132 30.066 102.537 2.7819 1.00246
0.000303 2.787952 0.004074 30.063 102.523 2.78458 1.00195
0.000241 2.789388 0.003235 30.065 102.538 2.78671 1.00155
0.000191 2.790527 0.002568 30.065 102.525 2.7884 1.00123
0.000152 2.791432 0.00204 30.064 102.54 2.78974 1.00098
0.00012 2.792151 0.001619 30.066 102.525 2.79081 1.00078
= 90, = 18
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.188564 0.642943 1 37.975 62.913 -1.27856 2.4999
0.091726 1.084919 1 29.801 96.563 0.17502 1.52113
0.065624 1.482466 0.841528 28.584 104.266 0.75096 1.39857
0.054515 1.810979 0.688196 28.379 103.816 1.21397 1.32252
0.044737 2.080312 0.563211 28.345 103.735 1.59188 1.26348
0.036634 2.300769 0.46088 28.339 103.711 1.90109 1.21556
0.029982 2.481269 0.377276 28.341 103.722 2.15409 1.17647
0.024542 2.629036 0.308866 28.339 103.729 2.36119 1.14446
0.020092 2.750006 0.252872 28.339 103.729 2.53072 1.11827
Appendix A Limit analysis results
A.6
0.016449 2.84904 0.207029 28.338 103.729 2.66951 1.09682
0.013468 2.930129 0.169497 28.341 103.728 2.78314 1.07928
0.011026 2.996522 0.138768 28.341 103.731 2.87618 1.06491
0.009027 3.050871 0.113613 28.34 103.73 2.95235 1.05314
0.00739 3.095374 0.093017 28.339 103.733 3.01471 1.0435
0.006051 3.131789 0.076115 28.339 103.709 3.06578 1.0356
0.004952 3.161597 0.062307 28.341 103.721 3.10756 1.02914
0.004053 3.186 0.051008 28.341 103.727 3.14177 1.02386
0.003318 3.205978 0.041761 28.342 103.728 3.16976 1.01953
0.002716 3.222335 0.03419 28.34 103.731 3.19269 1.01599
0.002224 3.235727 0.027992 28.338 103.732 3.21145 1.01309
0.001821 3.24669 0.022918 28.338 103.731 3.22682 1.01072
0.001491 3.255667 0.018763 28.338 103.731 3.2394 1.00877
0.001221 3.263017 0.015362 28.34 103.731 3.2497 1.00718
0.000999 3.269034 0.012577 28.341 103.73 3.25813 1.00588
0.000818 3.27396 0.010297 28.339 103.731 3.26503 1.00482
0.00067 3.277993 0.008431 28.341 103.729 3.27068 1.00394
0.000548 3.281295 0.006902 28.338 103.73 3.27531 1.00323
0.000449 3.283999 0.005651 28.339 103.731 3.2791 1.00264
0.000368 3.286212 0.004627 28.341 103.729 3.2822 1.00216
0.000301 3.288024 0.003788 28.338 103.732 3.28474 1.00177
= 90, = 16
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.19566 0.668677 1 36.766 62.231 -1.30178 2.47228
0.097622 1.142176 1 28.183 97.332 0.19403 1.50802
0.071924 1.58307 0.861568 26.856 105.654 0.79606 1.39851
0.061803 1.959673 0.728797 26.639 105.156 1.29593 1.33294
0.052416 2.278821 0.616615 26.605 105.079 1.7175 1.28116
0.044366 2.548851 0.521619 26.603 105.055 2.07402 1.23781
0.037532 2.777267 0.44117 26.603 105.047 2.37566 1.20113
0.031744 2.970422 0.373106 26.6 105.044 2.6308 1.17007
0.026848 3.133876 0.315693 26.602 105.064 2.8465 1.14392
0.022715 3.272109 0.267024 26.603 105.047 3.02903 1.12174
0.019214 3.389031 0.225828 26.603 105.044 3.18346 1.10296
0.016249 3.487949 0.191078 26.602 105.063 3.31401 1.08711
0.013749 3.57162 0.16162 26.6 105.05 3.4245 1.07367
0.01163 3.642385 0.136688 26.601 105.043 3.51796 1.06231
0.009835 3.702263 0.115653 26.603 105.063 3.59698 1.05273
Appendix A Limit analysis results
A.7
0.008322 3.752906 0.097823 26.599 105.051 3.66386 1.04459
0.007039 3.795737 0.082733 26.603 105.041 3.72042 1.03772
0.005953 3.831975 0.070001 26.602 105.062 3.76825 1.03191
0.005037 3.862626 0.059208 26.603 105.047 3.80873 1.02699
0.00426 3.888551 0.050073 26.603 105.043 3.84297 1.02283
0.003603 3.910485 0.042368 26.603 105.063 3.87192 1.01932
0.003049 3.929037 0.035836 26.6 105.05 3.89642 1.01634
0.002579 3.944727 0.030308 26.604 105.04 3.91714 1.01382
0.002181 3.958003 0.025643 26.604 105.061 3.93466 1.01169
0.001845 3.969231 0.02169 26.603 105.047 3.94949 1.00989
0.001561 3.978727 0.018343 26.599 105.044 3.96203 1.00836
0.00132 3.986763 0.015521 26.603 105.063 3.97263 1.00708
0.001117 3.99356 0.013128 26.602 105.049 3.98161 1.00598
0.000945 3.999308 0.011103 26.6 105.044 3.9892 1.00506
0.000799 4.004171 0.009394 26.599 105.065 3.99562 1.00428
0.000676 4.008285 0.007946 26.603 105.046 4.00105 1.00362
0.000572 4.011764 0.00672 26.599 105.044 4.00565 1.00306
0.000484 4.014708 0.005686 26.602 105.062 4.00953 1.00259
0.000409 4.017197 0.004809 26.601 105.049 4.01282 1.00219
0.000346 4.019303 0.004067 26.6 105.044 4.0156 1.00185
0.000293 4.021085 0.003442 26.604 105.062 4.01795 1.00157
0.000248 4.022592 0.002911 26.602 105.048 4.01994 1.00133
0.000209 4.023866 0.002462 26.602 105.041 4.02163 1.00112
0.000177 4.024945 0.002083 26.605 105.061 4.02305 1.00095
0.00015 4.025857 0.001762 26.601 105.05 4.02425 1.0008
0.000127 4.026628 0.00149 26.602 105.044 4.02527 1.00068
0.000107 4.027281 0.001261 26.601 105.063 4.02613 1.00057
0.000091 4.027833 0.001066 26.601 105.049 4.02686 1.00049
0.000077 4.0283 0.000902 26.603 105.04 4.02748 1.00041
0.000065 4.028695 0.000763 26.604 105.063 4.028 1.00035
0.000055 4.029029 0.000645 26.6 105.051 4.02844 1.00029
0.000046 4.029312 0.000546 26.602 105.042 4.02881 1.00025
0.000039 4.029551 0.000462 26.602 105.062 4.02913 1.00021
0.000033 4.029753 0.000391 26.602 105.046 4.0294 1.00018
0.000028 4.029924 0.00033 26.602 105.041 4.02962 1.00015
= 80, = 45
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.072692 0.254054 1 52.982 81.263 -0.30317 1.97283
Appendix A Limit analysis results
A.8
0.030938 0.431052 0.945137 51.733 94.876 0.16743 1.55772
0.019613 0.542449 0.574649 51.405 94.438 0.44952 1.33736
0.012176 0.61152 0.353267 51.323 94.336 0.62208 1.20708
0.007523 0.654148 0.217486 51.303 94.292 0.72833 1.12751
0.00464 0.680445 0.134115 51.297 94.295 0.7938 1.0786
0.002861 0.696664 0.082703 51.296 94.294 0.83416 1.04847
0.001765 0.706665 0.050999 51.298 94.291 0.85904 1.02989
0.001088 0.712832 0.031448 51.294 94.293 0.87439 1.01843
0.000671 0.716636 0.019393 51.298 94.291 0.88386 1.01137
0.000414 0.718981 0.011959 51.299 94.292 0.88969 1.00701
0.000255 0.720427 0.007374 51.297 94.292 0.89329 1.00432
0.000157 0.721319 0.004547 51.297 94.292 0.89551 1.00267
0.000097 0.721869 0.002804 51.296 94.293 0.89688 1.00164
0.00006 0.722208 0.001729 51.296 94.292 0.89772 1.00101
= 80, = 40
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.086145 0.301548 1 49.335 79.522 -0.36677 1.98319
0.037562 0.517996 0.99358 47.628 96.1 0.17259 1.57193
0.024846 0.66 0.626748 47.192 95.531 0.50532 1.35735
0.016036 0.751372 0.399174 47.083 95.359 0.71645 1.22719
0.010278 0.80995 0.255357 47.06 95.339 0.85113 1.14523
0.006582 0.847445 0.163341 47.053 95.324 0.9373 1.0929
0.004212 0.871436 0.104478 47.055 95.315 0.99244 1.05944
0.002695 0.886782 0.066827 47.056 95.312 1.02772 1.03802
0.001724 0.896596 0.042744 47.051 95.312 1.05029 1.02431
0.001103 0.902874 0.027339 47.053 95.309 1.06472 1.01555
0.000705 0.906889 0.017487 47.053 95.308 1.07396 1.00995
0.000451 0.909458 0.011185 47.05 95.309 1.07986 1.00636
0.000289 0.911101 0.007154 47.056 95.307 1.08364 1.00407
0.000185 0.912151 0.004576 47.054 95.31 1.08606 1.0026
0.000118 0.912823 0.002927 47.055 95.309 1.0876 1.00167
= 80, = 35
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.100308 0.352224 1 45.815 77.811 -0.42874 1.98493
0.045188 0.607975 1 43.418 96.923 0.18972 1.56263
Appendix A Limit analysis results
A.9
0.030675 0.784303 0.664047 42.952 96.775 0.56728 1.36619
0.020635 0.90267 0.44182 42.847 96.613 0.81697 1.24304
0.013797 0.981776 0.294683 42.824 96.576 0.98329 1.16201
0.009212 1.034621 0.196822 42.82 96.584 1.09421 1.10818
0.006153 1.069877 0.131258 42.817 96.554 1.16832 1.07217
0.004106 1.093421 0.087662 42.822 96.572 1.21774 1.04819
0.002741 1.109141 0.058549 42.82 96.581 1.25074 1.03218
0.00183 1.119631 0.039045 42.821 96.553 1.27279 1.02147
0.001221 1.126633 0.026077 42.817 96.575 1.28749 1.01433
0.000815 1.131311 0.017417 42.822 96.58 1.29731 1.00957
0.000544 1.134431 0.011615 42.817 96.555 1.30387 1.00639
0.000363 1.136514 0.007757 42.819 96.574 1.30824 1.00426
0.000243 1.137905 0.005181 42.821 96.581 1.31116 1.00285
= 80, = 30
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.115276 0.406515 1 42.409 76.143 -0.48794 1.97807
0.054003 0.70712 1 39.222 97.855 0.21341 1.54691
0.037812 0.926374 0.70319 38.712 98.334 0.63726 1.37304
0.026708 1.080924 0.491683 38.609 98.155 0.93157 1.26006
0.018751 1.189391 0.344479 38.591 98.117 1.1375 1.18212
0.013147 1.26549 0.241651 38.59 98.13 1.28175 1.12773
0.009222 1.318812 0.169282 38.585 98.101 1.38298 1.08949
0.006463 1.356205 0.118741 38.584 98.122 1.45388 1.06275
0.004532 1.382438 0.083298 38.588 98.13 1.50359 1.04403
0.003179 1.40082 0.058352 38.585 98.101 1.53848 1.03085
0.002228 1.413709 0.04093 38.584 98.123 1.56292 1.02163
0.001562 1.422752 0.028713 38.587 98.13 1.58006 1.01518
0.001096 1.429088 0.020114 38.585 98.1 1.59209 1.01063
0.000768 1.433532 0.014109 38.587 98.122 1.60051 1.00746
0.000538 1.436649 0.009897 38.588 98.132 1.60642 1.00523
= 80, = 25
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.131168 0.46516 1 39.118 74.531 -0.54338 1.96354
0.064348 0.819447 1 35.051 99.058 0.24354 1.52772
0.046913 1.094999 0.748116 34.473 100.304 0.71829 1.3797
Appendix A Limit analysis results
A.10
0.035095 1.300759 0.554551 34.365 100.114 1.06714 1.28033
0.026099 1.453713 0.411669 34.345 100.076 1.32571 1.20795
0.019387 1.567278 0.30553 34.339 100.056 1.51772 1.15431
0.014392 1.651561 0.22673 34.339 100.044 1.66025 1.11453
0.010681 1.714102 0.168242 34.339 100.04 1.76603 1.08498
0.007926 1.760509 0.124839 34.339 100.038 1.84453 1.06306
0.005881 1.794946 0.092632 34.34 100.037 1.90278 1.0468
0.004364 1.820495 0.068734 34.336 100.038 1.946 1.03471
0.003238 1.839455 0.051001 34.335 100.039 1.97807 1.02576
0.002403 1.853526 0.037843 34.34 100.037 2.00187 1.01912
0.001783 1.863964 0.02808 34.336 100.038 2.01953 1.01418
0.001323 1.87171 0.020836 34.338 100.038 2.03263 1.01052
= 80, = 20
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.148129 0.528891 1 35.923 72.997 -0.5935 1.94091
0.076713 0.949434 1 30.884 100.625 0.28225 1.50451
0.05886 1.301685 0.800402 30.206 102.834 0.81512 1.38591
0.047132 1.582991 0.634911 30.091 102.588 1.23386 1.30448
0.03748 1.806487 0.503885 30.066 102.534 1.56585 1.24137
0.029762 1.983946 0.39997 30.064 102.523 1.8293 1.19158
0.023626 2.12486 0.317645 30.063 102.538 2.03835 1.15213
0.018762 2.23672 0.252145 30.063 102.521 2.2044 1.12077
0.014894 2.325562 0.200244 30.064 102.538 2.3362 1.09591
0.011827 2.396086 0.158952 30.063 102.524 2.44088 1.07613
0.009389 2.452093 0.126236 30.064 102.539 2.52397 1.06046
0.007456 2.496554 0.100206 30.066 102.523 2.58996 1.048
0.005919 2.531857 0.079581 30.064 102.538 2.64233 1.03812
0.0047 2.559887 0.063171 30.065 102.525 2.68394 1.03026
0.003731 2.582144 0.050169 30.067 102.537 2.71696 1.02403
0.002963 2.599811 0.039824 30.066 102.522 2.74318 1.01908
0.002352 2.613842 0.031627 30.063 102.54 2.764 1.01515
0.001868 2.62498 0.025105 30.064 102.522 2.78053 1.01203
0.001483 2.633826 0.019938 30.065 102.539 2.79366 1.00955
0.001178 2.640847 0.015827 30.062 102.524 2.80408 1.00758
0.000935 2.646424 0.012569 30.065 102.539 2.81235 1.00602
0.000742 2.650851 0.009977 30.064 102.524 2.81892 1.00478
0.000589 2.654367 0.007924 30.066 102.54 2.82414 1.0038
0.000468 2.657158 0.00629 30.064 102.525 2.82828 1.00301
Appendix A Limit analysis results
A.11
0.000372 2.659374 0.004995 30.066 102.539 2.83157 1.00239
0.000295 2.661133 0.003965 30.066 102.525 2.83418 1.0019
0.000234 2.66253 0.003149 30.064 102.539 2.83625 1.00151
0.000186 2.66364 0.0025 30.062 102.525 2.8379 1.0012
0.000148 2.66452 0.001985 30.067 102.537 2.8392 1.00095
0.000117 2.66522 0.001576 30.062 102.526 2.84024 1.00075
= 80, = 18
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.155253 0.556117 1 34.674 72.412 -0.61173 1.92985
0.082368 1.007939 1 29.218 101.384 0.30084 1.49409
0.064727 1.399123 0.823784 28.486 104.053 0.86007 1.38819
0.053454 1.721151 0.67399 28.362 103.783 1.31289 1.31559
0.043826 1.985075 0.551742 28.344 103.739 1.68292 1.2581
0.035888 2.201079 0.451509 28.34 103.714 1.98585 1.21118
0.029372 2.377928 0.369615 28.341 103.725 2.23372 1.17289
0.024043 2.522687 0.3026 28.339 103.729 2.43661 1.14152
0.019685 2.641205 0.247743 28.338 103.73 2.6027 1.11586
0.016116 2.738234 0.202832 28.339 103.728 2.73867 1.09486
0.013194 2.817687 0.16606 28.341 103.73 2.85001 1.07767
0.010802 2.88273 0.135957 28.338 103.733 2.94116 1.06358
0.008844 2.935979 0.111312 28.339 103.73 3.01578 1.05206
0.007241 2.979575 0.091133 28.34 103.728 3.07687 1.04262
0.005928 3.015272 0.074611 28.34 103.731 3.1269 1.0349
0.004853 3.044496 0.061086 28.341 103.73 3.16785 1.02857
0.003974 3.068421 0.050012 28.339 103.731 3.20138 1.02339
0.003253 3.08801 0.040946 28.341 103.73 3.22883 1.01915
0.002664 3.104048 0.033523 28.339 103.731 3.2513 1.01568
0.002181 3.11718 0.027447 28.341 103.731 3.2697 1.01284
0.001785 3.127931 0.022471 28.341 103.732 3.28477 1.01051
0.001462 3.136732 0.018398 28.34 103.731 3.2971 1.0086
0.001197 3.143938 0.015063 28.342 103.729 3.3072 1.00705
0.00098 3.149838 0.012332 28.342 103.73 3.31547 1.00577
0.000802 3.154669 0.010097 28.342 103.731 3.32224 1.00472
0.000657 3.158623 0.008266 28.343 103.731 3.32778 1.00387
0.000538 3.161861 0.006768 28.342 103.73 3.33232 1.00317
0.00044 3.164512 0.005541 28.341 103.732 3.33603 1.00259
0.00036 3.166682 0.004537 28.342 103.73 3.33908 1.00212
0.000295 3.168459 0.003714 28.338 103.733 3.34157 1.00174
Appendix A Limit analysis results
A.12
= 80, = 16
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.162592 0.584399 1 33.435 71.848 -0.62861 1.91732
0.088517 1.070943 1 27.542 102.237 0.32159 1.48275
0.071407 1.50738 0.848941 26.75 105.415 0.90937 1.39029
0.060974 1.878731 0.718065 26.625 105.109 1.40121 1.32777
0.051654 2.193094 0.607297 26.606 105.054 1.81657 1.27692
0.043696 2.458993 0.513631 26.6 105.048 2.16778 1.23412
0.036959 2.683885 0.434391 26.6 105.043 2.4648 1.19801
0.031257 2.87416 0.367544 26.602 105.061 2.71591 1.16756
0.026446 3.0351 0.310875 26.599 105.05 2.92845 1.1417
0.02237 3.171209 0.262918 26.603 105.04 3.1082 1.11986
0.018918 3.28638 0.222454 26.603 105.062 3.26021 1.10142
0.016006 3.383783 0.188157 26.602 105.048 3.38883 1.08578
0.013539 3.466165 0.159129 26.601 105.043 3.49764 1.07254
0.01145 3.535873 0.134641 26.603 105.063 3.58964 1.06138
0.009688 3.594828 0.113883 26.604 105.047 3.66749 1.05192
0.008194 3.644689 0.096313 26.601 105.043 3.73334 1.04391
0.00693 3.686881 0.081492 26.603 105.063 3.78903 1.03715
0.005864 3.722565 0.068929 26.6 105.05 3.83615 1.03142
0.00496 3.752747 0.058296 26.603 105.042 3.87601 1.02658
0.004195 3.778282 0.049325 26.603 105.062 3.90971 1.02249
0.003549 3.799881 0.04172 26.602 105.049 3.93823 1.01902
0.003002 3.818147 0.035284 26.602 105.042 3.96235 1.01609
0.002539 3.833603 0.029854 26.604 105.063 3.98275 1.01361
0.002148 3.846676 0.025251 26.601 105.05 4.00002 1.01151
0.001817 3.857733 0.021356 26.602 105.044 4.01462 1.00974
0.001537 3.867087 0.01807 26.601 105.062 4.02697 1.00824
0.0013 3.875 0.015284 26.601 105.048 4.03741 1.00697
0.0011 3.881692 0.012926 26.601 105.043 4.04625 1.00589
0.00093 3.887354 0.010937 26.601 105.064 4.05373 1.00499
0.000787 3.892143 0.009251 26.602 105.047 4.06005 1.00422
0.000666 3.896193 0.007824 26.6 105.042 4.0654 1.00357
0.000563 3.89962 0.00662 26.601 105.064 4.06992 1.00302
0.000476 3.902519 0.005599 26.603 105.047 4.07375 1.00255
0.000403 3.90497 0.004735 26.602 105.042 4.07699 1.00216
0.000341 3.907045 0.004007 26.602 105.064 4.07972 1.00183
0.000288 3.908799 0.003389 26.604 105.047 4.08204 1.00155
Appendix A Limit analysis results
A.13
0.000244 3.910283 0.002866 26.602 105.044 4.084 1.00131
0.000206 3.911538 0.002425 26.605 105.062 4.08566 1.00111
0.000174 3.9126 0.002051 26.601 105.05 4.08706 1.00094
0.000148 3.913498 0.001735 26.604 105.041 4.08825 1.00079
0.000125 3.914258 0.001468 26.603 105.062 4.08925 1.00067
0.000106 3.914901 0.001241 26.604 105.046 4.0901 1.00057
0.000089 3.915444 0.00105 26.601 105.042 4.09082 1.00048
0.000076 3.915904 0.000888 26.603 105.062 4.09142 1.0004
0.000064 3.916293 0.000751 26.602 105.047 4.09194 1.00034
0.000054 3.916622 0.000635 26.603 105.042 4.09237 1.00029
0.000046 3.916901 0.000538 26.6 105.065 4.09274 1.00025
0.000039 3.917136 0.000455 26.6 105.048 4.09305 1.00021
0.000033 3.917335 0.000385 26.604 105.04 4.09331 1.00018
0.000028 3.917504 0.000325 26.604 105.061 4.09353 1.00015
= 70, = 45
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.045417 0.168205 1 53.425 88.946 -0.03238 1.76087
0.023339 0.302093 0.723738 51.863 95.023 0.32983 1.42825
0.014903 0.386823 0.438698 51.437 94.49 0.54542 1.25761
0.009275 0.439427 0.269255 51.332 94.34 0.67705 1.15789
0.005732 0.471913 0.165769 51.306 94.297 0.75804 1.09719
0.003536 0.491955 0.102225 51.299 94.297 0.80793 1.05991
0.002181 0.504315 0.063038 51.298 94.294 0.83868 1.03694
0.001345 0.511938 0.038872 51.298 94.292 0.85765 1.02278
0.000829 0.516639 0.023971 51.297 94.293 0.86935 1.01405
0.000511 0.519538 0.014782 51.296 94.293 0.87657 1.00866
0.000315 0.521325 0.009115 51.297 94.292 0.88102 1.00534
0.000194 0.522428 0.005621 51.297 94.292 0.88376 1.00329
0.00012 0.523108 0.003466 51.297 94.292 0.88545 1.00203
0.000074 0.523527 0.002137 51.297 94.291 0.88649 1.00125
0.000046 0.523785 0.001318 51.296 94.291 0.88714 1.00077
= 70, = 40
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.058303 0.214343 1 48.99 87.471 -0.07748 1.75412
0.029575 0.38447 0.774839 47.535 95.988 0.34114 1.44503
Appendix A Limit analysis results
A.14
0.019468 0.495623 0.489377 47.169 95.484 0.60097 1.27898
0.012541 0.567103 0.312157 47.082 95.359 0.76588 1.17766
0.008038 0.612912 0.199691 47.06 95.339 0.87119 1.11357
0.005147 0.642234 0.127734 47.057 95.322 0.93858 1.07266
0.003294 0.660993 0.081703 47.054 95.316 0.9817 1.04648
0.002107 0.672992 0.052259 47.053 95.311 1.00929 1.02973
0.001348 0.680668 0.033426 47.056 95.308 1.02694 1.01902
0.000862 0.685577 0.021379 47.053 95.309 1.03823 1.01216
0.000551 0.688717 0.013675 47.053 95.309 1.04545 1.00778
0.000353 0.690726 0.008746 47.057 95.307 1.05006 1.00498
0.000226 0.69201 0.005594 47.052 95.31 1.05302 1.00318
0.000144 0.692832 0.003578 47.054 95.31 1.05491 1.00204
0.000092 0.693357 0.002289 47.052 95.31 1.05612 1.0013
= 70, = 35
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.072167 0.26488 1 44.855 86.008 -0.12192 1.74699
0.036765 0.477678 0.824169 43.294 97.276 0.3551 1.45841
0.025214 0.622568 0.544392 42.926 96.738 0.66398 1.30001
0.016937 0.719772 0.362717 42.844 96.62 0.86866 1.19947
0.011326 0.784717 0.241927 42.827 96.575 1.00516 1.13303
0.007563 0.828104 0.161587 42.826 96.584 1.09623 1.08884
0.005051 0.85705 0.107761 42.821 96.556 1.15708 1.05926
0.00337 0.876379 0.07197 42.822 96.575 1.19765 1.03957
0.00225 0.889286 0.048069 42.823 96.581 1.22474 1.02642
0.001503 0.897896 0.032057 42.82 96.553 1.24284 1.01763
0.001003 0.903646 0.021409 42.823 96.572 1.25491 1.01177
0.000669 0.907485 0.014299 42.82 96.582 1.26297 1.00786
0.000447 0.910047 0.009536 42.818 96.554 1.26836 1.00524
0.000298 0.911757 0.006369 42.821 96.572 1.27195 1.0035
0.000199 0.912899 0.004254 42.818 96.582 1.27435 1.00234
= 70, = 30
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.087074 0.320305 1 40.964 84.586 -0.16463 1.73699
0.045213 0.58478 0.872502 39.107 98.966 0.37246 1.46845
0.032645 0.77403 0.605879 38.691 98.299 0.73692 1.32122
Appendix A Limit analysis results
A.15
0.02303 0.907374 0.42418 38.607 98.167 0.99042 1.22432
0.016175 1.000945 0.297201 38.589 98.121 1.16803 1.15711
0.011343 1.066596 0.208489 38.586 98.132 1.29247 1.11018
0.007957 1.112603 0.146051 38.584 98.101 1.3798 1.07721
0.005576 1.144867 0.102446 38.583 98.123 1.44098 1.05414
0.00391 1.167498 0.071867 38.584 98.13 1.48387 1.03798
0.002743 1.183355 0.050344 38.582 98.098 1.51397 1.02661
0.001922 1.194477 0.035313 38.584 98.121 1.53506 1.01866
0.001348 1.202279 0.024772 38.588 98.129 1.54984 1.01309
0.000945 1.207744 0.017353 38.582 98.1 1.56022 1.00917
0.000663 1.211578 0.012172 38.583 98.121 1.56749 1.00643
0.000465 1.214267 0.008539 38.587 98.129 1.57258 1.00451
= 70, = 25
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.103131 0.381398 1 37.273 83.236 -0.20432 1.72277
0.055335 0.709722 0.91739 34.938 101.095 0.39531 1.47392
0.042452 0.959155 0.676256 34.452 100.29 0.82329 1.34291
0.031738 1.145189 0.501236 34.36 100.101 1.13861 1.25334
0.023594 1.283448 0.372062 34.345 100.069 1.37235 1.18796
0.017522 1.386086 0.276125 34.342 100.052 1.5459 1.13948
0.013007 1.462258 0.204905 34.339 100.045 1.67473 1.1035
0.009653 1.518776 0.152048 34.337 100.041 1.77033 1.0768
0.007163 1.560716 0.112823 34.334 100.04 1.84127 1.05697
0.005316 1.591837 0.083716 34.336 100.036 1.89391 1.04228
0.003944 1.614929 0.062118 34.336 100.037 1.93297 1.03137
0.002927 1.632065 0.046092 34.34 100.036 1.96195 1.02328
0.002171 1.644778 0.0342 34.338 100.037 1.98346 1.01727
0.001611 1.654212 0.025377 34.338 100.037 1.99942 1.01282
0.001196 1.661212 0.01883 34.338 100.037 2.01126 1.00951
= 70, = 20
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.120492 0.449288 1 33.764 81.988 -0.23985 1.70405
0.067738 0.860108 0.960365 30.77 103.872 0.42529 1.4756
0.055915 1.194516 0.759017 30.183 102.788 0.92997 1.36555
0.044717 1.461373 0.602225 30.082 102.583 1.32698 1.28868
Appendix A Limit analysis results
A.16
0.035555 1.673513 0.478202 30.069 102.549 1.64178 1.22909
0.028242 1.841943 0.379625 30.066 102.53 1.89177 1.18185
0.022423 1.975707 0.301502 30.064 102.543 2.0902 1.14441
0.017808 2.081896 0.239339 30.063 102.526 2.24782 1.11463
0.014137 2.166225 0.19008 30.062 102.54 2.37292 1.09103
0.011227 2.233164 0.150887 30.063 102.522 2.47228 1.07227
0.008913 2.286323 0.119829 30.064 102.536 2.55114 1.05739
0.007078 2.32852 0.095118 30.064 102.521 2.61378 1.04556
0.005619 2.362036 0.075539 30.067 102.537 2.6635 1.03619
0.004462 2.388639 0.059962 30.066 102.523 2.70299 1.02872
0.003542 2.409765 0.04762 30.063 102.54 2.73433 1.02281
0.002813 2.426536 0.037801 30.062 102.524 2.75923 1.0181
0.002233 2.439853 0.030021 30.063 102.537 2.77898 1.01438
0.001773 2.450426 0.02383 30.063 102.523 2.79468 1.01141
0.001408 2.458822 0.018925 30.064 102.537 2.80713 1.00906
0.001118 2.465487 0.015023 30.062 102.524 2.81703 1.0072
0.000887 2.470781 0.011931 30.064 102.539 2.82488 1.00571
0.000705 2.474982 0.009471 30.063 102.522 2.83112 1.00454
0.000559 2.478319 0.007521 30.064 102.538 2.83607 1.0036
0.000444 2.480967 0.00597 30.063 102.521 2.84 1.00286
0.000353 2.483071 0.004741 30.062 102.538 2.84312 1.00227
0.00028 2.484741 0.003764 30.066 102.522 2.8456 1.0018
0.000222 2.486067 0.002989 30.063 102.538 2.84756 1.00143
0.000177 2.487119 0.002373 30.063 102.522 2.84913 1.00114
0.00014 2.487955 0.001884 30.062 102.538 2.85037 1.0009
0.000111 2.488619 0.001496 30.063 102.523 2.85135 1.00072
= 70, = 18
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.127845 0.478691 1 32.402 81.53 -0.25241 1.69502
0.073531 0.929656 0.976379 29.082 105.209 0.44025 1.47464
0.062753 1.308692 0.797365 28.459 104.008 0.98043 1.37521
0.051762 1.620323 0.652137 28.358 103.753 1.41865 1.3053
0.04241 1.875648 0.533728 28.344 103.726 1.77675 1.24969
0.034718 2.084688 0.436959 28.342 103.729 2.06972 1.2044
0.028423 2.255835 0.357745 28.338 103.733 2.30958 1.16731
0.023272 2.395963 0.292898 28.339 103.731 2.50594 1.13698
0.019053 2.510695 0.239803 28.34 103.732 2.66671 1.11216
0.015599 2.604625 0.196334 28.338 103.733 2.79834 1.09182
Appendix A Limit analysis results
A.17
0.012772 2.681493 0.160658 28.34 103.71 2.90614 1.07515
0.010451 2.74441 0.131514 28.337 103.725 2.99434 1.0615
0.008555 2.795924 0.107668 28.339 103.728 3.06653 1.05036
0.007004 2.838095 0.088149 28.341 103.729 3.12562 1.04123
0.005734 2.872623 0.072169 28.34 103.732 3.17401 1.03375
0.004695 2.900892 0.059086 28.341 103.731 3.21362 1.02764
0.003843 2.924036 0.048375 28.344 103.73 3.24605 1.02263
0.003147 2.942984 0.039606 28.341 103.731 3.27261 1.01853
0.002576 2.958498 0.032427 28.342 103.732 3.29435 1.01517
0.002109 2.971198 0.026549 28.339 103.731 3.31215 1.01242
0.001727 2.981597 0.021736 28.341 103.73 3.32672 1.01017
0.001414 2.99011 0.017796 28.339 103.731 3.33865 1.00832
0.001158 2.99708 0.01457 28.337 103.732 3.34842 1.00681
0.000948 3.002787 0.011929 28.34 103.73 3.35641 1.00558
0.000776 3.007459 0.009766 28.339 103.729 3.36296 1.00457
0.000635 3.011285 0.007996 28.339 103.731 3.36832 1.00374
0.00052 3.014417 0.006546 28.34 103.731 3.37271 1.00306
0.000426 3.016981 0.00536 28.337 103.732 3.3763 1.00251
0.000349 3.01908 0.004388 28.341 103.729 3.37924 1.00205
0.000285 3.020799 0.003593 28.338 103.732 3.38165 1.00168
= 70, = 16
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.135456 0.509632 1 31.067 81.101 -0.26386 1.68526
0.079908 1.006253 0.991429 27.391 106.698 0.45749 1.47294
0.070686 1.437841 0.838524 26.724 105.342 1.03725 1.38493
0.060245 1.804646 0.709167 26.621 105.092 1.5229 1.32365
0.051016 2.115203 0.599986 26.605 105.067 1.93299 1.27356
0.04317 2.377939 0.507502 26.604 105.05 2.27992 1.23139
0.036516 2.60014 0.429215 26.6 105.046 2.57341 1.19565
0.030885 2.788168 0.363173 26.601 105.064 2.82155 1.16556
0.026132 2.947201 0.307185 26.601 105.049 3.03154 1.14003
0.022104 3.081698 0.259797 26.599 105.045 3.20919 1.11842
0.018694 3.195507 0.219821 26.6 105.064 3.35938 1.1002
0.015817 3.29177 0.185932 26.603 105.048 3.48648 1.08477
0.013379 3.373175 0.157247 26.599 105.045 3.59401 1.07167
0.011315 3.442055 0.133051 26.602 105.062 3.68491 1.06065
0.009574 3.500313 0.112537 26.6 105.048 3.76184 1.0513
0.008098 3.549583 0.095175 26.602 105.04 3.82692 1.04339
Appendix A Limit analysis results
A.18
0.006848 3.591271 0.080528 26.602 105.061 3.88194 1.03671
0.005794 3.626532 0.068112 26.604 105.046 3.9285 1.03105
0.004901 3.656354 0.057603 26.602 105.043 3.96789 1.02626
0.004145 3.681587 0.048739 26.601 105.065 4.00119 1.02222
0.003507 3.702929 0.041225 26.602 105.047 4.02937 1.01879
0.002966 3.720979 0.034865 26.602 105.043 4.05321 1.01589
0.002509 3.736251 0.0295 26.602 105.062 4.07337 1.01345
0.002123 3.749168 0.024952 26.603 105.046 4.09042 1.01138
0.001795 3.760093 0.021102 26.602 105.043 4.10485 1.00962
0.001518 3.769337 0.017855 26.603 105.064 4.11705 1.00814
0.001285 3.777156 0.015102 26.6 105.051 4.12738 1.00688
0.001087 3.783768 0.012773 26.602 105.041 4.13611 1.00582
0.000919 3.789363 0.010807 26.602 105.063 4.1435 1.00493
0.000778 3.794095 0.009141 26.603 105.048 4.14974 1.00417
0.000658 3.798097 0.007731 26.603 105.043 4.15503 1.00352
0.000556 3.801484 0.006541 26.603 105.062 4.1595 1.00298
0.000471 3.804348 0.005533 26.603 105.047 4.16328 1.00252
0.000398 3.80677 0.004679 26.603 105.042 4.16648 1.00213
0.000337 3.80882 0.003959 26.603 105.064 4.16919 1.0018
0.000285 3.810554 0.003349 26.601 105.05 4.17148 1.00153
0.000241 3.81202 0.002832 26.602 105.041 4.17341 1.00129
0.000204 3.81326 0.002396 26.602 105.062 4.17505 1.00109
0.000172 3.814309 0.002027 26.603 105.048 4.17643 1.00092
0.000146 3.815197 0.001714 26.603 105.044 4.17761 1.00078
0.000123 3.815948 0.00145 26.603 105.064 4.1786 1.00066
0.000104 3.816583 0.001227 26.601 105.05 4.17944 1.00056
0.000088 3.81712 0.001038 26.602 105.042 4.18015 1.00047
0.000075 3.817574 0.000878 26.6 105.065 4.18075 1.0004
0.000063 3.817959 0.000743 26.601 105.049 4.18125 1.00034
0.000053 3.818284 0.000628 26.6 105.044 4.18168 1.00029
0.000045 3.818559 0.000531 26.601 105.064 4.18205 1.00024
0.000038 3.818792 0.000449 26.602 105.047 4.18235 1.0002
0.000032 3.818988 0.00038 26.603 105.04 4.18261 1.00017
0.000027 3.819155 0.000322 26.601 105.064 4.18283 1.00015
= 60, = 45
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.022861 0.096642 1 58.769 95.308 0.1701 1.831
0.016095 0.192541 0.606484 53.738 96.694 0.49303 1.37743
Appendix A Limit analysis results
A.19
0.011284 0.257433 0.353928 51.957 95.139 0.67063 1.20978
0.007246 0.298639 0.213855 51.462 94.515 0.7759 1.12565
0.004515 0.324262 0.131266 51.337 94.359 0.84004 1.07696
0.002793 0.340094 0.080818 51.308 94.306 0.8795 1.04738
0.001723 0.349864 0.049838 51.298 94.301 0.90381 1.0292
0.001063 0.35589 0.030733 51.3 94.295 0.91881 1.01801
0.000656 0.359606 0.018952 51.295 94.295 0.92806 1.01111
0.000404 0.361898 0.011687 51.296 94.292 0.93376 1.00685
0.000249 0.363312 0.007207 51.295 94.293 0.93728 1.00422
0.000154 0.364183 0.004444 51.295 94.293 0.93945 1.0026
0.000095 0.364721 0.00274 51.296 94.292 0.94078 1.00161
0.000058 0.365052 0.00169 51.296 94.292 0.94161 1.00099
0.000036 0.365256 0.001042 51.3 94.292 0.94212 1.00061
= 60, = 40
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.034584 0.139378 1 52.597 94.389 0.13514 1.76061
0.022545 0.271946 0.658687 48.633 97.176 0.50637 1.38942
0.01569 0.362033 0.406897 47.438 95.869 0.72526 1.23317
0.010274 0.420684 0.257737 47.147 95.459 0.8618 1.14685
0.006612 0.45835 0.164386 47.074 95.342 0.94869 1.09355
0.004235 0.48248 0.105155 47.058 95.33 1.00417 1.05981
0.002711 0.497922 0.067262 47.053 95.319 1.03966 1.03826
0.001735 0.507801 0.043022 47.053 95.313 1.06237 1.02447
0.00111 0.514119 0.027518 47.05 95.312 1.0769 1.01565
0.00071 0.518161 0.017601 47.05 95.31 1.08619 1.01001
0.000454 0.520747 0.011258 47.056 95.306 1.09213 1.00641
0.00029 0.5224 0.007201 47.049 95.311 1.09594 1.0041
0.000186 0.523458 0.004606 47.054 95.309 1.09837 1.00262
0.000119 0.524134 0.002946 47.055 95.31 1.09992 1.00168
0.000076 0.524567 0.001884 47.054 95.31 1.10092 1.00107
= 60, = 35
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.047757 0.187925 1 47.087 93.377 0.10115 1.71436
0.029917 0.363167 0.709585 43.897 98.03 0.52327 1.40148
0.021159 0.485143 0.463691 43.064 96.959 0.78821 1.25634
Appendix A Limit analysis results
A.20
0.014331 0.567419 0.307786 42.874 96.661 0.96231 1.1694
0.009598 0.622428 0.205004 42.828 96.572 1.07815 1.11274
0.006409 0.659192 0.136924 42.823 96.584 1.15533 1.07527
0.00428 0.683719 0.091313 42.82 96.554 1.20688 1.05021
0.002856 0.700097 0.060984 42.822 96.573 1.24127 1.03353
0.001907 0.711035 0.040732 42.825 96.581 1.26422 1.02239
0.001273 0.718331 0.027163 42.819 96.554 1.27956 1.01494
0.00085 0.723203 0.018141 42.822 96.572 1.28979 1.00997
0.000567 0.726456 0.012117 42.823 96.58 1.29662 1.00666
0.000379 0.728627 0.00808 42.821 96.554 1.30118 1.00444
0.000253 0.730076 0.005397 42.822 96.574 1.30422 1.00297
0.000169 0.731044 0.003604 42.825 96.581 1.30626 1.00198
= 60, = 30
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.062363 0.242922 1 42.093 92.351 0.06891 1.67875
0.038513 0.469349 0.760644 39.383 99.336 0.54459 1.41219
0.028185 0.632997 0.526185 38.753 98.403 0.86216 1.27948
0.019953 0.748441 0.367514 38.616 98.161 1.08236 1.19444
0.014013 0.829507 0.257494 38.591 98.123 1.23628 1.13613
0.009827 0.886382 0.180636 38.585 98.133 1.34409 1.09546
0.006894 0.926239 0.12654 38.582 98.1 1.41975 1.06689
0.004831 0.954194 0.08876 38.584 98.122 1.47275 1.04691
0.003388 0.973802 0.062266 38.585 98.129 1.50991 1.03291
0.002376 0.987542 0.043618 38.586 98.099 1.53599 1.02306
0.001665 0.997178 0.030595 38.587 98.122 1.55426 1.01617
0.001168 1.003937 0.021463 38.585 98.132 1.56707 1.01134
0.000819 1.008673 0.015035 38.584 98.101 1.57606 1.00795
0.000574 1.011995 0.010546 38.586 98.122 1.58236 1.00557
0.000403 1.014325 0.007398 38.588 98.131 1.58677 1.00391
= 60, = 25
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.078474 0.305355 1 37.521 91.37 0.0394 1.64759
0.048759 0.595127 0.812266 34.997 101.193 0.57179 1.42055
0.037524 0.815674 0.598348 34.465 100.311 0.95077 1.30355
0.028072 0.980279 0.44355 34.363 100.116 1.22975 1.22419
Appendix A Limit analysis results
A.21
0.020875 1.102628 0.329271 34.346 100.076 1.43656 1.16633
0.015506 1.193458 0.244377 34.341 100.054 1.59013 1.12344
0.011511 1.260871 0.181347 34.341 100.044 1.70413 1.09161
0.008543 1.310892 0.134567 34.339 100.04 1.78874 1.06797
0.006339 1.348012 0.099851 34.337 100.04 1.85153 1.05043
0.004704 1.375555 0.074092 34.337 100.038 1.89812 1.03742
0.003491 1.395992 0.054977 34.338 100.037 1.93269 1.02777
0.00259 1.411157 0.040793 34.336 100.039 1.95835 1.0206
0.001922 1.42241 0.030269 34.337 100.037 1.97738 1.01529
0.001426 1.430759 0.02246 34.339 100.036 1.9915 1.01135
0.001058 1.436954 0.016665 34.338 100.037 2.00198 1.00842
= 60, = 20
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.09624 0.376691 1 33.284 90.504 0.01422 1.61697
0.061289 0.747502 0.862631 30.668 103.698 0.60797 1.42511
0.050324 1.048288 0.682033 30.166 102.744 1.06091 1.32823
0.040197 1.288032 0.540966 30.08 102.557 1.41769 1.25931
0.031943 1.478531 0.429479 30.066 102.536 1.70049 1.20573
0.025367 1.629783 0.340915 30.063 102.524 1.92503 1.16328
0.020137 1.749901 0.270747 30.063 102.539 2.10322 1.12967
0.015992 1.845256 0.214919 30.062 102.524 2.24477 1.10293
0.012695 1.920981 0.170684 30.066 102.536 2.35709 1.08176
0.010081 1.981093 0.135487 30.065 102.524 2.44633 1.0649
0.008003 2.028836 0.1076 30.066 102.54 2.51715 1.05154
0.006355 2.066729 0.085414 30.064 102.525 2.5734 1.04091
0.005045 2.096827 0.067834 30.065 102.54 2.61805 1.03249
0.004006 2.120719 0.053847 30.065 102.525 2.65351 1.02579
0.003181 2.139691 0.042764 30.066 102.539 2.68166 1.02048
0.002526 2.154753 0.033946 30.068 102.522 2.70401 1.01626
0.002005 2.166712 0.026959 30.064 102.539 2.72175 1.01291
0.001592 2.176207 0.0214 30.062 102.525 2.73585 1.01025
0.001264 2.183747 0.016996 30.068 102.537 2.74703 1.00814
0.001004 2.189733 0.013491 30.065 102.525 2.75592 1.00646
0.000797 2.194486 0.010714 30.064 102.539 2.76297 1.00513
0.000633 2.198259 0.008505 30.062 102.525 2.76857 1.00407
0.000502 2.201256 0.006754 30.067 102.537 2.77302 1.00324
0.000399 2.203635 0.005362 30.066 102.525 2.77655 1.00257
0.000317 2.205524 0.004258 30.062 102.54 2.77935 1.00204
Appendix A Limit analysis results
A.22
0.000252 2.207024 0.00338 30.065 102.525 2.78158 1.00162
0.0002 2.208215 0.002684 30.065 102.54 2.78334 1.00129
0.000159 2.209161 0.002131 30.065 102.525 2.78475 1.00102
0.000126 2.209911 0.001692 30.066 102.54 2.78586 1.00081
0.0001 2.210507 0.001343 30.062 102.525 2.78675 1.00064
= 60, = 18
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.103859 0.408218 1 31.673 90.204 0.00572 1.60454
0.067139 0.818399 0.882447 28.937 104.935 0.62573 1.42573
0.056845 1.161552 0.72095 28.438 103.954 1.11313 1.33889
0.046819 1.443414 0.589706 28.355 103.746 1.50929 1.27604
0.038352 1.67424 0.482604 28.34 103.72 1.83307 1.22573
0.031394 1.86325 0.395087 28.337 103.729 2.098 1.18475
0.025702 2.018009 0.323461 28.341 103.728 2.31485 1.1513
0.021041 2.144713 0.26482 28.341 103.731 2.49241 1.12387
0.017226 2.248439 0.216815 28.338 103.733 2.63778 1.1014
0.014104 2.333311 0.177419 28.338 103.708 2.7568 1.08298
0.011542 2.402786 0.145231 28.339 103.72 2.85419 1.06792
0.009448 2.459667 0.118894 28.339 103.727 2.93391 1.05561
0.007734 2.506241 0.097339 28.342 103.729 2.99918 1.04553
0.006332 2.544366 0.079693 28.34 103.731 3.05261 1.03727
0.005184 2.575579 0.065247 28.339 103.731 3.09635 1.03052
0.004244 2.601134 0.053419 28.339 103.73 3.13216 1.02498
0.003475 2.622058 0.043735 28.339 103.731 3.16148 1.02045
0.002845 2.639189 0.035807 28.338 103.732 3.18549 1.01675
0.002329 2.653213 0.029316 28.339 103.729 3.20514 1.01371
0.001907 2.664696 0.024002 28.339 103.731 3.22123 1.01123
0.001561 2.674097 0.019651 28.337 103.733 3.23441 1.00919
0.001278 2.681795 0.016089 28.341 103.73 3.24519 1.00753
0.001047 2.688097 0.013172 28.34 103.732 3.25402 1.00616
0.000857 2.693256 0.010784 28.342 103.73 3.26125 1.00504
0.000702 2.69748 0.008829 28.339 103.731 3.26717 1.00413
0.000574 2.700938 0.007229 28.342 103.729 3.27202 1.00338
0.00047 2.703769 0.005918 28.339 103.731 3.27599 1.00277
0.000385 2.706088 0.004846 28.341 103.732 3.27924 1.00227
0.000315 2.707986 0.003967 28.342 103.732 3.2819 1.00186
0.000258 2.70954 0.003248 28.34 103.733 3.28407 1.00152
Appendix A Limit analysis results
A.23
= 60, = 16
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.111798 0.441731 1 30.108 89.938 -0.00171 1.59192
0.073586 0.896805 0.901983 27.202 106.347 0.6461 1.4256
0.064444 1.290131 0.763105 26.693 105.291 1.17198 1.3497
0.054847 1.624045 0.645414 26.614 105.085 1.61378 1.29443
0.046433 1.9067 0.546012 26.603 105.062 1.98701 1.24893
0.039287 2.145798 0.461832 26.603 105.049 2.30274 1.21055
0.033231 2.348013 0.390585 26.602 105.044 2.56981 1.17806
0.028105 2.519102 0.330483 26.603 105.063 2.79561 1.15067
0.02378 2.66382 0.279531 26.603 105.048 2.98671 1.12744
0.020113 2.78621 0.236408 26.601 105.044 3.14836 1.10777
0.017011 2.88977 0.20003 26.601 105.064 3.28504 1.09118
0.014393 2.97736 0.169193 26.603 105.047 3.40069 1.07714
0.012174 3.051432 0.14309 26.6 105.042 3.49853 1.06523
0.010296 3.114114 0.121069 26.601 105.064 3.58126 1.05519
0.008712 3.16713 0.102404 26.601 105.049 3.65126 1.04668
0.007369 3.211965 0.086607 26.603 105.041 3.71048 1.03948
0.006232 3.249902 0.073278 26.602 105.063 3.76055 1.03341
0.005273 3.28199 0.061981 26.6 105.05 3.80292 1.02825
0.00446 3.309127 0.05242 26.602 105.041 3.83876 1.0239
0.003772 3.332089 0.044352 26.602 105.063 3.86907 1.02022
0.003191 3.351512 0.037514 26.603 105.048 3.89471 1.0171
0.002699 3.367937 0.031727 26.604 105.043 3.91641 1.01447
0.002283 3.381834 0.026845 26.602 105.063 3.93475 1.01224
0.001932 3.393589 0.022706 26.6 105.05 3.95027 1.01035
0.001634 3.403531 0.019203 26.6 105.044 3.9634 1.00875
0.001382 3.411943 0.016248 26.602 105.063 3.9745 1.00741
0.001169 3.419058 0.013743 26.6 105.05 3.9839 1.00626
0.000989 3.425075 0.011623 26.6 105.044 3.99185 1.0053
0.000836 3.430167 0.009835 26.603 105.062 3.99856 1.00448
0.000708 3.434473 0.008318 26.602 105.047 4.00425 1.00379
0.000599 3.438115 0.007035 26.599 105.044 4.00906 1.00321
0.000506 3.441197 0.005952 26.602 105.062 4.01313 1.00271
0.000428 3.443803 0.005035 26.6 105.049 4.01657 1.00229
0.000362 3.446008 0.004258 26.602 105.041 4.01948 1.00194
0.000306 3.447873 0.003603 26.604 105.061 4.02194 1.00164
0.000259 3.44945 0.003047 26.604 105.047 4.02403 1.00139
0.000219 3.450784 0.002577 26.599 105.045 4.02579 1.00117
Appendix A Limit analysis results
A.24
0.000185 3.451913 0.002181 26.604 105.063 4.02728 1.00099
0.000157 3.452868 0.001844 26.603 105.048 4.02854 1.00084
0.000133 3.453676 0.00156 26.602 105.042 4.02961 1.00071
0.000112 3.454359 0.00132 26.602 105.063 4.03051 1.0006
0.000095 3.454937 0.001116 26.602 105.048 4.03127 1.00051
0.00008 3.455426 0.000944 26.602 105.041 4.03192 1.00043
0.000068 3.455839 0.000799 26.603 105.061 4.03246 1.00036
0.000057 3.456189 0.000676 26.601 105.05 4.03292 1.00031
0.000049 3.456485 0.000571 26.599 105.044 4.03331 1.00026
0.000041 3.456735 0.000483 26.602 105.062 4.03365 1.00022
0.000035 3.456947 0.000409 26.602 105.047 4.03392 1.00019
0.000029 3.457126 0.000346 26.603 105.043 4.03416 1.00016
0.000025 3.457277 0.000293 26.603 105.062 4.03436 1.00013
= 50, = 35
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.02557 0.117872 1 54.089 99.221 0.30786 1.89637
0.021959 0.252775 0.634403 46.401 100.142 0.7193 1.39125
0.017226 0.353332 0.401573 43.706 97.809 0.95607 1.22592
0.012069 0.422782 0.262971 43.016 96.869 1.10621 1.14525
0.008148 0.469538 0.174763 42.863 96.639 1.20501 1.09617
0.005453 0.500821 0.116575 42.832 96.59 1.27077 1.0641
0.003643 0.521702 0.077745 42.825 96.56 1.31467 1.04276
0.002431 0.535645 0.051924 42.824 96.577 1.34394 1.02855
0.001623 0.544956 0.034681 42.823 96.582 1.36348 1.01906
0.001084 0.551169 0.023128 42.818 96.555 1.37654 1.01272
0.000723 0.555317 0.015446 42.822 96.573 1.38525 1.00849
0.000483 0.558088 0.010317 42.824 96.581 1.39107 1.00567
0.000323 0.559936 0.00688 42.818 96.555 1.39495 1.00378
0.000215 0.56117 0.004595 42.822 96.573 1.39754 1.00253
0.000144 0.561994 0.003069 42.819 96.583 1.39927 1.00169
= 50, = 30
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.039351 0.171306 1 47.074 98.957 0.2812 1.784
0.031297 0.360808 0.692697 40.868 100.957 0.74219 1.39604
0.024456 0.503832 0.471097 39.086 98.937 1.03176 1.25276
Appendix A Limit analysis results
A.25
0.017639 0.606042 0.327086 38.686 98.279 1.22858 1.17341
0.012437 0.67804 0.228974 38.607 98.159 1.36548 1.1211
0.008732 0.728554 0.160425 38.59 98.118 1.46137 1.08481
0.006123 0.763989 0.112538 38.586 98.131 1.52855 1.05947
0.004295 0.78882 0.078835 38.582 98.099 1.57569 1.04167
0.00301 0.806237 0.055298 38.586 98.121 1.60871 1.02923
0.002111 0.818453 0.038791 38.587 98.131 1.63186 1.0205
0.00148 0.827013 0.027174 38.585 98.1 1.64811 1.01437
0.001037 0.833016 0.019061 38.585 98.121 1.65949 1.01007
0.000728 0.837227 0.013371 38.586 98.129 1.66747 1.00707
0.00051 0.840178 0.009367 38.585 98.099 1.67307 1.00495
0.000358 0.842247 0.00657 38.585 98.121 1.67699 1.00347
= 50, = 25
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.055254 0.234157 1 40.858 98.606 0.25804 1.70511
0.042379 0.490456 0.750962 35.843 102.317 0.77242 1.40246
0.033922 0.690472 0.548574 34.635 100.598 1.12406 1.28011
0.025608 0.840738 0.405774 34.396 100.176 1.37989 1.20534
0.019078 0.952526 0.300943 34.353 100.073 1.56912 1.15208
0.014171 1.03553 0.22335 34.341 100.056 1.7095 1.11282
0.01052 1.097143 0.165745 34.342 100.044 1.81369 1.08373
0.007808 1.142861 0.122989 34.339 100.041 1.89102 1.06212
0.005794 1.176786 0.091261 34.339 100.038 1.94841 1.0461
0.004299 1.201959 0.067717 34.339 100.037 1.99099 1.03421
0.00319 1.220637 0.050246 34.338 100.037 2.02258 1.02538
0.002367 1.234497 0.037283 34.339 100.037 2.04603 1.01883
0.001756 1.244781 0.027664 34.339 100.036 2.06342 1.01397
0.001303 1.252411 0.020527 34.338 100.036 2.07633 1.01037
0.000967 1.258073 0.015231 34.339 100.036 2.08591 1.00769
= 50, = 20
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.073378 0.30846 1 35.264 98.314 0.23991 1.64179
0.055897 0.649998 0.810574 31.101 104.393 0.81289 1.40793
0.0469 0.930864 0.639568 30.243 102.905 1.24014 1.3089
0.037633 1.155427 0.507013 30.095 102.589 1.57489 1.24321
Appendix A Limit analysis results
A.26
0.029928 1.3339 0.402385 30.068 102.535 1.84002 1.19277
0.023766 1.475594 0.319406 30.062 102.524 2.05039 1.15298
0.018867 1.588141 0.253664 30.065 102.538 2.21734 1.1215
0.014982 1.677486 0.201358 30.065 102.525 2.34996 1.09645
0.011894 1.748427 0.159915 30.062 102.539 2.45521 1.07659
0.009446 1.804751 0.126941 30.066 102.523 2.5388 1.06081
0.007498 1.849474 0.100813 30.064 102.539 2.60516 1.04828
0.005955 1.884982 0.080025 30.065 102.525 2.65786 1.03833
0.004727 1.913178 0.063554 30.063 102.54 2.69969 1.03044
0.003754 1.935562 0.05045 30.067 102.522 2.73291 1.02417
0.00298 1.953339 0.040066 30.066 102.54 2.75929 1.01919
0.002366 1.967449 0.031804 30.064 102.524 2.78023 1.01523
0.001879 1.978654 0.025258 30.064 102.538 2.79685 1.0121
0.001492 1.98755 0.02005 30.062 102.524 2.81006 1.0096
0.001184 1.994615 0.015923 30.066 102.538 2.82054 1.00763
0.00094 2.000223 0.01264 30.068 102.522 2.82886 1.00606
0.000747 2.004676 0.010038 30.065 102.537 2.83547 1.00481
0.000593 2.008211 0.007968 30.065 102.524 2.84072 1.00382
0.000471 2.011019 0.006328 30.066 102.539 2.84488 1.00303
0.000374 2.013247 0.005023 30.065 102.523 2.84819 1.00241
0.000297 2.015017 0.003989 30.065 102.54 2.85082 1.00191
0.000236 2.016423 0.003167 30.065 102.526 2.8529 1.00152
0.000187 2.017538 0.002515 30.067 102.538 2.85456 1.0012
0.000149 2.018424 0.001996 30.062 102.525 2.85587 1.00096
0.000118 2.019127 0.001585 30.065 102.539 2.85692 1.00076
0.000094 2.019686 0.001259 30.063 102.524 2.85774 1.0006
= 50, = 18
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.081299 0.342051 1 33.178 98.238 0.2344 1.61902
0.062206 0.725072 0.834825 29.249 105.48 0.83308 1.40941
0.053521 1.048597 0.6815 28.491 104.067 1.29585 1.32123
0.044217 1.314888 0.557339 28.363 103.77 1.67055 1.26099
0.036242 1.5331 0.456202 28.341 103.735 1.97658 1.21338
0.029675 1.711686 0.373314 28.338 103.711 2.22705 1.17459
0.024286 1.857874 0.305595 28.338 103.721 2.43197 1.14292
0.01988 1.977569 0.25018 28.341 103.726 2.59971 1.11702
0.016274 2.075562 0.204821 28.342 103.729 2.73704 1.09581
0.013323 2.155782 0.167689 28.339 103.732 2.84947 1.07843
Appendix A Limit analysis results
A.27
0.010908 2.221466 0.137292 28.339 103.732 2.94151 1.06421
0.008931 2.275247 0.112405 28.341 103.731 3.01687 1.05258
0.007312 2.319277 0.092029 28.342 103.731 3.07857 1.04305
0.005986 2.355318 0.075347 28.34 103.73 3.12908 1.03524
0.004901 2.384832 0.061688 28.34 103.731 3.17044 1.02885
0.004013 2.408996 0.050505 28.341 103.731 3.2043 1.02362
0.003285 2.428776 0.04135 28.34 103.73 3.23202 1.01934
0.00269 2.444974 0.033854 28.34 103.733 3.25472 1.01583
0.002202 2.458236 0.027718 28.341 103.733 3.2733 1.01296
0.001803 2.469087 0.022681 28.339 103.711 3.28852 1.01061
0.001475 2.477968 0.018567 28.338 103.722 3.30097 1.00868
0.001208 2.48524 0.0152 28.34 103.725 3.31116 1.00711
0.000989 2.491193 0.012444 28.342 103.727 3.3195 1.00582
0.000809 2.496067 0.010188 28.342 103.73 3.32633 1.00477
0.000663 2.500057 0.008341 28.34 103.732 3.33192 1.0039
0.000543 2.503324 0.006829 28.339 103.732 3.3365 1.00319
0.000444 2.505999 0.005591 28.34 103.729 3.34025 1.00262
0.000364 2.508189 0.004577 28.34 103.731 3.34332 1.00214
0.000298 2.509982 0.003748 28.338 103.732 3.34583 1.00175
0.000244 2.51145 0.003068 28.341 103.729 3.34789 1.00144
= 50, = 16
c/(*H) L/H h/H x [deg.] y [deg.] X
O
/H Y
O
/H
0.089635 0.378324 1 31.169 98.202 0.23022 1.59714
0.06916 0.808318 0.859121 27.407 106.734 0.85629 1.41021
0.061242 1.1824 0.726864 26.728 105.362 1.35871 1.33376
0.05222 1.500411 0.614833 26.619 105.104 1.7797 1.28057
0.04423 1.769588 0.519963 26.606 105.05 2.13534 1.23709
0.037412 1.997229 0.439753 26.603 105.043 2.43602 1.20048
0.031642 2.189852 0.372083 26.602 105.063 2.69025 1.16963
0.026773 2.352779 0.314718 26.604 105.047 2.90538 1.14349
0.022645 2.490567 0.266163 26.599 105.044 3.08739 1.12132
0.019152 2.607164 0.225206 26.6 105.064 3.24127 1.10266
0.016205 2.705776 0.190487 26.602 105.047 3.37148 1.08684
0.013707 2.789173 0.161099 26.602 105.041 3.48162 1.07344
0.011592 2.859737 0.136306 26.602 105.061 3.57476 1.06214
0.009808 2.919421 0.11529 26.603 105.046 3.65357 1.05256
0.008296 2.969897 0.097502 26.599 105.044 3.72024 1.04444
0.007016 3.012612 0.082499 26.602 105.064 3.77661 1.03761
Appendix A Limit analysis results
A.28
0.005936 3.048733 0.06978 26.601 105.047 3.82431 1.03181
0.005021 3.079286 0.059015 26.6 105.043 3.86467 1.0269
0.004246 3.105137 0.049933 26.6 105.065 3.89878 1.02276
0.003593 3.127004 0.042235 26.602 105.048 3.92766 1.01925
0.003039 3.145495 0.03572 26.603 105.041 3.95208 1.01629
0.00257 3.161143 0.030222 26.603 105.064 3.97273 1.01378
0.002175 3.174377 0.025563 26.602 105.049 3.99021 1.01165
0.001839 3.185569 0.02162 26.6 105.043 4.00499 1.00986
0.001556 3.195039 0.018293 26.602 105.062 4.01749 1.00834
0.001316 3.203048 0.015472 26.601 105.047 4.02806 1.00705
0.001113 3.209822 0.013085 26.603 105.04 4.03701 1.00597
0.000942 3.215554 0.011071 26.604 105.062 4.04458 1.00505
0.000797 3.220402 0.009364 26.603 105.047 4.05098 1.00427
0.000674 3.224502 0.00792 26.604 105.042 4.05639 1.00361
0.00057 3.227971 0.006701 26.601 105.063 4.06097 1.00305
0.000482 3.230905 0.005668 26.601 105.048 4.06485 1.00258
0.000408 3.233386 0.004793 26.602 105.04 4.06812 1.00219
0.000345 3.235486 0.004056 26.603 105.061 4.07089 1.00185
0.000292 3.237262 0.00343 26.599 105.05 4.07324 1.00156
0.000247 3.238764 0.002901 26.602 105.042 4.07522 1.00132
0.000209 3.240035 0.002455 26.601 105.062 4.0769 1.00112
0.000177 3.24111 0.002076 26.603 105.049 4.07832 1.00095
0.000149 3.242019 0.001756 26.601 105.044 4.07952 1.0008
0.000126 3.242788 0.001486 26.599 105.066 4.08054 1.00068
0.000107 3.243438 0.001257 26.6 105.05 4.08139 1.00057
0.00009 3.243989 0.001063 26.601 105.042 4.08212 1.00048
0.000076 3.244454 0.000899 26.601 105.064 4.08274 1.00041
0.000065 3.244848 0.000761 26.602 105.047 4.08326 1.00035
0.000055 3.245181 0.000643 26.602 105.043 4.0837 1.00029
0.000046 3.245463 0.000544 26.603 105.062 4.08407 1.00025
0.000039 3.245701 0.00046 26.601 105.049 4.08438 1.00021
0.000033 3.245903 0.000389 26.599 105.045 4.08465 1.00018
0.000028 3.246073 0.000329 26.603 105.064 4.08487 1.00015
0.000024 3.246218 0.000279 26.602 105.048 4.08506 1.00013
A.2. Linear interpolation of cohesion - crest retreat
relationships
The results, obtained in terms of cohesion-crest retreat, have been
linearly interpolated excluding the first two failure mechanisms.
Appendix A Limit analysis results
A.29
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
L/H
c
/
(
*
H
)
= 18 = 20
= 15
= 45
= 5
Fig. A.2: dimensionless normalised cohesion vs. crest retreat obtained by linear
interpolation for different values ( = 90). The first two failure mechanisms
have not been considered. Figure shown in Ch. 3 (Fig. 3.21).
0
0.02
0.04
0.06
0.08
0.1
0.12
0 5 10 15 20 25 30 35 40 45 50
[degrees]
s
l
o
p
e
o
f
l
i
n
e
s
Fig. A.3: slopes of lines achieved in Fig. A.2 vs. friction angles ( = 90). The curve obtained
is a parabola.
Appendix A Limit analysis results
A.30
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0 5 10 15 20 25 30 35 40 45 50
[degrees]
i
n
t
e
r
c
e
p
t
o
f
l
i
n
e
s
Fig. A.4: intercept of lines achieved in Fig. A.2 vs. friction angles ( = 90). The curve
obtained is another parabola.