Professional Documents
Culture Documents
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/274720930
CITATIONS READS
5 548
5 authors, including:
Andrei V. Lyamin
University of Newcastle
104 PUBLICATIONS 1,897 CITATIONS
SEE PROFILE
All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Scott William Sloan
letting you access and read them immediately. Retrieved on: 28 June 2016
Tschuchnigg, F. et al. (2015). Géotechnique 65, No. 4, 249–257 [http://dx.doi.org/10.1680/geot.14.P.022]
In practical geotechnical engineering the factor of safety is still determined by means of simple limit
equilibrium analysis in many cases. However, because displacement finite-element analysis is routinely
applied for assessing displacements and stresses for working load conditions, this technique is
increasingly being used to calculate ultimate limit states and, consequently, factors of safety, usually
by means of the so-called strength reduction technique, and results which are comparable to those
obtained with limit equilibrium methods have been reported in the literature. However, owing to the
inherent assumptions of limit equilibrium analyses, they do not always provide unique factors of
safety. The purpose of this paper is on the one hand to compare the strength reduction method with
rigorous limit analyses which are based on collapse theorems of plasticity, and on the other hand to
investigate if a shortcoming of the strength reduction method, namely possible numerical instabilities
for non-associated plasticity, can be overcome. Two examples are considered, namely slope stability
and tunnel face stability. Finally an important note on the definition of the factor of safety for
effective and total stress analysis under undrained conditions is provided.
249
250 TSCHUCHNIGG, SCHWEIGER, SLOAN, LYAMIN AND RAISSAKIS
complex algorithm is required, as pointed out by Potts & Factor of safety obtained from finite-element limit analysis
Zdravkovic (2012). Actually it is one of the goals of this (FELA)
paper to show that the simple strength reduction procedure The upper- and lower-bound theorems of plasticity are
works for classical failure criteria by comparing it with powerful tools for predicting the stability of geotechnical
rigorous limit analysis solutions. The factor of safety (de- problems. Finite-element formulations of these theorems
noted FoS) obtained from the procedure is defined by have developed markedly over the last two decades and it is
now possible to apply them to a wide variety of complex
tan 9 c9 engineering problems (see, for example, the summary in
FoS ¼ ¼ (1)
tan 9mobilised c9mobilised Sloan, 2013).
Finite-element limit analysis is particularly powerful when
both upper- and lower-bound estimates are calculated so that
where c9 is the effective cohesion, 9 is the effective friction the true collapse load (for the idealised material) is
angle, and the ‘mobilised’ subscript denotes mobilised bracketed from above and below. The difference between the
values. An issue which must be addressed in displacement two bounds then provides an exact measure of the error in
finite-element analysis of failure is the definition of the flow the solution, and can be used to refine the meshes until a
rule. In general, a non-associated flow rule with a dilatancy suitably accurate estimate of the collapse load is found. The
angle ł9 smaller than the friction angle 9 is employed, but formulations used in this paper stem from the methods
this may lead to numerical instability with no clear indica- originally developed by Sloan (1988, 1989) and Sloan &
tion of the failure mechanism. This issue has been investi- Kleeman (1995), and further improved by Lyamin & Sloan
gated by Nordal (2008) in the context of an earth pressure (2002a, 2002b) and Krabbenhoft et al. (2005, 2007). A
problem. When using structured finite-element meshes with detailed description of the formulation of the FELA methods
a non-associated flow rule, it has been observed that the used in this paper, including the process for adaptive mesh
failure surface tends to propagate along element boundaries refinement and strength reduction, is given in Sloan (2013).
(Krabbenhoft et al., 2012). This is accompanied by strong If a safety factor based on the loads is desired, which is
oscillations of the resulting factor of safety during the defined as the ratio between limit load and actual load, the
strength reduction procedure, which is the consequence of a solution can be obtained from a single pair of upper- and
non-unique failure mechanism, thus making it difficult (or in lower-bound analyses. However, if the safety factor needs to
some cases even impossible) to define a unique value for be expressed in terms of the material strength, which is
this quantity. In the finite-element code employed in this defined as the ratio between mobilised material strength and
study the flow rule in the strength reduction procedure is actual material strength, a strength reduction process must
handled as follows: for associated plasticity the dilatancy be performed as described in Sloan (2013). This involves
angle ł9 is reduced incrementally in the same way as the several upper- and lower-bound analyses, each with different
friction angle 9, while for the non-associated case with strength parameters. Once a state is found where the com-
ł9 , 9, ł9 is kept constant as long as the reduced value puted collapse load matches the actual applied load, the limit
for 9 is larger than ł9. Once 9 falls to the value of ł9, strength parameters are derived.
both are then reduced simultaneously in subsequent itera-
tions. This is only relevant in the following for analyses
under drained conditions, where extreme cases have been EXAMPLE SLOPE STABILITY – DRAINED
considered rather than using values based on experimental CONDITIONS
evidence. Undrained analyses are generally performed with Problem description and meshes
ł9 ¼ 0. This first example to be analysed is that of a simple
To model non-associated plasticity in limit analysis, Davis homogeneous slope, shown in Fig. 1, with a slope height HS
(1968) suggested the use of reduced strength parameters, c equal to 10 m and a varying slope angle ÆS of 158, 308, 458
and , in combination with an associated flow rule of the and 608. Additionally, two depth factors DS ¼ 1 and DS ¼ 5,
form which quantify the distance to a firm base (bedrock), are
considered. The dimensions of the slope are identical to the
c ¼ c9 (2)
tan ¼ tan 9 (3)
Hs
where αs Refinement along
the slope line
cos ł9 cos 9
¼ (4) Additional cluster with
1 sin ł9 sin 9 finer mesh properties
Hs·Ds
Thus, in the following, all analyses denoted as the Davis
approach have c and as input parameters.
Davis argued that the flow rule will not have a major
influence on the ultimate limit load unless the problem is
kinematically constrained, and only for these situations
would his approach need to be applied. As discussed by Y
Sloan (2013), however, it is not straightforward to identify Total element number: 1542
such cases in practice, but it is generally accepted that in
slope stability analysis the flow rule should not have a X
significant influence on the computed factor of safety (e.g.
Cheng et al., 2007). This assumption will be studied later, Fig. 1. Homogeneous slope geometry and mesh properties of the
where it will be shown to be questionable for steep slopes homogeneous slope model for SRFEA analysis (DS 5 and
with high angles of friction. ÆS 458)
FINITE-ELEMENT LIMIT ANALYSIS VERSUS STRENGTH REDUCTION TECHNIQUES 251
case studied previously by Yu et al. (1998) who also used the strength reduction is performed in the finite-element
FELA. analysis. The material sets include two different soil types, a
Examples of the meshes used in the displacement finite- purely frictional material (no cohesion) and a cohesive-
element calculations and FELAs are shown in Figs 1 and 2. frictional material. Table 1 summarises the material param-
In FELA the mesh refinement is performed adaptively as eters common to both methods (FELA and SRFEA), while
part of the analysis, whereas in the SRFEA the mesh is fixed Table 2 gives the additional parameters that are required for
and a sensitivity analysis is required to ensure that it is the SRFEA analyses. In the latter, four different materials
sufficiently fine (see Fig. 3). These results suggest that about are specified to allow for an associated flow rule (9 ¼ ł9)
1500 15-noded triangles are needed to give accurate esti- and a non-associated flow rule with zero plastic volume
mates of the factor of safety. change (ł9 ¼ 0).
Slope inclination, SRFEA SRFEA (Davis) FELA LB FELA UB FELA ¼ Percentage difference ¼ Percentage difference Davis ¼
ÆS (UB + LB)/2 100(FELA-SRFEA)/ 100(FELA SRFEA(Davis))/
FELA FELA
A similar picture is obtained for the cohesive-frictional 458), which is of practical relevance in alpine environments
cases (material set 2 of Tables 1 and 2) which are sum- where steep slopes with an inclination of 408 or higher are
marised in Tables 5 and 6. Table 5 indicates that the factors quite common. For these cases where the factor of safety is
of safety for the slopes with an associated flow rule compare usually low, any disturbance of the slope due to construction
again well for both methods, with a maximum difference of activities has to be carefully examined and the assessment of
5%. For most of the examples, the slopes with associated a reliable factor of safety becomes crucial. A slope with a
flow rules and non-associated flow rules give similar factors height Hs of 10 m and an inclination ÆS of 458 is therefore
of safety. This confirms that, for these cases, the flow rule considered in this section, with effective strength parameters
does not have a major influence on the stability. Although of 9 ¼ 458 and c9 ¼ 6 kPa. The dilatancy angle in this study
the results from the non-associated SRFEA are not identical is varied between ł9 ¼ 08, 158 and 458, purposely covering a
to those from the FELA technique with the Davis approach range of extremes, with ł9 ¼ 158 being probably a realistic
(Table 6), the maximum difference is just below 9%, which value.
is considered to be acceptable. The failure mechanisms Figure 6 shows the finite-element mesh used for this case
obtained from the FELA and SRFEA methods are shown in which comprised 1157 15-noded triangles. Fig. 7 shows the
Figs 4 and 5 and are, as expected, very similar. factor of safety from the SRFEA, and it is immediately clear
that the flow rule does have a significant influence on the
factor of safety, with the associated case giving a value of
Note on influence of flow rule on factor of safety about 1 .54 and the zero dilatancy case (ł9 ¼ 08) giving a
It has been mentioned above that only a slight influence value of approximately 1 .32. It is also apparent that the non-
of the flow rule on the factor of safety was observed in associated case gives highly erratic results, which makes it
some cases when employing a strength reduction technique. impossible to infer a precise value for the factor of safety.
However, this effect appears to be more important for This type of behaviour is well known and is a result of the
materials with high friction angles (in the range of 9 ¼ 40– large difference between the friction angle and the dilatancy
Table 6. SRFEA (non-associated and Davis approach) and FELA (Davis approach) results for material set 2 – Ds 1
Slope inclination, SRFEA SRFEA (Davis) FELA LB FELA UB FELA ¼ Percentage difference ¼ Percentage difference Davis ¼
ÆS (UB + LB)/2 100(FELA SRFEA)/ 100(FELA SRFEA(Davis))/
FELA FELA
Fig. 4. SRFEA failure mechanism for DS 5, ÆS 458, material set 2_A (deformed mesh on the left
side and deviatoric strains on the right side)
FINITE-ELEMENT LIMIT ANALYSIS VERSUS STRENGTH REDUCTION TECHNIQUES 253
Fig. 5. FELA upper bound results for DS 5, ÆS 458, material set 2_A (deformed mesh on the left
side and power dissipation on the right side)
10·0 m 15·0 m
Point A
15·0 m
20·0 m
45°
10·0 m
35·0 m
Fig. 8. Failure mechanism for 9 ł9 458 (incremental devia-
toric strains)
Fig. 6. Finite-element model for SRFEA analysis
ł9 ¼ 158. In this case the flow rule is associated and there-
angle. Figs 8 and 9 compare the failure mechanisms for both fore the analysis is numerically stable, but the factors of
cases by means of contour plots of incremental deviatoric safety are significantly below those for the non-associated
strains. In the non-associated case the slip plane does not go flow rule. This suggests that, for cases with high friction
through the toe of the slope and, compared with the asso- angles, the Davis (1968) approach is a conservative alter-
ciated case, is less clearly defined. If a more reasonable native for calculating factors of safety by means of SRFEA.
dilatancy angle of ł9 ¼ 158 is assumed, the factor of safety The Davis approach, however, has the significant practical
increases to about 1 .47, which is again not too different advantage of removing numerical instabilities which may
from the associated case. Fig. 7 shows two more curves for lead to an incorrect assessment of the factor of safety.
the Davis approach with SRFEA, assuming ł9 ¼ 08 and Figure 7 also illustrates the respective factors of safety
1·6
FELA (φ ψ)
FoS 1·57
1·5
1·4
Factor of safety
FELA (Davis
approach ψ 15°)
FoS 1·31
1·3
1·2
1·0
0 0·2 0·4 0·6 0·8 1·0
Displacement of point A: m
SRFEA – Associated flow rule (φ ψ) SRFEA – Davis approach (ψ 15°)
SRFEA – Non-associated flow rule (ψ 15°) SRFEA – Davis approach (ψ 0°)
SRFEA – Non-associated flow rule (ψ 0°)
Fig. 7. Computed factors of safety with SRFEA analysis including results obtained with the
Davis (1968) approach
254 TSCHUCHNIGG, SCHWEIGER, SLOAN, LYAMIN AND RAISSAKIS
initial stresses are assumed as v ¼ ªz(ª ¼ 19 .0 kN/m3) and
h ¼ K0v with K0 ¼ 1 sin9. This means that K0 is different
for each effective stress analysis, but this is reflected in the
calculation of the equivalent undrained shear strength.
The face of the tunnel is stabilised with a pressure T and
its roof and floor are supported by a rigid, smooth lining.
Two different overburdens HT (8 m and 24 m) are considered,
being equivalent to a ratio of HT /DT ¼ 1 and HT /DT ¼ 3.
Two supporting pressures T ¼ 25 kPa and T ¼ 75 kPa are
assumed.
Figure 11 shows an example for a finite-element mesh used
for the SRFEA analysis consisting of 2260 15-noded trian-
Fig. 9. Failure mechanism for 9 458, ł9 0 (incremental gles, which has been chosen on the basis of a sensitivity
deviatoric strains) analysis. The mesh refinement technique implemented in the
FELA is demonstrated in Fig. 12, where the development
obtained from FELA. These values agree well with those from the initial mesh towards the final discretisation is shown.
obtained from SRFEA with associated flow.
Material parameters
EXAMPLE TUNNEL FACE STABILITY – UNDRAINED The undrained shear strength su increases with depth at
CONDITIONS different ratios rDT /su0, where r is the strength gradient
Problem description and meshes with depth and su0 is the undrained shear strength at the
This example considers the stability of a tunnel face under surface (Fig. 10). The undrained shear strength at the tunnel
plane strain conditions as shown in Fig. 10. The focus in this crown (su,cr) is the same for all analyses, but the undrained
example is to examine the differences in modelling undrained shear strength at the surface su0 is different for all calcula-
conditions (effective stresses as opposed to total stresses) and tions depending on the strength gradient r. The ratio be-
to compare the results from SRFEA and FELA. The tunnel has tween soil stiffness and undrained shear strength is assumed
an overburden depth of HT and a diameter DT. In this example to be Eu ¼ su 3 300. The limiting case with constant soil
strength and constant soil stiffness is also considered. Table
su0 7 summarises the input parameters for the analyses in terms
su: kPa of total stresses with an associated flow rule (ł9 ¼ 0) and is
valid for both SRFEA and FELA.
1 The undrained shear strength resulting from a Mohr–
HT ρ Coulomb criterion can be derived from the effective strength
parameters due to the fact that the centre of the Mohr’s
Rigid smooth lining 9 ) remains at the same position under undrained
circle ( m
loading. Table 8 gives the input parameters for the SRFEA
su,cr const.
in terms of effective stresses using the relation
19 þ 39
DT 8 m su ¼ c9 cos 9 þ sin 9
Supporting 2
pressure σT (5)
¼ c9 cos 9 þ m
9 sin 9
su(z) ¼ c9 cos 9 þ s9 sin 9
Rigid smooth lining su0 ρz
Fig. 11. Mesh properties for the SRFEA analysis (HT /DT 1)
FINITE-ELEMENT LIMIT ANALYSIS VERSUS STRENGTH REDUCTION TECHNIQUES 255
Fig. 12. Mesh refinement procedure for the tunnel face (FELA)
rDT /su0 0 .00 0 .40 1 .20 2 .00 4 .00 6 .00 8 .00 12 .00
HT /DT ¼ 1 su0: kPa 60 .80 43 .43 27 .64 20 .27 12 .16 8 .69 6 .76 4 .68
r: kPa/m 0 .00 2 .17 4 .15 5 .07 6 .08 6 .51 6 .76 7 .02
HT /DT ¼ 3 su0: kPa 182 .40 82 .91 39 .65 26 .06 14 .03 9 .60 7 .30 4 .93
r: kPa/m 0 .00 4 .15 5 .95 6 .51 7 .02 7 .20 7 .30 7 .39
rDT /su0 0 .00 0 .40 1 .20 2 .00 4 .00 6 .00 8 .00 12 .00
HT /DT ¼ 1 c9: kPa 60 .80 43 .75 28 .54 21 .37 13 .27 9 .67 7 .62 5 .36
9: degrees 0 .00 6 .99 14 .43 18 .48 23 .58 26 .06 27 .55 29 .25
HT /DT ¼ 3 c9: kPa 182 .40 85 .61 43 .03 29 .01 16 .08 11 .14 8 .53 5 .81
9: degrees 0 .00 14 .43 22 .86 26 .06 29 .25 30 .53 31 .22 31 .96
As the Lode angle changes during strength reduction, some and total stress analyses yield identical results (Fig. 14 for
differences between the two cases can be expected (Potts, T ¼ 25 kPa, Table 9 for T ¼ 75 kPa, HT /DT ¼ 1).
2005). However, when evaluating the change of the Lode
angle in the analysis it varied by only a few degrees; that is,
the influence on the calculated factor of safety is minor in CONCLUSIONS
this case. Also in the effective stress analyses the dilatancy Finite-element limit analysis provides rigorous upper and
angle ł9 ¼ 0, thus a non-associated flow rule is defined. The lower bounds on the factor of safety and can therefore
unit weight for all parameter sets is 19 .0 kN/m3. estimate the error in the solution (for the idealised material
adopted). Because displacement finite-element analysis is
increasingly used to calculate factors of safety by means of
Results the strength reduction technique (SRFEA), results from this
Figure 13 shows that the failure modes are similar for the method are compared with those from FELA in order to
different analyses. The factors of safety from SRFEA and prove that strength reduction techniques can be safely
FELA compare extremely well for the total stress analysis applied in practice. This has been shown to be the case by
(Fig. 14), whereas the SRFEA in terms of effective stress comparing the results for slope and tunnelling face stability
yields higher factors of safety as compared to the total stress problems when adopting an associated flow rule, an intrinsic
analysis, which will be explained in the following. assumption of limit analysis. In SFREA one would hardly
During SRFEA the strength parameters are reduced incre- adopt an associated flow rule; however, the influence of the
mentally until failure occurs. Figs 15 and 16 show the dilatancy angle on the calculated factors of safety obtained
Mohr’s circles at failure for both an effective stress and a from SRFEA is considered to be minor for slope stability
total stress analysis, indicating also in which terms the factor problems. Although true in many cases, it has been shown
of safety appears. Because the strength reduction in an in this paper that for high friction angles (.408) and steep
effective stress analyses applies to c9 and tan 9 simultane- slopes with low factors of safety, this is no longer the case
ously, there is clearly a difference in the definition of the and the flow rule may have a significant influence on the
factor of safety, which is not always appreciated in practice. results. Importantly, these cases also lead to numerical
This comes on top of the slight difference due to the change instabilities, making an accurate determination of influence
of the Lode angle as described above. of the flow rule difficult. It has therefore been investigated
Only in the limiting case of a constant su do the effective whether the approach suggested by Davis (1968), which
256 TSCHUCHNIGG, SCHWEIGER, SLOAN, LYAMIN AND RAISSAKIS
1·50
1·45
1·40
B
1·35
Factor of safety
1·30
1·25
B
1·20 SRFEA – Effective stress analysis
1·15 SRFEA – Total stress analysis
1·10 FELA – Upper bound (total stress analysis)
SRFEA (effective stress analysis) FELA – Lower bound (total stress analysis)
1·05
FELA – (LB UB)/2 (total stress analysis)
1·00
0 2·0 4·0 6·0 8·0 10·0 12·0
ρDT/su0
tan φ
FELA – Upper bound
te
l sta tan φ/FoS
Initia
r
state
c Limit su,limit
c/FoS
σ3,f σm σ1,f σ
Fig. 13. Failure mechanisms of the tunnel face (HT /DT 1 and
T 25 kPa) τ
Table 9. Factor of safety from different calculation methods (HT /DT 1 and T 75 kPa)
rDT /su0 SRFEAeff – effective SRFEAtot – total stress FELA – LB FELA – UB FELA ¼ Percentage difference ¼
stress analysis analysis (UB + LB)/2 100(FELA SRFEAeff)/FELA