Sie sind auf Seite 1von 12

09_016-027 Benz (1237).qxd:000-000 Benz (1237).qxd 30.07.

2009 13:25 Uhr Seite 16

Thomas Benz DOI: 10.1002/bate.200910038


Radu Schwab
Pieter Vermeer

Small-strain stiffness in geotechnical analyses

Nonlinear soil behaviour at small strains is often neglected in geo- In German literature, dynamic stiffness is often used as
technical analyses. Doing so often leads to an overestimation of a synonym for small-strain stiffness. This is presumably due
foundation settlements and retaining wall deflections. Settlement to the small strain amplitudes commonly used in dynamic
troughs behind retaining walls or above tunnels, on the other hand, testing, as static tests with local strain gauges reveal that
may be analysed as too flat and extended. Comparing measured inertia effects only slightly increase a soil’s very-small strain
displacements of piles or anchors within the working load range, stiffness, if it is increased at all (e.g. Stokoe et al. [30]).
to those calculated without considering small-strain stiffness, Small-strain stiffness is mostly found to be a manifold
shows a considerably too soft response. This paper is concerned of the stiffness obtained in classical laboratory testing.
with a qualitative and quantitative discussion of the small-strain Therefore, not accounting for it in geotechnical analyses may
stiffness phenomenon. In combination with the commercially potentially result in overestimating foundation settlements
available small-strain stiffness model introduced, it provides the and retaining wall deflections. The gradient of settlement
basics for incorporating small-strain stiffness into routine design.
troughs behind retaining walls or above tunnels may be
underestimated. Piles or anchors within the working load
range may show a considerably too soft response. The im-
1 Introduction portance of small-strain stiffness to the engineering prac-
tice has been highlighted, for example, in the Rankine lec-
The maximum strain, at which soils exhibit almost fully ture by Simpson [29], or the Bjerrum Memorial lecture by
recoverable behaviour, is found to be very small. The very Burland [7]. As yet however, small-strain stiffness has not
small-strain stiffness associated with this strain range, i.e. been widely implemented in engineering practice.
shear strains lower than 1 × 10–5, is believed to be a funda- Settlement analyses, which introduces a limit depth,
mental property of all types of geotechnical materials. represents an exception to this statement. The limit depth
With increasing strain, soil stiffness decays non-linearly. is defined as a depth, where the soil is only slightly subjected
On a logarithmic scale, stiffness reduction curves exhibit a to additional stress due to external loading. These analyses
characteristic S-shape similar to the one shown in Fig. 1. indirectly consider small-strain stiffness in introducing an
The borderline between small strains and larger strains, active compression zone and an incompressible zone be-
outlined in Fig. 1, is commonly drawn at the limit of classi- low the limit depth. The incompressible zone remains in
cal laboratory testing, that is triaxial or oedometer testing the range of small-strains and small-strain stiffness and
without special instrumentation as e.g. local strain gauges. hence, responds relatively stiffly (see Section 4.3).
Strains, at which soils exhibit almost fully recoverable or If in numerical settlement analyses, a constitutive
almost elastic behaviour, are termed very small strains. model is used which is not capable of simulating small-
strain stiffness, a similar approach as the one described
above has to be chosen: Either the lower mesh boundary is
chosen at an assumed limit depth or the stiffness of deeper
soil layers is manually increased. However, both methods
are dissatisfactory. A lower mesh boundary that is chosen
in a distance, where it will influence calculation results, is
generally to be avoided and more important the limit
depth is not known a priori. Manually increasing the stiff-
ness of deeper soil layers requires knowledge about the
strain level in these layers after loading, which is also not
known a priori. Therefore the only satisfactory means of
accounting for small-strain stiffness in numerical analysis
is to adapt the constitutive model. To date, such constitu-
Fig. 1. Characteristic stiffness-strain behaviour in logarith- tive models are rarely found in commercially available
mic scale according to [3]. codes.

16 © Ernst & Sohn Verlag für Architektur und technische Wissenschaften GmbH & Co. KG, Berlin · Bautechnik Special issue 2009 – Geotechnical Engineering
09_016-027 Benz (1237).qxd:000-000 Benz (1237).qxd 30.07.2009 13:25 Uhr Seite 17

Th. Benz/R. Schwab/P. Vermeer · Small-strain stiffness in geotechnical analyses

Constitutive models, that can be applied to engineering ably different though: Small-strain stiffness was first dis-
problems, should adequately simulate soil behaviour not covered in soil dynamics. Shear strain γ s is defined as:
only at small strains, but also at larger strains in the sense
of Fig. 1. The Hardening-Soil model [26], for example, is a
γs = 1 ⎡⎣(ε1 − ε2 )2 + (ε2 − ε3 )2 + (ε3 − ε1 )2 ⎤⎦ (1)
model that is often applied to geotechnical problems. The 2
Hypoplastic law [15] with intergranular strain [20], includ-
ing – in contrast to the Hardening-Soil model – a small- and therefore simplifies in triaxial conditions (ε2 = ε3) to
strain stiffness formulation, is less used in routine design. In γ s = |ε1 – ε3|.
fact, the use of a constitutive model in engineering prac- In the following, secant moduli are used to describe
tice demands user-friendly models with input parameters, experimental results. Initial small-strain shear and Young’s
which are easy to understand in their physical meaning, moduli are denoted as G0 and E0 respectively. Assuming
and easy to quantify, based on test data or experience. This that soils at very-small strains are elastic, the Young’s
paper presents a possibility to enhance such user-friendly modulus can be related to the shear modulus as follows:
elastoplastic formulations for non-linear stiffness variation
at small-strains. E0 = G0(1 + 2ν) (2)
Although the paper’s main concern is with the devel-
opment of a small-strain stiffness model, it also provides where ν is Poison’s ratio. Next, the shear modulus G0 is
information which might be found helpful during its appli- discussed in more detail. After that, the focus will be on
ance. Firstly, small-strain stiffness is discussed from an ex- the relationship between stiffness reduction and applied
perimental point of view, with a focus on empirical corre- shear strain.
lations. Secondly, the new model is introduced, before it
finally is applied to various boundary value problems with 2.1 Small-strain shear modulus G0
a particular emphasis on the impact of small-strain stiff-
ness on calculation results. Such an impact is to be ex- The small-strain shear modulus G0 in all types of soils is
pected in most problems, including those expected to mainly affected by void ratio e and mean stress p. Cementa-
cause larger strains (Fig. 1); during construction, small- tion, which is a very important influencing factor, too, will
strain stiffness is always to be considered. be discussed later. In literature, many experimental results
are presented in form of the modified Hardin [11] equation:
2 Soil stiffness at small strains
⎛ p ⎞m
G0 = A f (e) OCR k ⎜ ⎜ p ⎟⎟ , (3)
Small-strain stiffness is believed to be a fundamental property ⎝ ref ⎠
of geotechnical materials including clays, silts, sands, gravels,
and rocks (Tatsuoka [33]) under static and dynamic loading where A, k, and m are material constants, f gives the func-
(Burland [7]) and for drained and undrained loading con- tional dependency of G0 on void ratio, OCR is the over-
ditions (Lo Presti et al. [16]). With increasing strain, soil consolidation ratio, and pref is a reference pressure. Fig. 2
stiffness decays non-linearly. On a logarithmic scale, stiffness depicts the results of Equation 3 for more than 30 parame-
reduction curves exhibit a characteristic S-shape (Fig. 1). ter sets (A, k, m) for different soils published in literature.
Experimental evidence shows that after a load reversal the All references to the test data shown in Fig. 2 are provided
very small-strain stiffness is recovered. in [4]. It turns out, that m for cohesionless soils equals ap-
These experimental observations can easily be ex- proximately 0.5 and slightly increases in cohesive soils. In
plained with a simple Coulomb-type frictional law, assumed cohesive soils, G0 increases with OCR. Viggiani & Atkinson
to exist in-between the particles of granular material. Ini- [35], for example, found 0,20 < k < 0,25 for clays with plas-
tially, all inter-particle contacts are in a sticking mode. Due
to contact-elasticity, most contacts remain in the sticking
mode for a finite strain amplitude. This finite strain ampli-
tude can be considered to be equivalent to the quasi-elas-
tic very small-strain regime. A further increase of the shear
strain will then cause an increasing number of particles to
slip. The global material stiffness, a function of the stiffness
at all local contacts, thus decreases. Strain reversals allow
the soil skeleton’s inter-particle contacts to revert to stick-
ing mode so that the maximum small-strain stiffness can
be recovered.
Assuming all particles of the granular matter to be
rigid (psammic behaviour [8]), volumetric strains cannot
decrease stiffness as it does not change the ratio of normal
and tagential contact forces. Clearly, psammic behaviour
is, like the frictional contact model, an abstraction of real-
ity but shear strains actually influence small-strain stiff-
ness degradation to a larger extent than volumetric strains.
The reason, why stiffness reduction curves are commonly
presented in literature as function of shear strains, is prob- Fig. 2. Correlations between G0 and void ratio e

Bautechnik Special issue 2009 – Geotechnical Engineering 17


09_016-027 Benz (1237).qxd:000-000 Benz (1237).qxd 30.07.2009 13:25 Uhr Seite 18

Th. Benz/R. Schwab/P. Vermeer · Small-strain stiffness in geotechnical analyses

ample of a correlation between G0 and cone penetration


tip resistance.
The chart published by Alpan [1] shown in Fig. 4
which – as he stated – relates static to dynamic soil stiff-
ness (see axis labels Es, Ed), can provide estimates for the
small-strain modulus E0, too. Interpreting the dynamic
modulus Ed as the initial or very small-strain modulus E0,
and the static modulus Es as the apparent elastic modulus
in conventional soil testing, e.g. the secant modulus Eur in
larger triaxial unloading- reloading loops, E0 can be esti-
mated from triaxial test results. To the authors’ experience,
Alpan’s chart provides reasonable estimates for the stiff-
ness of soils, when interpreted in the way described above.

2.2 Modulus reduction – the reference shear strain γ 0,7

Fig. 5 shows exemplary stiffness reduction curves for sand


Fig. 3. Correlation between G0 and cone penetration tip and cohesive soils, which were originally published by
resistance after Lunne et al. [17] Seed & Idris [28] and Vucetic & Dobry [36], respectively.
According to these sources, stiffness reduction curves of
cohesive soils with IP < 15 % are comparable to those of
ticity Index 10 < IP < 40, which seems a relatively small sands. However, with increasing soil plasticity, soil stiff-
variation for practical applications. ness decays more gradually with applied strain.
As a rule of thumb, either the relationship by Hardin & In soil dynamics, experimentally acquired stiffness re-
Black [9]: duction curves were first approximated with mathematical
expressions. The best known models of soil dynamics are
(2,97–e)
2
⎛ p ⎞0,5 the Hardin-Drnevich model [10], the Ramberg-Osgood
G [MPa ] = 33
0 ⎜⎜ ⎟⎟ (4) model [23], and the bilinear model. Fig. 6 gives a compari-
1+ e ⎝ p ref ⎠
son of these models in a loading-unloading-reloading loop.
or the more simple expression by Biarez & Hicher [6]: The bilinear model gives only a very rough approximation
of real soil behaviour and is therefore not discussed in any
⎛ p ⎞0,5 more detail. Excellent approximations, on the other hand,
E0[MPa ] = 140 ⎜⎜ ⎟⎟ (5) are obtained with hyperbolic laws as used in the Ramberg-
e ⎝ p ref ⎠
Osgood model and in the Hardin-Drnevich model. The ba-
can be used to estimate small-strain stiffness moduli E0 sic difference between these models is, that they are formu-
and G0 in various soils. Here again, e is void ratio, p is the lated in stress space and strain space, respectively:
mean effective stress in kPa, and pref = 100 kPa is the ref-
G = 1
erence pressure equal to atmospheric pressure. The rela- κ
Ramberg-Osgood (6)
G0
tionship by Biarez & Hicher was proposed for soils with wl
< 50%, whereas the one by Hardin & Black was derived
1+ α τ
τy
for undisturbed clayey soils and crushed sands.
The scatter band highlighted in Fig. 2 suggests, that G = 1
these correlations should be used only if no direct tests are Hardin-Drnevich (7)
G0 γ
available. In this case, small-strain shear moduli can be 1+ s
derived from indirect test results, too. Fig. 3 gives an ex- γr

Fig. 4. Alpan chart: Original publication [1] (left) and its interpretation after [5] with test results by [37] (right)

18 Bautechnik Special issue 2009 – Geotechnical Engineering


09_016-027 Benz (1237).qxd:000-000 Benz (1237).qxd 30.07.2009 13:25 Uhr Seite 19

Th. Benz/R. Schwab/P. Vermeer · Small-strain stiffness in geotechnical analyses

Fig. 5. Stiffness reduction curves after Seed & Idris [28] (left) and Vucetic & Dobry [36] (right)

t t t
G0 G0 G0
G0
1 1 1 tf 1
ty ty

gy g gy g gr g

a b c
Fig. 6. One-dimensional models known from soil dynamics: (a) Bilinear model, (b) Ramberg-Osgood model,
(c) Hardin-Drnevich model

where G0, α, κ, τ y are the material parameters of the Ram- impact on γ 0,7. Vucetic & Dobry [36] proposed the modu-
berg-Osgood model and G0, γ r are the material parameters lus reduction chart shown in Fig. 5. Stokoe et al. [31] pro-
of the Hardin-Drnevich model. Hence, the Hardin-Drnevich posed a linear dependency of γ 0,7 on the plasticity index
model is well suited for practical application not only be- (IP) of the form:
cause of its data fitting qualities, which are often pointed
out in literature, but also because of its simple structure γ0,7 = (γ0,7)ref + 5 × 10–6 IP (OCR)0,3, (9)
requiring two input parameters only.
Next, the hyperbolic Hardin-Drnevich model is adopted where (γ 0,7)ref is the reference shear strain for IP = 0, which
for quantifying stiffness reduction curves. For this, the ref- is about 1 × 10–4. Compared to the chart published by
erence shear strain γ r, introduced above, is replaced by the Vucetic & Dobry, the correlation by Stokoe et al. suggests
reference shear strain γ 0,7 = 3γ r/7: lower reference shear strains for higher plasticity indices.
In all soils, the reference shear strain γ 0,7 is addition-
G = 1 ally influenced by confining stress: The reference shear
. (8)
G0 3 γs stress increases with confining stress.
1+
7 γ0,7
2.3 In-situ tests
The reference shear strain γ 0,7 is the shear strain at which
the shear modulus G has been reduced to 70 % of its in- Comparisons of laboratory and in-situ small-strain stiff-
itial value G0. In the following, this reference shear strain ness measurements for a given material mostly yield
is utilized to quantify the importance of several influen- smaller stiffness in the laboratory. Fig. 7 shows experimen-
cing factors on the decay of small-strain stiffness. tal results from studies conducted in Japan and in the
The influence of void ratio on the reference shear strain USA. Correlations derived from laboratory tests generally
in sand is very limited. At a reference pressure of 100 kPa, indicate the lower limit of in-situ stiffness. When adopting
it is tested within the limits of 6 × 10–5 < γ 0,7 < 3 × 10–4 (see modulus reduction curves derived in the laboratory to in-
Fig. 5). Soil plasticity, on the other, hand has a significant situ measured shear moduli G0, they are usually scaled in

Bautechnik Special issue 2009 – Geotechnical Engineering 19


09_016-027 Benz (1237).qxd:000-000 Benz (1237).qxd 30.07.2009 13:25 Uhr Seite 20

Th. Benz/R. Schwab/P. Vermeer · Small-strain stiffness in geotechnical analyses

Modulus ratio G0,Lab/G0,Field


0.10 0.25 0.50 0.80 1.00 1.50 2.00

In-situ shear wave velocity vS,Field[m/s]


2 0
Holocene sand
Modulus ratio G0,Lab/G0,Field

(thin-wall sampling) 150


In-situ frezzing
300
method Remolded Range from
450 cemented ROSRINE
Soft rock
1 Holocene clay Pleistocene clay Gravel sandy soils study
(block) 600
Pl
e
(tu isto
be ce Soft rock 750
sa ne (rotary core)
General trend
m sa
pl n
in d 900 Range often found
g)
with rock cores
0 1050
10 100 1000 0.0 0.5 1.0 1.5
G0,Field [MPa] Velocity ratio vS,Lab/vS,Field

Fig. 7. Differences in the ratio of laboratory-to-field stiffness. Left: Data from Japanese case studies after Toki et al. [34].
Right: Results from the US American ROSRINE study after Stokoe & Santamarina [32]

stiffness only [32]. Thus, the reference shear strain γ 0,7 is reloading cycles as a consequence. Therefore, their elastic
not believed to be much affected by diagenesis. stiffness is generally taken as a secant stiffness of larger
unloading-reloading loops (see Fig. 8). Small-strain stiffness
3 Small-strain stiffness models is neglected. Unlike the stress-strain path sketched in Fig. 8,
unloading-reloading loops, which are based on a constant
Small-strain stiffness was not applied to static problems secant stiffness, show no hysteretic behaviour.
until the 1980’s. Among the first small-strain stiffness models A straight forward method to incorporate small-strain
were models that used one or more kinematic yield surfaces stiffness in elastoplastic frameworks is to define their elastic
e.g. Mroz et al. [19]. The idea of a very small inner yield stiffness as a function of the actual strain amplitude. At small
surface (bubble) was invented by Al-Tabbaa in 1987. It was strain amplitudes, the large strain secant unloading-reload-
afterwards that Al-Tabbaa & Wood [2] published their bubble ing stiffness is increased. The MIT S1 model [21] is an ex-
extension of the CamClay model. In his 1992 Rankine lec- ample, where strain dependent elastic moduli are imple-
ture, Simpson [29] made the analogy between soil behaviour mented. In contrast to the MIT S1 model, the approach
and a man pulling bricks behind him. Simpson’s analogy, proposed here is based on a generalization of the Hardin-
together with his interpretation in strain space, became Drnevich model:
well known by the term Simpson brick model. Within the
Hypoplastic framework, the intergranular strain concept G = 1 H* e
[20] was invented. The previously discussed models from mit γHist = 3 (10)
G0 γ e
soil dynamics were applied to static problems, too. In multi- 1 + 3 Hist
axial constitutive models, these one-dimensional models 7 γ0,7
need to be extended. An example for a multi-axial formu-
.
lation is given by the MIT S1 model [21]. where e is the devaitroric strain rate and H* is a tensor which
As yet, however, small-strain stiffness has not widely memorizes deviatoric strain history [4]. In proportional
been implemented in routine design. Considering numerical loading, H* equals the integral of deviatoric strain rates, so
analyses, this may be due to the complexity of the available
modeling concepts summarized above and/or the fact that
they lack some substantial features of real soil behaviour. q
Models used in engineering practice today, rarely have more qf
than one elastic domain. The Cam-Clay or the Hardening
Soil models are two well known examples of this model
class. A possibility to incorporate small-strain stiffness in
them is discussed in this section. The proposed model ex- E0
tension is based on a multi-axial formulation of the 0,5 qf
E0
Hardin-Drnevich model.
E0
3.1 A small-strain stiffness extension for elastoplastic models E50

It is impossible for small-strain stiffness to be associated with


the entire elastic domain of isotropically hardened elasto- 0
Eur eaxial
plastic models as e.g. the Cam-Clay, or the Hardening Soil
model, as they would respond too stiff in larger unloading- Fig. 8. The elastic secant modulus Eur

20 Bautechnik Special issue 2009 – Geotechnical Engineering


09_016-027 Benz (1237).qxd:000-000 Benz (1237).qxd 30.07.2009 13:25 Uhr Seite 21

Th. Benz/R. Schwab/P. Vermeer · Small-strain stiffness in geotechnical analyses

that Equation 10 simplifies to Equation 7. In non-propor- Given that a functional form exists which can describe
tional loading, however, H* can be partially or fully reset the initial loading curve, the above rules can likewise be
according to the following transformation rule: applied to construct the hysteresis loops of soils for symmet-
⎛T rical or periodic loadings. However, if the loading is irreg-
0 0 ⎞
* −1 ⎜ 11 ⎟ ular, i.e. not symmetrical or periodic, the current loading or
H = T HT mit T = ⎜0 T 22 0 ⎟ und unloading curve may intersect a previous one. Fig. 9 shows
⎜0 0 T 33 ⎟⎠
⎝ three possible extensions of the original Masing rules, that

( )
1 ⎡ ⎤ were proposed in literature: Path A: Rosenblueth & Her-
T11 = ⎢⎣1 + u(λ H11 ) H11 + 1 − 1 ⎥⎦
(1)
rera [25], path B: Jennings [14], path C: Richart [24]).
H11 + 1
All extended Masing rules require an internal mem-
T22 = 1
H22 + 1

⎢1 + u(λ H22 ) H22 + 1 − 1 ⎦⎥

(1) ⎤
( ) ory of previous changes in the stress-strain path, i.e. a loci
in stress space. Such an internal memory in stress space is

( )
⎡ ⎤ (11) not a practicable constitutive ingredient for the extended
T33 = 1
⎢⎣1 + u(λ H33 ) H33 + 1 − 1 ⎥⎦
(1)
Hardin-Drnevich model: In combination with an elasto-
H +1 33 plastic model, the memory would be highly influenced by
plastic yielding, so that the paraelastic stiffness could only
where u(x) is the Heaviside function and λ(i) are the be determined in an iterative scheme. The yield loci of a
Eigenvalues of the deviatoric strain rate. Underlined ten- combined isotropically hardened elastoplastic model can
sorial quantities are transformed in the strain rate’s Eigen- enforce the memory in stress space instead. The resulting
system. Poison’s ratio is assumed to be constant. stress-strain path is equivalent to that proposed by Richart
The resulting stress-strain law is not continuous in strain (Fig. 9, path C).
(load reversals), so that by definition it does not qualify as
an elastic stress-strain law. Nevertheless, strain on a closed 3.3 HS-Small – a small-strain extension of the
stress cycle starting from a stress reversal point is typically Hardening Soil model
recovered. In the following, the new stress-strain law is
therefore denominated paraelastic (Hueckel & Nova [13]). The combination of the multiaxial small-strain formula-
However, the terminology “elastic stiffness” is here main- tion, introduced above, with the Hardening Soil model is
tained for the use of the above defined stiffness in an elasto- discussed next. In the finite element code Plaxis, this com-
plastic framework as a reminder that it applies to all stress bination is referred to as HS-Small model. Its stress and
states, within as well as on the yield surface, similar to the strain history dependent elastic stiffness is controlled by
elastic stiffness known from traditional elastoplastic models. two additional material parameters. These are the initial
shear modulus Gref,0 defined for the reference pressure pref
3.2 Primary loading, unloading, and reloading – and the shear strain γ 0,7, at which the shear modulus has
the Masing rules decreased to 70 percent of its initial value. The shear mod-
ulus G is calculated from:
To systematically describe the behaviour of brass under cyclic
loading, Masing [18] proposed the following two rules: 7γ0,7 ⎛ σ + c cot ϕ ⎞m
G = G0, ref ⎜⎜ 3 ⎟⎟ , (12)
1. The shear modulus in unloading is equal to the initial 7γ0,7 + 3γHist ⎝ p ref + c cot ϕ ⎠
tangent modulus for the initial loading curve. 2. The shape
of the unloading and reloading curve is equal to the initial where σ 3 is the minor principal stress and c cot ϕ is the max-
loading curve, except that its scale is enlarged by a factor imum allowable isotropic tensile stress, which can be limited
of two. with an additional input parameter. The exponent m gives

1.0 t/ty 1.0 t/ty C

0.8 0.8 1
A B
0.6 0.6
3
0.4 0.4

0.2 0.2
g/ gy g/ gy
-8.0 -6.0 -4.0 -2.0 2.0 4.0 6.0 8.0 -8.0 -6.0 -4.0 -2.0 2.0 4.0 6.0 8.0
-0.2 4 -0.2

-0.4 -0.4

-0.6 -0.6

-0.8 2 -0.8
a b 0
-1.0 -1.0

Fig. 9. Hysteresis loops in symmetric (left) and irregular (right) loading according to the extended
Masing rules after Pyke [22]

Bautechnik Special issue 2009 – Geotechnical Engineering 21


09_016-027 Benz (1237).qxd:000-000 Benz (1237).qxd 30.07.2009 13:25 Uhr Seite 22

Th. Benz/R. Schwab/P. Vermeer · Small-strain stiffness in geotechnical analyses

the stress dependent increase of stiffness. A single exponent 4 Model application


m is used for scaling all stiffness moduli in the model. Poi-
son’s ratio is assumed constant at all strain amplitudes. All problems discussed in this section are analyzed using
A lower cut-off limit in the hyperbolic small-strain the finite element method. A more detailed description of
stiffness reduction curve is introduced at the shear the various problems including material data sets, model
strain, where tangent stiffness is reduced to the unload- geometry, loading history etc. is given in Benz [4]. In or-
ing-reloading stiffness Gur in larger strain cycles. From der to identify effects that can be attributed to small-strain
then on, the small-strain formulation is inactive. It is ac- stiffness, all calculations were run with and without small-
tivated again whenever a significant change in strain rate strain stiffness formulation, i.e. with the HS-Small model
is detected (Equation 11). The resulting small-strain stiff- and the Hardening Soil model, respectively. In the spread
ness behaviour of the HS-Small model is illustrated in footing example, presented in Section 4.3, the Hypoplastic
Fig. 10. model with intergranular strain is considered as an alter-
Stress relaxation erases a soil’s memory of previous native small-strain model.
applied stress. Soil ageing in the form of particle (or as-
sembly) reorganization during stress relaxation and for- 4.1 Triaxial test
mation of bonds between them can erase a soil’s strain
history. Considering that the second process in a naturally In triaxial testing, small-strain stiffness does not exessively
deposited soil develops relatively fast, the strain history affect the overall stress-strain curve. Fig. 11 illustrates this,
should start from zero in most boundary value problems. with a drained triaxial test on dense Hostun sand and its
This is the default setting in the HS-Small model. numerical back calculation. Only in up-scaling the very
However, sometimes an initial strain history may be first part of the stress-strain curve on the right hand side
desired. In this case, the strain history can be triggered of Fig. 11, the differences become visible.
by applying an extra load step before starting the actual
analysis. Such an additional load step might also be 4.2 Excavation problem 2D
used to model overconsolidation. Usually the overcon-
solidation’s cause has vanished long before the start of Simpson introduced his brick model in a case study of the
calculation, so that the strain history should be reset af- British Library deep excavation. The Houses of Parliament
terwards. underground car park case study is another deep excava-

Fig. 10. Stiffness-reduction curve of the HS-Small model. Left: Elastic secant shear modulus. Right: Elastic tangent
shear modulus

s1/s3 s3 = 300 kPa CD GSecant [kN/m2]

160000
4
120000

Experiment
80000
2
Hardening-Soil
40000
HS-Small
0 0
0.00 0.02 0.04 0.06 0.08 e1[-] 0.0001 0.001 0.01 e1-e3[-]

Fig. 11. Triaxial test on dense Hostun sand

22 Bautechnik Special issue 2009 – Geotechnical Engineering


09_016-027 Benz (1237).qxd:000-000 Benz (1237).qxd 30.07.2009 13:25 Uhr Seite 23

Th. Benz/R. Schwab/P. Vermeer · Small-strain stiffness in geotechnical analyses

tion study, that is often cited along with small-strain stiff- Fig. 12, and the soil parameters are taken from this refer-
ness. The tilt of the Big Ben clock tower right next to this ence solution. However, the bottom soil layer defined by
excavation nicely illustrates the effects of advanced (small- Schweiger could be omitted in the analysis when using the
strain stiffness) models. These can predict the total exca- HS-Small model. In the reference solution, the only pur-
vation heave more correctly and thus do not suggest that pose of this layer is the simulation of small-strain stiffness
the clock tower tilts away from the pit, which less refined due to a lack of small-strain stiffness constitutive models
models may do. Beyond the constitutive model however, back then.
there are a lot more things to consider in the analysis of Figs. 13, 14, and 15 show results from the finite element
deep excavations, for example: the interface between soil calculations. The small-strain stiffness formulation accu-
and retaining structure, initial stress conditions, structural mulates more settlements right next to the wall, whereas the
elements, time effects, and pore water pressures. Some of settlement trough is smaller. The triple anchored retaining
these issues should always be thoroughly considered in wall is deflected less when using the HS-Small model
excavation analysis (interface, initial stresses, structural el- (Fig. 13). Excavation heave is considerably less (Fig. 14)
ements), others are less critical under less difficult soil and the elastic stiffness expressed as ratio G/Gur reduces a
conditions. An example with less critical soil conditions is lot around the excavation pit (Fig. 15). In the example G0
chosen here: An excavation in Berlin sand. equals 3 Gur.
The working group 1.6 „Numerical methods in Geo-
technics“ of the German Geotechnical Society (DGGT) 4.3 Spread footing problem
has organized several comparative finite element studies
(benchmarks). One of these benchmark examples is the The next boundary value problem considered is a strip
installation of a triple anchored deep excavation wall in footing on dense Hostun sand, first published by Hintner
Berlin sand. The reference solution by Schweiger [27] is et al. [12]. The strip footing has a width of 1 meter, a thick-
used here as the starting point: Both, the mesh shown in ness of 1 meter and an embedded depth of 1 meter. The

Fig. 12. Geometry and anchor detail

Fig. 13. Surface settlement trough (left) and lateral wall deflection (right) at the end of excavation

Bautechnik Special issue 2009 – Geotechnical Engineering 23


09_016-027 Benz (1237).qxd:000-000 Benz (1237).qxd 30.07.2009 13:25 Uhr Seite 24

Th. Benz/R. Schwab/P. Vermeer · Small-strain stiffness in geotechnical analyses

0
3.0
20

40

G/Gur [-]
2.0
60

80
1.0
100

-20 0 20 40 60 80 100 120


Fig. 14. Vertical displacements in section A-A at the end of Fig. 15. Ratio of the elastic moduli at the end of excavation
excavation

Fig. 16. Calculated settlements (right) and ratio G/Gur (left) at p = 150 kPa/m

vertical load of 150 kPa is distributed over the width of the to the excavation pit. Fig. 18 shows the geometry, struc-
footing. The homogeneous soil layer’s material parameters tural features, and excavation stages modeled within the
are identical to those back calculated from the triaxial test finite element analysis. The structural features include two
discussed in Section 4.1. In the analysis, three different con- abutments on pile foundations, back-anchored and strut-
stitutive models are employed: The original Hardening ted retaining walls, sheet pile walls, and a floor slab in the
Soil model, the HS-Small model, and the Hypoplastic model strutted section. Embedded elements are used to model
with intergranular strain. All models were calibrated using the floor slab, sheet pile walls and piles. All other struc-
triaxial and oedometer test data. For details on the calibra- tural features and the subsoil are discretized by volume el-
tion of the constitutive models the reader will refer to [12]. ements. Four excavation stages are introduced: One be-
The results of the strip footing case study are illustrated fore wall construction (pre-excavation), and three after
in Fig. 16. Apparently, both models incorporating small-
strain stiffness give similar results. They both predict lower
settlements than the Hardening Soil model. As the vertical
displacements with depth are decaying much faster in the
small-strain models, the assumption of a limit depth for the
active compression zone, at the point where the additional
vertical stress due to external loads is not more than twenty
percent of the effective vertical pressure of the overburden
(e.g. German code of practice), seems justified. The ratio
G/Gur plotted in Fig. 16 indicates elastic stiffness at the
end of loading. In the example G0 equals 3 Gur.

4.4 Excavation problem 3D

The HS-Small model is deployed last in a 3D finite ele-


ment analysis of an interaction between the excavation pit
for the lock Sülfeld (Fig. 17) and two railway bridge abut- Fig. 17. Strutted and back-anchored Sülfeld excavation near
ments which are located in a distance of approximately 15 m the railway bridge

24 Bautechnik Special issue 2009 – Geotechnical Engineering


09_016-027 Benz (1237).qxd:000-000 Benz (1237).qxd 30.07.2009 13:25 Uhr Seite 25

Th. Benz/R. Schwab/P. Vermeer · Small-strain stiffness in geotechnical analyses

Fig. 18. Geometry of the Sülfeld excavation in the 3D FE model

wall construction (excavation step 1 to 3). Total excava-


tion depth is 14 m. Construction of the retaining walls is
not modeled accurately. Instead, they are placed as a
whole. The phreatic level is adjusted several times in the
calculation, due to groundwater discharge in the bedrock
or groundwater lowering in the fill. After the final excava-
tion step, the groundwater table is lowered due to works 0.0
near the pumping station. This change in phreatic level is
the final load step considered in the analysis.
A two-step model calibration procedure was employed: -5.0
First, all laboratory test results were statistically averaged
uz - Hardening-Soil
before secondly, selected tests were back analyzed in numer-
ical element tests. Properties of the sand layer were esti- -10.0
mated from in-situ heavy dynamic penetration test results.
The obtained material parameters were finally calibrated
in a 2D finite element model by comparing calculated dis-
-15.0
placements of the existing lock during operation to meas-
uz [mm]
ured ones. In the calibration procedure, the confidence in
the correlated small-strain stiffness parameters could be
specifically increased.
The Sülfeld analysis’ main focus is on displacements
of the railway bridge abutments and their backfill. The well uz - HS-Small
equipped site allows for various other comparisons, as well.
Fig. 19. Vertical displacements of the bridge abutments (here
From extensometer readings in the excavation pit, the final
shown with backfill)
heave during excavation step 3 can be quantified as 2–3 mm.
In the analysis, heave < 4 mm is obtained when using the
HS-Small model. The HS analysis over-predicts the excava- additionally derived that the abutments are not tilting. Ex-
tion heave by more than 8 mm. The measured settlement actly the same behaviour is found in the HS-Small analysis:
(extensometer) of the abutment closest to the excavation pit Almost no tilt of the abutments occurs with a total settle-
is shown in Fig. 19. From geodetic measurements, it can be ment of approximately 2 mm. The HS calculation on the

Bautechnik Special issue 2009 – Geotechnical Engineering 25


09_016-027 Benz (1237).qxd:000-000 Benz (1237).qxd 30.07.2009 13:25 Uhr Seite 26

Th. Benz/R. Schwab/P. Vermeer · Small-strain stiffness in geotechnical analyses

other hand shows settlements of up to 10 mm in combina- [9] Hardin, B. O., Black, W. L.: Closure to vibration modulus of
tion with tilt. normally consolidated clays. ASCE: Journal of the Soil Me-
In summary, the small-strain stiffness formulation in chanics and Foundations Division 95 (1969) SM6, 1531–1537.
the HS-Small model could be successfully deployed in the [10] Hardin, B. O., Drnevich, V. P.: Shear modulus and damping
in soils: Design equations and curves. ASCE: Journal of the
Sülfeld 3D finite element analysis. It significantly improved
Soil Mechanics and Foundations Division 98 (1972) SM7,
all results obtained from the Hardening Soil model. As in
667–692.
the previous calculations, the improvement is achieved with [11] Hardin, B. O.: The Nature of Stress-Strain Behaviour of
zero strain history at the onset of loading. Soils. Proc. Earthquake Engineering and Soil Dynamics,
Pasadena, Vol. 1, 1978, 3–90.
5 Summary [12] Hintner, J., Vermeer, P. A., Baun, C.: Advanced FE versus
classical settlement analysis. Proc. NUMGE06, Graz, 2006,
The hyperbolic Hardin-Drnevich model was generalized 539–546.
to multi-axial strain space, so that it can easily be com- [13] Hueckel, T., Nova, R.: Some hysteresis effects of the behav-
bined with many existing elastoplastic constitutive models ior of geological media. International Journal of Solids and
that are based on isotropic elasticity and hardening. In Structures 15 (1979) 8, 625–642.
[14] Jennings, P. C.: Periodic Response of a General Yielding
fact, these are the kind of models most commonly used in
Structure. ASCE: Journal of the Engineering Mechanics Divi-
engineering practice today. One of these is the Hardening
sion 90 (1964) EM2, 131–166.
Soil model, as implemented in the finite element code [15] Kolymbas, D.: An outline of hypoplasticity. Archive of Ap-
Plaxis V8. The small-strain stiffness enhanced version of plied Mechanics 61 (1991), 143–151.
the Hardening Soil model, introduced in this paper, is [16] Lo Presti, D. C. F., Jamiolkowski, M., Pallara, O., Caval-
named HS-Small model. laro, A.: Rate and Creep Effect on the Stiffness of Soils. In:
Application of the HS-Small model is straightforward. Sheahan, Kaliakin (Hrsg): Measuring and Modeling Time
Its Small-Strain Overlay extension requires only two addi- Dependent Soil Behavior. ASCE Geotechnical Special Publi-
tional parameters. These are the initial small-strain stiffness cation, 1996, 166–180.
G0 and the threshold shear strain γ 0.7 at which stiffness is [17] Lunne, T., Robertson, P. K., Powell, J. J. M.: Cone Penetra-
reduced to 0,7 G0. In the case of no small-strain stiffness tion Testing in Geotechnical Practice. London: E & FN Spon,
1997.
in-situ or laboratory test data being available, these two
[18] Masing, G.: Eigenspannungen und Verfestigung beim
parameters can be correlated with various other soil or test
Messing. Proc. 2nd Int. Congr. Appl. Mech., Zürich, 1926.
parameters. [19] Mroz, Z., Norris, V. A., Zienkiewicz, O. C.: An anisotropic
The HS-Small model was applied in various bound- critical state model for soils subject to cyclic loading.
ary value problems. In these, all effects that are commonly Géotechnique 31 (1981) 4, 451–469.
attributed to small-strain stiffness in soil-structure interac- [20] Niemunis, A., Herle, I.: Hypoplastic model for cohesionless
tion could be recognized: The width and shape of settlement soils with elastic strain range. Mechanics of Cohesive-Fric-
troughs, for example, are more precisely modeled and ex- tional Materials 2 (1997) 4, 279–299.
cavation heave is reduced to a realistic value. The overall [21] Pestana, J. M., Whittle, A. J.: Formulation of a unified con-
reliability of numerical displacement analysis is consider- stitutive model for clays and sands. Int. J. Numer. Anal. Meth.
ably increased. Analysis results are also less sensitive to 23 (1999), 1215–1243.
[22] Pyke, R.: Nonlinear soil models for irregular cyclic load-
the choice of proper boundary conditions. Large meshes
ings. ASCE: Journal of the Geotechnical Engineering Divi-
no longer cause extensive accumulation of displacements,
sion 105 (1979) GT6, 715–726.
because marginally strained mesh parts are very stiff. [23] Ramberg, W., Osgood, W. R.: Description of stress-strain
curve by three parameters. Technical Note 902, National Ad-
References visory Committee for Aeronautics, Washington, DC, 1943.
[24] Richart Jr., F. E.: Some effects of dynamic soil properties
[1] Alpan, I.: The geotechnical properties of soils. Earth-Sci-
on soil structure interaction. ASCE: Journal of the Geotech-
ence Reviews 6 (1970), 5–49. nical Engineering Division 101 (1975) GT12, 1193–1240.
[2] Al-Tabbaa, A., Muir Wood, D.: An experimentally based [25] Rosenblueth, E., Herrera, I.: On a kind of hysteretic damp-
‘bubble’ model for clay. Proc. NUMOG III, Vol. 1, 1989, 91–99. ing. ASCE: Journal of the Engineering Mechanics Division 90
[3] Atkinson, J.H., Sallfors, G.: Experimental determination of (1964) EM4, 37–47.
soil properties, Proc. 10th ECSMFE, Florence, Vol. 3, 1991, [26] Schanz, T., Vermeer, P. A., Bonnier, P. G.: The hardening
915–956. soil model – formulation and verification. In: Brinkgreve, R.
[4] Benz, T.: Small-strain stiffness of soils and its numerical con- (Hrsg.): Beyond 2000 in computational geotechnics, Rotter-
sequences. Dissertationsschrift. Mitteilung 55 des Instituts dam: Balkema, 1999, 281–296.
für Geotechnik, Universität Stuttgart, 2007. [27] Schweiger, H. F.: Influence of soil parameters on numerical
[5] Benz, T., Vermeer, P. A.: Zuschrift zu: Über die Korrelation analysis of a deep excavation. Proc. Int. Symp. Ident. and
der ödometrischen und der dynamischen SteiFig.keit nicht- Det. of Soil and Rock Para. for Geot. Design, Paris, 2002,
bindiger Böden. Bautechnik 84 (2007) 5, 361–364. 573–580.
[6] Biarez, J., Hicher, P-Y.: Elementary Mechanics of Soil Be- [28] Seed, H. B., Idriss, I. M.: Soil moduli and damping factors
haviour. Rotterdam: A.A.Balkema, 1994. for dynamic response analysis. Report 70-10, EERC (Berke-
[7] Burland, J. B.: Small is beautifull – the stiffness of soils at ley, California), 1970.
small strains, 9th Laurits Bjerrum memorial lecture. Canadian [29] Simpson, B.: Retaining structures: displacement and de-
Geotechnical Journal 26 (1989), 499–516. sign. The 32nd Rankine Lecture. Géotechnique 42 (1992) 4,
[8] Dietrich, T.: A comprehensive mechanical model of sand at 541–576.
low stress level. Proc. Speciality Session 9, 9th ICSMFE, Tokyo, [30] Stokoe, K. H., Darendeli, M. B., Andrus, R. D., Brown, L.
1/A/12, 1977, 33–43. T.: Dynamic soil properties: laboratory, field and correlation

26 Bautechnik Special issue 2009 – Geotechnical Engineering


09_016-027 Benz (1237).qxd:000-000 Benz (1237).qxd 30.07.2009 13:25 Uhr Seite 27

Th. Benz/R. Schwab/P. Vermeer · Small-strain stiffness in geotechnical analyses

studies. Proc. 2nd Int. Conf. on Earthquake Geotech. Eng., [36] Vucetic, M., Dobry, R.: Effect of Soil Plasticity on Cyclic
Vol. 3, 1999, 811–845. Response. Journal of Geotechnical Engineering 117 (1991) 1,
[31] Stokoe, K. H., Darendeli, M. B., Gilbert, R. B., Menq, F-Y, 89–107.
Choi, W. K.: Development of a new family of normalized [37] Wichtmann, T., Triantafyllidis, T.: Abschluss der Diskussion
modulus reduction and material damping curves. Int. Work- zu: Über die Korrelation der ödometrischen und der dy-
shop on Uncertainties in Nonlinear Soil Properties and their namischen SteiFig.keit nichtbindiger Böden. Bautechnik 84
Impact on Modeling Dynamic Soil Response, Berkeley, 2004. (2007) 5, 364–366.
[32] Stokoe, K. H., Santamarina, J. C.: Seismic-wave-based test-
ing in geotechnical engineering. Proc. GeoEng 2000: An In-
ternational Conference on Geotechnical and Geological En-
gineering, Melbourne, Vol. 1, 2000, 1490–1536.
[33] Tatsuoka, F., Shibuya, S., Kuwano, R.: Advanced Laboratory Authors:
Stress-Strain Testing of Geomaterials. Rotterdam: Balkema, Prof. Dr.-Ing. Thomas Benz, Geotechnical Division, Dept. of Civil and
2001. Transport Engineering, NTNU, Høgskoleringen 7a, N-7034 Trondheim,
[34] Toki, S., Shibuya, S., Yamashita, S.: Standardization of lab- Norway, thomas.benz@ntnu.no
oratory test methods to determine the cyclic deformation Dr.-Ing. Radu Schwab, Federal Waterways Engineering and Research
properties of geomaterials in Japan. In: Shibuya, S., Mitachi, Institute, Kußmaulstraße 17, D-76187 Karlsruhe, Germany,
T., Miura, S. (Hrsg.): Pre-failure Deformations of Geomateri- radu.schwab@baw.de
als, Vol. 2, 1995, 741–784. Prof. Dr.-Ing. Pieter A Vermeer, Institute of Geotechnical Engineering,
[35] Viggiani, G., Atkinson, J. H.: Stiffness of fine grained soils University Stuttgart, Pfaffenwaldring 35, D-70569 Stuttgart, Germany,
at very small strains. Géotechnique 45 (1995) 2, 249–265. pieter.vermeer@igs.uni-stuttgart.de

Bautechnik Special issue 2009 – Geotechnical Engineering 27

Das könnte Ihnen auch gefallen