Assessing Rock Mass Behaviour for Tunnelling

VASSILIS P. MARINOS1
Faculty of Sciences, School of Geology, Aristotle University of Thessaloniki,
GR-541 24 Thessaloniki, Greece

Key Terms: Tunnel, Rock Mass Type, Classification, Nowadays, knowledge and understanding related
Behaviour Type, Support to the role of the geological material and its
implication in design are reinforced with advances
in site investigation methods, the development of
ABSTRACT geotechnical classification systems, and the conse-
quent quantification of rock masses. The RMR
In the last few decades a rapid development in almost (Bieniawski, 1976) and Q (Barton et al., 1974) systems
all stages of geotechnical design for tunnels can be were developed in order to provide tunnel support
discerned. Analysis and computational methods are estimates through a rating of rock masses. The
fields in which significant progress has been made. development of powerful microcomputers and of
Nevertheless, regardless of the capabilities offered user-friendly software prompted a demand for data
by the numerical tools, the results can still involve related to rock mass properties, required as input for
uncertainties when parameters are used directly without numerical analysis or close-form solutions for design-
considering the actual failure mechanism of the rock ing tunnels. This necessity preceded the development
mass in tunnelling. This work analyzes the procedure of a different set of rock mass classifications, and the
that treats tunnel behaviour type as fundamental Geological Strength Index (GSI) is one such classifi-
information for the excavation and support measures. cation. The Hoek and Brown failure criterion (Hoek
The research is based on the analysis of design and et al., 2002) is closely connected to the GSI, covering
construction data from 62 non-urban tunnels of the a wide range of geological conditions and affecting
Egnatia Highway in Northern Greece. Within this the quality of the rock masses, including heavily
framework, one classification scheme and one detailed sheared weak rock masses (Hoek et al., 1998).
characterization scheme for assessing the rock mass The current work argues that use of the classifica-
behaviour in tunnelling are presented. The understand- tion ratings must be accompanied by an understand-
ing of the failure mechanism enables the selection of the ing of the actual rock mass behaviour in tunnelling.
appropriate design parameters and the definition of The behaviour in tunnelling may differ from one rock
the support philosophy. Support principles for every mass to another, even if the rock masses have the
behaviour type are proposed. same classification rating in the same stress field and
the same groundwater conditions. An example of two
rock masses with the same classification rating but
INTRODUCTION different tunnel behaviour is presented in Figure 1.
In the last few decades rapid development in almost The two systems in Figure 1 illustrate that the
all stages of geotechnical design can be discerned. selection of the temporary support measures cannot
Great progress has been made in the area of be based only on a classification rating (either GSI or
analytical and computational methods. Nevertheless, RMR) but that such selection also requires an
the results may still involve errors and uncertainties understanding of the tunnel failure mechanism. For
when these methods are used without consideration instance, it is clear that in the process of design, the
for the actual failure mechanism of an excavated rock structure of the rock mass has to be considered along
mass. Understanding the rock mass behaviour in with the classification index.
tunnelling can ensure selection of the appropriate The classification proposed in this article is focused
design parameters (for the rock mass and/or the on the evaluation of the tunnel behaviour in order to
discontinuities) to be used in numerical analysis and guide the selection of the appropriate design param-
the correct consideration of the principles of tunnel eters and to establish the temporary support philos-
support. ophy. The support measures associated with this
classification are discussed in terms of the specific
mechanism of failure. Hence, attention should be
1
Corresponding author email: [email protected] given to the evaluation of the failure mechanism that

Environmental & Engineering Geoscience, Vol. XVIII, No. 4, November 2012, pp. 327–341 327
 Marinos

Figure 1. Example of two equally rated rock masses with the GSI or RMR system but with completely different behaviour in tunnelling
(modified from Marinos [2010]). The selection of the temporary support measures should not be based only on the classification ratings but
also on the understanding of the tunnel failure mechanism, which is greatly dependant on the rock mass structure.

328 Environmental & Engineering Geoscience, Vol. XVIII, No. 4, November 2012, pp. 327–341
 Assessing Rock Mass Behaviour for Tunnelling

‘‘fits’’ the rock mass after its excavation. Following the evaluation of the geotechnical classification meth-
the evaluation of the mechanism of failure, one can be ods, and the relationships between the rock mass
more confident either in using the rating of the behaviour and the temporary support. Analysis of the
applied classification system or in investigating the database included the correlation between the design
specific geological or in situ characteristics—‘‘keys’’— parameters (such as the GSI, the strength of the intact
that govern the tunnel behaviour of the rock mass. rock [sci], the petrographic factor [mi], the cohesion [c],
Rock mass behaviour evaluation in tunnelling and the angle of friction [phi], the deformation modulus of
its connection to the design process have been the the rock mass [Erm], and the joint properties), the
subject of significant research. Goricki et al. (2004), geotechnical behaviour (such as the critical type of
Schubert (2004), Poschl and Kleberger (2004), and failure: deformations, structurally dependant and
Potsch et al. (2004) have studied rock mass behaviour gravity failures, types of overbreak, groundwater
from the design and construction experience of inflows), and the temporary support data (type,
Alpine tunnels, and Palmstrom and Stille (2007) have geometrical characteristics, and capacity of support
studied this behaviour from other tunnels. measures). The processing and evaluation of this
The present work proposes a classification tool for information contributed to assessing the behaviour of
assessing the rock mass behaviour in tunnelling; A the ground and the formulation of correlations between
‘‘Tunnel Behaviour Chart’’ and, consequently, the this behaviour and the temporary support require-
temporary support principles and measures to con- ments. The use of the TIAS database permitted the
tain each behaviour type are suggested. Additionally, determination of the possible rock mass types (tecton-
a ground characterization methodology for assessing ically disturbed sediments such as flysch, tectonically
the site-specific particularities in detail and, conse- undisturbed clastic sediments such as molasses, ophio-
quently, the behaviour mechanisms and support lites, weathered gneisses, and cataclastic limestones)
measures is proposed. and the engineering geological characterization of weak
or complex rock masses in terms of their behaviour in
SOURCE OF INFORMATION AND A underground construction (Marinos, 2007).
DATABASE TOOL
TUNNEL BEHAVIOUR ASSESSMENT IN THE
This research is based on experiences from the DESIGN PROCESS
design and construction of 62 non-urban two-lane
tunnels of the Egnatia Highway in Northern Greece. Design Procedure and Case Histories in a Variety of
The cross section of these tunnels is 100–120 m2, Rock Masses
constructed conventionally using the top heading and
bench method. In this context, a database termed The excavation of the 62 tunnels along the Egnatia
‘‘Tunnel Information and Analysis System’’ (TIAS) Highway was carried out using conventional drill and
was created by Marinos et al. (2010). The tunnels run blast methods. The design process involved the
across the entire width of Greece, traversing a wide definition of the different rock mass types along the
variety of geological conditions and displaying tunnel alignment, tagged with a classification rating
diversity in terms of the occurrence of rock masses, (mainly GSI and/or RMR), and of the joint
geological structures, and in situ stresses. The data properties. The excavation and support classes were
processed by TIAS are from geological mapping based on numerical analyses that incorporated
(from design records and from 5,872 tunnel face stresses and groundwater conditions. The decision
mapping records), boreholes (659 records), laborato- on which support class should be implemented during
ry tests (7,248 records), site testing, geotechnical construction was based on the geotechnical classifi-
classifications (11,800 design and construction rec- cation rating performed at the tunnel face. From the
ords), and designation of design parameters (1,784 investigation of all these cases it became apparent that
records). Data were also collected and processed with the overall process benefits from an accurate predic-
regard to geotechnical behaviour such as conver- tion of the ground behaviour type before commencing
gence, overbreak, and groundwater inflow (2,781 the main design analysis and the final selection of
design and 16,438 construction records), as well as the the support measures. After the evaluation of the
detailed information on temporary support measures behaviour type the classification ratings can be used
(280 support categories and thousands of construc- with greater confidence or to determine which
tion records) and tunnel construction cost. classification ratings are ‘‘keys’’ that predict the
In addition to compiling information, the purpose of tunnel behaviour of the rock mass in more detail.
the database was to compare the anticipated and the As mentioned above, there has been a serious effort
encountered geological and geotechnical conditions, to develop guidelines and procedures for tunnel

Environmental & Engineering Geoscience, Vol. XVIII, No. 4, November 2012, pp. 327–341 329
 Marinos

design in which the observation on the rock mass cause serious deformation, even under low to medium
behaviour is incorporated into the determination of overburden. A detailed classification of flysch rock
excavation and support classes (Schubert et al., 2003; masses based on their geotechnical behaviour is
Goricki et al., 2004; and Austrian Society for presented in Marinos et al. (2011). When a flysch
Geomechanics, 2010). The first step in this method- rock mass is undisturbed or slightly disturbed,
ology involves the definition of the possible rock mass independent of the siltstone or sandstone propor-
type, the second the evaluation of rock mass tions, high GSI ratings have to be considered. The
behaviour in tunnelling, the third the type of tunnel rock mass behaviour in undisturbed to moderately
excavation-support system, and, finally, the definition undisturbed structures is highly anisotropic and is
of tunnel length with equal support requirements and controlled by the orientation of discontinuities,
the appraisal of time and cost for incorporation in the mainly the bedding, in relation to the orientation of
tender documents. the tunnel. As a result, there is a possibility of wedge
This methodology is used as a model for the detachment and sliding along thin siltstone layers
procedure proposed in this article. The design with low shear strength. The behaviour of the
methodology discussed here incorporates the assess- disturbed structures and, even more significantly, of
ment of the tunnel behaviour type in the selection of the heavily sheared rock mass types is generally
design parameters and the definition of temporary isotropic, controlled by their low strength and low
support measures. A flowchart in Figure 2 (based on modulus of deformability. These masses may develop
Schubert et al. [2003], with modifications) presents noticeable deformation, even under low to medium
this design methodology. As shown in this flowchart, overburden, while at greater depths squeezing condi-
the fundamental link between the rock mass model tions prevail.
and the excavation and support classes is the
definition of the tunnel behaviour type. The assess-
Molassic Rock Masses
ment of the tunnel failure mechanism is the basic
outcome of this work. The investigation of the tunnel behaviour of
The variety of geological formations under differ- molassic formations was based on the excavation
ent in situ stress conditions in both mildly and heavily and support of 12 tunnels along the Egnatia
tectonized rock masses provided a significant amount Highway. Molasse is quite different from flysch,
of information regarding the engineering geological although both consist of the same lithological types,
behaviour of several rock mass types. This behaviour since molasse is formed after orogenesis and has not
was evaluated and analyzed based on the geotechnical suffered the consequent deterioration of the rock
characteristics of every rock mass type in association mass quality as a result of compressional tectonics.
with the temporary support measures (designed and Molassic rocks have been categorized into five rock
implemented). These investigations included hetero- mass types according to their sandstone-siltstone
geneous rock masses, such as flysch or molassic proportions and their structure and fabric. In depth,
formations, as well as sound, disturbed, and weath- where the molassic rocks are confined and restricted
ered gneiss; ophiolites; and limestones. from the slaking process, the rock mass is continuous
The general characteristics and the behaviour of the and massive, with little sign of stratification. If no
studied rock masses are briefly presented in the discontinuities are present, as is typically the case,
following sections. GSI values are very high, and the rock mass can be
treated as intact, with engineering parameters pro-
vided by direct laboratory testing. For the maximum
Flysch Rock Masses
depth investigated (150 m), the mollasic siltstones
In order to investigate the rock mass properties and presented minimal deformations because of the intact
tunnel behaviour of flysch, 12 tunnels with depths of nature of their structure. Only gravity-driven failures
up to 500 m were examined. Flysch formations are were encountered when the siltstone bedding planes
generally characterized by strong heterogeneity, by were unfavorably oriented.
the presence of members with low strength, and by
tectonically disturbed structures that may contribute Ophiolitic Rock Masses
to heavily sheared and chaotic masses. Flysch
formations were classified into 11 rock mass types The study of ophiolites is based on the design and
according to the siltstone-sandstone proportions and construction data from seven tunnels. The classifica-
their tectonic disturbance (Marinos, 2007). Thus, tion and tunnel behaviour of ophiolites are described
flysch can be stable (even under high overburden), by Marinos et al. (2005). Rock masses in an ophiolitic
exhibit wedge sliding and chimney type failures, or complex exhibit a wide range of engineering behaviour

330 Environmental & Engineering Geoscience, Vol. XVIII, No. 4, November 2012, pp. 327–341
 Assessing Rock Mass Behaviour for Tunnelling

Figure 2. Flowchart of the design procedure for tunnelling using conventional drill and blast excavation. Based on Schubert et al. (2003),
as modified by the author.

Environmental & Engineering Geoscience, Vol. XVIII, No. 4, November 2012, pp. 327–341 331
 Marinos

in tunnelling. This is the result of their petrographic (ranging from a few cubic millimeters to 80 m3).
variety and structural complexity. A high degree of These failures were provoked by the presence of
serpentinization together with the intensity of shearing fractured gneiss, the low shear strength characteristics
may result in a mass in which it is difficult to identify of the surfaces of discontinuities due to weathering or
any initial texture or fabric. Consequently, behaviour alteration, and the presence of clay material in
can vary from stable to severe squeezing conditions in brecciated fault zones. These failures were prevented
cases involving ophiolites that are associated with when pre-support systems, such as spilling, were
overthrusts. The main rock mass types are peridotites, implemented and were strongly linked to the length of
gabbros, pillow lavas, peridotites more or less serpen- each of the tunnel advances. A large tunnel advance
tinized, serpentinites, schisto-serpentinites, sheared per round activated wider zones, disrupting the
serpentinites, and chaotic masses in melanges. Perido- interlocking between the pieces of the rock mass.
tites are strong and behave as typical brittle materials.
However, serpentinization can be present on the
Limestone Rock Masses
surface of discontinuities, and the conditions of the
joints are dramatically reduced to very poor with Two tunnels in heavily fractured limestones and
coatings of ‘‘slippery’’ minerals, such as serpentine or three others under normal geological conditions were
even talc. In a disturbed ophiolitic mass, the serpenti- examined in this study. Limestones were classified
nization process often loosens and disintegrates parts into eight rock mass types according to the tectonic
of the rock matrix itself, which not only contributes to disturbance, the bedding pattern, and the presence of
a lower GSI value but also reduces the intact strength. frequent intercalations of thin pelitic layers (Marinos,
In good-quality masses of peridotite, simple and 2007). When limestones are overstressed by thrusting,
straightforward tunnelling conditions can be expected, disintegrated rock masses with a mixture of angular
in which the attention has to be concentrated on pieces can be produced. The RQD of such rock
avoiding structural instabilities from wedges. In the masses is zero. However, a good frictional strength
case of a more fractured peridotite, schistose, or good may be present and may give ‘‘assistance’’ to the
serpentinite, the behaviour is controlled by sliding and support measures. Ravelling was observed immedi-
rotation on discontinuity surfaces with relatively little ately after the excavation of the heavily disintegrated
failure of the intact rock pieces. In this case the control cases, and pre-support support elements such as
of stability can be improved during excavation of the grouted spiles were applied. Thin to medium bedded
tunnel by keeping the rock mass confined. In poor- limestones, when tectonically undisturbed, present a
quality serpentinite, due either to alteration or tight structure with tight bedding planes at depth, and
shearing, blockiness may be almost completely lost, light support measures were used (GSI design value
and clayey sections with swelling materials may be .60). Structural instability, in which the weak
present. Tunnel instability will then be due to stress- bedding planes allowed movement of blocks and
dependant rock mass failure with significant squeezing wedges, was the main design issue in low-stress
at depths. environments. In cases of intercalations of clayey
rocks, in a folded highly disturbed rock mass, these
plastic members are often sheared and the limestone
Gneissic Rock Masses
beds are broken into small blocks, and chimney-type
Six tunnels in a gneissic environment along the failures may be provoked. Under high overburden the
Egnatia Highway in Northern Greece were investi- total rock mass deforms in response to stress-
gated (Marinos et al., 2010). Fresh gneiss forms very dependant behaviour.
competent rock masses with minor problems in tunnel
works. However, in an environment resulting from Tunnel Behaviour Types
intensive and successive tectonic disturbance, gneiss
can produce poor to very poor rock masses if The term ‘‘failure mechanism–behaviour type,’’ as
brecciation associated with weathering and alteration referenced here, involves all of the mechanisms that
occurs at various degrees and depths. In such endanger the tunnel section when the rock mass has
conditions, the intact rock and the rock mass strength not yet been supported after excavation.
present a wide range in values, and the tunnel This section presents the tunnel behaviour types as
behaviour can vary from good, requiring only simple they have been described by Terzaghi (1946) and
support, to extremely problematic. Accordingly, the Schubert et al. (2003) but also as they have been
rock mass has been categorized into eight rock mass configured by the author from the experiences of the
types. The main instability encountered in the tunnels construction of the 62 Egnatia Highway tunnels and
was the frequent but irregular chimney-type failure from other cases in Greece. The behaviour types,

332 Environmental & Engineering Geoscience, Vol. XVIII, No. 4, November 2012, pp. 327–341
 Assessing Rock Mass Behaviour for Tunnelling

apart from ‘‘stable’’ conditions, are divided into that cannot be described through a standardized
structural (gravity)–controlled failures (wedge and classification.
chimney-type failures, ravelling, and flowing ground)
and stress-controlled failures (shear failures, squeez-
Tunnel Behaviour Chart
ing and swelling, anisotropic deformations, and
brittle failures). All of the presented behaviour types The TBC links the rock mass information directly
were experienced in the investigated tunnels, with the with the design and the tunnel support principles and
exceptions of the rock bursting (Br), swelling (Sw), covers a wide range of cases. TBC is a classification
and anisotropic deformation (San) types. In addition, for the estimation of tunnel behaviour and requires
the limits and ranges, for which every behaviour type only three parameters: the rock mass structure, the
is applied, are briefly described and shown in overburden (H), and the intact rock strength (sci).
Figure 3. This is an integrated classification based on the TIAS
The behaviour types are grouped based on the database and the experiences from the design and
investigation of tens of rock mass types, their rock construction data of the 62 recently driven tunnels in
mass and joint strength properties, and their observed Greece, as mentioned earlier.
behaviours under different stress conditions (from 30- The suggested system is presented in Figure 4. The
m to 500-m overburden). To define the stress-induced purpose of this diagram is to predict the critical failure
failures, the ratio of the rock mass strength to the in mechanism of several common rock mass types and
situ stresses, scm/po (Hoek and Marinos, 2000), is conditions. The examined cases involved intact
used. Gravity-driven failures are generally distin- strengths to about 100 MPa and depths not exceeding
guished according to the rock mass structure (original 500 m, while the majority of cases were under 300 m. It
conditions and tectonic transformation) and the is noted that the quantified ranges of the sci and the H
confinement conditions. Of course, there are cases used in the chart are reasonable trends but should only
in which both stress- and gravity-induced failures can be considered as indicative. TBC cannot be used at
be met in a rock mass. In those cases, special attention very significant depths, where brittle failures such as
should be given to the most critical failure mode for spalling and rock burst can be generated.
the selection of the appropriate support measures. The philosophy of the TBC becomes obvious if one
acknowledges that the rock mass structure is a basic
parameter with which to estimate its immediate
TUNNEL BEHAVIOUR ASSESSMENT
response in underground excavation. From the
General structure of the rock mass one can ‘‘read’’ the
intensity of tectonic disturbance, the blockiness of
A basic analysis in the TIAS database involved the the mass, the probable size of blocks, the shape of
correlation among rock mass parameters, behaviour rock elements (massive, blocky, foliated, or sheared),
type, and temporary support data. The processing and the ability of the rock blocks to rotate. Rock
and evaluation of this information allowed the mass structures were categorized according to the
assessment of the actual behaviour for a number of GSI system (Marinos and Hoek, 2000).
rock mass types and the identification of the These characterizations (intact, blocky, very blo-
temporary support requirements required to control cky, blocky/disturbed/seamy, disintegrated, sheared/
the failure mechanism type. This procedure resulted foliated/laminated) are well known and can provide a
in a methodology and classification. good description for the rock mass. Other descrip-
The assessment of the rock mass behaviour in tions, such as ‘‘good,’’ ‘‘medium,’’ or ‘‘poor’’ to ‘‘very
tunnelling is completed using two classification and poor,’’ are very general and are not adequate for the
characterization methods. The first, called the Tunnel purpose of identifying a tunnel behaviour. For
Behaviour Chart (TBC), is a classification based on example, a poor structure can be either a foliated-
three basic parameters. The second, termed Ground sheared material with clayey fractions or a disinte-
Characterization, Behaviour and Support for Tunnels, grated brecciated rock. In tunnelling, the first may
is a step-by-step detailed assessment of the tunnel behave as a squeezing material while the second will
behaviour. suffer from gravity-driven failures (ravelling).
A classification, like TBC, has the great benefit of In addition to the rock mass structure, tunnel cover
offering an immediate, easily assessed result using a H has a significant influence on the failure mechanism
common worldwide language between the users. since it relates to the in situ stress and the general
However, it is not to be used as a ‘‘cookbook’’ confinement conditions. The behaviour types that
without considering the ranges of its application and were examined are under a cover of 30 m to 500 m.
the possibility of the presence of geological features For the gravity-driven failures, the tunnel depth

Environmental & Engineering Geoscience, Vol. XVIII, No. 4, November 2012, pp. 327–341 333
 Marinos

Figure 3. Brief description and schematic presentation of tunnel behaviour types (modified from Marinos [2010]). (Based on data from
Schubert et al. [2003]; Terzaghi [1946], and on the work of the author). (Photos from the author, except for ‘‘Sq,’’ from E. Hoek [personal
communication], and ‘‘San,’’ from Seingre [2005].)

334 Environmental & Engineering Geoscience, Vol. XVIII, No. 4, November 2012, pp. 327–341
 Assessing Rock Mass Behaviour for Tunnelling

Figure 4. Tunnel Behaviour Chart (TBC): An assessment of rock mass behaviour in tunnelling (modified from Marinos [2010]).

Environmental & Engineering Geoscience, Vol. XVIII, No. 4, November 2012, pp. 327–341 335
 Marinos

influences the extent of a failure, since the degree of Marinos, 2000). In particular, when scm/po is between
interlocking between the rock blocks changes and the 0.3 and 0.6, shear failures can propagate in a shallow
confinement pressure differs with depth. For example, zone around the tunnel perimeter (Sh behaviour). Such
ground may ravel (Rv) close to the ground surface, cases involve rock masses with poor to very poor
but under higher cover a chimney-type (Ch) failure structures and low intact rock strength (,10 MPa to
may be experienced. As far as the stress-controlled 15 MPa) under medium overburden or with good
behaviour is concerned, H defines when shear failures structure and low intact rock strength under high cover.
and deformations are generated. These limits are Squeezing conditions (Sq behaviour) with severe tunnel
estimated to be 150 m for good structure (intact and deformations may be developed when scm/po , 0.3.
blocky-seamy undisturbed), 100 m for very blocky Brittle failure or spalling of very strong brittle rocks is
structure, and 70 m for the very poor structure not dealt with in this article since it was not encountered
(seamy-disturbed, disintegrated, foliated-sheared). in the Egnatia tunnels covered by this study, given that
These values are mainly determined by back-analysis the overburden was limited to 500 m.
and by the calculated values of the ratio scm/po, with
po indicating the in situ stress that is considered
Gravity-controlled Failures
isotropic. For instance, if scm/po , 0.3, squeezing
conditions are likely; if 0.3 , scm/po , 0.6, minor to Gravity-driven failures can occur in rock masses
medium deformations may be occurring; and if scm/ controlled by discontinuities. When the rock masses
po . 0.6, minor or no deformations are anticipated. are exposed after the excavation, wedges may fall or
The sci values that were considered in the design of slide, according to the tunnel geometry, the orien-
these tunnels ranged between 5 MPa and 100 MPa. tation, and the shear strength characteristics of the
The selected values considered to define the rock mass discontinuity planes. Wedge failures (Wg), Ch, or
behaviour classes were based on the value when shear Rv types can take place in rock masses with poor
failures and deformations initiate. This limit is interlocking of blocks or rock pieces. The rock mass
estimated at 15 MPa. cannot arch after the initiation of the fall, and the
As an example of how the TBC works, consider overbreak may be large and irregular. The volume
that the ground structure is described as ‘‘disintegrat- and frequency of these behaviour types depend on
ed’’ (e.g., brecciated limestone), the intact strength is the structure of the rock mass (‘‘blocky-disturbed’’
above 15 MPa (e.g., limestone), and the overburden is and ‘‘disintegrated’’), its relaxation (‘‘open struc-
less than 70 m; in this case, the likely behaviour is ture’’), and the tunnel depth. With increasing tunnel
Type 18 (i.e., ‘‘ravelling ground’’). For the same depth the rock mass quality is improved and the
material, if the overburden is greater than 70 m, then confinement pressure ‘‘tightens’’ the structure of the
the likely behaviour is Type 20 (i.e., ‘‘chimney-type mass.
failure–shear failure’’).
The surface condition of the discontinuities, the
TUNNEL SUPPORT MEASURES
second component of the GSI system, mainly affects
the intensity and the probability that the failure Following the assessment of the rock mass behav-
phenomenon will occur, but it does not define the iour in tunnelling, the temporary support system can
behaviour type. There are only a few cases in which be discussed. The design of the temporary support
the surface quality of the discontinuities can influence categories is composed of two stages: the selection of
a behaviour type. For example, an abundant presence the proper support elements and their detailed
of clay along the discontinuities in the rock mass may analysis. The selection of the support elements relies
shift the gravity-driven behaviour types down the heavily upon the behaviour of the rock mass. The
vertical axis of the chart (e.g., from Wg [9] to Ch [13]). choice of support should be based equally on
The presence of groundwater does not affect the experience and geotechnical appreciation and on the
behaviour type but it does affect the intensity of the analytical solutions.
failure and the factor of safety. However, in some The temporary support philosophy and principles
cases, as in a ‘‘disintegrated’’ rock mass, the presence for every tunnel behaviour type are illustrated in
of groundwater may shift a Ch or Rv behaviour type Figure 5. It is noted that these suggestions are focused
to a flowing ground (Fl) type. on containing and controlling the specific mechanism
of failure. These proposals are derived from the
Stress-controlled Failures experiences of the design and construction of the 62
tunnels of the Egnatia Highway in a great variety of
The development of significant deformations around rock masses and geological conditions and from the
a tunnel is characterized by the scm/po ratio (Hoek and numerous support systems implemented to control

336 Environmental & Engineering Geoscience, Vol. XVIII, No. 4, November 2012, pp. 327–341
 Assessing Rock Mass Behaviour for Tunnelling

Environmental & Engineering Geoscience, Vol. XVIII, No. 4, November 2012, pp. 327–341
Figure 5. Indicative support measures and principles for each type of rock mass behaviour and associated mode of failure (for a top heading and bench method) (modified from
Marinos [2010]). The illustrations of the tunnel are sketches; this shape corresponds to the top heading in a weak rock mass tunnel.

337
 Marinos

Figure 6. (A) Ground Characterization, Behaviour and Support for Tunnels (Sheet 1/2). Illustrated, in light characters, by an example of
tunnelling in a serpentinized peridotite.

338 Environmental & Engineering Geoscience, Vol. XVIII, No. 4, November 2012, pp. 327–341
 Assessing Rock Mass Behaviour for Tunnelling

Figure 6. (B) Ground Characterization, Behaviour and Support for Tunnels (Sheet 2/2). Illustrated, in light characters, by an example of
tunnelling in a serpentinized peridotite.

Environmental & Engineering Geoscience, Vol. XVIII, No. 4, November 2012, pp. 327–341 339
 Marinos

stress-induced failures and/or to confine gravity- choose if the behaviour is controlled by the overall
driven instabilities. The support suggestions for every rock mass or by the discontinuities or both. The rock
behaviour type are established for the following mass ‘‘Behaviour’’ during excavation of the tunnel is
conditions: (1) overburden depth up to 500 m; (2) then assessed (Step IV).
tunnel excavation with a horseshoe section of 100– After the identification of the failure mechanism,
120-m2 area; and (3) top heading and bench the suitable ‘‘Design Parameters’’ are entered,
excavation (mechanical or drill and blast). according to the principles of the failure mechanism
It is emphasized that the proposals for support (Step V). If the behaviour of the rock mass can be
presented in Figure 5 do not replace the proper design considered as isotropic and governed by stress-
methodology but rather allow for an early assessment induced failures (Sh, Sq; see Figure 2), the user must
of the principles of the appropriate support measures be focused on rock mass parameters (section Va). On
and their basic dimensioning, as they are dictated by the other hand, if the principal behaviour type is
the ground behaviour and the associated failure gravity-controlled failures (e.g., Wg, Ch, Rv; see
mode. Figure 2), the user must focus on parameters related
to the discontinuities (section Vb). If the rock mass is
weak but also anisotropic (e.g., as a result of
GROUND CHARACTERIZATION FOR
schistosity or well-defined bedding planes), both the
TUNNELS—A DETAILED
rock mass parameters and the persisting joint
STEP-BY-STEP PROCEDURE
properties must be considered (San; see Figure 2).
Apart from a classification such as TBC that covers Parameters of geotechnical classification systems
a wide range of engineering geological and tunnel RMR, Q, and GSI can be reported here (Step V).
conditions, the estimation of the tunnel behaviour Since most tunnel designs now involve numerical
and the philosophy of the support measures should analyses, the question is whether to use rock mass
also be based on a detailed ground characterization. parameters (shear strength of rock mass, cmass, Qmass,
This detailed characterization cannot bypass the Emass, etc.) when the rock mass behaves isotropically
geological and/or in situ characteristics dictating or or to include the discontinuity parameters (orienta-
influencing the tunnel behaviour compared with a tion, distribution, persistence, shear strength, cjoint,
standardized classification that could miss the specif- Qjoint) when the behaviour is controlled by the joints
ics and particularities of and around a tunnel section. or influenced by the resulting anisotropy (anisotropic
When tunnelling, for example, in a sheared siltstone deformations, San).
flysch formation under several tens of meters of Next, the ‘‘Tunnel Support Philosophy’’ (support
overburden (e.g., 100 m), deformations are expected. type) is discussed (Step VI), and, finally, the
However, when there is a ‘‘package’’ of good-quality ‘‘Remaining Risk’’ (after the installation of the
beds (e.g., not sheared flysch with sandstone beds) temporary support) is reported (Step VII).
close to the tunnel roof, the deformations could be
significantly fewer. In this case, there may be no need CONCLUSIONS
for a very heavy support category.
This characterization is named ‘‘Ground Charac- The assessment of ground behaviour in tunnelling
terization, Behaviour and Support for Tunnels’’ and is based on the identification of the rock mass type
prompts the user to evaluate the data in detail in under the site conditions of the tunnel, which leads to
order to assess the tunnel behaviour and the an understanding of the mechanism of failure. This
appropriate support measures. This characterization procedure assists the designer in the analysis of tunnel
is performed on two sheets. The first sheet (Sheet 1/2) behaviour and the selection and support measures
is presented in Figure 6A and the second sheet (Sheet and also in the establishment of the contractual
2/2) in Figure 6B. The characterization is presented documents and guidelines for the construction. This
by means of an example (light characters). study was based on a large set of data from the design
This approach starts with the definition of the and construction of 62 tunnels through a wide variety
‘‘Geological Model’’ (Step I) of the site-specific area. of geological conditions.
This is based on lithological data, tectonism, weath- Two classification and characterization schemes
ering, and permeability. This is followed by ‘‘In Situ have been developed. The first, labelled the TBC, is a
Conditions’’ (stresses, hydrogeological conditions, classification system for assessing the rock mass
and other boundaries) and ‘‘Tunnel Geometry’’ (Step behaviour in tunnelling and covers a wide range of
II). The next step identifies the rock ‘‘Characteristic– rock mass conditions. This assessment is based on the
‘Keys’,’’ which dictates the stability or instability of structure of the rock mass, the strength of the intact
the tunnel (Step III). At this point, the user must rock, and overburden depth. The second method,

340 Environmental & Engineering Geoscience, Vol. XVIII, No. 4, November 2012, pp. 327–341
 Assessing Rock Mass Behaviour for Tunnelling

called Ground Characterization, Behaviour and MARINOS, P. AND HOEK, E., 2000, GSI: A geologically friendly tool
Support for Tunnels, is an assessment of how a rock for rock mass strength estimation. Technomic publishers,
Lancaster. In Proceedings of the GeoEng2000 at the Interna-
mass type will behave and of its support requirements. tional Conference on Geotechnical and Geological Engineering:
This characterization is based on the geological model Technomic Publishers, Lancaster, U.K., pp. 1422–1446.
and on the appreciation of the rock mass characteristics MARINOS, P.; HOEK, E.; AND MARINOS, V., 2005, Variability of the
controlling the potential for instability of the rock mass. engineering properties of rock masses quantified by the
Having defined the most critical failure mechanism, geological strength index. The case of ophiolites with special
emphasis on tunnelling: Bulletin Engineering Geology Environ-
the temporary support philosophy and principles for ment, Vol. 65, No. 2, pp. 129–142.
every behaviour type are discussed in order to contain MARINOS, V., 2007, Geotechnical Classification and Engineering
and control the specific failure mechanism upon Geological Behavior of Weak and Complex Rock Masses in
excavation. The appropriate design parameters can Tunnelling [in Greek]: : Doctoral Thesis, National Technical
then be selected. These proposals allow an early University of Athens (NTUA).
assessment of the principles for the choice of MARINOS, V., 2010, Engineering geological behaviour of rock
masses in underground excavations. In Proceedings of the 11th
appropriate support measures and their basic dimen- International congress of IAEG, Auckland, New Zealand.
sioning, as dictated by the ground behaviour and the MARINOS, V. AND DROSOS, G., 2010, Tunnelling through gneiss. A
associated mode of failure. competent or a problematic rock mass? In Proceedings of the
11th Congress of the International Association of Engineering
Geology and the Environment, Williams, A. L., Pinches, G.
ACKNOWLEDGMENTS M., Chin, C. Y., McMorran, T. J., and Massey, C. I. (eds),
TRC Publishers, (on CD), Auckland, New Zealand.
The author would like to thank Egnatia Odos S.A. MARINOS, V.; FORTSAKIS, P.; AND PROUNTZOPOULOS, G., 2011,
for its support and the data provided. Dr. Evert Hoek Estimation of geotechnical properties and classification of
is gratefully acknowledged for reviewing the final geotechnical behaviour in tunnelling for flysch rock masses. In
Anagnostopoulos, A. (Editor), Proceedings of the 15th European
version of this article. He would like also to Conference on Soil Mechanics and Geotechnical Engineering,
acknowledge the valuable contributions of Dr. Rick ISO publisher, Part 1. Athens, Greece, pp. 435–440.
Nolting, Dr. Eric S. Lindquist, and an anonymous MARINOS, V.; KORKARIS, K.; FORTSAKIS, P.; PROUNTZOPOULOS, G.;
reviewer for their constructive comments during the MIRMIRIS, K.; PAPOULI, D.; AND MARINOS, P., 2010, TIAS
review process. Finally, the author expresses his database: A tunnel information and analysis system. In
Proceedings of the 11th Congress of the International Association
appreciation to Ms. D. Papouli, Geologist, M.Sc.,
of Engineering Geology and the Environment, Williams, A. L.,
for her assistance in editing the article. Pinches, G. M., Chin, C. Y., McMorran, T. J., and Massey, C.
I. (eds), TRC Publishers, (on CD), Auckland, New Zealand.
PALMSTROM, A. AND STILLE, H., 2007, Ground behaviour and rock
REFERENCES engineering tools for underground excavations: Tunnelling
AUSTRIAN SOCIETY FOR GEOMECHANICS, 2010, Guideline for the Underground Space Technology, Vol. 27, pp. 363–376.
Geotechnical Design of Underground Structures with Conventional POSCHL, I. AND KLEBERGER, J., 2004, Geotechnical risks in rock
Excavation: Translated from version 2.1, 29 p, 7-page Appendix. mass characterization: Tunnels & Tunnelling International,
BARTON, Y. O.; LIEN, R.; AND LUNDE, J., 1974, Engineering Part 1, May Issue, pp. 37–39.
classification of rock masses for the design of tunnel support. POSCHL, I. AND KLEBERGER, J., 2004, Geotechnical risks in rock
Rock Mechanics, Vol. 6, No. 4, pp. 189–239. mass characterization: Tunnels & Tunnelling International,
BIENIAWSKI, Z. T., 1976, Rock mass classification in rock Part 2, October Issue, pp. 36–38.
engineering. In Bieniawski, Z. T. (Editor), Exploration for POTSCH, M.; SCHUBERT, W.; GORICKI, A.; AND STEIDL, A., 2004,
Rock Engineering, Proceedings of the Symposium: Balkema, Determination of Rock Mass Behaviour Types: A Case
Cape Town, South Africa, pp. 97–106. Study: In Rock Engineering - Theory and Practice, EUROCK
GORICKI, W.; SCHUBERT, G.; AND RIEDMUELLER, G., 2004, New 2004 and 53rd Geomechanics Colloquium, Schubert ed., VGE
developments for the design and construction of tunnels in publisher, pp. 75–80.
complex rock masses: International Journal Rock Mechanics SCHUBERT, W., 2004. Basics and Application of the Austrian
Mining Sciences, Vol. 41, No. 3, pp. 497–498. Guideline for the Geomechanical Design of Underground
HOEK, E.; CARRANZA-TORRES, C.; AND CORKUM, B., 2002, Hoek– Structures, Schubert ed., VGE publisher, EUROCK 2004 and
Brown criterion—2002 edition, 2002. Mining and Tunnelling 53rd Geomechanics Colloquium.
Innovation and Opportunity, Proceedings of the 5th SCHUBERT, W.; GORICKI, A.; AND RIEDMULLER, G., 2003, The guideline
NARMS-TAC Conference, Hammah, R., Bawden, W., for the geomechanical design of underground structures with
Curran, J., and Telesnicki, M. (eds), University of Toronto conventional excavation: Felsbau, Vol. 21, No. 4, pp. 13–18.
press, Toronto, Canada, Vol. 1, pp. 267–273. SEINGRE, G., 2005, Tunnel de base du Lötschberg—Bilan de
HOEK, E. AND MARINOS, P., 2000, Predicting tunnel squeezing in l’excavation aux tunneliers: In Proceedings of Geology and
weak heterogeneous masses: Tunnels Tunnelling International: Linear Developments 2005, Arnould, M., Ledru, P. (eds), (on
Part 2, December Issue, pp. 34–36. CD), May 23–25. BRGM editions, Lyon, France.
HOEK, E.; MARINOS, P.; AND BENISSI, M., 1998, Applicability of the TERZAGHI, K., 1946, Rock defects and load on tunnel supports. In
Geological Strength Index (GSI) classification for weak and Proctor, R. V. and White, T. L. (Editors), Introduction to
sheared rock masses: Bulletin Engineering Geology Environ- Rock Tunnelling with Steel Supports: Commercial Shearing &
ment, Vol. 57, No. 2, pp. 151–160. Stamping Co., Youngtown, OH.

Environmental & Engineering Geoscience, Vol. XVIII, No. 4, November 2012, pp. 327–341 341