TRANSPORT RESEARCH LABORATORY

Ground classification systems in tunnelconstruction

Prepared for Quality Services (Civil Engineering), HighwaysAgency

G I Crabb

TRL REPORT 280

Transport Research Foundation Group of CompaniesTransport Research Foundation (a company limited by guarantee) trading as TransportResearch Laboratory. Registered in England, Number 3011746.

TRL Limited. Registered in England, Number 3142272.Registered Offices: Old Wokingham Road, Crowthorne, Berkshire, RG45 6AU.

First Published 1997ISSN 0968-4107

Copyright Transport Research Laboratory 1997. All rights reserved.

The information contained herein is the property of the TransportResearch Laboratory and does not necessarily reflect the views orpolicies of the customer for whom this report was prepared. Whilstevery effort has been made to ensure that the matter presented in thisreport is relevant, accurate and up-to-date at the time of publication,the Transport Research Laboratory cannot accept any liability for anyerror or omission.

CONTENTS

Page

Executive Summary 1

1 Introduction 3

2 Ground classification systems 3

2.1 Introduction 3

2.1.1 Classification by strength 3

2.1.2 Choice of support 3

2.1.3 The concept of classification systems 4

2.2 Classification parameters 4

2.2.1 Intact rock strength 4

2.2.2 Number of joint sets 4

2.2.3 Discontinuity spacing 5

2.2.4 Discontinuity condition 6

2.2.5 Discontinuity orientation 6

2.2.6 Groundwater conditions 6

2.3 Classification systems 6

2.3.1 Rock load classification 6

2.3.2 ‘Stand-up time’ concept 7

2.3.3 Rock quality designation (RQD) 7

2.3.4 Rock structure rating (RSR) 7

2.3.5 Geomechanics classification or Rock Mass Rating system (RMR) 8

2.3.6 Q - system 9

2.3.7 New Austrian Tunnelling Method (NATM) classification 11

2.3.8 Contractual matters 12

3 Literature review 13

3.1 Comparisons of and observations on ‘Q’ and RMR systems 13

3.2 Face logging 15

3.3 Block fallout 15

3.4 Norwegian method of tunnelling (NMT) 16

3.5 Probabilistic methods 17

3.6 Numerical modelling 17

4 Case histories 19

4.1 Round Hill tunnels 19

4.2 Penmaenbach tunnel 20

4.3 Pen-y-Clip 21

4.4 Carsington Aqueduct 23

4.5 Tyne-Tees Aqueduct 24

iii

iv

5 Conclusions 24

6 Recommendations 25

7 Acknowledgement 26

8 References 26

8.1 Rock Classification 26

8.2 Case history references 28

Appendix: Other useful references 29

Abstract 30

Related publications 30

1

classification system can be applied to the groundconditions encountered and used to assist with selection ormodification of the support to be used for each advance ofthe face. Payment and settlement of any disputes betweenthe contract parties can be facilitated by the comparison ofthe ground classifications agreed at the tunnel face, andspecified ‘ground reference conditions’ based on the pre-construction site investigation data. A form of Contractwhich allows flexibility in support choice and an equitablesharing of risk, is needed to facilitate this.

The most popular systems in use are the ‘Q’ system,developed at the Norwegian Geotechnical Institute(Barton, 1974), and the Rock Mass Rating (RMR) system,or Geomechanics’ system developed by Bieniawski(1973). The ‘Q’ system has been shown to be generallyless conservative and more sensitive to changes in rockquality than the RMR system. The system was originallydeveloped using just over two hundred case histories,mainly in Scandinavia. It is now backed by a database ofover one thousand case histories in a wide variety ofground types and support systems, from many differentcountries. The ‘Q’ system utilizes the classificationparameters RQD (a measure of joint spacing), number ofjoint sets, joint roughness, joint condition or alteration,groundwater inflow and stress condition. The individualparameters are multiplied to give the rock quality ‘Q’,which ranges from about 0.001 to 1000. This figure can beused to give an expected stand up time for specificunsupported spans and together with the effective span anda factor of safety, used to select an appropriate supportdesign. Correlations with ‘Q’, which allow estimation ofdeformation modulus and mass strength have beendeveloped. These are properties which are needed for theuse of analytical methods and are otherwise difficult todetermine, requiring large scale tests.

A good correlation between ‘Q’ and seismic velocityhas been observed in many types of ground. Themeasurement of seismic velocity during site investigationand when probing ahead from the tunnel face is apromising technique. However more data is required, inparticular on the effect of overburden stress, before thistechnique can be used with confidence.

Executive Summary

Ground classification systems play an important role insetting up and managing tunnel construction contractsbecause they may determine the level of payment to thecontractor. However because of the rapid rate ofdevelopment in tunnel design, currently employedmethods may no longer be the most efficient. Theclassification of the ground for tunnel support is an areawhere improvements may be possible in the planning andexecution of future tunnelling contracts.

This study was commenced in December 1994 to reviewthis area and identify the best route forward. The objectiveis to give guidance to the Highways Agency by indicatingan approach to ground classification which will assist insetting-up cost-effective contracts for new tunnellingworks.

This report reviews the development and use ofempirical ground classification systems for the selection oftunnel support. Recent UK and other practice is examinedto identify new developments and note particular problemareas. The concept and nature of the systems is discussedin the first section. Recent literature is then examined anddiscussed, followed by a review of UK case histories oftunnelling where ground classification schemes were used.Final sections draw together the most importantconclusions to be gleaned from the evidence and somerecommendations are made on the most profitabledirection of future research.

Ground classification systems have been developed overthe past fifty years and have become increasinglysophisticated. They provide an empirical method, based onthe analysis of case histories, of assisting with the selectionof tunnel support systems. They can be of use during thesite investigation, design and construction phases of atunnelling project. During site investigation a classificationsystem can be used to define the appropriate parameters ofthe ground to be measured to enable an adequateclassification to be carried out. During the design phasepreliminary selection of the support for sections of tunnelcan then be made by the use of the chosen classificationsystem. This should always be checked and refined by theuse of an analytical method. During construction the

2

3

1 Introduction

This report reviews the development and use of empiricalground classification systems for the selection of tunnelsupport. Recent UK and other practice is examined toidentify new developments and note particular problemareas. The basic concept and nature of the systems arediscussed in the first section. The most recent literature onthese systems is then examined and discussed, followed bya review of UK case histories of tunnelling where groundclassification schemes were used. Final sections drawtogether the most important conclusions to be gleanedfrom the evidence and some recommendations are madeon the most profitable direction of future research.

2 Ground classification systems

2.1 Introduction

The process of design of a tunnel support system to resistground loading includes several steps:

1 Characterisation of the ground mass.

2 Definition of the ideal geometry of the scheme.

3 Consideration of the ground/lining interaction effect onsupport load.

4 Selection of lining type and design of lining.

The process will often be an iterative one, particularly asat early stages in the design process there may be verylittle site investigation data available. Where theobservational method is used the design of the lining maybe changed as construction proceeds, based on measuredloads and deformations. The type of lining chosen mayadditionally be influenced by constructional andoperational requirements. The former could beenvironmental restrictions on the construction method andphysical or economic limitations on plant, and the latter arequirement for a smooth bore. Ground classificationsystems provide an empirical method, based on theanalysis of case histories, of applying the informationobtained in steps 1 and 2 above to the designconsiderations in steps 3 and 4. In any major tunnellingscheme the empirical design should be backed up by theuse of analytical methods and engineering judgement. It ispossible that such analytical methods will eventuallysupersede the empirical methods. However the complexnature of the problem makes this unlikely in the nearfuture. Ground classifications can be used both at thedesign stage using data from the site investigation, andduring construction to assist the selection or modificationof support at the tunnel face, based on the conditionsactually encountered.

It should be noted that the application of these systemsat the design stage is strongly dependant on the quality ofthe site investigation data available. If the site investigationis inadequate then it is inevitable that any design usingrock classification, or any other method, will be inaccurate.

2.1.1 Classification by strengthIn tunnelling there is a traditional classification of theground into soft ground and rocks of various strengths.Soft ground can usually be excavated by hand ormechanical digging equipment and requires immediate orvery rapid support. Soft ground mainly comprises recentalluvium and glacial drift deposits and also the stifffissured clays of the Eocene, Cretaceous and Jurassicperiods. The scales of strength given in BS5930 (BritishStandards Institution, 1981), for clays and rocks, can alsobe useful in discriminating between soft ground and rock,although there is undoubtedly some overlap. In the BritishStandard ‘Hard clay (or Very weak mudstone)’, is definedas having an undrained shear strength greater than 300kPa.In the section on rocks ‘very weak rock’ is defined ashaving an unconfined compressive strength of less than1.25 MPa.

2.1.2 Choice of supportTunnel support systems may be chosen in several ways,the choice of which will depend on the type of ground, thesize and purpose of the excavation, and the pre-existingstresses in the ground. The latter are mainly dependent onthe depth of overburden, but may be affected by adjacenttunnels.

2.1.2.1 Soft groundIn ground classified as soft, tunnels are normallyconstructed by the use of a tunnelling shield and asegmental lining of cast iron or concrete, although recentlyunshielded methods using sprayed concrete linings havebeen used in the London Clay. Research has shown that insoft ground hoop loads in linings rarely exceed a valueequivalent to the full overburden pressure, and typicallyreach about 75% of it in the long term. Design in theseconditions is therefore usually based on the fulloverburden pressure. This may include an allowance fordead loading at the surface. It will not normally benecessary to consider live loading unless there is very littlecover. The ‘full overburden pressure’ criterion is based onmeasurements made in a number of segment lined tunnelsin the London Clay and in particular measurements madeby TRL at Regents Park on the Jubilee line (Barratt et al,1994). This paper concluded that it is highly unlikely thatthe ring loads on segmental linings of either cast iron orconcrete ever exceed those corresponding to an all-roundpressure equal to that due to the overburden. More recentmeasurements supporting this view have been made byTRL at St James’ Park on the Jubilee line extension(Bowers and Redgers,1996). The hoop load reached, andits rate of increase in individual cases,will depend on thetime between excavation and support, the stiffness of thesupport and its degree of contact with the ground. Anoverriding design consideration is likely to be the controlof surface settlement, particularly as most soft-groundtunnels are at shallow depth in urban areas. This willusually require heavier support than that required tostabilize the opening when settlement control is notimportant. The highest and earliest hoop loads will occur

4

when minimization of settlement is required and a stifflining is erected quickly after excavation, with intimateground contact. In order to provide a stiff lining it is likelythat it will have a large cross-sectional area which willresult in relatively low lining stresses.

2.1.2.2 RockIn rock the situation is somewhat different because,depending on the engineering characteristics of the rockand the size and depth of the tunnel, the support neededmay vary from none to a heavy mass concrete lining,depending on the ability of the ground to support itself.Great strides have been made in recent years in therefinement of analytical methods for tunnel design basedon the principles of rock mechanics. In particular thesevere shortcomings in the use of finite element and othernumerical methods to analyse discontinuous rocks arebeing reduced by the development of discrete elementmethods. However, due the variable nature of the groundand difficulties with adequately defining the parameters ofa ground model, empirical rock classification systems fortunnel support estimation are often resorted to both forpreliminary design and to guide support decisions duringconstruction. In effect such systems allow the systematicapplication of past experience to new tunnels. It istherefore important that the classification used is based ona range of case histories which encompass the conditionsto be encountered. It should also be recognised thatbecause of the highly variable nature of the ground suchempirical design cannot be more than an approximateguide to support selection and should never be totallyrelied upon. Analytical studies, field observations andengineering judgement by experienced personnel must alsobe used. It should also be noted that because it isimpossible to determine the factor of safety in the casehistories upon which these empirical methods are basedtheir use may tend to perpetuate excessive conservatism.

2.1.3 The concept of classification systemsBieniawski’s book ‘Engineering Rock MassClassifications - A complete manual for engineers andgeologists in mining, civil and petroleum engineering’(Bieniawski, 1989) provides a good overview of thehistory and use of rock classification systems. The conceptembodied in rock classification systems for tunnel supportis to identify the set of parameters of the rock mass whichare best correlated with its engineering behaviour and toassign numerical weights to the likely range of eachparameter. The parameters chosen must have values whichare readily obtainable from site investigation data. Theparameter weights are derived by adjusting them to best fitavailable case history data. Finally the values ofparameters are combined to produce a single ratingparameter representing the ‘quality’ of the rock. The fullrange that is possible for this quality rating figure can thenbe used to assign descriptive classes to the rock eachassociated with, in conjunction with the size and purposeof the excavation, a suitable type of support. In Barton(1988) the ‘Q’-system of rock classification is likened to

an ‘expert system’ based on the knowledge and experienceof an expert tunnelling consultant.

2.2 Classification parameters

A rock mass generally comprises intact blocks separatedby discontinuities such as joints, bedding planes and faults.The strength and stiffness of the rock mass, and its stabilitywhen excavated, are reduced by the presence of thediscontinuities. The properties of the intact rock aretypically less important than the properties of thediscontinuities in determining the overall properties of therock mass, except in cases where rock stresses areapproaching the intact rock strength. In shallow tunnels thestrength/stress ratio is only likely to be low in the weakestmaterials. The quality and type of classification dataavailable will depend on whether the work is at the siteinvestigation phase or the construction phase. Adescription follows of each of the properties of a rock masswhich may be quantified as part of a rock classificationsystem, with an indication as to the method ofmeasurement. The determination and recording of each ofthese parameters is covered in detail by the publications ofthe International Society for Rock Mechanics (ISRM)Commission on Testing Methods, most of which havebeen brought together in one volume (International Societyfor Rock Mechanics, 1981).

It should be remembered that even with the best siteinvestigation there will be considerable uncertainty as tothe values of the parameters due to the inevitablevariability of the ground, the small size of the sample takenand errors in measurement. Probabilistic methods attemptto accommodate this variability by calculating theprobability distribution of the output of a classificationsystem from the measured or assumed probabilitydistributions of the inputs. This subject will be examined insection 3.5.

2.2.1 Intact rock strengthThe intact rock strength can be assessed by uniaxialcompressive strength tests on prepared cores in thelaboratory, or by the point load test (ISRM, 1985) onsections of cores or irregular lumps in the field. The lattertest produces an index value which can be correlated touniaxial compressive strength. The Schmidt reboundhammer and sonic velocity tests may also be used to giveindex values related to strength and elastic modulus.

2.2.2 Number of joint setsThere is normally more than one set of joints in a rockmass (Figure 1). Spacing data for each significant joint setis required to determine the block size. The size andorientation of blocks relative to the size and orientation ofthe tunnel is an important factor in determining tunnelstability. It is difficult to assess this parameter fromborehole core, an exposure is needed.

5

2.2.3 Discontinuity spacingThe mean discontinuity spacing can be obtained from

both exposures and borehole core. The inverse is oftenquoted as the Fracture Index or Frequency for lengths ofcore with a similar intensity of fracturing. The FractureIndex or Frequency is defined as the number ofdiscontinuities per unit length of core. In the case of coresit is important that fractures caused by the drillingoperation are not counted. When measuring discontinuityspacing on an exposure it is usual to count thediscontinuities intersecting a straight line or lines,

commonly denoted a ‘scanline’, defined on the surface(Figure 2). It is important to realise that the orientation ofthe borehole or scanline relative to the orientation of thesets of discontinuities present is likely to affect the result.Ideally the spacing of each set of discontinuities should bemeasured normal to its plane. This is likely to requireboreholes or scanlines in more than one direction. Priestand Hudson (1979) suggest that scanlines, and therefore byimplication borehole core, should be at least 50 timeslonger than the mean discontinuity spacing to yield results‘to a reasonable precision’. In many cases, particularly inboreholes, the length available will be less than this,introducing a degree of error into the result.

Figure 1 Examples of one and three joint sets (afterBrown, 1981)

One joint set

Three joint sets

Tape

Set 1

Set 2

Set 3

Figure 2 Measurement of joint spacing using a 'scanline'

The Rock Quality Designation (RQD) introduced byDeere et al (1967) is a very commonly used index of rockfracturing, which was developed for core logging. TheRQD is the percentage of length of core or scanlineconsisting of intact pieces longer than 0.1m. Priest andHudson (1976) have observed that the distribution ofdiscontinuities in most rock closely follows the negativeexponential distribution, so in this case there is afunctional relationship between RQD and mean spacing:

RQD = 100e-0.1λ(0.1λ + 1); where λ = mean discontinuityspacing

They also observed that the standard RQD with athreshold value of 0.1m is insensitive to variations in rockquality when the average discontinuity spacing exceeds0.3m. A additional RQD

1.0 index with a threshold of 1.0m

is suggested to extend the range to a discontinuity spacingof about 2.5m, although this idea does not seem to havebeen adopted in any classification system.

6

2.2.4 Discontinuity conditionSome of the most important characteristics ofdiscontinuities are aperture, filling (Figure 3), roughness(Figure 4) and wall strength. Persistence may also beimportant. Unfortunately filling and aperture in particularcannot be obtained from cores, although downholeinspection with a miniature TV camera may yield thisinformation.

2.2.5 Discontinuity orientationThe orientation of discontinuities can be an importantparameter in tunnel stability. This can be obtained fromcore if special precautions are taken, but more reliablyfrom exposures or orehole inspection. The InternationalSociety for Rock Mechanics (1981) describe threemethods for obtaining orienting core. One depends onmatching successive sections of a run of core, one pre-reinforcement of core with a grouted reinforcing rod andthe last scribing of the core and the use of a compass photodevice. Borehole inspection with a CCTV camera, yieldingorientation data, is now possible to depths exceeding1000m. A more recent technique is the ‘Acoustic BoreholeTeleviewer’ (Siddans, 1995) which utilises a scanningacoustic reflection technique in a water filled hole,together with computer processing of the data to providean ‘unwrapped’ oriented chart on which, for example, aplane discontinuity shows as a sinusoidal trace.

2.2.6 Groundwater conditionsGroundwater conditions depend on the position of thewater table and the permeability of the rock. The latter isdependent on the properties of the discontinuities althoughthe intact rock may be permeable in some cases. Forclassification purposes groundwater conditions are oftenexpressed as the ratio of water pressure to major principalstress or inflow rate per unit length of tunnel.

2.3 Classification systems

The empirical rock classification systems discussed in thisreport are intended to assist with the selection of tunnelsupport at both the initial design phase and duringexcavation. They are attempts to provide a) a standardisedmethod, b) data based on the accumulated experience oflarge numbers of case histories and c) a method of definingand processing the site investigation data necessary for theselection of tunnel support. For this reason it is importantthat they are not used outside the range of the casehistories analysed in their development, and that they arenot used in isolation without appropriate engineeringjudgement. They are introduced here in chronologicalorder of their development.

2.3.1 Rock load classificationTerzaghi’s rock load classification (Terzaghi, 1946) isgenerally accepted as the first rock classification systemformulated to evaluate rock loads in tunnels supported bysteel arches. Rock load was expressed in terms of theheight and the width of the tunnel for nine classes of rock.

Closed discontinuity

Open discontinuity

Aperture

Filled discontinuity

Width

Figure 3 Examples of discontinuity condition (after Brown, 1981)

7

These classes were originally purely qualitative, rangingfrom ‘1) Hard and intact’, which contains nodiscontinuities, to ‘9) Swelling rock’ which is rockcontaining clay minerals which expand on wetting. Latermodified versions included an RQD range for each class.This system has been extensively used in the USA but hasbeen shown to be excessively conservative in manysituations. More recently developed systems are moresatisfactory because they permit a more qualitativeassessment of the range of factors which control rockbehaviour and also permit the selection of support systemsbased on rock bolts and sprayed concrete as well as steelarches.

2.3.2 ‘Stand-up time’ conceptThe idea that the stand-up time for an unsupported headingis dependent on the rock mass class and the unsupportedspan is attributed to Lauffer (Lauffer, 1958). The larger theunsupported span and the poorer the rock class, the shorteris the stand-up time. This concept is central to the morerecent rock mass classification systems.

2.3.3 Rock quality designation (RQD)The RQD concept was introduced in section 2.2.3. It

was introduced by Deere (1967) as an improvement on the‘percentage core recovery’ as an index of rock quality. TheInternational Society for Rock Mechanics recommendsthat core should be at least NX (54.7mm) diameter anddrilled with double-tube core barrel. Descriptive termswere assigned as shown in Table 1. Some early attemptswere made to correlate RQD with support requirement.However, because of the importance of the other

properties of discontinuities (see sections 2.2.4, 2.2.5) it isnow used as an index property in more comprehensiveclassification systems. Two main drawbacks to the use ofRQD identified by Priest and Hudson (1976) are:

1 calculating RQD is a time consuming process, which,when applied to borehole core, can give resultsunrepresentative of the rock mass;

2 the conventional RQD is insensitive to variations in rockquality when the average discontinuity spacing is greaterthan 0.3m.

2.3.4 Rock structure rating (RSR)Wickham et al (1972) introduced the Rock StructureRating (RSR), as a partly quantitative, nine-parameter,weighted classification system for determining rockquality. Support prediction is a function of the RSR andthe steel arch support given by Terzaghi’s rock load. Thissystem is essentially an enhancement to Terzaghi’s rockload method, which allows the incorporation of the qualityof the rock mass and a corresponding reduction inconservatism of support predictions. The rock massparameters included are grouped as in Table 2.

Figure 4 Examples of joint roughness classifications

Rough

Smooth

Slickensided

Rough

Smooth

Slickensided

Undulating

Planar

B

C

D

E

F

G

Table 1 Rock quality designation

RQD% Rock Quality

< 25 Very poor25-50 Poor50-75 Fair75-90 Good90-100 Excellent

Table 2. Rock Structure Rating parameters

Group Parameter Range

A Rock type (3 grades)(general appraisal Hardness (4 grades) 6 - 30of rock structure) Geological structure (folding,

faulting, 4 grades)

B Joint spacing (6 grades)(effect of joint Joint orientation relative to drive 7 - 45structure) to drive (8 grades)

C Joint condition (3 grades)(effect of Water inflow (4 grades) 6 - 25groundwater)

All the previously considered parameters (see sections2.2.1 to 2.2.6) are included. Each group is given a singlenumerical value selected from a matrix of values for all theparameters in the group. The three group values whichresult are then added to give the RSR, which has amaximum value of 100. An adjustment factor is tabulatedto convert the standard RSR to a value applicable tomachine bored tunnels. Possible criticisms are that thejoint condition parameter only allows three ratings, ‘good’,‘fair’ and ‘poor’, and the ‘rock type’ classes are qualitativeand could be open to misinterpretation. Ninety percent ofcase histories were for steel rib supported tunnels, so the

8

predictions are based on this type of support. To correlateRSR with support the Rib Ratio (RR) is defined. This is theratio of the actual support spacing, in the 190 case records,to that given by standard steel arch tables, derived usingTerzaghi’s rock load method, for the same tunnel diameter.The plot of RSR against RR for the case records yields themean relation (RSR+30)(RR+80)=8800, although there isconsiderable scatter, suggesting the use of a cautiousapproach in its use.

Use of this method therefore requires determination ofthe RSR and the corresponding RR. The support given byTerzaghi’s method, is then modified by the RR to give thefinal support. Although it is possible to calculateequivalent rockbolt and/or shotcrete support from thepredicted rock load this is not supported by sufficient casehistory data of these support types for the use of thisclassification system to be recommended with them.

2.3.5 Geomechanics classification or Rock Mass Ratingsystem (RMR)

The Rock Mass Rating (RMR) system, sometimes alsoknown as the Geomechanics System (Bieniawski, 1973), isone of the two systems which are currently the most used.A full description including various enhancements whichhave been developed since and taking account of furthercase histories (numbering 351 in 1989) is given inBieniawski’s book (Bieniawski, 1989). Extensions to theoriginal system by various authors covering mining,ripping, foundation design and slope stability, aredescribed in the book. There are six parameters in thecurrent system:

1 Uniaxial compressive strength.

2 RQD.

3 Discontinuity spacing.

4 Discontinuity condition.

5 Orientation of discontinuities.

6 Groundwater conditions.

The range of each of these parameters is subdivided intofive classes, each of which is assigned a tabulatednumerical weighting depending on its relative importance.The ranges of the rating parameters used conform to theISRM (1978, 1981) method. The weights were originallyderived from those used by Wickham et al (1972) in theRSR system. To use the system the ratings for eachparameter are determined from a table and added to givethe RMR, which lies in the range 0-100. A separate table,based on Wickham et al (1972), is given to assist with theclassification of discontinuity strike and dip directionsrelative to the tunnel axis. The joint orientation ratings aredifferent for tunnels, foundations and slopes and are givenas a negative adjustment to the RMR for each of fiveclasses. It is assumed that there are three joint sets, themost unfavourable of which should be considered. It issuggested that when there are only two joint sets that theratings for joint spacing should be increased by 30%. Asubsidiary chart is presented which subdivides the ratingsfor joint condition across the sub-parameters; persistence,aperture, roughness, infilling and weathering. Other charts

present smooth curves for the interpolation of values ofstrength, RQD and spacing between those given in themain table. For mining it is stated that further adjustmentsfor stress, blasting damage and major faults or fracturesmay be called for.

A peculiarity of this classification system is that bothRQD and discontinuity spacing are included as parameters,although they are strongly correlated, as described byPriest and Hudson (1976). Bieniawski goes as far as topresent a chart based on Priest and Hudson’s relationshipwhich is to be used if either RQD or spacing is missing.

The final rating (0-100) puts the rock into one of fivecategories (Table 3), each with a span/stand-up-timecharacteristic which is plotted on a span-stand-up timechart (Figure 5). This chart could be used, for example toassess the maximum unsupported advance which could bemade in a given class of rock. Guidelines for excavationand support are tabulated for steel ribs, rock bolts andsprayed concrete, for each of the five classes. It isspecifically stated that these support recommendations arefor permanent support in tunnels constructed byconventional drill and blast. A similar stand-up time/unsupported span chart is presented for tunnel boringmachine (TBM) excavation, adjusted for the difference inrock damage between these two excavation modes. Thishighlights a potential problem with the use of rockclassification systems, that the rock quality assessed fromsite investigation boreholes and surface exposures is likelyto differ from that measured during construction,particularly by drill and blast. Bieniawski suggests ablasting damage adjustment to RMR of 0.8 to 1.0 formining applications which should perhaps also be appliedto civil engineering tunnelling.

Table 3 RMR classes

Class Rating Rock Averageno. Description stand-up time

1 100-81 Very good 20 years (15m span)2 80-61 Good 1 year (10m span)3 60-41 Fair 1 week (5m span)4 40-21 Poor 10 hours (2.5m span)5 <20 Very Poor 30 min (1m span)

The following useful correlations are also presented:

1 between deformation modulus and RMR;

2 between rock mass strength (Hoek and Brown,1980,1988) and RMR for both ‘disturbed’ (eg. blastdamaged) and ‘undisturbed’ (eg. machine bored) rock;

3 between rock load and a function of RMR rock densityand tunnel width;

4 between RMR and ‘Q’ based on 111 case histories;

5 between RSR and RMR based on 7 case histories inNew Zealand.

These correlations should be used with caution as thescatter is considerable. The output of the RMR method isstated to tend to be conservative, which should beaddressed by monitoring deformations during constructionand adjusting the design as necessary.

9

2.3.6 Q - systemBarton et al (1974) developed the ‘Q’ system basedoriginally on an analysis of 212 mainly Norwegian casehistories of tunnelling in hard rock. Since that time therehas been considerable development and revision to includenew support types and about 1050 extra case historyrecords (Barton and Grimstad, 1994a), bringing the total toabout 1262. These cover a wide range of rock types andsupport methods, in many different countries. However thesystem has required virtually no re-calibration of the rockquality ratings since its original development. There are sixparameters in this scheme.

1 RQD.

2 Number of joint sets (Jn).

3 Joint roughness (Jr).

4 Joint condition - or alteration (Ja).

5 Groundwater inflow (Jw).

6 Stress condition. (Stress reduction factor(SRF)).

The value for each rating parameter is selected from thetable shown in Figure 6. The rock quality, ‘Q’, iscalculated as follows:

and lies in the approximate range 0.001 to 1000. The threeratios represent block size, inter-block shear strength andactive stress. Barton (1988) considers that the ‘Q’ systemprovides a much more detailed assessment of jointroughness, filling (alteration) and relative orientation thanany other system. The joint roughness is rated on a sevenpoint scale from ‘discontinuous joints’ to ‘slickensided

planar’ and the joint condition is rated on a sixteen pointscale from ‘tightly healed’ to ‘thick continuous’ zones ofswelling clay’. Although joint orientation is not included itis implicit because the joint parameters are to be applied tothe most unfavourable joint set.

The ‘stress reduction factor’ has 16 classes falling intofour groups:

1 zones of weakness causing loosening or fallout.

2 rock stress problems in competent rock.

3 squeezing or flow of incompetent rock.

4 swelling rock.

The ratio of rock stress to rock strength is onlyconsidered in the ‘competent rock - rock stress problems’group. This is appropriate because, as previously discussedin section 2.2, the rock stress is only important when it is asignificant proportion of the rock strength.

The ‘excavation support ratio’ (ESR) reflects the desiredstand-up-time and can be thought of as a risk-related safetyfactor. It is selected from a table according to the use of theexcavation. Suggested figures range from approximately 3to 5 for ‘temporary mine openings etc.’ to approximately0.8 for public facilities, although the confidence in theseextreme values is stated not to be high because there areonly two case histories relating to each. The ‘equivalentdimension’ (ED) is then given by

Figure 5 RMR system - Stand up time versus span for a range of rock mass classes (after Bieniawski, 1989)

QRQD

J

J

J

J

SRFn

r

a

w= . .

EDspan or height

ESR=

In this equation the choice of span (or diameter) is

appropriate to roof support, and height to wall support.Support is then chosen from 38 tabulated basic categories

10

Figure 6 Ratings for the six 'Q' - system parameteres (after Barton and Grimstad, 1984a)

11

which depend on ‘Q’ and ED. The support categories arefurther subdivided depending on tabulated critical valuesof combinations classification parameters. Barton et al(1974) and Barton, (1988) both include a chart of EDversus ‘Q’ on which the 38 support categories are plotted.Barton and Grimstad (1994a) give a similar chart (Figure 7)showing the applicable regions for the various generalsupport types including the more recent fibre reinforcedsprayed concrete. The chart clearly shows the boundarybeyond which no support is required. Relationships forbolt length and maximum unsupported span in terms of‘Q’, and roof pressure in terms of ‘Q’ and J

r. can be found

in Barton et al (1974) and Bieniawski (1989).Several authors have correlated ‘Q’ values with RMR

values with widely differing results. A graph of theregression RMR=9 ln ‘Q’ + 44 for the data from 111 casehistories is given by Bieniawski (1976), from which it isapparent that there is considerable scatter. The 90%confidence limits are at ±18. It is evident that a the use ofsuch correlations is not likely to yield a very accurateconversion from one system to the other.

2.3.7 New Austrian Tunnelling Method (NATM)classification

Bieniawski (1989) includes a section on the ‘NATMclassification’. The NATM is stated to be a designphilosophy rather than a ‘method’. An essential componentis the continuous observation of lining loads anddeformations and subsequent modification of the support.Thus the NATM is an application of the ‘Observationalmethod’ to tunnelling. The NATM also aims to mobilisethe strength of the rock mass by placing suitable support atthe correct time. These support measures are determinedby the application of a rock mass classification system. Inorder for such a system to be workable contractually,payment must be based on rock mass classification at theworking face after each advance.

It is stated that when using the NATM the groundshould be classified according to a scheme drawn up foreach individual site and which should form part of thecontract. There would be a small number of classes, eachwith its appropriate basic support type. An example ofseven classes is tabulated in which the ground is described

Figure 7 ‘Q’-system support type predictions (after Barton and Grimstad, 1994a)

12

qualitatively and a basic support type assigned to each.This approach, which is still relatively novel in the UK,would be facilitated by a type of contract which defines asimple method for the resolution of any disagreementsover support. Contractors and engineers experienced inthese techniques are also essential.

Austrian Standard Önorm B2203 (ÖsterreichischesNormungsinstitut, 1994) contains a purely qualitative rockclassification. This describes three main classes of rock,subdivided into ten sub-classes from ‘Stable Rock’ to‘Flowing Ground’. This classification system is based onAustrian, and in particular Alpine, tunnelling conditions. Itis the only formalised system specifically developed forNATM tunnels. For each of the classes qualitative adviceis given on the generic type of support required, and theeffect of the support on progress. No information is givenon how the classifications should be measured, either atthe site investigation stage, or during tunnelling.

Sauer (unpublished report to TRL) considers thatthere are three subjects to be classified for NATM tunnels.These are rock or soil, excavation and support. It is statedthat classes are normally developed which are projectspecific. Parameters relevant to support classes are given.The summary of all the site investigation data should leadto a reliable estimate of stand-up-time, deformation,overbreak, support load, settlement and ground water flow.However no methods are proposed for the determination ofthese from the site investigation data. During construction,continuous geotechnical documentation is recommended.This should comprise groundwater observation and facelogging. However no instructions are given on how suchface logging should be carried out. Especially in the caseof soft ground tunnels (as noted in section 2.1.2.1),settlement may be an overriding factor affecting therequirement for support. This is not considered in any ofthe classification systems. It is stated that it is notpracticable to produce a universally applicable, formalisedclassification system for tunnelling in general and softground tunnelling in particular.

Sauer also recommends that, if a general rockclassification system is used, both of the two most popularsystems, RMR and ‘Q’, should be considered. It isemphasised that these systems should be used with greatcaution, and preferably should not be included in thecontract specification. Defects in these systems areconsidered to be as follows:

1 Inherently conservative.

2 Each omit key elements of rock mass characterisation.

3 Certain measurements can be subjective and notrepeatable.

4 The values obtained depend on how and where themeasurements are taken, and on the technique andexperience of geotechnical staff.

5 There may be a tendency to accept the values obtainedin favour of the actually observable condition of theground.

6 The values obtained are subject to statistical uncertainty.

It is suggested that the results of the two systems,obtained by both the contractor’s and the consultant’sgeologists, should be compared, to take account of theneglect of some of the possible parameters by theindividual systems. Most importantly the results should bechecked against the experience of as many tunnelling staffworking at the face as possible. It is particularlyemphasised that measurements on a relatively smallnumber of scanlines may not reveal as much about theground as a good look at the whole face.

2.3.8 Contractual mattersThe successful use of rock classification systems for theselection of tunnel support requires both sufficient goodquality site investigation data and a Contract form which isflexible enough to allow rapid variation in the support typeused according to the conditions found in the tunnel. TheContract specification should allow the parties to theContract to rapidly agree on the support to be provided foreach advance of the face without costly disputes anddelays.

The Institution of Civil Engineers (ICE) 6th edition ofthe Conditions of Contract (1991) in a modified form, theICE Design and Construct Conditions (1992), the NewEngineering Contract (NEC)(1993, 1994), and theInstitution of Chemical Engineers (IChemE) Form ofContract (1992) (the ‘Green Book’) may each beappropriate depending on the type of project management.This subject is fully examined by Attewell (1995), in acomprehensive book on tunnelling contract and siteinvestigation practice, which includes an appendix on theuse of rock classification systems.

A key document is Construction Industry Research andInformation Association (CIRIA) Report 79 (1978),‘Tunnelling - improved contract practices’. This report isaimed to increase efficiency and reduce costs of tunnellingby recommending practices which would identify,eliminate or reduce risk connected principally with thebehaviour of the ground. It also established appropriatemeans to evaluate and allocate the responsibility for therisk that remains.

It is recommended that a possible tunnelling systemshould be defined at the time of Tender and the limitsdetermined for the ground conditions expected.Subsequent variations can then be compared with these‘ground reference conditions’. It is emphasised that theEngineer’s role is crucial in safeguarding the interests ofboth parties to the Contract. Areas and incidence ofpossible risk should be explicitly defined in the contractdocuments, along with liabilities in the event of theiroccurrence.

Site investigation is identified as fundamental, theground being the source of the largest risk in tunnelling.The recommendation of the Harris committee (NationalEconomic Development Council, 1968) that the Engineershould be responsible for the selection of the groundinvestigation contractor on the basis of expert knowledgeof ground investigation firms, rather than purely by pricecompetition, is endorsed. The investigation should bedesigned and controlled by engineers or geologists

13

conversant with the difficulties of tunnelling. The siteinvestigation should ideally (although it is currentlyuncommon) provide the information required by both theDesigner and the Contractor. To this end, the CIRIA reportsuggests that allowing expected Tenderers to influence thesite investigation could avoid this deficiency.

It is recommended that the Engineer ‘should have theresponsibility for supplying the Tenderer with the fullreport of the site investigation and for making the finalassessment and judgement of any specialists findings’. It isalso suggested in the CIRIA report that, where aninterpretation of the site investigation data has been madewhich is fully accepted by the Engineer and has aparticular bearing on the design and construction of atunnel, that this interpretation should be made available totenderers. It is also recommended that no disclaimers orwarranties should be applied under the Contract to any siteinvestigation information, and that Clauses 11 and 12 ofthe ICE Form should be fully accepted. The Tenderer mustuse the information provided to make its own judgementof the construction method and expected rate of progress.If any site investigation information is factually inaccurate,especially due to the negligence of the Promoter orEngineer, then the Contractor may have a valid claim. TheTenderer’s assessment of the site investigation should befully reflected in his detailed method statement, and theEngineer should be satisfied that the Tenderer has fullyunderstood and utilised the information.

The ground encountered during construction should bemonitored, recorded, agreed and compared with thereference conditions previously determined by theEngineer. It should then be possible to apply Clause 12without dispute. The reference conditions may containelements from one or more of: a) geological: b) method ofconstruction; c) response of the ground; d) rate of progress.

Commenting on the Round Hill project (see section 4.1),Mott MacDonald (1994) noted that the Client was meetingmost of the risk in that Contract, through the instructedsupport clauses. The suggestion is made that the NEC orIChemE forms of Contract may hold some advantages forthis type of work, but that a modified 5th edition of theICE Conditions of Contract should also succeed. It is alsocommented that the alternative approach of ‘design andbuild’ contracts does not necessarily guarantee that therewill be no claims. This is stated to be particularly trueunless the client imposes no design changes, which israrely the case. There will also be a danger under thisarrangement that there will be problems with ensuringquality. Napthine and Smart (1995), reviewing the lessonsto be learned from the use of design and build at the UKChannel Tunnel Terminal, state that the main disadvantageis a constant tendency to look for minor cost savings whichgradually erode the final quality. It is also worth notingthat lack of Client supervision in tunnel construction mayallow defects to be hidden which do not come to light formany years.

Recommendations are also made by Mott MacDonaldfor beneficial changes to future tunnelling specifications.In the area of ground support it is recommended that allparameters for rock mass classification should be defined

in detail. The split of support types should not be shown onthe drawings but quantified in the Bill of Quantities, thusgiving the Engineer more freedom to design final usage.

3 Literature review

A large number of papers relevant to the subject of rockclassification systems have been published worldwide.This section reviews the most recent of these to provide anup-to-date discussion of the state of development of thesystems. The reviews are grouped under six subheadingsaccording to the topic covered.

3.1 Comparisons of and observations on ‘Q’ and RMRsystems

Goel, Jethwa and Paithankar (1995) have evaluated the‘Q’ and RMR systems with reference to measured supportpressures on steel arch ribs at 25 tunnel sections in India.Results are claimed to show that ‘Q’ is unsafe for theprediction of rock loads for large tunnels in squeezingground (squeezing being the plastic flow of the groundwhere the stress is high compared with the strength).However this finding is based on a very small sample inwhich the observed pressure exceeded that predicted inonly two of 10 squeezing cases. Support pressurepredictions based on Unal’s (1983) correlation betweensupport pressure and RMR are even poorer, with observedpressures exceeding those predicted in all the squeezingcases and four of the non-squeezing cases. Equationsbased on the rock mass number (N), a ‘stress-free Q’ valueas defined below, tunnel depth and radius, are presentedfor both squeezing and non-squeezing cases. Theseequations are shown to be much better predictors of thesupport pressure for the small number of case historiesexamined.

In order to improve the comparability between RMRand ‘Q’, new parameters are defined which omit theratings for joint orientation and intact rock strength fromRMR and rock stress in the ‘Q’ system. The modifiedRMR, is designated RMR

mod, and the modified ‘Q’ is

designated Rock Mass Number (N). Correlations betweenRMR and ‘Q’ by Bieniawski (1976) and Rutledge andPreston (1978) are claimed unsafe and an improvedcorrelation between RMR

mod, and N from 61 case histories,

is demonstrated.Tarkoy (1995) considers the problems associated with

the use of the RMR system with tunnel boring machines(TBMs) and provides guidance as to acceptable ranges ofRMR for their use. He states that rock masscharacterisation systems have often been used blindly inthe past, whereas they should only be used as a design aid.They are useful for assessment of support requirementsbased on past experience, mainly from drill and blasttunnels. Although the predictions of support required maybe reliable it should not be assumed that the use of a TBMis appropriate for the ground conditions. There is no easyway to adjust the RMR reliably for TBM excavation. Themain problems are the distance from the face that supportis installed and the resulting delay, the effect of gripperloads on the tunnel walls, and the effects on the excavation

14

system of the face condition. It is observed that the wallsof a TBM tunnel may initially look more stable than theyare. Limitations of the RMR are that it does not apply tosoil and is insensitive in the lower ranges of values,particularly of RQD and intact strength.

It is recommended that RMR only be used for:

1 Drill and blast excavation;

2 TBM support assessment, but only if RMR>20;

3 TBM excavation assessment if RMR<45 and then onlywith judgement and experience and an appropriate TBMdesign.

Barton and Grimstad (1994a) present an updated ‘Q’-system with 1050 new case records and the inclusion offibre-reinforced shotcrete as support. This stainless-steelfibre-reinforced shotcrete is corrosion resistant, dense andof low permeability with compressive strengths of 35MPato 45MPa. To avoid corrosion problems, triple-sleevedepoxy-coated rock bolts can be used, which are groutedafter shotcreting to fully protect them. Single-pass liningsare proposed using this material, which has been shown tosave costs in 160km of Norwegian road tunnels. ‘Q’-system tables are presented. These are identical to the 1974tables except for some changes to the SRF for rock stressproblems. A graphical ‘Q’-logging chart is presentedwhich is convenient for recording geotechnical data duringa survey. An updated Equivalent Span (ED) v. ‘Q’-chart ispresented, with support-type zones, including bolts andfibrecrete, superimposed (Figure 7). A new feature of thischart is that contours of shotcrete thickness are included.As in the original system the safety factor is governed bythe value of ESR chosen by the user (high for temporaryopenings, low for permanent). A table of suggestedadjustments to ESR and ‘Q’ to determine temporarysupport requirements and wall support requirements isgiven. This suggests that for temporary support ‘Q’ may beincreased by a factor of 5 and the ESR increased by afactor of 1.5.

Equations relating critical depth for squeezing andcompressive strength to ‘Q’, after Singh et al (1992) aregiven. An approximate relation between ‘Q’ and P-wavevelocity (V

p) is also given, based on case histories from

several countries in a range of rocks, but not corrected fordepth. More data is required to refine this concept. Thissuggests the possible use of ‘design-as-you-drive’ using‘Q’ derived from V

p obtained by probing ahead of the face,

particularly with long and deep tunnels where siteinvestigation boreholes are sparse.

The use of the ‘Q’ system with TBMs is discussed. Theuse of a TBM is considered likely to reduce the need forsupport only in the mid range of ‘Q’ from 3 to 30. Anexample is given in which the support needs increasedabout 70% when a tunnel originally driven by TBM waswidened by blasting.

Choubey and Dhawan (1990) present a case history ofa 300m long x 7m span, D-shaped access tunnel throughquartzitic phyllite, for a hydroelectric project. Some 40exposures 300m-500m from the tunnel were mapped,giving 219 joint observations. Stereographic plots revealedfour prominent joint sets. Each set was characterised and

the rock mass classified according to both the ‘Q’ and theRMR systems. The tunnel faces were logged during tunnelconstruction. There were some differences between thesurface exposures and the in-tunnel measurements. Table 4shows the resulting average ratings for both ‘Q’ and RMR.

Table 4 ‘Q’ and RMR ratings

RMR ‘Q’

Surface exposure 51 0.888Tunnel face 45 0.922Description fair rock very poor

In either system the surface to tunnel range falls withinone support class but overall the ‘Q’ predictions forsupport were nearer to the actual requirement than thosebased on RMR. It is stated that the number of rockboltscould be reduced by attention to potential loose blocks. Bythis means bolt spacing was gradually increased up to 3min rock for which ‘Q’ suggested 1m spacing.

Kirsten (1988a) emphasises that classification systemsare no substitute for engineering judgement. He highlights,in this discussion on Bieniawski’s (1988) paper, a lack ofsensitivity of the RMR system, particularly to jointcondition. This is largely due to the additive nature of theRMR rating system, which is overcome in Barton’s ‘Q’-system by the multiplication of individual parameters.Scatter in correlations between RMR and ‘Q’ is stated tobe largely due to this lack of sensitivity. RMR alsomeasures discontinuity spacing twice by including RQDand spacing as separate parameters, which gives block sizean undue predominance in the final rating. The prioritygiven to rock strength in the RMR system is alsodenigrated as it is only critical in cases of high stressrelative to strength. This is much better accommodated inthe ‘Q’ system. It is considered that the RMR system,which predicts final support requirements, is not adequatefor the design of immediate support as required by theNATM.

Tallon (1982) compares RSR, RMR and ‘Q’ systems inthe construction of seven tunnels on the Campomanes-Leon highway in Spain. The circular tunnels had adiameter of 11.6m with up to 600m of cover of Palaeozoicrocks (shale limestone, sandstone and quartzite).Classification indices were established at the tunnel facesand used to assist with the selection of support. A problemidentified was that where rock quality is variable over thetunnel face, no standard method of face logging is laiddown by any of the classification systems. Should zones beaveraged, or the worst taken? This will depend on whetherthe poorer zones are judged significant in terms ofstability. Problems with evaluating the variousclassification parameters are discussed. For example,uniaxial compressive strength can only easily be assessedat the face by the Schmidt hammer index, which is likelyto be highly unreliable. A great deal of experience isrequired to perform a reliable assessment of most of theparameters. One hundred and fifty determinations of eachof the indices were made. The RSR was found not to

15

differentiate clearly between average quality rocks andthose of better and poorer quality. The RMR showed abetter distinction between average and good quality rocks.The ‘Q’ index was found to be the most sensitive in itsdefinition of rock quality, particularly for rocks of otherthan average quality. Correlations are presented betweenthe three systems for four tunnels and compared withcorrelations published by Bieniawski (1976) and Rutledgeand Preston (1979). These correlations are quite goodconsidering that the data was from different places, rocksand observers, suggesting that all three systems are broadlycomparable.

Cameron-Clarke and Budvari (1981) examined theapplicability of both ‘Q’ and RMR systems to the selectionof tunnel support, for three South African tunnels, basedon both borecore and in-tunnel observations. The tunnelcross sections were of a 3m wide x 3m high inverted ‘U’shape. A wide variety of rock types and support systemswere covered. In both cases borecore tended to indicatepoorer rock than in-situ observations. For both systems82% of the results from borecore and in-situ measurementswere within one rock class. For the in-situ measurementsof discontinuities 10m long scanlines were used. Bothclassification systems were considered to be useful butshould not be regarded as providing more than apreliminary assessment of the support required. Great careis necessary when interpolating between boreholes. Thecorrelation between borecore and in-situ rock classes wasfound to be better for the ‘Q’ system than for the RMRsystem. The tables presented in the paper show that the ‘Q’system’s predictions generally agreed much better with thesupport actually installed than those of the RMR system,which were often excessively conservative. In most casesthe ‘Q’ system was slightly conservative. The correlationbetween RMR and ‘Q’ for these sites was poor and verydifferent from that presented by Bieniawski (1976).

3.2 Face logging

Rock mass classification systems are often used duringconstruction of tunnels to guide the choice of supportsystem. For this purpose the rock exposed in the headingmust be assessed. In a drill and blast tunnel some or all ofthe face and a short length of sidewall will normally beavailable, whereas in a full-face machine bored tunnel theonly available exposure may be the tunnel peripherybehind the shield. Unless the geotechnical assessment(‘logging’) of the ground is comprehensive, and reflectsthe condition of the most critical areas of the face, thesupport predicted by the application of a rock classificationsystem may not be appropriate.

An example of this is given by Garrett (1993) whodescribes the construction of the twin Cumberland Gaphighway tunnels through the Appalachian mountains. Thenorthbound bore is 11.43m wide, 7.8m high, 1249m longand the southbound bore is 12.1m wide, 9.55m high and1259m long. A pilot tunnel 3m wide x 3m high formed thecrown of the southbound bore. The geology was shale,mudstone, limestone, sandstone and coal in which voids upto 27.4m high were encountered. Geological mapping ofboth the face and the periphery was carried out during each

shift. Excavation was by heading and bench using drill andblast. RMR and ‘Q’ were both evaluated and thetemporary support was intended to be chosen from 5 pre-determined temporary support categories (supposedlyNATM). The support details for each category aretabulated, but not the appropriate ‘Q’ - range. However thesupport used was generally chosen daily by the tunnelproject engineer and the Client’s consultant. TheConsultant’s assessment of the support required oftendiffered from the result of using the rock classificationsystems on the shift geologists’ interpretations of thegeology. The two shift geologists often rated the sameground very differently. It is not recorded how the‘experienced NATM consultants’ decided on the supportcategory. Sprayed concrete was by the wet process sprayedby robot. Both steel fibre reinforced and non-reinforcedsilica fume mixes were used together with dowels andbolts. Lattice girders were used in the lowest grade ground.The final support was cast in-situ concrete.

3.3 Block fallout

Several authors have considered the problem of thestability of potentially unstable blocks in a tunnelexcavation and how this might be taken account of in rockclassification systems.

Barrett and McCreath (1995) consider that empiricalrules such as RMR and ‘Q’ are a useful starting point inthe design of tunnel support but that uncertainty over thefactor of safety achieved is an inherent problem. Theyexamine the action of shotcrete in stabilizing the ground.Wet-mix shotcrete reduces rebound and increases itscompressive strength, particularly with the addition ofsilica fume. The replacement of weldmesh by steel fibresin the mix saves the time required for its erection, with noloss of post-peak strength. There are, however, doubtsabout its performance at large deformations. A basis fordesign is presented based on four mechanisms of failureapplicable to cases where rock mass failure is controlledby discontinuities rather than rock stress:

1 Adhesive strength.

2 Direct shear.

3 Flexure.

4 Punching shear.

The structural supporting ring function of a shotcretelining in weak ground is not considered here. The mainfunction of shotcrete and bolts in rock is to restrain ‘keyblocks’ which if allowed to fall out can lead to progressivecollapse. Shotcrete should be applied as soon as possibleexcept in high stress zones where large deformations areallowed to occur. Particularly in blasted tunnels fibrecretecan be applied more quickly than mesh reinforcedshotcrete. The worst case shotcrete load from blockloosening is considered to be defined by weight of the 60°apex prism with base defined by the rockbolts. No frictionis assumed as this might be reduced by blastingdisturbance.

Estimates are given of various strengths of unreinforcedsilica fume shotcrete at a range of ages. Fibre

16

reinforcement is stated to impart no greater peak strengthbut gives it some post-peak strength in flexure andpunching shear (when cracking has occurred). Calculationmethods for the rock load between bolts are given basedon the four modes listed above. Stability charts for thesemodes of failure are plotted based on bolt spacing andfactor of safety for assumed values of lining thickness andstrength.

Hatzor (1993) discusses the difficulties in definingcritical blocks in a tunnel through jointed rock, which leadrock engineers to resort to empirical rock classificationschemes for support design. A major drawback ofconventional rock classification systems is considered tobe their ‘disregard for block motion into the newly createdspace’. Not all the removable blocks identified by blocktheory actually fall out in construction, because ofdifferences between idealised geometry and reality,leading to a concept of block failure likelihood. Currentrock mass classification systems do not adequately accountfor the interaction between the tunnel orientation and rockstructure. The block failure likelihood P (B) was originallydefined as the product of the three parameters: JointCombination Probability (P (JC)), Block InstabilityParameter (F) and Shape Parameter (K). Correlationsbetween the prediction parameters and observations in twotunnels suggest that the inclusion of shape parameter is notjustified. ‘B’, the cumulative block failure likelihood, isdefined as the area of the P (B) histogram and representsthe overall tendency of the rock mass to produce blockfailures from a free face of known attitude. It is proposedthat B, which will vary with tunnel orientation, couldreplace parameters such as spacing, orientation andcondition of joints in rock classification schemes.Insufficient information is given on the practical use of thesystem, in particular it is not at all clear how the factor F iscalculated from the field data, and no suggestions are madeas to how the concept would be integrated with anyspecific rock classification system.

Ward (1978) Discusses three-dimensional blockloosening mechanisms when tunnelling in weak rock, theimportance of restraining ‘key’ blocks and the use ofphysical models and field observations. In cases wheresurface settlements are not critical he highlights the need toallow a degree of rock yield combined with monitoringand to consider the placement of support in relation to theadvance of the face. ‘Proper assessment of and control ofthe many variables which combine to produce asatisfactory and efficient support system cannot be done atthe design stage. It is strongly recommended thatconstruction should be aided and guided by monitoring ofperformance carried out by engineers relieved ofcontractual responsibilities as the only positive means oftaking sensible, economic and safe decisions during theprogress of the work.’ Much reference is made to theKielder Water experimental tunnel in support of thesearguments. He denigrates the rock classification approachto design because it perpetuates existing practice, whichmay be excessively conservative, and takes no account ofthe fact that the same supports can be both satisfactory orunsatisfactory in the same rock depending on construction

procedures. This is true however many case histories arecorrelated. The problem with the alternative analyticalapproach is incomplete information on the rock mass andthe supports. The ‘characteristic line’ or ‘ground reactioncurve’ is discussed in detail. It is stated that the ‘risingportion of this curve postulated by Pacher (1964) andothers has not been found in the field’. He considers thatsprayed concrete applied close to the face may prove to betoo stiff and attract excessive loading even though it has ahigh creep and shrinkage capacity when green. A methodof controlling the circumferential stiffness of a shotcretelining by the use of longitudinal slots is described. Thiswas used in the Tauern and Arlberg tunnels in Austria, inconjunction with steel ribs with frictional joints.Construction monitoring is highlighted as essential, alsoload and stress measurements are of little value withoutdisplacement measurement.

3.4 Norwegian method of tunnelling (NMT)

Barton and Grimstad (1994b) define a Norwegian Methodof Tunnelling, which they claim to be more appropriatethan NATM in harder, jointed rocks driven by drill andblast, where overbreak makes mesh erection difficult andcauses excessive concrete consumption. In Norway wetprocess steel fibre reinforced shotcrete totally supplantedmesh reinforced shotcrete by 1984, due to its improvedcharacteristics and ease of robot application. It is quickerto apply, has less rebound and thus reduces concretevolume. It has also replaced cast concrete, even in faultand clay bearing zones. Here sprayed fibrecrete with rebarreinforcement is used, with cost savings of approximately50%. Drill and blast driving rates of 40m to 70m a weekare achieved compared to 60m to 100m, with only minorreinforcement being necessary. 160km out of a total of460km of road tunnels in Norway have an unreinforced orfibre-reinforced final lining. Fibre corrosion does notappear to be a problem, even in 10 year old sub-seatunnels.

In Barton and Grimstad (1994a) the Norwegian Methodof Tunnelling (NMT) is defined as following the followingprinciples:

1 Usually applied to jointed rock, with or without claybearing zones and or stress slabbing. ‘Q’ from 0.001 to10 or more.

2 Excavation usually by drill and blast or hard rock TBM.

3 Support using fibrecrete, possibly with reinforcedshotcrete ribs, bolting or cast concrete. No lattice girdersor dry process shotcrete.

4 Contractor chooses temporary support.

5 Owner/Consultant chooses permanent support, which isoften fibre reinforced shotcrete and not cast concrete.

6 Rock mass classification is used to predict rock massquality and support needs, and both are updated duringconstruction with monitoring of lining loads anddeformations only in critical cases.

In contrast the NATM is quoted as being suitable forsoft ground (‘Q’ from 0.001 to 0.01), machine excavatedtunnels, with a closed invert in very weak ground. NMT is

17

most appropriate in drill and blast tunnels in harder rockswhere overbreak is a problem. Design is often based on the‘Q’- system of rock classification. Monitoring is generallynot performed unless ‘Q’ is less than 0.01, unless the spanis particularly great, as at the Gjøvik cavern. On the newsupport chart (Figure 7) NATM overlaps with NMT incategories 8 and 9. A possible system for soft rockcombining the NMT and NATM principles is given as:

1 support prediction using ‘Q’.

2 temporary support close to the face with bolts andfibrecrete.

3 adjustment of support class for final support well backfrom the face.

3.5 Probabilistic methods

The inherent variability of the ground, the relatively smallamount of ground sampled in even the best siteinvestigation and measurement errors inevitably result in adegree of uncertainty in both the assessed rockclassification at the measurement points and thatinterpolated between them. This can be addressed by theuse of risk analysis techniques by which the probabilitydistribution of the output of a classification system can beestimated from the measured or assumed probabilitydistributions of the inputs. The technique may be useful atthe design stage, as detailed below. Alternatively it couldbe used during construction to provide a continuouslyupdated prediction of the probability of encounteringvarious rock classes in the tunnel yet to be constructed,based on the measured distributions of the classificationparameters encountered to date. A detailed appraisal of thepotential application of risk analysis to all aspects ofhighway tunnel construction is provided byConway et al (1995).

Conway (1993) describes risk assessment techniquesapplied to the choice of alignment of a tunnelled rivercrossing. As it was a feasibility study there were nodetailed investigation results. Minimum, most likely andmaximum values were assigned to the geologicalparameters by a ‘best-guess’ approach, using a triangularprobability density function (PDF) in the @RISK add-infunction to the Lotus 1-2-3 spreadsheet. The trialalignments were divided into 100m long segments, eachassumed to have constant characteristics. Triangular PDFswere used to model a) the depth of weathering ofvolcaniclastic rocks, b) thickness of alluvium c) the widthof faults and d) depth to water table. Locations of faultswere modelled using a Poisson distribution. The PDFs forthe cost of ground treatment to reduce water inflow atfaults was based on combining estimates of theprobabilities of intersecting a fault, the specific discharge,the length and the wetted perimeter.

Tunnel support prediction was based on a probabilisticimplementation of the ‘Q’ system. Estimated triangularPDFs were substituted for each ‘Q’-parameter. Asimulation method using Latin Hypercube sampling withinthe spreadsheet was then used to generate a PDF for ‘Q’.Ninety-five percent confidence limits profiles for ‘Q’could then be plotted along each alignment. Four support

classes were developed from the ‘Q’ systemrecommendations. The cost of each alignment wascalculated, based on a bill of quantities model usingestimated PDFs for unit costs, and the extra costs due tofault zones. Cumulative frequency distributions of cost foreach alignment were plotted, based on the ‘Q’ profiles. Asensitivity analysis could then have been performed to givefurther confidence in the model.

This method provides a systematic way of applyingengineering judgement to the assessment of geologicaldata for tunnel construction. Where the number ofvariables is small this approach may not be necessary, butwith a larger number this approach provides a consistentbasis for the application of engineering judgement.

A risk based prediction model using these techniqueswas developed and applied during the construction of thePen-y-Clip tunnel, which is described in section 4.3.

3.6 Numerical modelling

A number of authors consider that improved analyticalmethods will in time supplant the empirical rockclassification schemes currently in use. Most major tunneldesigns will involve the use of both methods. However,until recently the techniques available for numericalmodelling did not include the analysis of discontinuousmedia, which was a severe handicap as in most practicalcases it is the properties of the discontinuities whichpredominate in controlling the behaviour of the rock mass.Because of such difficulties with the characterization of therock mass all numerical analyses for tunnelling need verycautious application. Ways in which conventionalnumerical analyses are being developed to make themmore reliable are described in this section. It is likely thatsuch calculations will always need checking by othermeans, including empiricism.

Pan and Trenter (1992) discuss the application of finiteelement analysis (FEA) and discrete element analysis(DEA), including block theory, to tunnels. It is assertedthat a numerical modelling technique can assist theengineer in many practical problems, but failure to selectthe correct model could lead to expensive or evendangerous mistakes. However no specific numericalmodelling packages are named. The basic idea is that therock mass can be modelled as a continuum if the block sizeof the rock is either very large or very small relative to thesize of the excavation. It is suggested that rockclassification systems could be useful in determiningwhether to use a continuum (FEA) or discontinuumapproach (DEA). On the ‘Q’ rating scale it is asserted thatDEA is applicable in the central band of rock classes from‘poor’ to ‘good’, while FEA is restricted to ‘very good’and above, or in a pseudo-continuous form of FEA, toconditions worse than ‘poor’. Hudson has introduced therepresentative elemental volume (REV) concept, which isdesigned to help with this problem (see below). Thestrength and stiffness properties of the rock mass anddiscontinuities are required for the DEA approach andthose of the rock mass for the FEA approach, together withthe properties of the support system. These properties canbe estimated from laboratory tests or rock mass

18

classification systems by making use of the Hoek-Brownfailure criterion (Hoek and Brown, 1980, 1988).

Barton et al (1990) in ‘Tunnelling by numbers’describe the use of UDEC-BB, a discrete element methodusing the Barton-Bandis (BB) (Barton and Bandis, 1990)joint model, with qualitative examples. It is stated thatnumerical modelling may be required in some cases as asupplement to the use of the ‘Q’ system. There are severeshortcomings in the use of continuum analyses such as thefinite element method (FEM), finite difference method(FDM) or boundary element method (BEM) whendeformation and failure are controlled by discontinuities.The Norwegian Geotechnical Institute (NGI) has madeextensive use of the code UDEC by Peter Cundall of ItascaConsulting. This has been extended by the use of the BBnon-linear joint model. The input data required for themodel can be acquired from drill core and/or outcrops, bythe use of index tests. The case history used is described inmore detail in Makurat et al (1990) (see section 3.6below).

Barton and Bandis (1990) introduce the ‘jointroughness coefficient (JRC) - joint compressive strength(JCS)’ model for rock joints for use with the UDEC-BBprogram.

2 simple index tests on samples. (Point load, Schmidthammer, tilt tests, in-situ profiling, strain-gaugeduniaxial compression; for the JRC-JCS joint model).Lugeon tests to determine joint conducting apertureusing Snow’s method;

3 generate joint performance curves of normal load/(aperture or conductivity) and shear strain/ (shear stress,dilation, conductivity) from above data using the onedimensional program Lotus-BB;

4 build numerical model UDEC-BB. It is practicallyimpossible to model every joint so a representative jointpattern is chosen and the elastic moduli correspondinglyreduced. The model can be run in several stages with theeffects of, for example, bolting and lining applied atappropriate times.

Output from the model includes:

1 changes in joint conducting aperture;

2 principal stresses;

3 deformations;

4 joint shear deformation.

There is no comparison in the paper between thepredicted effects of tunnelling and those which actuallyoccurred so it is not useful as a case history.

Wood (1991) discusses the estimation of Hoek-Brownrock mass strength parameters from rock massclassifications. This would allow rock mass strength andfailure conditions to be assessed from a knowledge ofintact strength and rock classification. Hoek and Brown(1980) developed a relationship between the principalstresses at failure in rock and the uniaxial compressivestrength of intact rock which involved the empiricalconstants m and s, very approximately analogous to φ' andc' in the Mohr-Coulomb failure criterion. The values of mand s for intact rocks are determined from curves fitted totriaxial test results. A table of values of m and s fordiscontinuous rock, from Hoek and Brown (1988), is givenwhich was generated using a set of empirical relationshipswith RMR which are also given. The RMRs in this caseare not adjusted for discontinuity orientation. Experiencehas since shown that these relationships underestimate thestrength of the rock mass at low confining stresses.

For this reason Wood proposes that the RMR should besubdivided into partial classification parameters RMR

m

and RMR s and uniaxial compressive strength σ

c. RMR

m

contains the discontinuity condition term only and isrelated to m. RMR

s contains the discontinuity spacing and

RQD terms and is related to s. A correlation betweenRMR

m and m

b /m

i is given, based on Pangua Andesite

data. The correlation is extended to give an RMR m

of 40for intact rock (m

b /m

i=1). A correlation between RMR

s

and s is also proposed and plotted. The correlation isextended to a rating of 25 for unbroken, intact rock.Correlations are also presented between ‘Q’ ratings J

r /J

a

and m b /m

i and between RQD/J

n and s.

The process of deriving the Hoek-Brown parameters isdescribed and illustrated with a worked example. If a fullrock testing programme is not possible, m

i must be taken

from published tabulated values and σ c can be obtained from

The derivation of the constitutive model and the inputdata required for UDEC-BB is described. A laboratory tilttest on jointed cores is described which has been found agood index test for finding peak strength of joints. Theresults show remarkably good agreement with direct sheartests at five orders of magnitude higher normal stress. JRCcan be evaluated directly from tilt tests. Both shearstrength and stiffness are affected by scale effects,confirmed by model tests. Scale correction curves arepresented for JRC and JCS for extrapolation fromlaboratory tests to in-situ block sizes. Methods are given ofobtaining JRC by the use of asperity depth measured witha straight edge or by using J

r from the

‘Q’ system. A

method of estimating the peak friction angle (φpeak

) from Jr

and Ja is also given.

The special version of UDEC used at NGI incorporatesthe joint friction model, a joint dilation model, the scaleeffect, shear stress/displacement curve model and theeffects of load reversals. The practical use of this model isdiscussed in Makurat et al (1990), below.

Makurat, et al (1990) discuss the practical applicationof the two dimensional UDEC-BB discontinuum code tothe design of a cross section of the Fjellingen road tunnelsunder Oslo. The determination of the input data isdescribed. Steps involved in numerical modelling are:

1 surface exposure mapping of discontinuities andborehole logging;

( )[ ]τ σ σ φ

τσφ

= +n r

r

where

tan JRC log JCS /

= shear strength

= normal stress

= joint friction angle

n

n

19

point load tests or a field estimate based on ISRMrecommendations.

Hudson (1989) introduced the concept of arepresentative elemental volume (REV) as an aid todetermining whether the rock in a rock engineeringproblem should be analysed as a continuum or adiscontinuum. The REV is the minimum volume of adiscontinuous rock which must be tested to reduce thescatter in the measurements of a particular property to aneffectively constant value. It is suggested that in modellingstructures in rock that the near-field, up to the REV, betreated as a discontinuum and the far-field as a continuum.

4 Case histories

4.1 Round Hill tunnels

The twin bores of the Round Hill tunnels were the firstroad tunnels in the UK to be designed using the basicprinciples of the NATM. They were constructed betweenNovember 1990 and their opening in December 1993 totake the re-aligned A20 through the Lower Chalkescarpment north of Folkestone. There are two tunnelsapproximately 350m long, with a centreline separationwhich varies but is never less than the 30m separation atthe portals. The excavated cross-section is about 11.7mwide and 10.3m high, with an arched crown and a fairlyflat invert. The faces were excavated by top heading andbench with the benches following after completion of thetop headings. Primary support was by rockbolts and meshreinforced shotcrete with or without lattice arches. Finalsupport was by a cast concrete arch. A drainage layer andwaterproofing membrane was placed between the twolinings.

Ground investigationGround investigation borehole logs and the results of

laboratory testing formed part of the Contract documentsavailable to tenderers. The ground investigation providedfive boreholes near the tunnel alignment, using rotarycoring, and one shell and auger borehole. The tunnel wasentirely within the Grey Chalk and the Upper Chalk Marlof the Lower Chalk. Initially both ‘Q’ and RMR systemswere used for rock mass classification, based on theborehole data (Table 5). However it was later decided torely on the ‘Q’ system alone. The Lower Chalk at thislocation comprises 15m of grey/white massive to blockychalk known as the White Chalk underlain by some 25mof Grey Chalk which is in turn underlain by around 40m ofthe Chalk Marl. The Chalk Marl is a dark grey marly

limestone with a high clay content. At the south-westernportals the tunnel cross-section is entirely within the UpperChalk Marl with the cover in the Grey Chalk, whereas atthe north eastern portals both the tunnel and the cover arewithin the Grey Chalk.

DesignDesign of the support systems was by continuum analysis(Murphy and Buttfield, 1991) using both two-dimensionalfinite element programs and a finite difference analysisprogram for soils and rocks. Rock mass strengthparameters for this purpose were derived from the rockmass classification systems using the Hoek-Brown (1980,1988) failure criterion. (This technique allows for thedegradation of rock mass strength caused bydiscontinuities but cannot model the kinematic behaviourof jointed rock.) Four combinations of ground and primarysupport were analysed using this model, covering the threesupport classes (Table 6). Support from the primary liningwas ignored in the design of the 300mm thick castconcrete secondary lining (Murphy et al, 1993).

Contract and specificationThe conditions of contract were the ICE 5th edition.Specifications, method of measurement and cost controlsystems were those of DOT. Three types of primarysupport had been designed by the Engineer. Each supporttype was assigned to a range of rock mass quality asdetermined by the ‘Q’ system. For tendering purposes thelength of tunnel in each rock class was estimated. Thecontract provided for the final support type to be chosenby the Engineer, based on rock mass quality anddeformation measurements in the tunnel (the latter beingan essential component of NATM). The Engineer was alsoempowered to vary the details of any of the support typesto suit the ground conditions.

ConstructionBoth top headings were driven from the eastern portals.Excavation of the westbound tunnel started first and wasalways considerably in advance of the eastbound. The topheading in the westbound tunnel was completed before itsbench was excavated, as was the top heading in theeastbound tunnel. Table 7 below compares the proportionsof each support type actually installed with those originallyanticipated from the site investigation data.

Difficulties were encountered with overbreak andinstability of the crown which necessitated the use ofspiling and reductions in the advance length with bothType 2 and Type 3 support. Type 1 was little used becausethe lattice girder was necessary to facilitate spiling andmaintain the profile of the crown. As seen from the tableabove it was generally found possible to use a lower classof support in the bench than in the crown. This may in parthave been due to better quality rock in the lower part of thetunnel, but also highlights that block fallout problems arelikely to be more critical in the crown than the sidewalls.

There was initially a problem with the contract indistinguishing temporary (cost allowed for in the

Table 5 Rock mass classification

Chalk RMR Geomechanics Q Description SupportGrade Classification Class

II 40-50 Fair 0.7-2 Poor 1III 25-40 Poor 0.12-0.7 Very poor 2IV 17-25 Very poor - poor 0.05-0.12 Extremely poor 3

20

contractor’s bid) and permanent (measured costreimbursed), overcome with the Client’s agreement by theEngineer directing most of the spiling, which then becamepart of the measured permanent support.

The design model produced expected displacements forcomparison with those measured in the tunnel duringconstruction, but not those at the ground surface. Thepredictions of crown settlement were 15% to 45% less thanthose actually measured for top heading excavation and1.4 to 5.6 times greater than those measured for excavationof the bench. However the overall crown settlementpredictions for both heading and bench were much closerfor three of the sections, being generally within 10%,except for the section modelled with type 1 support wherethe total crown settlement was over predicted by 42 to 48%.

Seismic investigationAs discussed in section 3.1 Barton and Grimstad (1994a)propose an approximate correlation between compressive(P) wave velocity (V

p) and rock quality ‘Q’ which shows

promise as a site investigation tool. The relationship isaffected by stress (depth) and porosity and more data isneeded to refine it. To this end a seismic survey wascarried out during the construction of the Round Hilltunnels by TRL (Bowers and Hiller, 1994) to assess thevariation of V

p along the drive and compare it with the

assessed ‘Q’ values predicted by the site investigation andmeasured in the tunnels. This permitted an independentmeasure of the extent to which the support types usedreflected the ground conditions measured by V

p. The

investigation was carried out by measuring the time takenfor a seismic wave to travel between the two bores at anumber of locations along the tunnels. A critical depth ofcover of 22m was found above which the seismic velocitywas approximately proportional to the overburden. Atshallower depths, the depth of cover had little influence onvelocity, probably due to insufficient overburden stress to

seismically close the joints. The results tended to confirmthe distribution of ground quality indicated by the siteinvestigation. There was also a reasonable correlation withthe support used. However the differences between supporttype predicted and that used did not only reflect groundquality variation. Lattice girders were installed at somelocations at which the ground conditions did notnecessitate their use since the efficiency of the constructiontechnique benefitted from their presence.

This was the first time that the NATM had been usedunder British highway contract conditions. Tunnellinggenerally went well, with the support methods providingrapid stabilisation in poor ground conditions.

4.2 Penmaenbach tunnel

The Penmaenbach tunnel is one of a series ofimprovements to the North Wales Coast Road constructedin 1986-7 through a headland of strong igneous rock(Ordovician Rhyolite). Excavation was by drill and blastexcept for a short length at the Eastern portal through looseground. Temporary support was to comprise either steelribs or rockbolts and sprayed concrete with and withoutfabric reinforcement depending on the ground conditionsencountered. This temporary support was defined asforming part of the permanent works.

Site investigationThe contract documents provided a wealth of siteinvestigation information including the results of mappingof discontinuities shears and folds in outcrops and theparallel unlined rail tunnel. Good access for geologicalmapping was available, in the existing unlined rail tunnelrunning roughly parallel about 100m distant from the newtunnel, and surface outcrops. These data were provided inthe form of tables, drawings and polar discontinuity plots.The site investigation interpretive report reveals thattunnelling conditions were assessed by both ‘Q’ and RMRrock mass classification systems. The indices werecalculated from both borehole logs and surface mapping.Ranges of ‘Q’ and RMR are given in detail for boreholesin the portal areas and the tunnel zone, at tunnel level, at3m above the crown and 3-8m above the crown. They arealso given for coastal exposures and at intervals along therail tunnel. Expected ground water inflows were reportedas 35 L/min/10m length at the east portal zone, negligiblein the west portal zone and less than 50 L/min/10m in thecentral part. The hardness of the Rhyolite was expected to

Table 7 Round Hill tunnel-actual v. predicted supportclass

(Percentages are Actual Actualof total length) (Eastbound) (Westbound)

Support type Predicted Heading Bench Heading Bench

1 20% 1% 54% 7% 7%2 64% 92% 43% 64% 91%3 16% 7% 3% 29% 2%

Table 6 Support description

Heading Bench

Support Lattice Advance Dowels Dowel Mesh Shotcrete Dowelstype girder (m) (m x mm) spacing (m) (mm)

1 No 1.5 3 or 4 x 25 1.5 1 layer 150 None2 No 1.5 3 or 4 x 25 1.5 1 layer 200 None3 Yes 1.0 2 x 25 1.5 2 layers 200 4m x 25mm

21

necessitate drill and blast excavation, and high rates ofdrill tool wear were expected. An unfavourably orientatedjoint set striking parallel to the tunnel with a steep dip wasidentified as a likely cause of overbreak and roof-falls, as itwas in the rail tunnel. The rail tunnel also allowed theidentification of the position of shear zones at which rapidsupport would be necessary in the new tunnel. Designadvice was given for three possible east portal schemesand for the west portal in general terms for both temporaryand permanent support. The expected lengths of the near-portal support type were given. Design advice for the maintunnel support drew heavily on observation of the 100 yearold rail tunnel which is largely unlined. Rock dowels ortensioned bolts with shotcrete were suggested for thefracture zones, together with spot bolting of unstableblocks.

Contract and specificationThe Conditions of Contract were based on the ICE 5thedition form to the general requirements of theSpecification for Highway Works. It was specified that theexcavated face was to be geologically mapped afterblasting and scaling of the face and the installation oftemporary support. Initial rockbolt and sprayed concretesupport was to be installed between 5m and 7m from theface to avoid blasting damage. Steel rib support was fullyspecified and shown on the drawings. Full details ofrockbolt installation tensioning and proof testing werespecified. A detailed specification covering all materialsand admixtures, mix design, setting times and strengths,operators, test panels and testing was given. Steel fibrereinforcement was allowed at the Resident Engineer’sdiscretion.

Alternatives for the permanent lining, required tominimize maintenance and for safety from minor falls,were suggested as shotcrete with dowels or bolts and meshreinforcement where required to control cracking indiscontinuity zones, or shuttered mass concrete to give asmooth profile. In the event the unreinforced massconcrete lining was constructed using a rail mounted steelshutter.

4.3 Pen-y-Clip

The Pen-y-Clip tunnel is also on the North Wales CoastRoad and was the next to be constructed afterPenmaenbach. It was opened in October 1993 at thecompletion of a 4 year contract period of which tunnellingtook 2 years. It passes through an intrusion of Ordovicianmicrodiorite into mudstone which is covered with fossilscree and quarry debris. The total length is 930m includingportal rock shelters, and is of ‘D’-section 11m high and 9mwide. The primary lining comprises rock bolts withshotcrete or steel ribs which was designed to provide thefull structural support. The final lining is a minimum of300mm thick cast in-situ concrete.

Site investigationArber and Selley (1991) give details of the Pen-y-Clipsite investigations, which took place in 13 stages between

1974 and 1986. Discovery in 1983 of a fissure largeenough for man entry, after cleaning to rockhead at theportal site of the originally proposed northern tunnel,resulted in the abandonment of this line and the adoptionof a line further into the headland. Access for drilling onthe tunnel line was very difficult due to the very steepslope covered with scree. Drilling difficulties (laterrepeated in the tunnel) were such that one borehole tookeight weeks to complete 22.2m through fossil scree andwas aborted before reaching bedrock, due to the nature ofthe ground. Due to the complexity of the geotechnics itwas still necessary to incorporate further site investigationworks into the main tunnel construction contract. It wasexpected from the site investigation data that theexcavation would be by drill and blast in hard rock formost of the tunnel length.

As had been the case for the Penmaenbach tunnel thetender documents included detailed factual siteinvestigation data. Again the adjacent unlined rail tunnelprovided ready access to exposures of similar rock. Jointmeasurements from outcrops in nine areas were presentedas stereograms and tables. The tables include data on rocktype, RQD, weathering, groundwater, intact strength, jointspacing of each set, orientation of each joint set,continuity, separation, infilling, shape, roughness andstrength of joint wall rock for each set and the presence offaults or shear zones. RQD measurements were madealong scanlines in several directions using the techniquesof Priest and Hudson (1976). RMR values were calculatedfrom the collected data and were tabulated for 11 zones.The nature of the jointing is described in detail.

Discontinuity logging of borehole core from 46boreholes in the region of the tunnel was also used toassess both ‘Q’ and RMR values. The records include, corerecovery, RQD, fracture index (discontinuities/m), position(depth), orientation, roughness and condition and numberof sets. The entire West Quarry, close to the tunnel line,was mapped in 60m2 strips to provide RQD and fractureindex measurements and computer generated stereoplots.

Contract and specificationAs at Penmaenbach the Contract was based on the ICE 5thedition form, to the general requirements of theSpecification for Highway Works. Exploratory fully-coreddrilling in the tunnel was specified to be carried out asdetermined by the Engineer. Full details of the coringmethod, storage and core logging were specified. It wasalso specified that the arrangement of permanent supportsdescribed in the Contract might be modified aftergeological mapping of the exposed and cleaned bedrock.The initial supports were specified and were part of thepermanent works. Rockbolts and shotcrete were to beinstalled between 3m and 7m from the face. Othertemporary support would not be accepted as part of thepermanent works. Details of advance probe drilling weregiven as were procedures for dealing with water ifdetected. Full details of the installation of steel rib supportwere given. Details of the support systems are givenbelow.

Geological mapping of the face was to be completedbefore drilling for the next round, by the Contractor’s

22

specialist engineering geologist, in the presence of theEngineer’s geologist. ‘This shall comprise a detailedexamination of the tunnel face, crown and sidewalls for thepurpose amongst other things of determining the rock massclass’. The geological logs were to include, lithology,degree of weathering and details of discontinuities. RQD,RMR and ‘Q’ values were to be determined and the rockmass class identified in accordance with the classesdescribed in the Contract. The Engineer’s assessment wasto prevail in the event of a disagreement.

Subsequent clauses similar to those for Penmaenbachdescribed the materials and processes for the varioustunnel support elements and grouting. Instrumentation wasspecified to be installed at prescribed intervals to measureconvergence, rock strain and support load, and the times ofreading.

ConstructionIn September 1991, after about 12 months of excavation,the tunnel was 25 weeks behind the Clause 14 programmeand extrapolation suggested a total expected delay of 88weeks. Eight weeks extensions to time had already beengranted. The main difficulty was in tunnel support. Thecontract specified 3 rock classes in terms of ‘Q’ and RMRwith associated support for each class (see Table 9 inCatling and Scholey, below). The expected occurrences foreach class, as a percentage of the total length of tunnelwere 36% in Class 1, 56% in Class 2 and 8% in Class 3.

Tunnel support was reviewed and conditions comparedwith those expected, as work progressed. The disparitybetween the conditions expected and those found in thetunnel resulted in a need to revise the predictions for theremainder of the work. Revised support classes are shownin Table 8. A risk-based prediction model was developed.This approach used the distributions of the rockclassification system input parameters, measured in thetunnel completed to date, to predict the probability of theoutput distributions, using a numerical simulationtechnique. In this case the output required was the ‘Q’value at any point of interest in the tunnel. In certainmixed-face conditions consideration of infilled fissurewidths overrode the ‘Q’ value, so the model included thisas an extra parameter to the six input parameters to ‘Q’.The model divided the tunnel into 168 five metre segmentsand calculated the probability of each support class foreach segment, based on the input data. In the tunnel, thesupport category was based on an agreed overall ‘Q’ valuefrom ‘all exposed rock faces’. In mixed face conditions acontrolling ‘Q’ value was identified, usually for the worstconditions at the face, or extrapolated just ahead of it.

Unexpected geotechnical problems encountered, givingrise to Clause 12 claims, included: infilled fissures; highlydisturbed rock with clay-filled joints; loose ground andshear zones. These conditions often required a spiledcanopy or in some cases, a micro-heading. Small blockshad given difficulties with rock-bolting. Loose ground andwide jointing were expected from the ground investigation,but some ‘multiple wide fissures’ had been accepted asunforseen conditions. An assessment of the need to orderfurther steel ribs was based on a ‘broad overview of thegeological structures exposed in viaduct bay and the WestQuarry and ground investigation boreholes’. It wasassumed that Class 3 could continue to the easternboundary structure where the more competent rock wasexpected at chainage 1250. The revised requirement forsteel ribs was based on this. This chainage was found toagree quite well with the 65% confidence limit for Class 3support predicted by the most recent risk-based analysis,and was not seen as unduly conservative. Further detailsare provided by Haycock (1996).

Operational problems had given rise to much lower ratesof advance than anticipated, mainly caused by difficultieswith achieving clear drilled holes for rockbolts and in theface similar to those experienced in drilling from thesurface during the ground investigation. These werecaused by small fragments of rock jamming the drill string,and the resulting holes being unsuitable for resin capsuleswithout further treatment, such as grouting and re-drilling.These problems were mainly in the lower Class 2 rock (‘Q’= 0.01 to 0.1). Delays caused by difficulties withshotcreting were not related to the geological conditions.Extensive trials of rockbolting were performed to identifythe best procedures for overcoming the problems. Varioussupport modification options were considered to speed upconstruction in the light of these difficulties. A modifiedClass 3 support (Class 3m) was adopted from chainage1040, with steel ribs at 1.5m spacing and 100mm meshreinforced shotcrete between the ribs. An alternativedesign was considered comprising 370mm shotcrete withone layer of mesh and lattice arches at 750mm centres,installed by heading and bench working, as for the Class 3steel support. However frequent switching between Class 2support (necessarily constructed full-face) and thisalternative would incur excessive delay. The varioussupport options throughout the remainder of the tunnelwere assessed on a time and cost calculation based on boththe 50% and 65% confidence limits of expected rockclasses and locations. It was concluded that theintroduction of lattice girder and shotcrete support inClasses 2 and/or 3 should produce a significant reductionin excavation times even if conditions were worse thanthen expected. It was expected that, although the directcosts would increase, the cost of prolongation wouldsubstantially reduce, giving a net gain. In the event thiswas not accepted and the Class 3 part of the tunnel wascompleted in steel arches.

Catling and Scholey (1994) provide a detailed accountof lining the tunnel. The ‘Q’ system and the RMR wereboth used for rock classification. The ‘Q’ system proved tobe the more accurate and sensitive, particularly in poor

Table 8 ’Q’ based support selection

‘Q’ value Support Class

<0.01 Class 30.01 to 0.1 Class 2 lower0.1 to 0.4 Class 2 upper> 0.4 Class 1Widely fissured Class 3

23

rock. The Contract specified the primary support systemsdescribed in Table 9 below. These were designed toprovide the full permanent support.

The length of bolts was determined from wedge analysis(the specific method is not specified) and the steel archeswere designed for the full overburden pressure at theportals. Class 3 sections were driven by heading and benchand Classes 1 and 2 full-face. Permanent works were notallowed closer than 3m to the face in Classes 1 and 2.Sprayed concrete was by the dry process. Some steel fibrereinforced shotcrete was used for re-profiling but was notpursued for long. Re-profiling to fill overbreak wasintended to be by shotcrete, but problems were expectedwith obtaining the required tolerance (25mm) on thesurface profile. Wet process shotcrete was rejected afterobservation of European practice. Finally this concrete wasplaced behind a 10m long shutter. Rockbolts and meshwere used to tie this concrete to the shotcrete so that noload was transferred to the secondary lining. Thissecondary final lining was cast in-situ, over a full-profiledrainage membrane, to provide a smooth surface but wasnot necessary as structural support. This lining is 300mmthick in the rockbolted sections and 600mm thick in thesteel rib sections.

On the completion of the tunnel Fowler (1993) reportedthat 68% of the tunnel was built with steel ribs, rather thanthe 8% originally envisaged. This was caused by a muchgreater length of poor quality rock than expected from thesite investigation. This ground was mostly able toexcavated without blasting and the tunnel had to be drivenby heading and bench to avoid face instability. Progress inthe rib sections averaged 12m per week, less than half thatachieved in the full-face central section. Arches at1500mm spacing were reverted to after 35m of difficultieswith drilling holes for bolts. The Contract was granted aten week extension.

The difficulties with poor rock are highlighted byWatson (1991) who reported a crown collapse 60m fromthe west portal. Loose rock flowed from the crown andface of the top heading, when miners were excavating upto 2m ahead of the heavy steel ribs at 750mm centres usedto support continuous poor ground. The geotechnicalconsultant insisted that the conditions were consistent withthe site investigation.

One of the reasons for the failure of the groundinvestigation to detect the full extent of the poor rock isdescribed by Slavid (1991). The partial collapse describedabove occurred at a location where a borehole had passed

through solid rock, between two wide clay-filled fissures.This is a common problem with ground investigation fortunnels, because even the largest practical number ofboreholes will only have sampled a very small proportionof the ground to be excavated.

4.4 Carsington Aqueduct

This was an example of a scheme where a site specificrock classification scheme was used to determine supportrequirements. Davey and Eccles (1983) report that anaqueduct including a tunnel 8.5km long and 2.3mdiameter, at a maximum depth of 160m, was required toconnect Carsington reservoir to the Derwent. Thepermanent lining is by smooth-bore pre-cast concretesegments.

Ground investigation boreholes were generally at 1kmintervals, which is typical of long tunnels. These indicatedinterbedded mudstones, sandstones and siltstones at thetunnel horizon. A geophysical survey and in-situ watertests was carried out.

The main recommendations of ‘Tunnelling, improvedcontract practices’ (CIRIA, 1978) were followed in thecontract documents. It was intended that the tunnel bedriven by a cutting boom shield, erecting the lining toprovide immediate roof support. The contract was an ICEstandard admeasurement form. Probing ahead of the facefor 5m and subsequent grouting, where necessary tocontrol water inflow, was specified. A provisional rockclassification into 6 groups was appended to thespecification, based on rock strength and discontinuities.Items for each rock group were included in the Bill ofQuantities. Two rules were applied:

1 The length of tunnel in each rock group was defined bythe characteristic rock group at the tunnel face.

2 Where two or more rock groups were exposed at theface, the characteristic rock group was to be taken as thehighest ranking of those present, provided its totalthickness at the face was not less than 0.6m.

The expected proportions of each group were inserted inthe Bill with allowed pre-grouting time for each.

This case history highlights the need for adequate siteinvestigation, without which no rock classification schemeis likely to be of any assistance. After one year the workwas 6 months behind, mainly due to hard and massive rockand water ingress at higher rates than expected. Boommachines had to be replaced by drill and blast in 3 of 4headings, and compressed air was necessary at two portals

Table 9 Support classes for the Pen-y-Clip tunnel

‘Q’ value RMR Support Class Support

< 0.01 <20 3 Steel arches at 750mm centres 305x305x158kg/m arch on wall beam;(exceptionally poor) (very poor rock) 305x305x137kg/m walls

0.01 - 0.4 2 Rockbolts (.85m x 1m array; typ. 6 x 7m long + 13 x 3.5m long) + full(extremely - very poor) mesh + shotcrete (100mm min)

> 0.4 1 Rockbolts (1.3m x 1.5m array; typ. 4 x 7m long + 13 x 3.5m long)+crown mesh + shotcrete (100mm min)

24

to control water. Additional boreholes were sunk todetermine the optimum measures to be adopted fortunnelling work.

4.5 Tyne-Tees Aqueduct

Davies et al (1981) examined the effectiveness of siteinvestigation in 32.2km of tunnelling, in three sections, forthe Tyne-Tees water transfer scheme. This tunnel was of3.5m excavated diameter through a varied sequence ofsedimentary carboniferous strata with limited occurrenceof intrusive igneous rocks. Most of the tunnel length wasdriven by full face tunnelling machine with limitedsections by roadheader and by drill and blast. Temporarysupport methods included: none; rockbolts and mesh orsteel sheets; sprayed concrete; and in the worst conditions,steel arches. The ground investigation for the Derwent-Wear-Tees tunnels (27.7km) included 18 main boreholescored in the lower sections (75 and 100mm diameter), with18 further boreholes in the portal areas. Spacing betweenboreholes in the deeper tunnels was up to 3km and manywere over 500m from the tunnel line because of routeuncertainty at the time of the investigation. Cores werelogged for RQD, Schmidt rebound and fracture spacing.Packer tests and piezometers gave indications ofpermeability and water pressures at tunnel level. Strength,deformation and swelling characteristics of over 350samples were measured in laboratory tests. Based on thesite investigation data, which included results from anexperimental tunnel at Rogerley Quarry, TBMs wasselected for the excavation of the two long tunnels anddrill and blast for the shorter ones. It was expected thatrockbolts, with or without steel mesh would suffice assupport for most of the tunnel, with lagged steel arches orshotcrete being necessary near portals, in some mudstoneareas and in faulted ground.

A five-zone rock classification system for payment wasdeveloped specially for the long tunnels of the scheme,based on the long sections derived from the investigationresults (Table 10). The system was designed to allowadjustment of the contract value to take account of theactual rock conditions.

Ground conditions were continuously logged in thetunnel by the RE’s inspectors. In the machine boredtunnels the face was not accessible so the sidewalls andcrown were logged on specially designed logsheets. Thesewere assessed by the Resident Engineer’s geologist whodetermined the classifications.

Drill and blast tunnel. Overall the site investigationpredictions were substantially correct with the mainstructural features occurring at the expected locations. Onemajor unanticipated fault zone was encountered, and minorvariations in the bedding dip caused greater lengths thananticipated to be excavated with unfavourable roofconditions.

Machine bored tunnels. Overall there was less Class 1and 2 and more Class 3 and 4 than predicted. However thisled to an increase of only 5.5% in payment, confirming thegenerally accurate predictions. There were somesignificant deviations including locally greater faultingthan expected, solution cavities in limestone into whichmudstone had collapsed, an upward transgression of a harddolerite sill, and some lengths in less favourable strata atthe crown than expected.

The level of support required was generallyunderestimated in the drill and blast and overestimated inthe machine bored lengths. It was concluded that there willbe uncertainties in detail in any investigation, due tolimitations on numbers and positions of boreholes, andlikely shortcomings in core and borehole logging. In thiscase the pre-defined rock classification system was flexibleenough to permit adjustment to the Contract value to takeaccount of deviations from the predicted geology. Themost important factor in improving site investigation wasstated to be the positioning of boreholes close to the tunnelline.

5 Conclusions

General

The empirical rock classification systems discussed in thisreport are useful to assist with the selection of tunnelsupport at both the initial design phase and duringexcavation. However they are not and cannot be precisetools and cannot replace experience and engineeringjudgement. In most cases it will also be necessary to useanalytical methods at the design stage to analyse supportoptions.

Choice of system

It is apparent that in recent years the use of general ratherthan site specific rock classification schemes has becomemore common. The general consensus appears to be thatthe ‘Q’ system of Barton et al (1974) is the mostsuccessful of all the systems. This is also the only schemewhich is being actively developed and extended by itsoriginator. It is now backed up by a database of over onethousand case histories covering a wide range of groundtypes and support systems. The RMR system ofBieniawski (1989) is the only other popular contender. In

Table 10 Site-specific rock classification for the Tyne Tees Aqueduct tunnels

Class Rock Quality Support

1 Sandstone Good Little or noneLimestone

2 Sandstone PoorLimestoneMudstone/Shale Very good

3 Mudstone GoodMixed beds

4 Mudstone PoorMixed bedsFault boundaries

5 Fault conditions Heavy ( eg. full circle steel arches)

25

many cases the ‘Q’ system has been shown to be lessconservative than the RMR system and more sensitive tochanges in rock quality. For example Barton (1988)reports that Einstein et al (1979) found the RMR system topredict considerable support in many of the unsupportedcases analysed by Cecil (1970) which were part of thedatabase upon which the ‘Q’ system was originally based.

Site investigation

It is vitally important that thorough site investigation iscarried out. Data should be obtained from as manyboreholes as possible on or very close to the tunnel line.The acquisition of adequate data for the use of rockclassification systems may require more detailed analysisof boreholes than is commonly made, such as the use oforiented downhole CCTV or ‘acoustic televiewer’ surveysto determine joint condition and orientation, which cannotnormally be obtained from cores. Full use should also bemade of exposures close to the tunnel line. Particularlywhere the number of boreholes is limited, investigationahead of the tunnel face by drilling may be justified.

Seismic investigation

Both at the site investigation stage and during tunnellingseismic velocity measurements show promise as a meansof deriving the rock classification. At the site investigationstage seismic cross-hole tomography has been used withsuccess (Barton, 1991; Barton et al, 1992). Barton (1994)has suggested the use of seismic velocity logging in probeholes drilled ahead of the tunnel face for this purpose.

Access for face logging

When rock classification schemes are used during tunnelconstruction to determine support at the face it is vitallyimportant that they are applied by engineering geologistsexperienced in their use. The whole face should be loggedwith attention to the critical areas near the periphery, inparticular near the crown. To this end it is important thatthe Contract specifies that the Contractor should providethe means of access to these areas.

Contracts

Because of the variability of the ground and the inevitablelack of complete knowledge of the ground conditions priorto tunnelling it is important that Contracts allow flexibilityin support selection based on the conditions actuallyencountered. The establishment of the ‘ground referenceconditions’ recommended by CIRIA report 79 (1978) (seesection 2.3.8) could be facilitated by the use of a specifiedrock classification system. The parameters for this rockclassification system should be defined in detail. Theclassification agreed at each advance during constructionwould then, in accordance with the CIRIArecommendations, be compared with the reference classfor the determination of payment. Flexibility would befacilitated by the split of support types not being shown onthe drawings but quantified in the Bill of Quantities, thusgiving the Engineer more freedom to design final usage.

As recommended by CIRIA, cost reimbursable andtarget type Contracts are likely to be the most appropriatewhen limited information is available about the ground,there is insufficient time to prepare an admeasurement typeof Contract, there is a wish to use innovative methods forwhich little cost experience is available, or Contractors arenot willing to respond to a high risk venture. PossibleContract forms include a modified version of the ICE 6thedition, the ICE New Engineering Contract and theIChemE (Green Book) Contract.

Generating analysis parameters

As previously discussed it is vitally important that at thedesign stage the use of rock classification systems for thedetermination of support requirements should be supportedby analytical methods. Where the scale of the discontinuitystructure relative to the excavation is such that theproperties of the discontinuities control the behaviour ofthe ground, it is likely that the use of a discrete elementmethod which can model the discontinuities will provide amore reliable solution than continuum methods. Asdiscussed, a major difficulty in the use of these methods isthe determination of appropriate ground properties andimprovements need to be made. However methods havebeen developed of estimating some of the groundproperties needed for input to these methods using rockclassification schemes.

6 Recommendations

Further work to refine the relationship between rockquality and seismic velocity could render this a very usefulpredictive tool. Measurement of seismic velocity couldthen be used during site investigation and construction toassess the likely rock quality. The development of astandard method of measuring the seismic velocity wouldfacilitate this.

A particular problem of which TRL staff have beenmade aware of is how such irregularities in a tunnel cross-section as recesses and cross-passage junctions might beaccommodated at the support selection stage of the use ofa rock classification system. As this is in effect a change inthe tunnel cross-sectional shape or span it does not appearlogical that the rock quality should be adjusted toaccommodate this. In the ‘Q’ system it would appear thatthe most appropriate parameters to adjust would be thespan or the ‘excavation support ratio’ (ESR) whichtogether are used to define the effective span to be takenfor the selection of the support. The appropriateadjustment could in principle be determined by thecollection and analysis of case history data. Howeverbecause of the difficulties of data collection and ofassessing the factor of safety in existing construction amore satisfactory route would be to calculate theadjustment which would be required for a representativeselection of cases using one of the numerical modellingtechniques described in this report.

A suggested outline sequence for use of rockclassification system in tunnelling projects is given below:

26

1 Site investigation and laboratory testing.

2 Preliminary design using both rock classification and anappropriate numerical modelling techniques to define arange of possible support types.

3 Contract specification to suggest an appropriatetunnelling method and specify ground referenceconditions and measurement of classificationparameters.

4 Continuous face logging, possibly with probe drillingahead of the face, to assess actual ground quality.

5 Selection or modification of support type based on 2 and4 above combined with engineering judgement andexperience.

6 Payment based on agreed support system actually used.

7 Acknowledgement

The author is grateful for the assistance of Dr J Perry inreviewing the draft of this document.

8 References

The references which follow are divided into two sectionscovering first rock classification and second, case histories.The Appendix contains further useful references whichhave been consulted but are not referenced in the text.

8.1 Rock Classification

Attewell P B (1995) Tunnelling Contracts and siteinvestigation. E&F N Spon, London.

Barrett S V L and McCreath D R (1995) Shotcretesupport design in blocky ground: towards a deterministicapproach. Tunnelling and underground space technology,Vol 10, No. 1, pp 79-89.

Barton N (1988) Rock mass classification and tunnelreinforcement using the ‘Q’-system. Rock classificationsystems for engineering purposes. ASTM special technicalpublication 984. ASTM, Philadelphia, pp59-88.

Barton N (1991) Geotechnical design. World Tunnelling,Nov. 1991.

Barton, N and Bandis S (1990) Review of predictivecapabilities of JRC-JCS model in engineering practice.Proceedings of the international symposium on rock joints,Leon, Norway. Eds. Barton and Stephansson, Balkema,Rotterdam.

Barton N and Grimstad E (1994a) The ‘Q’-systemfollowing twenty years of application in NMT supportselection. 43rd Geomechanics Colloquy Salzburg.

Barton N and Grimstad E (1994b) Rock mass conditionsdictate choice between NMT and NATM. Tunnels andtunnelling, Oct 94, pp39-42.

Barton N, Grimstad G, Aas G, Opsahl O A, Bakken Aand Johansen E D (1992) The Norwegian method oftunnelling. World Tunnelling, Part 1, June 1992, Part 2,July 1992.

Barton N, Lien R and Lunde J (1974) Engineeringclassification of rock masses for the design of tunnelsupport. Rock mech 1974-12 v6 n4 p189-236.

Barton N, Makurat A and Grimstad E (1990)Tunnelling by numbers. Tunnels and Tunnelling, July1990.

Bieniawski Z T (1973) Engineering classification ofjointed rock masses. Trans. S. Africa Inst. Civ. Eng. 15, pp335-344.

Bieniawski Z T (1976) Rock mass classifications in rockengineering. Proceedings of the symposium: exploration forrock engineering, Johannesburg. Balkema, Vol 1, pp 97-106.

Bieniawski Z T (1989) Engineering rock massclassifications. Wiley, New York.

Bowers K H and Hiller D M (1994) Seismic investigationof ground variation along the line of the Round HillTunnels. TRL Unpublished Project Report PR/GE/12/94,Transport Research Laboratory, Crowthorne.

British Standards Institution (1981) Code of practice forsite investigations. BS5930:1981, British StandardsInstitution, London.

Brown E T (ed) (1981) Rock characterization, testing andmonitoring, ISRM suggested methods. Pergamon Press,Oxford

Cameron-Clarke L S and Budvari S (1981) Correlationof rock mass classification parameters obtained fromborecore and in-situ observations. Engineering Geology,17, 19-53.

Cecil O S (1970) Correlations of rockbolt/shotcretesupport and rock quality parameters in Scandinaviantunnels. PhD thesis, University of Illinois.

Choubey V D and Dhawan G (1990) Correlations of rockjoints and subsurface for support assessment in aHimalayan tunnel. Proceedings of the internationalsymposium on rock joints, Leon, Norway. Eds. Barton andStephansson, Balkema, Rotterdam.

Construction Industry Research and InformationAssociation (1978) Tunnelling - improved contractpractices. CIRIA report 79.

Conway J J (1993) Application of risk assessment intunnel feasibility studies. Risk and reliability in groundengineering. (Ed. SKIPP, B O). Thomas Telford, London.pp 227-240.

27

Conway J J, Goodfellow R J F and Bowers K H (1995)The potential application of risk analysis to highwaytunnel construction. TRL unpublished project report PR/CE/110/95. Transport Research Laboratory, Crowthorne.(Unpublished report available on direct personalapplication only)

Deere D U, Hendron A J, Patton F, D and Cording E J(1967) Design of surface and near-surface construction inrock. Proc. 8th US Symp. Rock. Mech., AmericanInstitution of Mining Engineers (AIME), New York

Einstein H H, Steiner W and Baecher G B (1979)Assessment of empirical design methods for tunnels inrocks. Proc. 4th rapid excavation and tunnellingconference, AIME, New York, Vol 1,pp 683-706.

Garrett R (1993) Cumberland Gap, North AmericanTunnelling, June 1993, pp n65-n69.

Goel R K, Jethwa J L and Paithankar A G (1995)Indian experiences with Q and RMR systems. Tunnelling andunderground space technology, Vol 10, No 1, pp 97-109.

Hatzor Y (1993) The block failure likelihood: acontribution to rock engineering in blocky rock masses.Int. J. Rock. Min. Sci. & Geomech Abstr. Vol 30, No. 7, pp1591-1597.

Hoek E and Brown E T (1980) Empirical strengthcriterion for rock masses. Journal of GeotechnicalEngineering, ASCE. 106 (GT9), pp 1013-1035.

Hoek E and Brown E T (1988) The Hoek-Brown failurecriterion, a 1988 update. Proc. 15th Can. Rock. MechSymp, University of Toronto, Oct 1988.

Hudson J A (1989) Rock mechanics principles inengineering practice. CIRIA ground engineering report.Construction Industry Research and InformationAssociation, London.

Institution of Chemical Engineers (1992) Model form ofconditions of contract for process plant suitable forreimbursable contracts, second edition, IChemE, Rugby.

Institution of Civil Engineers, Association of ConsultingEngineers, Federation of Civil Engineering Contractors(1991) ICE conditions of contract and forms of tender,agreement and bond for use in connection with works ofcivil engineering construction, 6th edition (Jan, 1991),Thomas Telford Ltd, London.

Institution of Civil Engineers (1992). Design andconstruct conditions of contract and forms of tender,agreement and bond for use in connection with works ofcivil engineering construction, 52pp, Thomas Telford Ltd,London.

Institution of Civil Engineers (1993) The newengineering contract, 1st edition, Thomas Telford Ltd,London.

International Society for Rock Mechanics (1981) Rockcharacterization testing and monitoring. E T Brown (Ed.)for the Commission on standardisation of laboratory andfield tests. Pergamon.

International Society for Rock Mechanics (1985)Commission on standardisation of laboratory and fieldtests. Suggested methods for determining point loadstrength. Int. J. Rock. Min. Sci. & Geomech Abstr.Vol 22,No.2, pp.51-60.

Johnston I W (1994) Soil mechanics, rock mechanics andsoft rock technology. Proc Inst Civil Engrs. GeotechEngng, 107, Jan, 3-9.

Kirsten H A D (1988a) Discussion on rock mass ratingsystem. Rock classification systems for engineeringpurposes. ASTM special technical publication 984.ASTM, Philadelphia, pp32-33.

Kirsten H A D (1988b) Discussion on ‘Q’ system. Rockclassification systems for engineering purposes. ASTMspecial technical publication 984. ASTM, Philadelphia, pp85-87.

Lauffer H (1958) Gebirgsklassifizierung Füf denstollenbau. Geol. Bauwesen 74, 1958, pp 46-51.

Makurat A, Barton N, Vik G, Chryssankathis P andMonsen K (1990) Jointed rock mass modelling.Proceedings of the international symposium on rock joints,Leon, Norway, Eds. Barton and Stephansson, Balkema,Rotterdam.

Murphy and Buttfield (1991) A20 Round Hill. WorldTunnelling, May 1991.

Murphy P, Harrison N and Myers A (1993) Design andconstruction of the Round Hill Tunnels. Report on ameeting of the British Tunnelling Society; Snowdon, E,Rapporteur. Tunnels and tunnelling, April 1993.

Napthine R and Smart R (1995) Design and build -lessons from the UK Channel Tunnel terminal. Proc InstCivil Engrs, Civ Engng, 108, Aug, 123-130.

National Economic Development Council (1968) Reporton contracting in civil engineering since Banwell. HMSO,London.

Österreichisches Normungsinstitut (1994) UndergroundWorks - Works contract. ÖNORM B 2203.

28

Pacher, F (1964) Deformationsmessungen im versuchsstollenals Mittel zur Erforschung des Gebirgsverhaltens und zurBemessung des Ausbaues. Felsmechanik undIngenieurgeologie, Supplementum 1, 149-161.

Pan X D and Trenter N A (1992) Tunnelling in classifiedrock masses. Tunnels and Tunnelling, July 1992. IRRD850743

Priest S D and Hudson J A (1976) Discontinuity spacingsin rock. Int. J. Rock Mech. Min Sci. & Geomech. Abstr.Vol 13, pp. 135-148. Pergamon Press.

Rutledge J C and Preston R L (1978) New Zealandexperience with engineering classification of rock for theprediction of tunnel support. Proceedings of theInternational tunnelling symposium, Tokyo, A3.1-A3.7,Pergamon.

Siddans A W B (1995) A new digital accoustic televiewer.Robertson Geologging Ltd, Deganwy, Conwy, Gwynedd,LL31 9PX.

Singh B, Jethwa J L, Dube A K and Singh B (1992).Correlations between observed support pressure and rockmass quality. Tunnelling and underground spacetechnology, 7 (1): 59-74.

Tallon E M (1982) Comparison and application ofgeomechanics classification schemes in tunnel construction.Proceedings of the third int symp, Tunnelling 82, Institute ofMining and Metallurgy, London2, pp. 189-236.

Tarkoy P J (1995) Rock mass rating systems: to use ornot to use? Tunnels and Tunnelling, May 1995.

Terzaghi K (1946) Rock defects and loads on tunnelsupports. Rock tunnelling with steel supports, eds. R VProctor and T White, Commercial Shearing Co,Youngstown, OH, pp 15-99.

Unal E (1983) Design guidelines and roof controlstandards for coal mine roofs. PhD thesis, PennsylvaniaState University.

Ward W H (1978) Ground supports for tunnels in weakrocks. Géotechnique 28, No.2, 133-171.

Wickham G E, Tiedman H R and Skinner E H. (1972)Support determination based on geologic prediction. Proc.rapid excavation and tunnelling conference, AIME, NewYork, pp 691-707.

Wood D F (1991) Estimating Hoek-Brown rock massstrength parameters from rock mass classifications.Transportation Research Record 1330.

8.2 Case history references

Arber A W and Selley P J (1991) Pen-y-Clip siteinvestigation. World Tunnelling, Vol 4, No.2, p112-116.

Barratt D A; O’Reilly, M P and Temporal J (1994) Long-term measurements of loads on tunnel linings inoverconsolidated clay. Seventh International Symposium,‘Tunnelling’94’, of the Institute of Mining and Metallurgy andthe British Tunnelling Society. Chapman and Hall, London.

Bowers K H and Redgers J (1996) Discussion:Observations of lining load in a London clay tunnel. Proc.Int. Symp. Geotechnical aspects of undergroundconstruction in soft ground, eds Mair and Taylor.Balkema, Rotterdam.

Catling P and Scholey J (1994) Pen-y-Clip road tunnel:achieving a dry lining through a pervious rock formation.Seventh Int. Symp, Tunnelling 94. Chapman and Hall,London.

Davey P G. and Eccles P G (1983) The Carsingtonscheme - reservoir and aqueduct. The Journal of theInstitution of Water Engineers and Scientists, Vol 37, June1983.

Davies T P, Carter P G, Mills D A C and West G (1981)Kielder aqueduct tunnels - predicted and actual geology.TRRL Supplementary Report SR 676. Transport ResearchLaboratory, Crowthorne

Fowler D (1993) Peak performance. New Civil Engineer,No 1063, p14-15, 4/11/93.

Haycock D (1996) The Pen-y-clip tunnel. Report onBTS meeting; speakers Hilmy, A M, Hindle, D J,Scholey, J and Cockett, J. Tunnels and Tunnelling, Vol 28,No 1, Jan 1996.

Jones H (1987) Methods magnify Saltash backlash. NewCivil Engineer, 9 July 1987, No 747, p23-4. IRRD 808230.

Slavid R (1991) Fissures thwart a rocky operation.Construction Weekly, 22 May 1991.

Wallis S (1987) Caution the key to Cornish road tunnel.Tunnels and Tunnelling, July.

Watson R (1991) Mountain moves on Pen-y-Clip. Tunnelsand Tunnelling, Feb 1991.

White P (1988) The Cornish Underground movement. TheSurveyor, Nov, 1988, Vol 169, No 4981, 18-19.

29

Appendix: Other useful references

Brown L A (1987) Correlation of tunnel support loadswith geology. Bulletin of the Association of EngineeringGeologists. Vol XXIV, No 1, 1987, pp 15-26.

Francis T E (1991) Determination of the influence of jointorientation on rock mass classification for tunnelling usinga stereographic overlay. Quarterly Journal of EngineeringGeology. 24, 267-273.

Goel R K, Jethwa, J L, and Paithankar A G (1995)Tunnelling through the young Himalayas - a case historyof the Maneri - Uttarkashi power tunnel. EngineeringGeology

Henscher S R and McNicholl D P (1995) Engineering inweathered rock. Quarterly journal of Engineering geology.28, 253-266.

John M and Crighton G S (1989) Monitoring andinterpreting of results of geotechnical measurements forNATM linings design for the Channel Tunnel. Proceedingsof the conference, Geotechnical instrumentation in civilengineering projects, Nottingham. Thomas Telford,London.

Nakao, K, Iihoshi S and Koyama S (1983) Statisticalreconsideration on the parameters for GeomechanicsClassification. Proceedings of the International Congresson Rock Mechanics, Melborne, ISRM. Vol. 1, pp B13-B16.IRRD 271888.

Sen Z and Eissa E A (1991) Technical note - Cumulativesemivariogram technique for identification of intact lengthcorrelation structure. Int. J. Rock. Min. Sci. & GeomechAbstr. Vol. 28, No. 5, pp. 421-429. IRRD 844740

Scholey J and Ingle D G (1989) Monitoring tunnelsupport by convergence measurement. Geotechnicalinstrumentation in practice. Proceedings of the conference,Geotechnical instrumentation in civil engineering projects,Nottingham. Thomas Telford, London.

Scott G A and Kottenstette J T (1993) Tunnelling underthe Apache trail. Int. J. Rock. Min. Sci. & Geomech Abstr.Vol 30, No. 7, pp 1485-1489.

Sinha R S (1988) Discussion on rock structure ratingmodel. Rock classification systems for engineeringpurposes. ASTM special technical publication 984.ASTM, Philadelphia, pp 50-51.

Ulusay R, Aksoy H and Ider M H (1992) Geotechnicalapproaches to the design of a railway tunnel section inandesite. Engineering Geology, 34, 81-93.

Verman M, Jethwa J L and Singh B (1991) ‘Q’- systemmodified for squeezing ground. Tunnelling 91, Sixthinternational symposium, London. Elsevier, p117-22.

30

Abstract

Recent UK and other practice in the use of empirical ground classification systems for the selection of tunnelsupport is examined to identify new developments and note particular problem areas. The concept and nature of thesystems are discussed in the first section. Recent literature is then examined and discussed, followed by a review ofUK case histories of tunnelling where ground classification schemes were used. Final sections draw together themost important conclusions to be gleaned from the evidence and some guidance is provided on the practicalapplication of the systems. Recommendations are also made on the possible direction of future research.

Related publications

PR60 Study of the efficiency of site investigation practices by Mott MacDonald and Soil Mechanics Ltd.1994 (price code J)

TRL192 Sources of information for site investigations in Britain (revision of TRL Report LR403) byJ Perry and G West. 1996 (price code L)

TRL209 Field evaluation of the TRL load cell pressuremeter by P Darley, D R Carder, M D Ryley andP G Hawkins. 1996 (price code E)