Professional Documents
Culture Documents
The IES Journal Part A: Civil & Structural Engineering
The IES Journal Part A: Civil & Structural Engineering
To cite this article: P.L. Teo & K.S. Wong (2012) Application of the Hardening Soil model in deep excavation analysis, The IES
Journal Part A: Civil & Structural Engineering, 5:3, 152-165, DOI: 10.1080/19373260.2012.696445
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
The IES Journal Part A: Civil & Structural Engineering
Vol. 5, No. 3, August 2012, 152–165
TECHNICAL PAPER
Application of the Hardening Soil model in deep excavation analysis
P.L. Teo* and K.S. Wong
WKS Geotechnical Consultants, 60 Kallang Pudding Road #02-00 Tan Jin Chwee Industrial Building, 349320, Singapore
(Received 6 January 2012; final version received 16 April 2012)
The Mohr–Coulomb model (MC) is commonly used in deep excavation analysis for its simplicity. However, it
has shortcomings that may produce unrealistic soil behaviour. The Hardening Soil model (HS) is an advanced
soil model that is able to generate more realistic soil response in terms of non-linearity, stress dependency and
inelasticity. This article highlights some of the shortcomings of the MC model and presents a simplified
approach to determine the HS parameters for drained and undrained analysis of deep excavations. Three case
studies were back-analysed to validate the application of the HS model for practical excavation analysis. The
HS model suffers the same problems as the MC model in using effective stress parameters c0 and f0 to determine
Downloaded by [Princeton University] at 10:02 18 August 2014
Keywords: Hardening Soil model; Mohr-Coulomb soil model; excavation; constitutive model; nonlinearity; back-
analysis
1. Introduction It has two yield surfaces as shown in Figure 1(a). The first
The Mohr–Coulomb model (MC) is commonly used in one deals with yielding due to shear stress. The second
practice despite of its many shortcomings (Wong 2011). one handles the expansion of the cap due to changes in
The Hardening Soil model (HS) is an advanced soil mean effective stress p0 . Figure 1(b) shows the yield
model that can better simulate real soil behaviour. It is a surfaces in three-dimensional principle stress space.
big improvement over the MC model in several aspects. The hardening law for shear is given in Equation
This article presents an overview of the HS model and (2). For a given shear stress increment, the plastic shear
highlights its strengths over the MC model in excavation strain increment can be computed from this equation
analysis. Back-analysis of deep excavation case records and the corresponding plastic volumetric strain incre-
in Singapore are presented to compare the predicted wall ment can be computed from the flow rule in Equation
and ground movements between both soil models and (3). The hardening law for the cap is given in Equation
against measured data. This article also discusses short- (4). For a given increment in mean stress, the plastic
comings of the HS model. volumetric strain increment can be computed using
this equation with no shear strain generated.
2. The Hardening Soil model 1 q 2q
fs ¼ gp 0 ð2Þ
2.1. An overview E50 1 q=qu Eur
The HS model is an elastic-plastic soil model based on Depv ¼ sin cm Dgp ð3Þ
the classical plasticity theory (Brinkgreve et al. 2004).
For every stress increment, there is a corresponding Dp Dp 1
incremental elastic and plastic strain if the soil is Depv ¼ ¼ Dp ð4Þ
Kc Ks H
undergoing primary loading or only elastic strain if it is
under unloading-reloading. The main attraction of this where Kc is related to the compression index Cc and Ks
model is its ability to simulate the nonlinear, inelastic is related to the recompression index Cr.
and stress dependent behaviour of soil.
The model adopts the MC failure criterion.
0 2.2. Parameters of the Hardening Soil model
2c cos f0 þ 2s03 sin f0 The required soil parameters are summarised in Table
ðs1 s3 Þf ¼ ð1Þ
1 sin f0 1. Most of the soil parameters can be determined from
consolidated drained triaxial compression test (CD) modulus E50 can be determined from each sample. By
and consolidation test. In the absence of consolidation c0 cos f0 þs0 sin f0
plotting ln(E50) versus ln ðc0 cos f0 þpref3 sin f0 Þ on a natural
test, the parameter Eoedref can be set equal to E50ref. If
scale and fitting a straight line through the data, the y-
the unloading–reloading cycle is not carried out in the
intercept gives the value of ln(E50ref) and the slope
CD test, the parameter Eurref can be set equal to
gives the value of the parameter m. An illustration is
3E50ref.
shown in Figure 2(a). The parameter Eoedref can be
Once these parameters are known, the modulus of
determined from consolidation test. If the preconsoli-
a soil under any stress condition can be computed
dation pressure sc0 is not greater than 100 kPa, Eoedref
using the following equations in Table 2.
is the tangent modulus at s10 ¼ 100 kPa as illustrated
The parameter E50ref can be determined from the
in the stress–strain curves in Figure 2(b). If sc0 is
CD test. Assuming that the test is carried out on three
greater than 100kPa, Eoed can be determined at any s10
samples under different confining pressures, the secant
greater than 100 kPa as illustrated in Figure 2(c). The
corresponding Eoedref can be computed using the
equation in Table 2 (column 3).
Downloaded by [Princeton University] at 10:02 18 August 2014
Figure 2. Illustration on determination of: (a) E50ref and m; (b) and (c) Eoedref.
The IES Journal Part A: Civil & Structural Engineering 155
Proposed value/correlation
Parameter Sand Clay
c0 (kPa) 0 cu
f0 (8) 28 þ 0.15 Dr (%) 0
E50ref 6 e0.025 Dr (%) 0.8 Eu
(MPa) (Eu based on
Eu/cu ratio)
Eoedref E50ref E50ref
(MPa)
Eurref 3 E50ref 3 E50ref
(MPa)
m 0.45 þ 0.003 Dr (%) 0 m 1 (for f0 ¼ 0)
nur 0.2 0.2
c(8) f0 – 308 0
pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi
Downloaded by [Princeton University] at 10:02 18 August 2014
0
Figure 3. Plot of E50ref against relative density. Ko Ko ¼ (1 – sinf ) OCR Ko ¼ (1 – sinf0 ) OCR
Figure 6. (a) Case study to illustrate the importance of nonlinearity attribute; (b) response computed by MC model and (c)
response computed by HS model.
Figure 7. Rigid footing in sand (a) load-settlement curves; and (b) stress contours of vertical displacement.
divided into layers to simulate the increase in the initial in stress and the stress-path involved. While the MC
modulus with depth. During excavation, the modulus model produced larger wall deflection, toe movement,
remained unchanged. The modulus of a HS soil bottom heave and ground settlement, the HS model
however, changed continuously according to the change produced a more reasonable response.
The IES Journal Part A: Civil & Structural Engineering 157
Figure 9. Case study of excavation in sand (a) finite element model; (b) wall deflection; (c) ground settlement and (d) ground
heave.
3.4. MC model cannot generate the correct 1D 3.5. MC model may produce an incorrect response
compression behaviour under certain stress paths
The nonlinear stiffness of real soil is evident from There are certain stress paths where an elastic soil will
consolidation test. The stiffness increases with stress as produce an incorrect response. Figure 11 illustrates one
the soil becomes more compact. The MC model of them where a soil sample is subjected to a
produces a linear response as shown in Figure 10 simultaneous reduction in s30 and (s10 7 s30 ). An elastic
because it does not have a cap yield surface. The HS MC soil sample rebounded upwards whereas the real
model is able to capture the nonlinear behaviour soil actually moved downwards. The elastic-plastic HS
because of the presence of cap yield surface. model is able to simulate the correct response.
158 P.L. Teo and K.S. Wong
3.6. MC model under-estimates the horizontal stress in issue in the HS model. The Poisson’s Ratio only affects
certain stress paths the elastic strain but not the plastic strain. The unloading-
Consolidation tests were simulated to compare the reloading Poisson’s Ratio value of real soil is typically
horizontal stresses generated by the MC and HS models between 0.15 and 0.25. A value of 0.2 is commonly used
and the empirical relationship Ko,OC ¼ Ko,nc OCR0.5. to simulate the near elastic response during soil unloading
Figure 12 shows that the MC underestimated the (more reduction in sv0 compared to sh0 ).
horizontal stress during loading and unloading. The HS
gave better agreement with the empirical relationship.
4. Case studies
Back-analysis of the three case studies of deep
3.7. MC model results may be sensitive to Poisson’s excavations in Singapore was carried out using the
Ratio in a drained analysis finite element (FE) program Plaxis version 9.02. They
For some problems, results can be very sensitive to the
Poisson’s Ratio when using the MC model. Figure 13
shows an example of excavation in sand. By varying the
Poisson’s Ratio from 0.2 to 0.4, the wall deflection
doubled and strut forces increased by up to 1.5 times. The
Downloaded by [Princeton University] at 10:02 18 August 2014
Figure 10. Stress–strain curves for one-dimensional Figure 12. Comparison of horizontal stresses computed
consolidation. using empirical relationship, the MC and the HS models.
Figure 11. Behaviour in a drained triaxial test subjected to reduction in s30 and (s10 7 s30 ).
The IES Journal Part A: Civil & Structural Engineering 159
Figure 14. Cross-section of excavation at the Rochor Complex (Lim et al. 2003, Halim and Wong 2005).
Figure 15. Comparison of results from analysis using the MC and HS models at the final excavation level of Rochor Complex
(a) and (b) wall deflection; (c) ground settlement and (d) bottom heave.
160 P.L. Teo and K.S. Wong
Figure 16. Comparison of plastic points of the Rochor Complex at final excavation level for (a) HS model and (b) MC model.
Downloaded by [Princeton University] at 10:02 18 August 2014
Figure 17. Soil profile and cross section of excavation at the Lavender Station.
Figure 18. Comparison of results from analyses using the MC and HS models at the final excavation level of Lavender Station:
(a) and (b) wall deflection; (c) ground settlement and (d) bottom heave.
Figure 19. Soil profile and cross-section of excavation at Hougang–Buangkok cut-and-cover tunnel.
162 P.L. Teo and K.S. Wong
were the Rocher Complex, the Lavender MRT Station sheetpiles and three levels of preloaded struts. The
and the Hougang-Buangkok cut-and-cover tunnel. ground water level was at 1.5 m below ground level. A
Results of analysis using the MC and the HS model cross-section of the excavation and the soil profile are
were compared with the measured data. The HS shown in Figure 14 (Lim et al. 2003, Halim and Wong
parameters were determined according to section 2. 2005).
The total stress approach (Method C) was used in the In the FE model, the Upper Marine Clay and
MC analyses. The effective stress approach with Lower Marine Clay layers were divided into sub-layers
specified cu (Method B) was adopted in the HS analyses. (2–3 m thick) with increasing undrained shear strength
and stiffness. The HS stiffness parameters E50ref, Eoedref
and Eurref were determined for each Eu/cu ratio.
4.1. The Rochor complex Results shown in Figure 15 indicate that the Eu/cu
The Rochor complex excavation was 95-m wide and ratio of 250 for the HS model and 300 for the MC
6.3-m deep. The excavation was retained using model produced a reasonable match with the measured
Table 5. Parameters used in the Hougang–Buangkok cut-and-cover tunnel excavation case study.
weight
Soil type (kN/m3)
cu (kPa) f0 (deg) c0 (kPa) E50ref ¼ Eoedref (MPa) Eurref (MPa) m E0 (MPa) Eu (MPa)
Fill 19 – 33 0.1 14.4 3 E50ref 0.555 15.6 –
E 13 15 – 0.8 cu 0 – 300 cu;
Eu ¼ 250 cu; 350 cu;
300 cu 400 cu
F1 19 – 31 0.1 9.9 0.510 10.7 –
OA1 20 – 31.5 0.1 11.2 0.525 12.1
OA2 20 135 – 0.8 cu 0 – 300 cu;
OA3 20 400 Eu ¼ 250 cu; 350 cu;
OA4 20 600 300 cu 400 cu
OA5 20 900
Figure 20. Comparison of results from analysis using the MC and HS models at the final excavation level of Hougang–
Buangkok cut-and-cover tunnel (a) deflection; (b) ground settlement and (c) bottom heave.
The IES Journal Part A: Civil & Structural Engineering 163
Downloaded by [Princeton University] at 10:02 18 August 2014
Figure 21. Comparison of wall deflections from analyses using MC and HS models (a) and (b) deflection and (c) strut force.
Figure 24. Prediction of cu/p0 ratio under (a) different OCR and (b) different test conditions.
It is interesting to note that from tests on real soil (4) The HS model produced smaller toe movement
and simulations on HS soil, the moduli obtained from than the MC model.
the simple shear, triaxial extension and triaxial (5) The HS model predicted smaller bottom heave
compression tests are different. Test results also show than the MC model.
that the strength and modulus obtained under plane (6) The HS model generated lesser plastic points
strain condition are also different from those obtained because the model is able to simulate the softer
from triaxial test. Therefore, it should be reminded soil behaviour as the soil approaches failure.
that the parameters obtained from the conventional The MC model generated more plastic points
UU, CU, CD and consolidation tests can only be used which gave a false impression on the extent of
as a crude approximation. soil yielding.
data. Highlights from the back-analyses are sum- 27–29 Nov 2003. Singapore, 83–94.
marised below. Halim, D. and Wong K.S., 2005. Evaluation of modified cam
clay parameters for deep excavation analysis. In:
(1) For the HS model, the application of Eu/cu Proceedings of the conference on underground Singapore,
1–2 Dec 2005. Singapore, 188–200.
ratio of 250 to the clayey soils seemed to work
Li, W., 2001. Braced excavation in old alluvium in Singapore.
well for these three case studies. Thesis (PhD). School of Civil and Environmental
(2) To achieve a similar agreement with the Engineering, Nanyang Technological University,
measured deflection using the MC model, it Singapore.
would be necessary to vary the Eu/cu ratio, as in Wong, K.S. and Duncan, J.M., 1974. Hyperbolic stress–
strain parameters for nonlinear finite element analyses of
the case of the three back-analyses, to 300 or
stresses and movements in soil masses. Geotechnical
400. Engineering Report. Berkeley: Department of Civil
(3) The HS model produced a more realistic Engineering, University of California.
ground settlement profile. The consideration Wong, K.S., 2011. Things you should know about the Mohr–
of nonlinear and inelastic stiffness in the HS Coulomb Model. In: Seminar on infrastructure in soft
ground – challenges and solutions. 16 Aug 2011, ed.
model gave a better prediction of the settlement
Singapore.
near the excavation, as shown in the Hougang–
Buangkok cut-and-cover tunnel case study.