Tunnelling and Underground Space Technology 17 (2002) 263–274

Deformational characteristics of weak sandstone and impact to tunnel

deformation
Fu-Shu Jeng*, Meng-Chai Weng, Tsan-Hwei Huang, Ming-Lang Lin
Department of Civil Engineering, National Taiwan University, Taipei 106, Taiwan

Received 31 October 2001; received in revised form 10 March 2002; accepted 16 March 2002

Abstract

In northern Taiwan, a tunnel under construction along a segment where weak sandstone, the Mushan sandstone, was encountered
and an excess crown settlement (14–30 cm) has been reported. This paper studies the deformational characteristics of Mushan
sandstone and its impact on tunnel deformation. To distinguish the volumetric and the shear deformation of the sandstone,
experiments with controlled stress paths, including hydrostatic compression, pure shearing and conventional triaxial compression,
were conducted. The measured deformations were then decomposed into elastic and plastic components further exploring the
stress–strain behavior of weak sandstone. The results indicate that, similar to other soil-like geo-materials, this sandstone has
plastic strain before the stress path reaches the failure envelope and significant shear dilation is induced, especially when
approaching the failure envelope. Meanwhile, the distinct features of deformation have also been highlighted by comparing the
experimental results to the prediction, derived from existing constitutive models that were originally developed for other geo-
materials. These features include significant plastic volumetric strain at low levels of confining stress, suppression of plastic
volumetric strain at higher levels of confining stress, and the fact that the actual amount of shear compression is less than that
predicted by the model. Numerical analysis indicates that the weak rock leads to the greatest inward displacement, which results
from the shear dilation prior to failure state. 䊚 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Shear dilation; Weak sandstone; Pure shear stress path test

1. Introduction in several sections of the tunnels under construction, in

which the Mushan sandstone was encountered.
Most of the rocks in western Taiwan are tertiary weak In addition to the strength, other properties of weak
rocks, including sandstone, shale and mudstone, to rock should be considered in identifying the possible
which the juvenile rock forming process has resulted in sources inducing tunnel squeezing. For instance, some
weaknesses, not only in strength, but also in the capa- of the squeezing occurred after the top heading of the
bility of resisting weathering. As a result, the typical entire tunnel was fully excavated. This phenomenon
strength of these weak rocks ranges from 10 to 100 implies that the time-dependent deformation andyor the
MPa (Jeng et al., 1994). deterioration of the weak rock may also account for the
While tunneling through the weak rock strata, several squeezing. Jeng et al. (1996b) reviewed those unsuc-
unsuccessful cases had been reported (Jeng et al., cessful cases and found that the causes of squeezing are
1996b). Difficulties, including severe squeezing and related to, as a speculation at least, the following
raveling, were encountered during construction of these characteristics of weak rocks:
tunnels. For instance, a crown settlement of 180 cm of
a 12.4-m wide highway tunnel was reported. Meanwhile, a. Shear dilation—before the rock approaches its failure
a crown settlement ranging from 14 to 30 cm occurred state during excavation, it is often considered to be
elastic, and the inelastic deformation is negligible in
*Corresponding author. Tel.: q886-2-2363-0530; fax: q886-2- the pre-peak stage of loading. This point of view can
2363-1558. be adequate where hard rocks are concerned; however,
E-mail address: fsjeng@ce.ntu.edu.tw (F.-S. Jeng). for weak rocks, this assertion is not necessarily valid

0886-7798/02/$ - see front matter 䊚 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 8 8 6 - 7 7 9 8 Ž 0 2 . 0 0 0 1 1 - 1
264 F.-S. Jeng et al. / Tunnelling and Underground Space Technology 17 (2002) 263–274

and requires further study. Theoretically, shearing of

an isotropic elastic rock will not induce volumetric
dilation, which inherently contributes to the tunnel
squeezing. However, shear dilation of weak rock has
been reported in several literatures (Aristorenas, 1992;
Bernabe et al., 1994; Besulle et al., 2000; Dyke and
Dobereiner, 1991; Handin and Hager, 1957; Jeng and
Huang, 1998a). In addition to elastic shear dilation,
the plastic deformation, from both volumetric and
shear deformation, of weak rock requires further
exploration.
b. Creep of rock—based on an experimental study (Jeng
and Huang, 1998b), the rock, in the locations where Fig. 1. Schematic illustration for stress paths of pure shear test and
squeezing occurs, also exhibits significant creep conventional triaxial compression test. The abscissa and the ordinate
represent the hydrostatic stress and the shear stress, respectively.
behavior. Creeping of weak sandstone can be initiated
at low shear stress levels and becomes more signifi-
cant when the shear stress approaches the ultimate compressive strength of 37.1 MPa. Moreover, petro-
strength. This may directly relate to the excess crown graphic analysis is also applied in this research. The
settlement in addition to the instantaneous results of petrographic analysis show that Mushan sand-
deformation. stone is mainly composed of quartz, rock fragments,
c. Deterioration due to wetting process—the humidity and matrix with an average grain size of approximately
of the tunneling as well as seeping water significantly 0.72 mm.
reduces the strength of rock. Investigation of the
deterioration of rock (Jeng et al., 1996a) indicates 2.1. Pure shear stress path test
that, when dry sandstone is soaking in water for 120
min, the decrease in the compressive strength, shear
modulus and bulk modulus is 60%, 60% and 70%, The specimens were loaded in a triaxial cell, by
respectively. The softening of sandstone potentially which the axial pressure (s1) and the confining pressure
has a relevant influence on the excess crown settle- (s3) are applied, under room temperature. The triaxial
ment with a time lag after the excavation of the cell is able to sustain a confining pressure up to 175
tunnel. MPa and the confining pressure was monitored by a
pressure transducer with an accuracy of 0.1 MPa. The
Among those factors, this paper focuses on the short- axial load was provided by a servo-controlled high
term (time independent) deformational behavior, includ- stiffness machine, which has a maximum load and a
ing shear dilation, of weak sandstone through a series stiffness of 4448 kN and 13.1=109 Nym, respectively.
of experiments. The impact of the deformational behav- The load is applied at a rate of 5 MPaymin.
ior of weak rock to the deformation of tunnel was also Although shear dilation behavior can also be observed
assessed by using numerical analyses. in the conventional triaxial compression tests, to isolate
the influence of volumetric loading from the shear
2. Setup of experimental study loading, a pure shear stress path loading, which is also
termed as ‘triaxial compression’ (TC) test by Chen and
In this research, Mushan sandstone, the strata in which Saleeb (1982), was adopted (Aristorenas, 1992; Bernabe
squeezing has occurred during tunnel construction, is et al., 1994). As shown in Fig. 1, the stress path of a
adopted as the specimen. The physical and mechanical conventional triaxial compression test (CTC) is O–A–
properties of Mushan sandstone were measured accord- C, in which a hydrostatic stress (path O–A) is first
ing to the methods suggested by ISRM (ISRM, 1981). applied to the specimen and this is followed by an
The specimen size is 5.5 cm in diameter and 12.5 cm increase of axial stress (path A–C). In such a test, both
in height. The specimen is oven-dried (105 8C) to the hydrostatic stress, p, and the shear stress yJ2 increase
remove its natural water content. The longitudinal and due to the increase of axial stress. Therefore, the
transverse deformation were separately measured by a volumetric behavior is jointly affected by the shearing
full Wheatstone bridge consisted of four strain gauges, and the hydrostatic loading.
which are capable of measuring strains up to 2% with To separate the influence of hydrostatic loading during
an accuracy of "0.85 (mmym)y 8C. the stage of shearing, the confining pressure must be
The results show that this sandstone has a mean dry reduced according to the increase of axial pressure
density of 2.3 gycm3, a mean porosity of 8.27%, a mean 1
(Ds3sy Ds1), i.e. along stress path A–B, such that
saturated water content of 3.60% and a mean uniaxial 2
 F.-S. Jeng et al. / Tunnelling and Underground Space Technology 17 (2002) 263–274 265

Table 1 2. The volumetric deformation induced by shear stress;

Testing program of Mushan sandstone 3. The shear deformation induced by shear stress; and
Type of tests Test code Confining pressure No. of tests 4. The shear deformation induced by hydrostatic stress.
(MPa) Among the four possible types of deformation, it was
Pure shear stress MS20 20 3 found that the deformation of the fourth type is very
path test MS40 40 3q1* slight, and hence, can be neglected. The deformational
MS60 60 3q2* behavior of the first three types is presented as follows.
MS80 80 1
Conventional MSC02 2 1 3.1. Volumetric deformation induced by hydrostatic
triaxial test MSC10 10 1 stress
MSC18 18 1
MSC40 40 1
Fig. 3a shows a typical hydrostatic stress vs. volu-
Remarks: *Cyclic loadingyunloading was performed. metric strain relationship, in which cycles of loading
and unloading have been applied to identify the elastic
the hydrostatic stress remains unchanged during the and inelastic components of the total deformation. Note
shearing stage. The measure of shear loading is that a positive stress or strain represents a compression
1 condition. Fig. 3a indicates that an increase in confining
expressed in terms of yJ2, where J2s ŽSijSij. and Sijs
2 stress not only compresses the sandstone, but also
deviatoric stress tensor. Tests were carried out under improves the stiffness (represented by the slope of the
confining pressure ranging from 20 to 80 MPa to study curve ´tv,p) of the sandstone and results in a non-linear
the effect of confining stress. Meanwhile, cycles of curve. Similar volumetric behavior has also been
unloadingyreloading were conducted in several tests for observed by Bernabe et al. (1994). If all the cycles of
obtaining the elastic deformation, which will be used to unloading and reloading curve were put together, as
decompose the total deformation into elastic and plastic shown in Fig. 3b, the elastic component of deformation
components. Table 1 summarizes the corresponding can then be determined as indicated by the thick, solid
testing parameters of all tests conducted in this research. curve based on regression analysis. Also shown by Fig.
3b, cycles of loading and unloading seem to decrease
3. Experimental results—typical deformational the bulk modulus slightly, which is possibly the result
behavior of interior damage induced by these cycles of loadings.
The plastic component of deformation can then be
The failure envelope of the sandstone tested resembles determined by subtracting the elastic component from
a linear geometry, expressed in terms of stress invariants the total deformation. Both the elastic and the plastic
of three-dimensional stress components, as illustrated in components (´ev,p and ´pv,p) of the total deformation are
Fig. 2. Accordingly, the linear relationship shown in Fig. shown in Fig. 3a. It has been observed that the plastic
2, can be expressed in terms of the stress invariants, I1 deformation is suppressed under higher levels of confin-
and yJ2, as: ing stress as shown in Fig. 3a, and therefore, elastic
deformation plays a more dominant role in the volu-
yJ2s0.34I1q11.60 (1) metric deformation at higher levels of confining stress.
where I1 and yJ2 are stress invariants, representing the This phenomenon can be further illustrated if we observe
confining stress and the shear stress, with a unit of MPa,
and I1 is defined as:
I1s(s11qs22qs33)s3p (2)
where s11, s22 and s33 are the normal stresses, and p
is the hydrostatic stress. The failure envelope expressed
by Eq. (1) can be further converted to a conventional
Coulomb failure envelope defined in snyt coordinate
and results in a frictional angle f and a cohesion
intercept c of 36.18 and 10.0 MPa, respectively.
As long as a couple phenomenon exists between the
shear and the volumetric behavior, the deformational
behavior of sandstone tested should be observed in the
following aspects:
1. The volumetric deformation induced by hydrostatic
stress; Fig. 2. Failure envelope of sandstone tested.
266 F.-S. Jeng et al. / Tunnelling and Underground Space Technology 17 (2002) 263–274

Remarkably, Fig. 5b depicts that a pure-shear loading

would lead to an elastic dilation and a plastic compres-
sion at earlier stages of loading. Again, all the cycles of
unloading–reloading curves are again put together as
shown in Fig. 5c, in which the shear stiffness of
sandstone significantly drops upon application of greater
levels of shear stress. Since such a deterioration process
is in fact irreversible and the effect shown in Fig. 5c is
much more significant when compared to the deteriora-
tion of bulk modulus shown in Fig. 3b, it is no longer
valid to just obtain a mean elastic behavior based on all
loading cycles, especially the last cycle where significant
deterioration appears. Therefore, the last cycle of loading
is excluded in the regression analysis and a hypothetical
elastic curve is obtained as shown by the thick solid
curve in Fig. 5c. This elastic dilation induced by
shearing can also occur in a CTC test; however, it
cannot be clearly isolated and evaluated as the volumet-
ric deformation obtained from a CTC test in fact results
from both the volumetric compression and the shear
loadings.
As a result, the elastic and the plastic components are
shown in Fig. 5d. It reveals that the sandstone tends to
dilate elastically upon shearing. However, the plastic
deformation is comprised of a compression at an earlier
stage of shearing and a significant dilation at subsequent
stages of shearing, especially when the shear stress is
close to the shear strength of the sandstone. Consequent-
ly, when the applied shear stress approaches the shear
Fig. 3. (a) Relationship of volumetric strain with confining stress
strength of the sandstone, the total deformation shows a
during hydrostatic compression state of loading. (b) Loops of unload- significant dilation resulting from both the elastic and
ing–reloading curves. The plastic strains at zero hydrostatic stress of the plastic behavior.
all cycles have been removed, and a hypothetical elastic curve is
obtained by regression analysis of these cycles of curves. Therefore,
this hypothetical elastic curve, shown in thick solid line, in fact rep-
3.3. Shear deformation induced by shear stress
resents a ‘mean’ behavior of these cycles. The sample number, MS60-
噛1-60, represents that the first specimen stressed at maximum The typical relation of shear strain (gts2yJ92) vs.
hydrostatic pressure of 60 MPa. shear stress (yJ2) is shown in Fig. 6a, in which the

the variation of increments of ´pv,p (per increase of

hydrostatic stress, in units of 1 MPa) with the hydrostatic
stress illustrated in Fig. 4. Closure of micro-cracks may
account for the disappearance of plastic strain at higher
levels of confining stress and results in a greater bulk
modulus as well.

3.2. Volumetric deformation induced by shear stress

Volumetric deformation induced by shear stress does

exist as depicted by Fig. 5a, in which increasing of
shear stress first compresses the sandstone, followed by
dilation of sandstone. Test results can also be decom-
posed into elastic and plastic components, which are
Fig. 4. Variation of plastic strain increment magnitudes with applied
deduced from the unloading–reloading curve. This pro- hydrostatic stress. It shows that significant plastic, volumetric com-
cess is illustrated by a schematic drawing shown in Fig. pression occur at lower levels of confining stress and this plastic strain
5b. is suppressed upon greater levels of confining stress.
 F.-S. Jeng et al. / Tunnelling and Underground Space Technology 17 (2002) 263–274 267

Fig. 5. (a) Variation of volumetric strain with shear stress during shear loading. Cycles of unloading indicates that plastic deformation exists. (b)
Schematic illustration of the loading cycles. It indicates that, at an earlier stage of loading, the pure-shear loading tends to induce an elastic
dilation and, however, a plastic compression. (c) Loops of unloading–reloading curves. A hypothetical elastic curve (shown by thick, solid line)
was obtained by regression analysis of most cycles, but with the last cycle excluded, since significant deterioration of deformation modulus has
been induced in the last cycle. (c) Typical result of total, elastic and plastic deformation after separation of components. The sample number,
MS60-噛1-67, represents that the first specimen sheared at maximum shear stress of 67 MPa.

load–deformation curve approximates toward a horizon- 4. Characteristics of plastic deformation

tal asymptote (i.e. failure state) when shear stress
approaches the strength of sandstone. To identify the Remarkably, plastic deformation occurs before yield-
elastic component of deformation, all the cycles of ing of the material as shown in Figs. 3–6. If both of
unloading–reloading curves are put together, as shown the volumetric and the shear strains are considered, the
in Fig. 6b. It indicates that the initial elastic shear increments of plastic deformation can be presented as
modulus increases upon greater levels of hydrostatic vectors as shown in Fig. 7a. Each plastic deformation
stress applied as shown in Fig. 6c. To obtain a represen- vector has an angle b, as defined in Fig. 7a, which
tative initial elastic shear modulus (Go ), the later cycles indicates a compression or a dilation for b ranging from
of unloading–reloading curves, are therefore, excluded 08 to 908 and greater than 908, respectively. Fig. 7a
in regression analysis and a hypothetical elastic defor- reveals that the plastic compression occurs at earlier
mation curve is obtained as shown in Fig. 6b. Subtract- stages of shearing, followed by dilation after a ‘threshold
ing the elastic component from the total deformation, condition’ of shear stress.
the plastic component can be obtained as shown in Fig. The plastic deformation vector tends to be perpendic-
6a. It indicates that, when the shear stress approaches ular to the failure envelope of sandstone, i.e. plastic
the shear strength of sandstone, significant plastic shear deformation at failure state appears to follow the so-
strain occurs and it eventually dominates the total called associated flow rule of plastic deformation. Fur-
deformation. thermore, this pattern of plastic strain vector seems not
268 F.-S. Jeng et al. / Tunnelling and Underground Space Technology 17 (2002) 263–274

sub-parallel at the points where the stress paths cross

each other. Namely, the deformation vectors obtained
either from a pure shear test or a CTC test are, in
general, consistent with each other, except that there are
a few discrepancies under conditions of low confining
stress.

4.1. Effect of confining stress on threshold condition

initiating dilation

To represent the level of shearing corresponding to

the shear strength of the material, a normalized shear
stress measure h f is introduced:
B E
hfsyJ2yDyJ2Gf
C F
(3)

Fig. 6. (a) Variation of shear strain with shear stress during shear
loading. (b) Loops of unloading–reloading curves. A hypothetical
elastic curve (shown by the thick, solid line) was obtained based on
the cycles of lower levels of shearing. (c) The variation of initial
shear modulus with volumetric stress. The sample number, MS60-噛1-
67, represents that the first specimen sheared at maximum shear stress
of 67 MPa.

to be influenced by the stress paths, as shown in Fig.

7b. In Fig. 7b, the results of plastic deformation vectors Fig. 7. (a) Variation of vectors of plastic strain increments during
obtained from tests of pure shear loading under various shear loading. The final vector (when the stress path meet the failure
envelope) tends to be perpendicular to the failure envelope, which
confining stresses and from conventional traxial test indicates a typical associated rule of plastic flow. (b) Vectors of plastic
(CTC) are plotted together and compared. Comparison strain obtained from pure shear tests and CTC tests. The results are
indicates that, in most cases, the vectors are parallel or consistent for different stress paths.
 F.-S. Jeng et al. / Tunnelling and Underground Space Technology 17 (2002) 263–274 269

the plastic deformation. The magnitude of plastic strain

increment, d´p is defined as:
d´psyd´pijd´ijp (4)
where d´pij is the tensor of plastic strain increment
calculated on the basis of per 1 MPa increment of shear
stress (yJ2), while subscripts i and j range from 1 to 3.
The magnitude of plastic strains at various stages of
shearing is shown in Fig. 8b; it indicates that plastic
strain occurs even for a minor degree of shearing (h f s
0–0.4) and significantly increases, by approximately
two orders of magnitude, while approaching the failure
state (h f s0.8–1).

5. Characteristics of deformation—theoretical
concerns

Given the deformational behavior of sandstone

observed from experimental study, a constitutive model
can accordingly be developed to describe the behavior
found in this study. A preliminary model of sandstone
is presented in this paper. This preliminary model is
mainly based on the existing models, which were orig-
inally developed for other materials, and the predictions
derived from this proposed model are compared to the
observed deformational behavior. By doing this, the
discrepancy between the model’s prediction and the
actual behavior can be identified, and thus, the charac-
teristics of sandstone deformation behavior can be high-
Fig. 8. (a) Variation of b during shear loading. Dilation occurs when lighted. Meanwhile, the necessary features of a more
b is greater than 908. The threshold condition, at which dilation ini- sophisticated model for sandstone can accordingly be
tiated, varies with the confining stress applied. (b) Magnitude of plas- determined.
tic strain increment also varies with the levels of shearing. The greater
levels of shearing, the greater the magnitude of plastic strain, espe- 5.1. Elastic component of deformation
cially when h approaches to 1.
In view of the non-linear elastic behavior shown in
B E
where DyJ2Gf is the shear strength of the material and
C F
Fig. 3a,b, a Green elastic model (or hyperelastic model)
yJ2 is applied shear stress. Fig. 8a indicates that the
threshold condition (h f ), at which plastic dilation is
initiated (bs908), is affected by the confining stress
applied: a greater h f is required when the specimen is
confined by a greater hydrostatic stress as shown in Fig.
9. Meanwhile, since the hydrostatic stress applied by
CTC tests continuously increases during the loading
process, its b variation thus lies in between the curves
of constant hydrostatic stress as shown in Fig. 8a.
However, also shown in Fig. 9, if measured in terms
of total deformation, the confining stress seems to have
negligible effects on the threshold condition and the
threshold h f is maintained at an almost constant mag-
nitude (approx. h f s0.4).

4.2. Magnitudes of plastic strain increments

Fig. 9. Variation of threshold h magnitude with confining stress.
Increases of confining stress result in a greater threshold for plastic
Note that the vectors shown in Fig. 7 have a unit deformation, but however, have very little influence on the total
length, which thus does not represent the magnitude of deformation.
270 F.-S. Jeng et al. / Tunnelling and Underground Space Technology 17 (2002) 263–274

Fig. 10. Variation of volumetric strain with confining stress during Fig. 11. Variation of volumetric strain with shear stress during shear
hydrostatic loading for results from model prediction and experiments. loading for results from model prediction and experiments.

is thus adopted. On the basis of such a model, the strain mental curves shown in Fig. 3a, Fig. 5b and Fig. 6a.
tensor can be determined from an energy density func- For the Mushan sandstone tested, b1, b2 and b3 are
tion V and the stress tensor sij by the following found to be 1112 mØ(MPa)y1y2, y0.049 mØ(MPa)y2
expression: and 48.48 mØ(MPa)y1, respectively. The strains from
e e
model prediction, Ž´v,p .m, Ž´v,s.m and Žge.m are compared
≠V to the experimental results, ´ev,p, ´ev,s and ge, as shown
´eijs (5)
≠sij in Figs. 10–12. In general, the proposed model is
where ´eij is the elastic strain tensor and V the energy capable of predicting the non-linear elastic behavior,
density function. From the experimental results, V is including shear dilation of sandstone, and its prediction
selected to be: has a reasonably good agreement with the experimental
curves as well.
Vsb1I3y2
1 qb2I1J2qb3J2 (6)
5.2. Plastic component of deformation
where b1, b2 and b3 are material constants. Substituting
Eq. (6) into Eq. (5), the elastic strain tensor ´eij has the As illustrated by Fig. 7a, the vector of plastic defor-
following form: mation tends to be perpendicular to the failure envelope
≠V B 3 1y2 E when the applied shear stress reaches shear strength.
´eijs sC b1I1 qb2J2FdijqŽb2I1qb3.sij (7) Namely, this type of deformation is consistent with the
≠sij D 2 G
associated flow rule of plasticity. If an associated flow
According to Eq. (7), the volumetric strain can be rule and an imaginary plastic potential surface (or yield
obtained as: cap) can be assumed, the plastic potential surfaces for
different stages of shearing would have the geometry
9
´evs´e11q´22
e e
q´33 s b1I11y2q3b2J2 (8)
2
Eq. (8) indicates that both hydrostatic stress and shear
stress will induce volumetric strain. Therefore, the elastic
volumetric strain can be separated into two components,
´ev,p and ´ev,s, as the following expressions:
9
´ev,ps b1I1y2
1 (9)
2
´ev,ss3b2J2 (10)
e
The shear strain (g s2yJ92) can also be determined
based on Eq. (7) as:
ges2yJe92 s2Žb2I1qb3.yJ2 (11)
Accordingly, the material constants b1, b2 and b3 can be Fig. 12. Variation of shear strain with shear stress during shear loading
determined by a curve fitting derived from the experi- for results from model prediction and experiments.
 F.-S. Jeng et al. / Tunnelling and Underground Space Technology 17 (2002) 263–274 271

increases when the material is subjected to continuous

loading and leads to a propagation of the plastic potential
surface. ad is determined by the following expression:
ad
ads (13)
Žd´p.hd
where d´p is the magnitude of plastic strain increments,
and ad and hd are material constants.
Assuming that the associated plastic flow rule is
applicable, the plastic strain increment tensor can then
be expressed as:
≠F
d´pijsl (14)
≠sij
Accordingly, the magnitude of plastic strain increment
d´p, the plastic volumetric strain increment d´pv and the
plastic shear strain increment d´ps can be obtained as:
Fig. 13. Schematic illustration on the propagation of hypothetical plas-
tic potential surface during shear loading, provided that the defor- B ≠F ≠F E1y2 B ad E1yhd
mation follows the associated flow rule. d´pslC F sC F (15)
D ≠sij ≠sij G D ad G

B
≠F E
shown in Fig. 13. Remarkably, the plastic potential B ≠F E
3d´pC
F
D ≠I1 G
surface tends to have a tangential connection with the d´vps3lC Fs B
D ≠I1 G ≠F ≠F E1y2
failure envelope. In fact, similar yield surfaces have C F
been proposed (DiMaggio and Sandler, 1971; Mullenger D ≠sij ≠sij G

and Davis, 1981; Gens and Nova, 1993; Lagioia and B

ad E1yhdB ≠F E
Nova, 1995). This coincidence reveals the similarity in 3C F C F
D ad G D ≠I1 G
nature of weak rock and soil. s B (16)
≠F ≠F E1y2
An existing plastic model, which possesses a plastic C F
potential surface with the above-mentioned features of D ≠sij ≠sij G

sandstone, developed by Desai et al. (1986), Desai and

Salami (1987), Desai and Varadarajan (1987)) is, there- d´ps syŽd´p.2yŽd´pv.2 (17)
fore, adopted. If assuming that the material has a uniform Based on Eqs. (12) and (13), the adopted plastic model
strength on the deviatoric plane, this model has a has five material parameters (gd, T, n, ad and hd),
simplified form as: which can be obtained by a curve fitting derived from
FŽI1,J2.sJ2ywyadŽI1qT.nqgdŽI1qT.2xs0 (12) the experimental results. For the detailed process in
determining these material parameters, reference should
where T is the I1 stress at which shear strength vanishes, be made to Desai et al. (1986), Desai and Salami
n and gd are material constants controlling the geometry (1987), Desai and Varadarajan (1987)). Accordingly,
of the surface. In Eq. (8), ad is a state parameter; it these material were found to be 0.12, 37.1 MPa, 4.04,

Table 2
Corresponding parameters of Mushan sandstone

Type of model Parameter Remark Value

Elastic b1 Elastic constant 1112 mØ(MPa)y1y2
model b2 Elastic constant y0.05 mØ(MPa)y2
b3 Elastic constant 48.48 mØ(MPa)y1
Plastic gd Ultimate constant 0.12
model T The corresponding I1 stress 37.1 MPa
at which shear strength
vanishes
n Phase change constant 4.04
ad Hardening constant 2.04=10y5
hd Hardening constant 0.73
272 F.-S. Jeng et al. / Tunnelling and Underground Space Technology 17 (2002) 263–274

Table 3
Inward displacements predicted by different models

Displacement (cm) Crown Spring line Invert

Material model
Elastic model 5.2 5.9 13.6
Elasto-plastic model 6.7 9.2 14.6
Weak rock 12.2 11.8 18.8

ical analysis using finite element code. The analyzed

cross-section of a tunnel with a height and a width of
16 and 12.4 m, which is based on the geometry of a
squeezing tunnel (Jeng et al., 1996b), is illustrated in
Fig. 14. The material parameters required by elastic
model, elasto-plastic model and weak rock were deter-
mined based on the deformational behavior of the
sandstone studied.
The inward displacements of the sidewall calculated
based on the three types of material models are sum-
marized in Table 3 and shown in Fig. 15. It shows that
Fig. 14. Schematic illustration of analyzed tunnel. the inward displacements calculated based on the actual
behavior of weak rock have the greatest magnitudes all
2.04=10y5, and 0.73 for gd, T, n, ad and hd, respec- over the sidewall including the crown, spring line and
tively. All the parameters of this preliminary model for invert, when compared to those predicted by elastic
weak sandstone are also summarized in Table 2. model or elasto-plastic model.
A comparison of predictions from the model with What accounts for the significant amount of inward
experimental results of ´v,p, (´v,p)m vs. hydrostatic stress, displacement can be further explored by examining the
is shown in Fig. 10 and significant discrepancies have volumetric deformation predicted by the three models.
been observed. The experimental results show a signif- As shown in Fig. 16, a dilation zone around the sidewall
icant amount of plastic strain at earlier stages of loading; (the shaded areas) was developed based on elasto-plastic
however, only a minor amount of plastic strain was model (Fig. 16b), while the elastic model only predicts
predicted at later stages of loading. Therefore, this model a dilation zone beneath the invert (Fig. 16a). As a
is not representative in predicting the plastic volumetric
strain induced by hydrostatic stress, either in general
trend or in magnitudes.
A comparison of (´v,s)m with ´v,s is shown in Fig.
11. In general, the model gives a consistent trend:
compression at earlier stages followed by shear dilation.
However, a discrepancy in the magnitudes can still be
found. The amount of compression predicted by the
model is greater than that in reality.
A comparison of (g)m with g is shown in Fig. 12. It
shows that the model prediction is consistent with the
experimental results except that (gp)m is somewhat
smaller than gp.
Overall, the preliminary model discussed is capable
of describing well the elastic behavior, but requires
further modification for better describing the plastic
behavior of sandstone.

6. Impact to the tunnel deformation—numerical

assessment

Given the distinct features in the deformational behav-

ior of weak sandstone, the impact of such behavior to
the tunnel sidewall deformation was assessed by numer- Fig. 15. Inward displacements predicted by different models.
 F.-S. Jeng et al. / Tunnelling and Underground Space Technology 17 (2002) 263–274 273

7. Conclusion

The deformational behavior of weak sandstone is

presented in this paper, in which the elastic and the
plastic components are separated on the basis of the
cyclic loading curves. The deformation of weak sand-
stone at least possesses the following characteristics:
1. Plastic strain can be initiated at very low stress levels;
2. Significant plastic volumetric strain is induced by
hydrostatic stress at low confining stress levels and it
is suppressed while compressed at high confining
stress levels;
3. Shear compression occurs at low shear stress levels,
which is followed by much significant shear dilation
when shear stress approaches the shear strength of
sandstone; and
4. Higher levels of shearing significantly reduce the
elastic shear modulus.
The prediction from a preliminary model based on
existing models was compared to the actual behavior
observed from experiments. By doing this comparison,
the discrepancy in the deformational behavior of weak
sandstone from that of other geo-materials (e.g. hard
rock and soils) is highlighted. It is found that the elastic
behavior can be reasonably well described by a Green
elastic model based on a proposed energy density
function wEq. (6)x. Remarkably, the irreversible process
of the deterioration of elastic modulus, including shear
modulus and bulk modulus, at higher levels of loading
is, for the sake of simplicity, considered to be part of
plastic deformation. This deterioration process of these
elastic moduli can also be considered in a more sophis-
Fig. 16. Volumetric deformation predicted by different models. Dila- ticated elastic model. For instance, the concepts of the
tion zone (with a volume strain equal to or less than y2=10y3) is so-called damage model can probably be applied (Zhang
illustrated by the shaded areas.
et al., 1990).
As to the plastic model, additional adjustments to the
result, the elasto-plastic model predicts greater sidewall existing models are required to better describe the actual
displacements than the elastic model does. Moreover, plastic deformational behavior of sandstone, especially
the dilation zone of weak rock is even more significant in the following aspects:
than those of elasto-plastic model, especially around the
1. Significant plastic volumetric strain at low confining
crown and beneath the invert (Fig. 16c). Again, the
stress levels. Furthermore, higher levels of confining
greater the dilation zones, the greater inward displace-
stress tend to suppress plastic volumetric strains; and
ments, especially at crown and at spring line, when
2. The actual magnitudes of plastic strains differ from
weak rock and elasto-plastic model are compared. For
those that have been predicted from the existing
the weak rock, the shear dilation prior to failure state
model.
accounts for the greatest dilation zone, since both the
elastic model and the elasto-plastic model would not Last but not least, the hysteresis loops of unloading–
allow dilation in elastic range. loading curves exist, and these have so far not been
Remarkably, the crown settlement predicted by using incorporated in either the elastic model or the plastic
the deformational behavior of weak rock is approxi- model discussed above.
mately 12.2 cm, which is closer to, but still somewhat Overall, this research found that the existing model
away from the actual crown settlement, 14–30 cm (Jeng cannot properly describe the deformation behavior of
et al., 1996b). Therefore, the other factors, e.g. creep, weak sandstone, and therefore, further development of
should also be concerned to yield a more realistic sophisticated constitutive models for sandstone is thus
estimation of the sidewall displacement. needed. Numerical analysis indicates that the weak rock
274 F.-S. Jeng et al. / Tunnelling and Underground Space Technology 17 (2002) 263–274

leads to a greatest inward displacement, which is resulted References

from the shear dilation prior to failure state. Aristorenas, G.V., 1992. Time-Dependent Behavior of Tunnels Exca-
vated in Shale. Doctoral Thesis. Dept. of Civil Eng., Massachusetts
Institute of Technology, MA, USA.
8. Nomenclature Bernabe, Y., Fryer, D.T., Shively, R.M., 1994. Experimental obser-
vations of the elastic and inelastic behaviour of porous sandstones.
Geophys. J. Int. 117, 403–418.
ad Hardening constant of plastic model Besulle, P., Desrues, J., Raynaud, S., 2000. Experimental characteri-
C Cohesion strength of Coulomb failure criteria (MPa) zation of the localisation phenomenon inside a Vosges sandstone
eij Deviatoric strain tensor in a triaxial cell. Int. J. Rock Mech. Min. Sci. Geomech. Abstr.
I1 first stress invariant, I1sskk (MPa) 37, 1223–1237.
Chen, W.F., Saleeb, A.F., 1982. Constitutive Equations for Engineering
1
second deviatoric stress invariant, J2s sijsij Materials, Vol. I. John Wiley & Sons, elasticity and modeling’
J2 2 New York.
Desai, C.S., Salami, M.R., 1987a. A constitutive model and associated
Ž .
yJ2 f shear strength of the material (MPa) testing for weak rock. Int. J. Rock Mech. Min. Sci. Geomech.
J92 second deviatoric strain invariant Abstr. 24, 299–307.
n phase change constant of plastic model Desai, C.S., Varadarajan, A., 1987b. A constitutive model for quasi-
1 1 static behavior of rock salt. J. Geophys. Res. 92, 11445–11456.
Hydrostatic stress ps I1s skk (MPa)
P 3 3 Desai, C.S., Somasundram, S., Frantziskonis, G., 1986. A hierarchical
sij deviatoric stress tensor approach for constitutive modeling of geologic materials. Int. J.
T The corresponding I1 stress at which shear strength vanishes Numerical Anal. Methods Geomech. 10, 225–257.
(MPa) DiMaggio, M.L., Sandler, I.S., 1971. Material model for granular
ad parameter of plastic model soils. J. Eng. Mech. Div. ASCE 97, 935–950.
b angle of plastic flow, tanbsd´ps yd´pv Dyke, C.G., Dobereiner, L., 1991. Evaluating the strength and
´ev,p elastic volumetric strain due to hydrostatic stress deformability of sandstones. Q. J. Eng. Geol. 24, 123–134.
´ev,s elastic volumetric strain due to shear stress Gens, A., Nova, R., 1993. Conceptual bases for a constitutive model
e
Ž´v,p .m elastic volumetric strain due to hydrostatic stress from model for bonded soils and weak rocks. In: Anagnostopoulos, A., et al.
e
Ž´v,s .m elastic volumetric strain due to shear stress from model (Eds.), Geotechnical Engineering of Hard Soils—Weak Rocks.
´pv,p plastic volumetric strain due to hydrostatic stress Balkema, Rotterdam, pp. 485–494.
´pv,s plastic volumetric strain due to shear stress Handin, J., Hager, R.V., 1957. Experimental deformation of sedimen-
p tary rock under a confining pressure. J. Am. Assoc. Pet. Geol. 41,
Ž´v,p .m plastic volumetric strain due to hydrostatic stress from model
p 1–50.
Ž´v,s .m plastic volumetric strain due to shear stress from model
´tv,p total volumetric strain due to hydrostatic stress ISRM, 1981. In: Brown, E.T. (Ed.), Rock Characterization Testing
´tv,s total volumetric strain due to shear stress and Monitoring—ISRM Suggested Methods. Pergamon, New
t York,.
Ž´v,p .m total volumetric strain due to hydrostatic stress from model
t
Ž´v,s .m total volumetric strain due to shear stress from model Jeng, F.S., Huang, T.H., 1998. Shear dilational behavior of weak
d´p increment of plastic strain, d´psyd´ijp d´ijp sandstone. Proc. Regional Symp. Sediment. Rock Eng., Taipei. pp.
f frictional angle of Mohr–Coulomb failure criteria 262–267.
gd ultimate constant of plastic model Jeng, F.S., Huang, T.H., 1998. Time dependent behavior of weak
g shear strain, gs2yJ92sy2Øeijeji sandstone on Mushan Formation. Proc. 13th Southeast Asian
ge elastic shear strain Geotech. Conf. Taipei, pp. 75–80.
(ge)m elastic shear strain from model Jeng, F.S., Ju, G.T., Huang, T.H., 1994. Properties of some weak rock
gp plastic shear strain in Taiwan. Proc. ’94 Taiwan Rock Eng. Symp. pp. 259–267.
(gp)m plastic shear strain from model Jeng, F.S., Lee, I.T., Huang, T.H., 1996. Deterioration of weak rock
(gt)m total shear strain from model induced by wetting process. Proc. ’96 Taiwan Rock Eng. Symp.
gt total shear strain, pp. 373–382.
hd hardening constant of plastic model Jeng, F.S., Lin, M.L., Huang, T.H., 1996. Study of the Geological
hf Ž .
normalized shear stress measure, hfsyJ2y yJ2 f Barriers of the Tunnels in Northern Taiwan, Research Report
V strain energy density function MOTC-STAO-RD1501. Ministry of Transportation.
Lagioia, R., Nova, R., 1995. An experimental and theoretical study
of the behaviour of a calcarenite in triaxial compression. Geotech-
nique 45, 633–648.
Mullenger, G., Davis, R.O., 1981. An unified yield criterion for
Acknowledgments cohesionless granular materials. Int. J. Numerical Anal. Methods
Geomech. 5, 285–294.
Zhang, J., Wong, T.F., Davis, D.M., 1990. Micromechanics of pres-
The research is supported by the National Science sure-induced grain crushing in porous rocks. J. Geophys. Res. 95,
Council of Taiwan, Grant no. NSC-89-2211-E-002-152. 341–352.