You are on page 1of 6

Numerical Methods in Geotechnical Engineering – Schweiger (ed.

)
© 2006 Taylor & Francis Group, London, ISBN 0-415-40822-9

Simulation of a large excavation using a Hoek-Brown model

Thomas Benz
University of Stuttgart & Federal Waterways Engineering and Research Institute, Karlsruhe, Germany

Regina Kauther & Radu Schwab


Federal Waterways Engineering and Research Institute, Karlsruhe, Germany

ABSTRACT: The Hoek-Brown criterion is widely used in practice to evaluate the strength of rock masses. It
starts from the properties of intact rock and introduces factors to reduce these properties on the basis of joint
characteristics within the rock mass. In this paper, the 2002 edition of the Hoek-Brown criterion is used as the
failure envelope of a nonlinear elastic-perfectly plastic constitutive model. The nonlinear elastic part of the pro-
posed model takes into account the closure of cracks in the small stress domain, which is a significant feature
of a rock mass’ deformation characteristics in typical civil engineering applications. Numerical results from the
proposed model are presented in a case study. In this, the model is calibrated in back-analyzing laboratory and
in-situ borehole dilatometer test data obtained for the new Minden lock excavation. In the example, the influ-
ence of crack closure and rock mass strength on slope deformation and failure is studied.

1 INTRODUCTION

This paper is concerned with the numerical simula- σ1


tion of rock masses using the Hoek-Brown (HB) fail-
ure criterion in a nonlinear elastic – perfectly plastic peak strength
constitutive model. The brief description of the model σpeak
and its numerical implementation found in the first
crack damage σcd
part of this paper follows a practical application in its threshold
second part: The new Minden navigable lock excava-
tion in siltstone/claystone.
In the past, the deformation characteristics of rock
masses has been studied by numerous researchers
crack initiation
(Eberhardt et al. 1998). Based on the stress–strain threshold
σci
behavior shown in Figure 1, Bienawski (1967) for
example, defines five stages in the stress–strain behav-
crack closure σcc
threshold
ior of rock masses: (a) Crack closure gradually occurs total volumetric ε1
until the normal stress reaches a threshold value scc.
contraction

strain
During crack closure, stiffness increases as pre-existing
cracks successively close. Crack closure is particulary
important to near-surface structures; (b) Linear elastic
∆V/V

behavior is encountered once the majority of existing


cracks is closed; (c) Stable micro-fracturing is found crack volumetric
dilation

strain
after the crack initiation stress sci is reached; (d) Crack
growth becomes then unstable for stresses exceeding crack closure crack growth
the threshold value scd. The threshold stress scd marks ε1
at the same time the stress at which volumetric strains
reverse; (e) Rock masses may show either ductile Figure 1. Behavior of fractured rock in uniaxial compres-
or brittle failure depending on geology, confining sion (after Eberhardt et al. 1999)

513
is taken as positive. Tensile stress and strain is taken as
negative. All stresses are taken to be effective values.

2 THE CONSTITUTIVE MODEL AND ITS


IMPLEMENTATION

2.1 Governing equations of the HB criterion


The generalized HB criterion is expressed as:

(1)

where s1 and s3 are the maximum and minimum


effective stresses at failure respectively, mb is the value
of the Hoek-Brown constant mi for the rock mass,

(2)
Figure 2. Diagram showing the idealized transition from
intact to heavily jointed rock masses with increasing sample
size (modified after Hoek 2004). sci is the uniaxial compressive strength of intact rock
and s and a are constants which depend upon the rock
pressure, and temperature. In the proposed constitu- mass characteristics given by the following relation-
tive model, crack closure (a) is simulated by means of ships:
a simple nonlinear elastic law. After crack closure, the
model is linear elastic up to failure. After failure, the (3)
rock is assumed to be perfectly plastic. As failure cri-
terion serves the HB criterion.
The HB criterion was derived from studies on the
behavior of jointed rock masses (Hoek and Brown (4)
1980) and is often used in practice. The criterion starts
from the properties of intact rock and then introduces
factors to reduce these properties on the basis of joint The Geological Strength Index (GSI), introduced by
characteristics within the rock mass. The original cri- Hoek (1994) provides a system for estimating the
terion was subsequently changed in order to adapt it to reduction in rock mass strength under different geolog-
the behavior of weak rock masses, as well. In the fol- ical conditions. The Geological Strength Index takes
lowing, the generalized HB criterion (2002-Edition) into account the geometrical shape of intact rock frag-
presented in Hoek et al. (2002) is used. In their paper, ments as well as the condition of joint faces. Finally,
Hoek et al. also derive equivalent parameters for the D is a factor that quantifies the disturbance of rock
Mohr-Coulomb (MC) failure criterion using a best-fit masses. It varies from 0 (undisturbed) to 1 (disturbed),
procedure within a given stress domain. However, there depending on the amount of stress relief, weathering
is no direct theoretical relationship between the HB and and blast damage as a result of nearby excavations. For
the MC criterion. Any attempt to estimate equivalent the significance of the parameters and their values see
cohesion and friction values in gravitational stress fields (Hoek 2004).
will thus be inaccurate. The curvilinear HB criterion
automatically accommodates changing stress fields, 2.2 Yield surface and plastic potential
the MC criterion does not.
Although rock masses typically show both, inher- The Hoek-Brown failure surface in principal stress
ent anisotropy and stress induced anisotropy caused space is shown in Figure 3. Taking into account a
by the evolution of crack systems in an inhomoge- strength reduction factor (RF), which reduces mate-
neous stress field, the HB criterion was initially intro- rial strength for RF ⬎1.0, the model’s yield surface is
duced to predict failure in quasiisotropic rock masses written as:
(Figure 2). The constitutive model proposed here is
likewise isotropic. In the following, the sign convention (5)
of soil mechanics is used: Compressive stress and strain

514
Figure 4. Stress dependant stiffness in the open crack
domain.

For tensile stresses below the maximum allowable one,


Figure 3. Hoek-Brown criterion in principal stress space. the yield surface is undefined. Therefore, the first stress
return for trial stresses outside the yield surface is
always conducted to a circumscribed Drucker-Prager
Trial stresses outside the yield surface are returned to yield surface.
it with a Drucker-Prager potential:

3 MODEL APPLICATION – THE NEW


(6) MINDEN LOCK EXCAVATION

3.1 The new Minden navigable lock


where p and q are the Roscoe invariants and cmob is
the mobilized angle of dilatancy. With increasing minor The so called ‘Schachtschleuse’ Minden, built 1913,
principal stress, the initial angle of dilatancy c is located at Germany’s Mittelland Canal. The lock
is reduced to 0 in a linear manner. The minor princi- chamber has a length of 85 m and a width of 10 m. In
pal stress at which zero volumetric plastic flow is order to adapt the Minden lock to the new ship gener-
reached and the initial angle of dilatancy c are model ation (2100 t ships), a new lock chamber is projected
parameters. in close vicinity of the existing lock. For the con-
struction of the new lock a 25 m wide, 250 m long and
18 m deep excavation pit is necessary.
2.3 Elastic stiffness The soil profile (from top to bottom) is made up of
an up to 15 m thick layer of inhomogeneous fill, fol-
The simulated pre-failure rock mass behavior consists
lowed by a 3.5 m thick layer of quaternary sands and
of: (a) a nonlinear crack closure domain and (b) a lin-
gravels of sometimes very high density. The bottom
ear domain. Stable plastic crack propagation and the
layer is composed of tertiary fine grained clastic sed-
gradual stiffness degradation associated to it is neg-
imentary rocks (siltstone/claystone), which usually
lected in the model proposed. The nonlinear stress–
show a weathering profile.
strain behavior can be simulated by piecewise linear,
The excavation pit has a slope of about 10 m height
hyperbolic or exponential stiffness approximations
in the sedimentary rock. The face angle is 70°. After
(Lahaie et al. 2002, Kaselow and Shapiro 2003). The
the interpretation of the geological general information,
piecewise linear approximation shown in Figure 4 is
the examination of photos from the old lock’s excava-
adopted here. Poison’s ratio is assumed constant.
tion pit, and existing soil data that have been acquired
in the near vicinity, a preliminary exploration pro-
2.4 Implementation issues gram was carried out. The exploration program com-
prised 6 boreholes and 6 cone penetration tests. In the
The maximum allowable tensile stress for which the boreholes, 8 water pressure tests and 9 dilatometer
HB criterion is defined calculates as: tests were conducted at different depths to determine
the permeability and the stiffness parameter of the rock.
(7) In the laboratory, rock samples were tested for their
unconfined compressive strength (UC-tests).

515
Three sets of discontinuities were observed in the behavior at the onset of testing is significantly different.
drill cores, which results in a orthogonal joint struc- These differences might be explained and modelled
ture: the bedding planes which decline with values by a variation of the threshold stress scc at which
between 0 and 12, a vertical joint family and one which all pre-existing cracks are closed. From the back-
declines under an angle of 45. According to DIN EN calculation, scc is found to vary between 400 kPa and
ISO 14689-1, spacing of the bedding is characterized 1000 kPa.
as “thin” (200 to 60 mm) and joint spacing as “medium”
(600 to 200 mm). Back-analysis of dilatometer tests
Knowing most of the model’s strength parameters and
having calibrated all its stiffness parameters, the GSI
3.2 Model parameter selection
value can now be calibrated from dilatometer test data.
The new Minden lock excavation is located in a rela- Figure 6 shows two test results and their simulation
tively homogeneous rock layer (Figure 7). All param- with the strength index GSI ⫽ 55. The difference in the
eters and tests discussed in the following refer therefore two calculated stress–strain curves is solely due to the
to that specific rock layer only. Based on geological above discussed variation in the threshold stress scc.
evidence, the Hoek-Brown constant mi is assumed as 8.
The uniaxial compressive strength sci of the siltstone/
3.3 Analysis of the new Minden lock excavation
claystone is determined as 2000 kPa. The rock mass
disturbance D depends on the respective experiment or In the following, a simplified calculation of the new
calculation. It is assumed that D ⫽ 0.0 for unconfined Minden lock excavation is presented. The rock is actu-
compression tests, D ⫽ 0.2 for dilatometer tests, and ally not dry like assumed in the calculation. The actual
D ⫽ 0.7 for the excavation. The Geological Strength geometry is also more complex than the one shown
Index (GSI) equals 100 for samples subjected to in Figure 7. However, for the purpose of demonstrating
unconfined compression. In situ, the discontinuities the possibilities of the constitutive model, the simpli-
described above reduce the strength index to GSI fied calculation seems adequate. The finite element
⬇55–65. The GSI value is later verified in the numer-
ical simulation of dilatometer tests. First, however,
the model’s stiffness parameters are quantified in a
back-analysis of unconfined compression tests.

Back-analysis of unconfined compression tests


From unconfined compression tests, the linear elastic
stiffness for the model proposed is determined to
208 MPa (cracks closed). Its initial stiffness is quanti-
fied to 10 MPa (cracks opened). Figure 5 shows the
unconfined compression tests used in the stiffness
determination. These tests can be basically grouped in
two sets. Both sets have approximately the same stiff-
ness in the linear domain. However, their nonlinear
Figure 6. Simulation of dilatometer tests.

Figure 7. Geometry and finite element mesh used in the


Figure 5. Simulation of unconfined compression tests. excavation analysis.

516
code used in the 2D plane strain analysis is Plaxis V8 the HB analysis reaches failure first. Figure 8 shows
(Brinkgreve 2002). Calculation results are compared to the respective shear strain distributions at failure.
results obtained in an equivalent linear elastic – per- Obviously, the HB calculation gives a much steeper
fectly plastic analysis using the MC failure criterion. failure surface.
The equivalent MC strength parameters (w ⫽ 31.2°, Figure 9 highlights the failure mechanisms for dif-
c ⫽ 39 kPa) have been obtained by the method pro- ferent GSI values by visualizing incremental displace-
posed in Hoek et al. (2002). Note that these parameters ments at failure. The strength reduction factor (RF) is
are less conservative than these used in the preliminary now below 1.0 for GSI ⫽ 40 and 3.5 for GSI ⫽ 70.
design as presented in Kauther and Herten (2006). The failure mechanism for GSI ⫽ 70 resembles the
one observed in the equivalent MC calculation shown
Slope failure before.
Slope failure is reached in the HB calculation by fac-
torizing material strength by means of the factor Slope deformation
introduced in Equation 5. In successively decreasing Figure 10 illustrates calculated horizontal slope
material strength until failure, finally the limit state is deformations for unreduced material strength. Again,
found. This scheme is basically identical to the well results from the equivalent linear elastic – perfectly
known w–c reduction scheme used in the equivalent plastic MC analysis are compared to results from the
MC calculation. However, the level of safety, quanti- nonlinear elastic – perfectly plastic HB analysis. In
tatively expressed in the strength reduction factor and the latter, the threshold stress scc is set to the limiting
the factor of safety obtained in the w–c reduction are values found in the backanalysis of the unconfined
not directly comparable. In the example, the HB cal- compression tests. Both, deformation pattern and
culation gave a strength reduction factor of 1.2. The deformation magnitude is different in the HB and the
factor of safety found in the MC analysis is 1.5. As equivalent MC analysis. In the HB analysis, the vari-
the equivalent MC parameters can only poorly fit the ation of scc leads to a variation in horizontal dis-
HB criterion in the highly nonlinear domain around placements in the same order of magnitude as the
the apex of the yield surface, it is yet reasonable that input is varied.

Figure 8. Shear strain distribution at failure (a – MC crite-


rion, b – HB criterion). Figure 9. Influence of the GSI on the failure mechanism.

517
tests could be used to narrow down the uncertainties
of the applied in situ Geological Strength Index.
Analysis of a 10 m slope showed significant differ-
ences in the failure mechanism and deformation char-
acteristics obtained by the new model compared to
those of an equivalent linear elastic – perfectly plastic
analysis with the Mohr-Coulomb failure criterion. In
the comparison, the equivalent Mohr-Coulomb param-
eters, which were fitted in the governing stress region
turned out to be inadequate. For small minor principal
stress, cohesion is most likely overrated whereas the
angle of friction is taken too small. For design purposes,
strength reduction should thus be incorporated in the
Hoek-Brown analysis directly as for example shown
in Equation 5.
The siltstone/claystone mainly addressed in this
paper showed distinct nonlinear features at the onset
of loading (crack closure). The simple nonlinear elastic
law introduced was found sufficient to model the defor-
mation characteristics in the crack closure domain. The
non-linear law’s impact in the simulation of dilato-
meter tests and slope deformation problems is consi-
derable. Experimental scatter in dilatometer tests
could partly be explained by a scatter in the threshold
stress scc.

REFERENCES
Bienawski, Z. T. (1967). Mechanics of brittle rock fracture.
Int. J. of Rock Mech. Minn. Sci. 4 (4), 395–423.
Brinkgreve, R. (Ed.) (2002). PLAXIS, 2D Version 8. AA.
Balkema.
Eberhardt, E., D. Stead, B. Stimpson, and R. S. Read (1998).
Identifying crack initiation and propagation thresholds in
brittle rock. Can. Geotech. J. 35, 222–233.
Hoek, E. (1994). Strength of rock and rock masses. ISRM
News Journal 2 (2), 4–16.
Hoek, E. (2004). Practical rock engineering – an ongoing set
of notes available on the Rocscience website, www.
rocscience.com.
Figure 10. Horizontal slope displacements in [mm] (a –
Hoek, E. and E. T. Brown (1980). Empirical strength crite-
MC criterion, b – HB criterion with scc ⫽ 400 kPa, c – HB
rion for rock masses. J. Geot. Engn. Div, ASCE 106
criterion with scc ⫽ 1000 kPa).
(GT9), 1013–1035.
Hoek, E. C. Carranza-Torres, and B. Corkum (2002). Hoek-
brown failure criterion – 2002 edition. In Proc 5th North
4 CONCLUSIONS Amer. Symp.– NARMS-TAC, Toronto, Canada.
Kaselow, A. and S. A. Shapiro (2003). Elastic properties of
The nonlinear elastic-perfectly plastic constitutive anisotropic rocks under hydrostatic load. In Proc. 65th
model introduced in this paper could be successfully Conf. EAGE, Stavanger, Norway.
applied in the back-analysis of various test data. The Kauther, R. and M. Herten (2006). Optimierte
Baugrunderkundung für den Neubau des Weserschleuse
model’s strength parameters are easy to determine as
Minden. In Proc. 13 Darmstädter Geotechnik
its failure envelope is the well known Hoek-Brown cri- Kolloquium, Darmstadt, Germany, pp. 161–170.
terion. Representative unconfined compression tests Lahaie, F., C. J. Bean, and A. Khasar (2002). Nonlinearity in
are sufficient to determine its stiffness parameters as the elastic moduli-porosity. Geophysical Prospecting. in
well. In the example, the simulation of dilatometer press.

518

You might also like