Professional Documents
Culture Documents
This document provides description and theoretical background for the newly developed FLAC3D
constitutive model – Plastic Hardening. The document contains the list of properties and key-
words used by the model and a number of single zone and benchmark examples that illustrate
model usage, its features, and comparison with other models and lab/test data. The last section is
devoted to the description of material parameter calibration techniques. A list of references con-
cludes this manual.
Contents
1.0 INTRODUCTION
The new Plastic Hardening (PH) model is a shear and volumetric hardening constitutive model for
the simulation of soil behavior. When subject to deviatoric loading, soils usually exhibit a decrease
in stiffness, accompanied by irreversible deformation. In most cases, the plot of deviatoric stress
versus axial strain obtained in a drained triaxial test may be approximated by a hyperbola. This
feature was discussed by Duncan and Chang (1970) in their well-known “hyperbolic-soil” model,
which is formulated as a non-linear elastic model.
The new PH model is formulated within the framework of hardening plasticity (Schanz et al.,
1999), allowing it to remove the main drawbacks of the original non-linear model formulation
(e.g., detection of loading/unloading pattern, nonphysical bulk modulus).
Einstein notation (summation over repeated indices) is adopted in the document. Stresses are as-
sumed to be effective stresses; tension is taken to be positive, if not explicitly stated. For principal
stress, the assumed order is .
2.0 THEORY
∆ ∆
∆ 2 ∆ ,
where p is the mean pressure defined as /3, is the volumetric elastic strain defined as
, and and are the deviatoric stress tensor and deviatoric elastic strain tensor, respec-
tively. K and G are the elastic bulk and shear moduli, which can be derived from the unloading-
reloading Young’s modulus, Eur, and the elastic unloading-reloading Poisson’s ratio, , using the
relations
3 1 2
2 1
In the PH model, the Poisson’s ratio is assumed to be a constant parameter with a typical value of
0.2 (if otherwise not provided at input), while the Young’s modulus is a stress-dependent param-
eter:
∙ cot
∙ cot
PH model also employs another stiffness measure, , which defines the shape of the primary
shear hardening surface and takes the following power law:
∙ cot
∙ cot
Here is a material parameter, which could be estimated from multiple sets of triaxial com-
pression tests with various cell stresses.
2 2
0
1 2 sin
∙ cot
1 sin
The failure ratio should have a value smaller than 1, and 0.9 is chosen for most cases.
The ultimate deviatoric stress still abides by the MC failure law (Figure 1).
qa
qf
E50
axial strain
Figure 1. Hyperbolic stress-strain relation in primary shear loading.
∆ ∆ ∆ ∆
Due to the increase of , the shear yield surface will expand, but the ultimate surface is still the
conventional MC failure surface.
The PH model uses the following flow rule between plastic volumetric and shear strain:
Δ sin Δ
where is the mobilized dilation angle, which is distinct from the user-defined ultimate dilation
angle . The Rowe (1962) law used for the mobilized dilation angle is
sin sin
sin , sin sin
1 sin sin
sin sin
sin , sin sin
1 sin sin
where is a parameter denoting the scale factor of contraction, with an allowable range of 0 to
0.25 and a default value of 0.0 for most soils. The critical state friction angle is defined as
sin sin
sin
1 sin sin
sin
2 ∙ cot
As in the case for MC flow rule, a non-associated flow rule is proved to be consistent to the above
relation of the plastic volumetric and shear strain measurement. The shear potential function is
defined as
1 sin 1 sin
2 2
In order to avoid over-dilatancy when soil reaches its critical void state with , the dilation
angle needs a minor modification. One way proposed by Schanz et al (1999) is to set a cut-off rule,
so that
sin 0, if
The dilation rules with cut-off and the smoothing technique are compared with the case without
dilation cut-off in Figure 2.
No Cutoff
Cut‐off
Smoothing
Volumetric Strain
Axial Strain
Figure 2. Volumetric strain curve for a standard triaxial compression test with dilation cut-off
and smoothing.
where is an internal variable derived from other material parameters; is a shear stress measure
defined as 1 , and 1 sin / 1 sin . The hardening pa-
rameter , which denotes the pre-consolidation pressure, can be determined using the initial stress
state and an input material parameter ocr, so that
, ∗
If ocr has a very large value, the model nearly becomes a no-cap model.
The associated flow rule is adopted for volumetric hardening, which means that the potential vol-
umetric function is assumed to be the same as the volumetric yield function.
∙ cot
Δ Δ
∙ cot
where , again, is an internal parameter that can be derived from other material parameters.
Instead of taking and as input material parameters, another two parameters, and are
required as input. denotes normal consolidation coefficient and stands for the tangent
oedometer stiffness at the reference pressure . If is not provided by the user, the default
is calculated as 1 .
where is the tension limit. By default, is zero and user can provide value up to the upper
limit / tan .
Basic Properties:
Secant stiffness at 50%
of the ultimate devia- Material parameter determining the
E50_ref
toric stress when stress-strain behavior before failure.
1
Advanced property (y/n)
Notes: The example FISH function for FLAC3D to assign sig1, sig2, and sig3:
def iniprin
loop foreach local _z z_list
local pp = z_pp(_z)
z_prop(_z, 'sig1') = z_sig1(_z) + pp
z_prop(_z, 'sig2') = z_sig2(_z) + pp
z_prop(_z, 'sig3') = z_sig3(_z) + pp
endloop
end
@iniprin
Data- and project files for the examples in Sections 4-6 are provided separately.
Figure 3 shows deviatoric stress versus axial strain for both PH and MC models. It is easy to verify
from the figure that:
(1) The ultimate failure deviatoric stresses are the same for both models, as expected;
(2) For the pre-failure curve, PH and MC models are crossing at the half of the failure stress, which
is consistent with the concept of stiffness;
(3) The unloading stiffness in the MC model is the same as the loading stiffness while these stiff-
nesses are different in the PH model.
The resulting plots of mean stress versus axial strain are presented in Figure 4 and show the non-
linear behavior (corresponding to power law) of the cap hardening surface.
Figure 4. Mean stress vs. axial strain for dense, medium, and loose sands in an isotropic compression.
Figure 5. Drained triaxial deviatoric stress versus axial strain for dense, medium and loose sands.
Figure 6. Volumetric strain versus axial strain for dense, medium and loose sands.
A plot of deviatoric stress versus axial strain is shown in Figure 5. As expected, the plots show a
hyperbolic behavior. The unloading-reloading paths are also shown in the figure. The plot of vol-
umetric strain versus axial strain is shown in Figure 6. The dilatancy of the denser sands is clearly
present. The smooth decrease of the dilation angle when the void ratio is approaching the critical
state is distinct due to introduced dilation smoothing technique (compare with Figure 2).
Initially the model is in equilibrium with an isotropic stress state in each zone, 0.1 kPa.
Oedometer compression follows for 1000 steps. The results of stress ratio evolution due to com-
pression are shown in Figure 7, and it is seen that they correctly reproduce the expected evolution
path. Figure 8 presents vertical oedometer pressure versus vertical strain, which reproduces the
expected oedometer stiffness at the reference vertical pressure of 100 kPa.
The test is repeated this time using predefined values of 0.7, 0.6, and 0.5. The results for
are plotted in Figures 9 and 10. Again, correct paths and oedometer stiffness at the reference
vertical pressure are successfully reproduced.
Figure 7. path calculated from the oedometer test with friction angles of 30, 35 and 40 de-
grees and default values.
Figure 8. Vertical pressure versus vertical strain from the oedometer test with friction angles of
30, 35 and 40 degrees and default values.
Figure 9. path calculated from the oedometer test with friction angles of 30, 35 and 40 de-
grees and specified values.
Figure 10. Vertical pressure versus vertical strain from the oedometer test with friction angles of
30, 35 and 40 degrees and specified values.
The problem is modeled again using the PH model. The cohesion, friction, and dilation angles are
kept the same as for the MC model. In order to avoid possible numerical instability, a zero-degree
friction angle is replaced by 0.01 deg, which has insignificant impact on the final results. The
material parameters are summarized in Table 6 and the dimensions and boundary conditions for
the problem can be found in Example 3 of FLAC3D Verification Problems manual (velocity mag-
nitude is reduced to 10 m⁄step). Three cases, ocr = 100, 2, and 1, are considered for the PH
model. A high value of ocr = 100 approximately represents the effect of no cap hardening.
Table 6. Material parameters for the problem of a strip footing on cohesive frictionless material
Property \ Model PH MC
3
Density, kg/m 1500 1500
Young’s Modulus, MPa - 257
Poisson’s ratio 0.2857 0.2857
Friction angle, deg 0.01 0.01
Dilation angle, deg 0 0
Cohesion, MPa 0.1 0.1
Tension limit, MPa 107 107
E50_ref, MPa 257 -
Eoed_ref, MPa 205.6 -
p_ref, MPa 0.1 -
Rf 0.9 -
m 0.9 -
Knc 0.6 -
ocr 100, 2, 1 -
First, the initial equilibrium stress state under gravitational loading is calculated using the SOLVE
elastic command. After this calculations are proceeded using the MC and PH models. For the PH
model, the effective principal stresses are assigned using FISH function presented in Section 3.
The velocity fields at the collapse load are plotted in Figure 12 for the MC model and in Figure 13
for the PH model with ocr = 1. The velocity fields at the collapse load are very close for these two
models.
As it can be seen from Figure 14, the MC and PH models predict approximately the same bearing
capacity, which is expected as the PH model adopts the MC failure surface. However, the pre-
failure regions of the load-displacement curves are quite different. Two points can be noted on the
results. First, when using the PH model, the resulting load-displacement curves are smooth, while
same curve has distinct kinks when using the MC model. Second, using lower ocr values in the
PH model leads to lower initial tangent stiffness of the load-displacement curve.
This example illustrates the advantage of using the PH model when accurate predictions of the
pre-failure behavior is crucial for the problem, even though both MC and PH models predict sim-
ilar ultimate bearing capacity.
Figure 12. Velocity field at the collapse load for the MC model.
Figure 13. Velocity field at the collapse load for the PH model with ocr = 1.
Stage 1 – Initialize stress state, including groundwater table, 3 m below soil surface.
Stage 2 – Activate diaphragm wall and lower water level to -17.90 m in the pit.
Stage 3 – Excavation step 1 (to level -4.80 m).
Stage 4 – Activate anchor row 1 at level -4.30 m and pre-stress anchors.
Stage 5 – Excavation step 2 (to level -9.30 m).
Stage 6 – Activate anchor row 2 at level -8.80 m and pre-stress anchors.
Stage 7 – Excavation step 3 (to level -14.35 m).
Stage 8 – Activate anchor row 3 at level -13.85 m and pre-stress anchors.
Stage 9 – Excavation step 4 (to level -16.80 m).
The zone profile and the structural elements are shown in Figure 16. The soil is modeled with the
PH model and material properties are provided in Table 7. The wall is model by linear elastic liner
2
This problem is also solved using the Cysoil model in FLAC Example Application manual, Example 17
structural elements. The anchors are modeled by cable structural elements. The connection be-
tween the liner and the cables is through beams with rigid links representing anchored breasting
beams. Material and geometrical properties for the structural elements are provided in Table 8 and
in Figure 15.
The calculated wall deflection is plotted in Figure 17 to compare with the measured data (Schwei-
ger, 2002). A good match is observed between the FLAC3D simulation results and the measured
data.
The calculated surface settlement profile behind the wall is presented in Figure 18. It is seen from
the figure that the PH model predicts realistic lifting of the soil behind the wall. The axial force in
the anchors after the final excavation is shown in Figure 19.
Figure 15. Problem geometry and excavation stages (after Schweiger, 2002).
3
Adapted from Plaxis Material Models Manual (2002)
Itasca Consulting Group, Inc. Page 24 www.itascacg.com
Minneapolis, Minnesota (612) 371-4711
Plastic Hardening Model: Theory and Examples 7/1/2015
Ref. 8506 J. Cheng
Wall Deflection (m)
‐0.04 ‐0.02 0 0.02 0.04
0
‐5
‐10
‐15
Depth (m)
PH model
MC Model
‐20
Measured
‐25
‐30
‐35
Figure 17. Comparison of wall deflection between the field data and results from the PH and
MC models at the final excavation stage.
Itasca Consulting Group, Inc. Page 25 www.itascacg.com
Minneapolis, Minnesota (612) 371-4711
Plastic Hardening Model: Theory and Examples 7/1/2015
Ref. 8506 J. Cheng
Figure 19. Axial forces in the anchors after the final excavation.
It is notable to compare the simulation results between the PH and MC models. For this, the prob-
lem is solved using the MC model in which the Young’s modulus is taken equal to E50 of the PH
model and other parameters, including the friction and dilation angles, cohesion, Poisson’s ratio,
and tension limit are same as for the PH model (see Table 7). The comparison of the results is
presented in Figures 17 and 20-21.
Analysis of Figures 17 and 20-21 reveals the following advantages of the PH model over using the
MC model in the current example:
a) The MC model predicts wall deflection in the positive direction at the top portion of the
excavation, which is unrealistic. The deflection predicted by the PH model is in negative
direction and matches field data (Figure 17).
b) The PH model predicts expected ground settlement behind the wall (Figures 18, 20), while
the MC model predicts unrealistic ground lifting.
c) The PH model predicts more realistic lifting for the excavation base at the final stage of
excavation, while the MC model predicts over-lifting (Figure 21).
0.06
PH model
0.04 MC Model
Settlement (m)
0.02
‐0.02
0 10 20 30 40 50 60
Distance from the wall (m)
Figure 20. Comparison of the ground settlements calculated by the PH and MC models at the
final excavation stage.
0.3
PH model
MC Model
Settlement (m)
0.2
0.1
0
0 5 10 15 20 25 30
Distance from the plane of symmetry (m)
Figure 21. Comparison of the excavation base lifting calculated by the PH and MC models at the
final excavation stage.
The data are based on three sets of triaxial compression tests with confining pressures of 1.2, 0.6,
and 0.3 MPa. The initial void ratios are 0.783, 0.786, and 0.781, respectively.
3.5
2.5
2
q (MPa)
1.5
0.5
0
0 1 2 3 4 5 6
Axial Strain (%)
Figure 22. Original curves for the confining pressures of 1.2, 0.6, and 0.3 MPa of the tri-
axial compression tests of Monterey Sand (Lade, 1972).
‐0.8
‐0.7
‐0.6
‐0.5
Volumetric Strain (%)
‐0.4
‐0.3
‐0.2
‐0.1
0 1 2 3 4 5 6
0
Axial Strain (%)
0.1
0.2
0.3
Figure 23. Original curves for confining pressures of 1.2, 0.6, and 0.3 MPa of the triax-
ial compression tests of Monterey Sand (Lade, 1972).
y = 1.403x
R² = 0.9999
q
0 1 2 3
p
Figure 24. Determination of friction angle and cohesion from the triaxial compression test data.
Figure 25 plots the curves of / vs. / using triaxial compression test lab data of the Monte-
rey Sand with confining pressures 1.2, 0.6, and 0.3 MPa. The slopes of these lines are , the
intercepts are 100/ (as strain is given in %).
This figure determines three pairs of ( , ), which are summarized in Table 9. The average
is 0.957. Parameter needs no calibration and its value is assumed to be 0.1 MPa. Finally, plot
parameters ln / vs. ln , as shown in Figure 26 (remember that cohesion 0). The
slope of the trend line in Figure 26 determines 0.707 and the intercept determines
exp 4.63 102.5 MPa.
y = 0.9558x + 0.2459
R² = 0.9997
e1/q
y = 0.9678x + 0.1286
R² = 0.9997
y = 0.9476x + 0.093
R² = 0.9997
0 1 2 3 4 5 6
1/qf
Figure 25. Determination of and from three sets of triaxial compression tests with three
different confining stresses.
ln(E50)
y = 0.707x + 4.6302
R² = 0.971
0 1 2 3
ln(3/pref)
Figure 26. Determination and from three sets of triaxial compression tests with three dif-
ferent confining stresses.
6.3 Calibration of
For the example of Monterey Sand, is determined to be 320.0 MPa. If the unloading-reloading
moduli are not available, a value in the range of 3 5 can be used for most soils. The
PH model uses a default value of 4 ∙ if no input is provided for .
Summary of all material properties determined for the Monterey Sand is provided in Table 10.
Using these parameters, the triaxial compression tests can be reproduced by the PH model. The
results presented in Figures 27 and 28 reveal close match of simulated results and lab test data.
Figure 27. Deviatoric stress vs. axial strain for consolidated drained triaxial compression tests on
fine Monterey Sand.
Figure 28. Volumetric strain vs. axial strain for consolidated drained triaxial compressor tests on
fine Monterey Sand.
7.0 REFERENCES
Duncan, J.M. and Chang, C.Y. (1970). “Nonlinear analysis of stress and strain in soil”, J. Soil
Mech. Found. Div. 96(5), pp.1629-1653.
Lade, P.V. (1972). “The stress-strain and strength characteristics of cohesionless soils”, Ph.D.
Thesis, University of California, Berkeley, Berkeley, CA.
Rowe, P.W. (1962). “The stress-dilatancy relation for static equilibrium of an assembly of particles
in contact”, Proc. Roy. Soc. A. 269(1339), pp.500-527.
Schanz, T., Vermeer, P.A., Bonnier, P.G. (1999). “The hardening soil model: formulation and
verification”, in Beyond 2000 in Computational Geotechnics -10 Years of Plaxis, R.B.J.
Brinkgreve, Ed. Rotterdam: Balkema.
Schweiger, H.F. (2002) “Results from numerical benchmark exercises in geotechnics”, in 5th Eu-
ropean Conference Numerical Methods in Geotechnical Engineering, Vol. 1, pp. 305–314. Presses
de l’ENPC/LCPC, Paris.
Terzaghi, K. and Peck, R. B. (1967). “Soil Mechanics in Engineering Practice”, 2nd Ed., New
York: John Wiley and Sons.