loach, simon c. (1987) repeated loading of fine grained soils for...
TRANSCRIPT
Loach, Simon C. (1987) Repeated loading of fine grained soils for pavement design. PhD thesis, University of Nottingham.
Access from the University of Nottingham repository: http://eprints.nottingham.ac.uk/13693/1/376631.pdf
Copyright and reuse:
The Nottingham ePrints service makes this work by researchers of the University of Nottingham available open access under the following conditions.
This article is made available under the University of Nottingham End User licence and may be reused according to the conditions of the licence. For more details see: http://eprints.nottingham.ac.uk/end_user_agreement.pdf
For more information, please contact [email protected]
REPEATED LOADING OF FINE GRAINED SOILS FOR PAVEMENT DESIGN
by
SIMON C. LOACH, M.A. (Cantab)
Thesis submitted to the University of Nottingham for the degree ofDoctor of Philosophy
FEBRUARY 1987
(i)
ABSTRACT
The primary aim of this research was to investigate the behaviourof a clay subjected to a loading regime similar to thatexperienced by a road subgrade under traffic loading in GreatBritain. The material used was Keuper Marl. The samples wereanisotropically consolidated in a triaxial apparatus from aslurry which allowed careful control over the stress history andproduced uniform samples. The samples were fully instrumentedand the apparatus was capable of applying repeated axial andradial stresses. The test programme was designed to investigatethe resilient and permanent response of the samples to a varietyof stress pulse magnitudes and time periods.
The main conclusions were:
i) The material exhibited a marked stress softening.ii) The mean normal effective stress remained constant under a
variety of total stress paths over the range of frequenciestested.
iii) The resilient response was found to depend on the magnitudeof the applied stress pulse and the mean normal effectivepressure, and to be independent of the preconsolidationpressure.
iv) The material exhibited significant thixotropy.
A smaller parallel series of tests was carried out on compactedtriaxial samples of three clays (Keuper Marl, Gault clay andLondon clay) in a simple pneumatic repeated load triaxial rig.The test programme was designed to investigate the resilientresponse of the samples over a range of repeated deviatorstresses. The suction moisture content relationship for eachclay was established, and the resilient response of the clay wasfound to be controlled by the magnitude of the stress pulse andthe suction.
A series of California Bearing Ratio tests was carried out oncompacted samples of the three clays, and on anisotropically
(i i )
consolidated samples of Keuper Marl, to allow a comparison to bemade between the resilient modulus and CBR.
A review of previous work is presented.
.'(iii)
ACKNOWLEDGEMENTS
The author would like to thank the following people for theirhelp in this research project:
Professor R C Coates and Professor P SPell for making availablethe facilities of the Department of Civil Engineering. ProfessorS F Brown for his helpful supervision throughout the work.Mr B V Brodrick for his help and advice on running andmaintaining the equipment. Mr J Moodyfor his help, advice andworkmanship in the construction and maintenance of the apparatus.Mr G Hanley for his help, advice and workmanship in themodification and repair of the electrical equipment.Mrs P Elliott for typing the thesis. Miss C Brayley for tracingthe figures.
This research project would not have been possible without thegenerous financial support of the Transport and Road ResearchLaboratory and this is gratefully acknowledged.
ABSTRACT
.'
(iv)
CONTENTSPage
iiiACKNOWLEDGEMENTSLIST OF SYMBOLSCHAPTER 1CHAPTER 2
2.12.22.32.42.52.62.72.82.92.102.112.122.132.14
ix
INTRODUCTION 1
6LITERATURE REVIEWEquilibrium States under Cyclic LoadingPore Pressure-Strain RelationshipEffect of Frequency of LoadingEffect of Overconsolidation RatioEffect of Drainage PeriodStress-Strain ResponseAnisotropyLoading RegimesYield SurfacesThixotropyEnd RestraintSuctionRelationship between Suction andResilient ModulusPrediction of Resilient Modulus fromSoil Properties
66101012121314151719202326
CHAPTER 3 THE TEST PROGRAMME3.1 Introduction3.2 Consolidated Triaxial Samples
3.2.1 Initial Stress Conditions3.2.2 Loading Conditions
3.3 Compacted Triaxial Samples3.4 California Bearing Ratio Tests
2626
26283334
CHAPTER 4 THE EQUIPMENT4.1 The Servo Hydraulic Triaxial Test Facility 364.2 Existing Triaxial Equipment 36
4.2.1 The Test Equipment4.2.2 The Consolidation Equipment
3640
4.3 Modifications to the Triaxial Test Facility 404.3.14.3.2
ModificationsLoading FrameModificationsControl Unit
to the Control System and 40to the Consolidation 42
(v )Page
4.3.3 Modifications to the Measuring andRecording Systems
43
4.3.3.1 Axial Load4.3.3.2 Pressure Transducers4.3.3.3 Deformation Transducers4.3.3.4 Volume Change4.3.3.5 Data Processing Equipment4.3.3.6 Digital Data Recording System
434646464748
4.4 Pneumatic Triaxial Rig4.4.1 Existing Equipment4.4.2 Modifications to the Equipment
494949
4.5 Slurry Mould for CBR Samples4.6 Suction Measuring Apparatus4.7 Determination of Pre-consolidation Pressure
505355
CHAPTER 5 EXPERIMENTAL PROCEDURE5.1 The Materials5.2 Consolidated Triaxial Specimens
5757
5.2.1 Initial Soil Preparation 575.2.2 First Stage Consolidation - Slurry Moulds 585.2.3 Cell Base Preparation 585.2.4 Sample Installation 615.2.5 Installing the Instrumentation 645.2.6 Consolidation Procedure 655.2.7 Test Procedure 65
5.3 Consolidated CBR Samples5.4 Compacted Triaxial Samples5.5 Compacted CBR Samples5.6 Measurement of Soil Suction
5.6.1 Rapid Suction Apparatus5.6.2 Equilibrium Method
656869696970
CHAPTER 6 BEHAVIOUR OF OVERCONSOLIDATED SATURATED SAMPLESOF KEUPER MARL
6.1 Introduction6.2 Triaxial Consolidation Results 71
716.2.1 Consolidation Paths 716.2.2 Response of the Radial Drainage System 73
6.3 Effective Stress Response6.3.1 Total Stress Path Tests6.3.2 One Second Stress Pulse Tests6.3.3 One Tenth Second Stress Pulse Tests
77
798588
(vi)
6.4 Resilient ResponsePage94
6.4.1 One Second Stress Pulse Tests 946.4.2 One Tenth Second Stress Pulse Tests 1006.4.3 Total Stress Path Tests 1036.4.4 Comparison with Previous Research 107
at Nottingham6.4.5 Discussion of Resilient Model 110
6.5 Sample Behaviour under Repeated Loading 1126.5.1 Introduction 1126.5.2 Development of Permanent Strain 1136.5.3 Development of Pore Pressure 1196.5.4 Variation in Resilient Modulus with 121
Number of Cycles6.6 Undrained Strength6.7 Equipment Performance 124
1276.7.1 The Consolidation Control System 1276.7.2 Data Monitoring System 1296.7.3 Instrumentation 1296.7.4 The Control System 130
CHAPTER 7 BEHAVIOUR OF COMPACTED TRIAXIAL SAMPLES7.1 Introduction7.2 Suction Measurements7.3 Compacted Triaxial Samples
7.3.1 Introduction7.3.2 General Behaviour7.3.3 Resilient Behaviour
7.3.3.1 Axial Strain7.3.3.2 Poisson's Ratio7.3.3.3 Permanent Strain
1321321361361361411411501591597.4 Undrained Shear Strength
CHAPTER 8 CBR TEST RESULTS. 8.1
8.28.3
IntroductionConsolidated CBR SamplesCompacted CBR Samples
166166172
CHAPTER 9 UNDISTURBED SAMPLES9.1 The Samples 1909.2 Determination of National Overconsolidation 190Ratio9.3 Determination of Resilient Modulus 194
.' (vii)
PageCHAPTER 10 PAVEMENT ANALYSIS
10.1 Introduction 20010.2 Pavements Analysed 20010.3 The Computer Program 20010.4 The Soil Models 202
10.4.1 The Granular Model 20210.4.2 The Subgrade Model 202
10.5 Use of the Models in the Program 20310.6 The Finite Element Mesh 20410.7 Results and Discussion 206
CHAPTER 11 COMPARISON BETWEEN THE BEHAVIOUR OF CONSOLIDATEDAND COMPACTED SAMPLES
11.1 Introduction 21311.2 Comparison between the behaviour of Compacted 213
and Consolidated Triaxial Samples of Keuper Marl11.2.1 Resilient Modulus11.2.2 Thixotropy11.2.3 Development of Permanent Strain
213216216
11.3 Comparison between the behaviour of Compacted 217and Consolidated CBR Samples of Keuper Marl
11.4 Consolidated Samples of Keuper Marl 219- Resilient Modulus and CBR
11.5 Compacted Samples - Resilient Modulus and CBR 220CHAPTER 12 CONCLUSIONS
12.1 Consolidated Triaxial samples12.1.1 Consolidation12.1.2 Resilient Response12.1.3 Permanent Response12.1.4 Undrained Strength
12.2 Suction12.3 Compacted Triaxial samples
12.3.1 Resilient Response12.3.2 Permanent Response
12.4 CBR Tests12.5 Resilient Modulus and CBR
12.5.1 Consolidated Samples12.5.2 Compacted Samples
CHAPTER 13 RECOMMENDATIONS FOR FURTHER WORK
228228228229230230230230231231232232232
13.1 Consolidated Triaxial Samples13.2 Consolidated CBR Samples
233236
.'
(viii)
13.5 Equipment Development13.5.1 Consolidation Control System13.5.2 Servo Control System13.5.3 Data Acquisition
Page234234234
13.3 Compacted Triaxial Samples13.4 Compacted CBR Samples
234235235
REFERENCESAPPENDIX AAPPENDIX BAPPENDIX CAPPENDIX D
MATERIAL PROPERTIESPERMEABILITY OF FILTER PAPER DRAINSSUPPLEMENTARY TESTSCALIBRATIONS
237244247249254
(ix)
LI ST OF SYMBOLS
LV DrMrN
OCRV
VAVkcsCve
.Soil Suctionpore pressure coefficientspecific gravityHertz (!/seconds)bulk modulusstress ratio during consolidationstress ratio during consolidation giving no radialstrainlinear variable differential transformerresilient modulus of elasticitynumber of load cyclesoverconsolidation ratiospecific volume (1 + e)spec ific volume of A 1ine at p'=!specific volume of k 1ine at p'=!centistokescoefficient of consolidationvoids ratiocoefficient of permeabilitycoefficient of volume compressibilitypore pressurespecific volume of critical state line at p'=!small increment10-6
ratio of q/p' at failureslope of overconsolidated line in V - ln p' spaceslope of normally consolidated line in V - ln p'spacePoisson's ratioshear stressangle of internal frictionaxisymrnmetric axial stress - triaxial axial stressaxisymmetric radial stress - triaxial radial stressaxisymmetric axial strain - triaxial axial strainaXisymmetric radial strain - triaxial radial strain
k
rt::,
~MkA
.'(x)
triaxial stressesp normal stress 0 +20a r
3
q deviator stress °a -0 r
f.v volumetric strain c a + 2 c r
~s shear strain ~ (€ a - c r )3
Superscripts
p permanent
r res il ient
Stresses listed are total stresses.denoted by the use of primes.
Effective stresses are
Compressive stresses and strains are taken as positive. Othersymbols are defined and used in restricted conditions as the needarises.
.' 1
CHAPTER ONE
INTRODUCTION
The rapid growth in road traffic over the last few years,especially in the numbers and gross weights of commercialvehicles has led to premature failures of trunk roads andmotorways. This is expensive not only because of the cost ofrepair but also because of the extensive delays to private andcommercial road users. These failures and the need to designroads for greater and greater volumes of traffic has highlightedthe need for a fuller understanding of the behaviour of thedifferent materials in a road, both individually and collectivelyas a road structure. Once this has been achieved pavements canbe designed analytically, rather than empirically as is the caseat present where the results from existing pavements areextrapolated to design roads to carry ten or more times thetraffic flow of the original pavements.
Much research has been done on the properties of bituminousmaterials, unbound aggregates and concrete, but less relevantdata is available on cohesive soils under the stress conditionsoccurring in road subgrades. Hight and Stevens (1982) using theNottingham design charts, Brown (1980), have demonstrated thesensitivity of the life of a pavement to changes in the value ofthe modulus of the subgrade, as shown in Figure 1.1, and hencethe necessity of accurately estimating the value of the modulusof the subgrade at the design stage. The elastic properties ofsoil are needed for computing the response of a pavement to asingle wheel load as this controls the magnitude of the tensilestresses in the upper layers and hence the ability of a pavementto resist fatigue failure. However the most common form offailure of a road pavement in this country is by rutting, which isdue to the excessive build up of irrecoverable deformation causedby large numbers of load repetitions. Therefore an understandingof how plastic strains accumulate under repeated loading is alsorequired.
2
ASSUMED PAVEMENTDESIGN
SUBGRADEVERTCALSTRAIN
ASPHALT THICKNESS 220mmSTIFFNESS 7 GPo
SUBGRADE STIFRtSS 50 MPa237 J.1E
Z 180(!)v;w0 160u,0
~ 11.00
<~ 120OwWlL.Vl-Vl_J
~ .....100a..zXW
~W w 80w~lL.Q.:_J
~60
w~w 1.0~a.._J 20<:::>b<
o 20 1.0 60 80 100 120 11.0ACTUAL SUBGRAOESTIFFNESS EXPRESSED AS A %OF DESIGN SUBGRADE STIFFNESS
FIGURE 1·1 THE INFLUENCE OF SUBGRADESTIFFNESS ON PAVEMENT LIFE
PAVEMENTLIFE
107STANDARDAXLES
3
Conventionally, subgrades are characterised by a CaliforniaBearing Ratio (CBR), either determined insitu or in thelaboratory. This is an empirical penetration test which has beenrelated to stiffness but it is not of direct use in an analyticaldesign method which requires a stress strain model. However, asthe CBR test is so widely used, correlations with elasticstiffness have been proposed and are used at present in designcomputations in the absence of more accurate data. Thecorrelations with CBR include:
E = 10 x CBR MPa 1.1
proposed by Heukelomand Klomp (1962) and
E = 17.6 CBRO.64 MPa 1.2
developed by Powell et al (1984) at TRRL. The permanentdeformation of the subgrade is limited by a resilient straincriterion without any consideration of the type of soil or theinsitu conditions.
Previous work on soils, although not under the conditionsapplicable to road pavements, have indicated that the actualstiffness of soil is influenced by a number of variables. Theseinclude not only the physical conditions such as the stressstate, moisture content, and preconsolidation pressure but alsosuch variables as the rate of loading. The stress strain(response of soils is known to be non linear.
Soil, whether it is undisturbed or compacted as fill is veryvariable and a very extensive and expensive site investigationwould be required to provide sufficient undisturbed samples forlaboratory testing to determine the variations of the stiffnessof the subgrade along the length of the road. What is requiredthen is a simple and inexpensive test, which ideally can beperformed insitu, which allows the subgrade to be characterisedsuch that its stress strain response can be predictedsufficiently accurately to allow a proper analytical design of a
4
road pavement. To be able to design such a test requires adetailed' knowledge of the response of soils to traffic loadingand how this is influenced by the various insitu parameters.
The aim of the research described in this thesis was toreconstitute samples of a clay under laboratory conditions to astress state similar to that found at the top of road subgradesand to subject them to a comprehensive test programme. Thesamples were fully instrumented and subjected to stress pulses ofa similar magnitude and frequency to those acting on road.subgrades and the resilient response was recorded per load pulseand the permanent deformation recorded per number of load pulses.The test programme was extended to cover the application of avariety of lower stress level, lower frequency total stress pathsto investigate the effective stress response of the soil. Aservo hydraulic loading rig was used for the test programme andthe samples were overconso1idated anisotropica11y from a slurryin the laboratory. The results from this series of tests arereported in Chapter 6. The equipment had been developed over anumber of years at Nottingham during earlier research projectsand is described more fully in Chapter 4. In order to verifythat the samples were consolidated to representative stressessome undisturbed samples were obtained from road subgrades andthe preconso1idation pressures of these samples were determined.These results are described in Chapter 9.
In view of the importance of the CBR test in current pavementdesign practice some CBR samples of the same clay wereanisotropica11y consolidated to similar stress states and the CBRvalues determined.
A smaller less time consuming test series was performed on handcompacted triaxial samples in a simple repeated load pneumatictriaxial rig to investigate the response of samples which werenot necessarily saturated. The same clay was used in this testseries and the samples were manufactured to moisture contentssimilar to those achieved by the consolidated samples. The testseries was repeated on two other clays compacted to similar
" 5
liquidity indices as the first clay. These results are presentedin Chapter 7.
A parallel series of CBR test on compacted samples of the samethree clays was also carried out together with suctionmeasurements. This enabled the responses of the compactedsamples to be analysed in effective stress terms. These testsare detailed in Chapter 8.
The objective of this work therefore, was to determine the stressstrain response of soil samples with a known and controlledstress history and to determine which parameters had the mostsignificant effect on the resilient and permanent response. Thisresearch therefore provides a source of experimental data on thebehaviour of a particular soil under this type of loading whichis the first stage in identifying a much simpler and cheaper testthan the repeated load triaxial test which would give a betterestimation of soil stiffness than the standard CBR test ~ andwhich can be used in analytical design methods.
6
CHAPTER TWO
lITERATURE REVIEW
2.1 EQUILIBRIUM STATES UNDER CYCLIC lOADING
Most researchers reporting on the behaviour of clay under cyclicloading have found a threshold value of repeated deviator stressmagnitude, above which the sample eventually fails and belowwhich an equilibrium state is reached regardless of the number offurther cycles. Sangrey et al (1969) reported that if the peakdeviator stresses at equilibrium for all their tests were plottedin stress space then they fallon straight lines, the lines fordifferent overconsolidation ratios being at different angles asshown in Figure 2.1a. This was found to be valid for bothisotropiC and anisotropic normally consolidatedisotropically overconsolidated specimens.
specimens, andAnisotropically
overconsolidated specimens were not tested. Further work reportedby France and Sangrey (1977) on a different clay and at higheroverconsolidation ratios confirmed their earlier findings, thoughthe work was not extended to cover anisotropic overconsolidation.
Sangrey et al (1977) proposed that all cyclically loaded samplesat equilibrium would show this relationship if the pore pressurechange due to creep was eliminated. They proposed that the totalchange in pore pressure (6uT) was the sum of 6up' the creepcomponent, (depe~dent on time and pi), 6 uq, which is dependentsolely on q, and \6~y which is the pore pressure change generatedby cyclic loading. Therefore, if6 uT was corrected for the creepdependent part, 6 up' in each case, then the equilibrium pointswould move in effective stress space, as shown in Figure 2.1b, toform straight lines.
2.2 PORE PRESSURE-STRAIN RELATIONSHIP
la (1969) showed, mathematically, that the excess pore pressureinduced by shear may be expressed as a sole function of the majorprincipal strain. He carried out slow strain controlled cyclic
7
FAILURE
U')U')lJJa::.....U') ,
or/0"'=0.56
oa::o~>lJJCl
MEAN NORMAL EFFECTIVE STRESS p'
FIGURE 2·10 EQUILIBRIUM STATES DURING CYCLICLOADING (AFTER SANGREYet 01 1969)
FAILURE
er(/)(/)lJJa::.....U')
a::o~>lJJCl EQUILIBRIUM
STATE
MEAN NORMAL STRESS p
FIGURE 2·1b EFFECT OF CREEP CORRECTION ONEaUI LI BRIUM STATES (AFTER SANGREYet al 1977)
8
triaxial tests on isotropic and anisotropic normally consolidatedsamples, the results of which confirmed his theory. He showedthat the pore pressure-strain relationship varied for differentmodes of consolidation, but was independent of time of sustainedstress, magnitude of consolidation pressure or time ofconsolidation. Lo used a parameter called the pore pressure
•ratio defined as 1 - k + 6.u.s/ a vc where k is the principalstress ratio, 6.Us the pore pressure increment due to shearstress andavcl the vertical effective stress at preconsolidation.
Yasuhara et al (1982) working with isotropic and anisotropicnorma 11y conso1idated clay found that 6.u/ a c' where 6.u is theexcess pore pressure and a c the cell pressure, was related tothe axial strain by a unique hyperbolic function independent ofthe loading method, i.e. either static or repeated. However,Mitchell and King (1977) working with isotropically normallyconsolidated clay found no apparent relationship betweendistortional or axial strain and the development of porepressures.
Wilson and Greenwood (1974), testing normally consolidated clay,found that the permanent pore pressure developed was linearlyrelated to the plastic axial strain, for small strains, and thatthe cyclic pore pressure was linearly related to the elastic orresilient axial strains. The frequency they used was 0.017 Hz,which should have allowed the pore water transducer to follow thepore pressure response, though no positive evidence was presented.
Matsui et al (1980), testing normally consolidated andIover-consolidated clay, found that plots of ~r;ac ,where ~ is
the residual excess pore pressure and acl the radial or minorprincipal effective stress, against lntc' where ¥c is the maximumsingle amplitude cyclic shear strain, produced straight parallellines for each value of over-consolidation ratio tested (seeFigure 2.2). Koutsoftas (1978) produced results similar toMatsui et al but these were presented, as 6.u/oc• against ¥r(the peak to peak shear strain), so the relationship may not bequite the same if ln~r is plotted.
9
l1Ja::::::>(/')(/')l1Ja::a..l1Ja::~~ ~OCR=4
~ /~ /~ /~ ~a::~ ~ ~ _
1er3
OCR=1
OCR=2
FIGURE 2·2 RESIDUAL EXCESS PORE PRESSUREAGAINST MAXIMUM SINGLE AMPLITUDECYCLIC SHEAR STRAIN (AFTER MATSUIet 01 1980)
10
2.3 EFFECT OF FREQUENCY OF LOADING
Brown et al (1975) found that there was no apparent effect onresilient strains for a loading frequency range between 0.01 Hzand 10 Hz. Yasuhara et al (1982) reported little effect onundrained strength for loading frequencies between 0.1 Hz and 1Hz, and Sherif et al (1977), working with a hollow cylinderapparatus, reported little change in cyclic strength over thefrequency range 0.025 Hz to 10 Hz.
Matsui et al (1980) found that as the frequency increased overthe range 0.02 Hz to 0.5 Hz, there was less excess pore pressuregenerated for the same number of cycles and that the axial strainamplitude decreased. Wood (1980) states that it seems morelikely that the effect of frequency is greater for more plasticclays. The plasticity index of the silty clay tested by Brown etal was 18%, which supports Wood's assertion, but the remainingsoils all had plasticity indexes in the range 55 to 60.
The only way to subject a specimen to a large number of cycles ina feasible time, is to use frequencies of the order of 10 Hz,which, in the case of road subgrades, models the loading ratescorrectly (Croney, 1977). The majority of pore pressuretransducers will then only record the mean or permanent porepressure reliably. Therefore to monitor the pore pressureresponse to an individual cycle would mean reducing thefrequency. Matsui and Abe (1981) report a fully saturatedresponse from a miniature pore pressure transducer in the sampleup to frequencies of 3 Hz. Hight (1983) reported a response timeof 0.5 seconds for a small transducer placed against the side ofthe specimen, while Overy (1982) reported a loss in response atabout 0.1 Hz for a miniature pore pressure transducer in thecentre of the sample and 0.005 Hz for the base pore pressuretransducer.
2.4 EFFECT OF OVERCONSOLIDATION RATIO
The overconsolidation ratio reflects the type of behaviour
11
exhibited by a soil when it is sheared. Those samples on the wetside of the critical state have low overconsolidation ratios andtend to behave uniformly, while those on the dry side with higheroverconsolidation ratios tend to develop discrete failure planes.Under monotonic loading samples on the wet side of critical statedevelop positive pore pressures, while those on the dry sidedevelop negative pore pressures.
Brown et al (1975), testing a silty clay under cyclic loading,report similar results to the monotonic tests foroverconsolidation ratios of 2 and 10, namely positive porepressure development for OCR 2 and negative for OCR 10. Hyde(1974), testing the same clay, reported that initially positivepore pressures were developed up to 104 cycles for samples athigh overconsolidation ratios. Anderson et al (1980), working onDrammen clay, reported that all cyclic loading compression testsshowed initial positive pore pressure development except at anoverconsolidation ratio of 10. Togrol et al (1979) also reportedpositive pore pressure development when testing a clay at an OCRof 4.
France ,and Sangrey (1977) reported negative pore pressuredevelopment when testing a clay at an OCR of 8 and Koutsoftas(1978) and Knight and Blight (1965) also reported negative porepressure development for samples on the dry side of critical i.e.at OCR's of about 4 or more. Matsui et al (1980) and Taylor andBacchus (1969) reported their findings more fully and obtainednegative pore pressure development initially, which increased inmagnitude as the overconsolidation ratio increased. They bothalso report that, as the number of cycles increases, the negativepore pressure peaks and then starts going positive, even for anoverconsolidation ratio of 16.
Ladd et al (1977) presents results from monotonic loading testswhich indicate that the stiffness of a soil under a set value ofdeviator stress increases as the overconsolidation ratioincreases from 1 to approximately 3 and then decreases quite
12
rapidly as the overconsolidation ratio increases further.However, the results of Brown et al (1975) show that theresilient modulus of samples of silty clay after reachingequilibrium under 105 cycles seems to be independent ofoverconsolidation ratio. The results of Ogawa et al (1977)support this and they conclude, generally, that the effects ofoverconsolidation ratio decrease with increasing numbers ofcycles.
2.5 EFFECT OF DRAINAGE PERIOD
Knight and Blight (1965) point out that, even on heavilytrafficked roads, the total time actually under load is quitesmall and it is therefore reasonable to assume that there is somedrainage although individual load pulses themselves areundrained. France and Sangrey (1977) allowed some continuousdrainage during their cyclic tests, but not sufficient to inhibitthe build up of pore pressures. Brown et al (1977), Anderson etal (1976), and Overy (1982) have all considered this problem andthe general conclusion is that samples on the wet side ofcritical with positive excess pore pressures expel water andbecome stiffer while those on the dry side, with negativeresidual pore pressures will tend to imbibe water and soften. Theamount of water imbibed will depend on the slope of the swellback line, and if this is shallow, the effects of allowingdrainage will only have a small effect as reported by Brown et al(1977). There seems to be a tendency for the failure plane ofoverconsolidated samples to be at a water content some 3% higherthan the main bulk of the sample as reported by Togrol et al(1979) and France and Sangrey (1977).
2.6 STRESS-STRAIN RESPONSE
Ogawa et al (1977) presented some results of cyclic loading testsin the form of stress-strain hysteresis loops and they show thatfor small resilient axial strains the loops are symmetrical andelliptical in shape. For larger strains the loops degenerateinto S shapes. They also state that there is a much more marked
13
change of shape for the samples with higher overconsolidationratios. Sherif et al (1977) also reported symmetrical ellipticalhysteresis loops from their low strain cyclic tests using ahollow cylinder apparatus. Anderson et al (1980) obtained similarfindings from their work on Dramman clay. They also show that,for two way cyclic loading, the loops are not symmetrical betweenthe extension and compression zones. Taylor and Bacchus (1969),using strain controlled cyclic loading tests, also report adegeneration of elliptical hysteresis loops into S shapes withincreasing numbers of cycles. Macky and Saada (1984) also reporta similar finding from their tests on isotropically consolidatedand overconsolidated clay samples in a hollow cylinder apparatus.Overy (1982) did not report any change in the stress strainhysteresis loops with numbers of cycles from his work on repeatedtriaxial loading of Keuper Marl.
Sanchez and Sagaseta (1981) reported discontinuities in theeffective stress paths of triaxial specimens subjected to slowundrained strength tests. They occurred at strains of less than0.5% for normally consolidated samples, up to strains of about 9%for samples with an overconsolidation ratio of 3. Thisdiscontinuity was exhibited byanisotropically consolidated samples.
both isotropically andBjerrum (1973) reported
that when a cohesive soil is loaded the cohesive component of thestrength is mobilised and this reaches a peak at a low strain andthen tails off to a residual value, meanwhile as the strainincreases, more friction is mobilised but, by then, the particlestructure has started to change. This effect will lead to changein stress-strain behaviour under monotonic strength tests andsome similar mechanism may well be acting in samples under lowstrain amplitude cyclic loading.
2.7 ANISOTROPY
Most work on anisotropically consolidated samples has been onnormally consolidated material in the repeated load triaxialtest. The results which have been reported indicate somedifferences in behaviour caused by anisotropy. For example,
14
Yasuhara et al (1982) found that smaller excess pore pressureswere developed with increasing axial strain for anisotropicsamples. France and Sangrey (1977) reported that strength testson samples which had reached equilibrium under repeated loadingshowed a gain in strength of 30% for isotropic and only 15% foranisotropically consolidated samples.
Saada and Bianchini (1975) investigated the effects of differentstress paths to failure for anisotropically consolidated samplesusing a hollow cylinder apparatus. Their results show that thereis no unique ~01, relation for an anisotropic sample. Saada etal (1978), again using a hollow cylinder apparatus, report thatunder low frequency, low strain cyclic loading the behaviour ofanisotropic samples was totally different from isotropic ones,even after severe cyclic straining.
Graham and Houlsby (1983) reported a theoretical analysis and aseries of experimental results which demonstrate that some of theeffects of anisotropy can be observed in the triaxial cell,though the work was limited to a vertical axis of symmetry of thesample coinciding with that of the triaxial cell. The work wasconcerned with small amplitude stress paths within the yieldsurface and demonstrates that the elastic response of anisotropic soil to various total stress paths was unique, as forisotropic material, but that pi was not constant during the loadcycles. This is because, unlike isotropic material, there 1s across dependence of shear strain on the mean pressure pi, and ofvolumetric strain on the deviator stress q.
2.8 LOADING REGIMES
A large number of the cyclic load triaxial tests reported in theliterature are isotropically normally consolidated oroverconsolidated, and the pulsed loading starts from zerodeviator stress. Most tests involve one way loading(compression) while some are two way loading with equal positiveand negative peaks of deviator stress (compression - extension).
15
Anderson et al (1980) reported that much larger shear strainsdeveloped under two way loading than one way loading. Theyreporte~ ~hat, in triaxial samples, the behaviour is different incompression and extension and that a mean plastic strain willdevelop even if the loading is symmetrical. At large cyclicstrains, some dilatancy occurs and affects the cyclic porepressure changes. One way cyclic loading causes a change in bothpore pressure and strain. In general, though, cyclic porepressures in triaxial tests are caused by mean normal stresschanges, a fact which has been recognised by Matsui et al (1980)who report a series of tests where the cell pressure is pulsed1800 out of phase to the deviator stress at such an amplitude asto keep the total mean normal stress constant. They do not,however, report tests results where the mean normal stress changed
_for comparison purposes.
Most researchers have reported threshold values separatingfailure and equilibrium conditions for cyclic load triaxialtests. Most of these are pulses from isotropic conditions.However, Mitchell and King (1977) report that, for their normallyconsolidated clay, a repeated deviator stress pulse of 0-50% ofundrained strength was sufficient to cause failure, but if thesoil was allowed to reach equilibrium under a static deviatorstress of 20% of the undrained strength, then a cyclic deviatorstress of 50% was still required to fail the sample. Houston andHerman (1980), working mainly with normally anisotropicallyconsolidated soil demonstrated that soils under a static deviatorstress of about 30% of undrained failure require a larger cyclicdeviator stress to cause failure than when starting underisotropic conditions.
2.9 YIELD SURFACES
The yield surface for normally isotropically consolidated samplesunder slow monotonic loading has generally been found to beequivalent to the Roscoe surface, as predicted by the criticalstate theory, Schofield & Wroth (1968). Mitchell (1969)demonstrated that the yield surfaces were different for normally
16
consolidated isotropic and anisotropic samples. Crooks and Graham(1976) testing reconsolidated undisturbed site samplesdemonstrated that the yield envelope for anisotropic samples wascentred approximately on the Ko consolidation line in q-p' space.Tavernas et al (1979), testing three different naturallyoccurring clays, reported similar findings and concluded thattheir yield surface approximated to a surface of constant strainenergy. Later work by Folkes and Crooks (1985) working with twodifferent clays confirmed the above findings. The work coveredlow overconsolidation ratios, though still remaining on the wetside of the critical state.
Graham et al (1983) tested a natural clay, reconsolidated to theinsitu stresses, in a triaxial rig under a variety of effectivestress paths. They reported a nest of yield surfaces, dependingon the overconsolidation ratio and mean normal effective stress,which were not symmetrical about the Ko line, the critical stateline or the isotropic consolidation line. Normalising the q-p'axes by dividing by the preconsolidation pressure resulted in asingle yield locus. Graham et al pointed out that the positionof the yield locus would depend on the loading rates.
Banerjee and Stipho (1979) proposed a yield surface for heavilyoverconsolidated clays which involved adapting the Hvorslevsurface and using it as a yield surface. The model theyproposed was for isotropically overconsolidated samples wherethe original yield surface, consisting of the Roscoe and Hvorslevsurfaces, was set by the preconsolidation pressure. Subsequentloading beyond this surface caused plastic straining and resetthe yield surface to pass through the peak of the effectivestress path. The intercept of the surface on the deviator stressaxis also changed. The model was developed for monotonicallyloaded samples and test data was presented which showed that themodel gave reasonable results.
Hyde and Ward (1985) reported the results of a series of repeatedload triaxial tests on Keuper Marl and concluded that theHvorslev surface forms a yield and consequent failure surface for
17
cyclic effective stress paths of samples of all stresshistories. They found that normally consolidated sampleseventually failed by stress path migration caused by a build upof pore pressure in the same way as heavily overconsolidatedsamples.
2.10 THIXOTROPY
It has been found by most researchers that the response of clayssubject to repeated loading differs with different rates ofloading. Indeed one of the original assumptions in the criticalstate theory, Schofield and Wroth (1968), was that the rates ofloading were sufficiently slow for viscous effects to be ignored.The continual deformation of clays under a constant load, i.e.creep, is also well documented and some researchers have had somesuccess relating the behaviour of clays under repeated loading tothat under creep loading, Hyde (1974), Hyde and Brown (1976)..
Thixotropy is another aspect of the time dependent behaviour ofclays but reference to it in the literature is less common.Thixotropy is the stiffening of the soil sample with time underconstant stress conditions with no change in state. Seed et a1(1962) carried out a large number of repeated load triaxialtests on subgrade soils compacted by different methods to various,dry densities and moisture contents. They found that the lengthof time after compacting a sample before testing it causedsignificant changes in behaviour. Samples left for 50 daysbefore testing showed a much reduced permanent strain after50,000 stress pulses, and the samples were found to be stifferfor up to about 4000 stress applications. In fact the initialstiffness could be as much as 4 times that of a sample testedalmost immediately after compaction. The repeated loading wasfound to have effectively destroyed the thixotropic strength gainafter approximately 4,000 applications, and this was consideredlikely to have been caused by the deformations induced by thecyclic loading.
Pusch (1982) working with natural undisturbed clays and saturated
18
laboratory prepared clay samples also demonstrated that there wasa stiffening effect with time, present in clay samples storedundrained under constant stress conditions after disturbance ofthe clay structure by loading. Kim and Novak (1981) measured thestiffness of normally and overconsolidated, naturally occurringclay samples in a resonant column apparatus and demonstrated thatthe low strain amplitude shear modulus measured immediately afterhigher amplitude shear straining was lower than if it wasmeasured prior to the high amplitude shear straining. They notedthat the reduced stiffness was only temporary and not due tochanges in stress or state brought about by excess pore pressureor drainage.
Mitchell (1960) postu1 ated that the interparti culate forcebalance between soil particles is such that soil will tend toflocculate and become stiffer. He stated that the energy ofinteraction between soil particles is at a level commensuratewith externally applied forces, and that therefore shearing thesoil is capable of disrupting the thixotropic bonds andreorganising the platey clay minerals into a uniform parallelstructure. When shearing stops the interparticulate forces becomedominant again and the structure readjusts itself. Since thesestructural changes are dependent on actual physical movement ofparticles and ions they are time dependent.
Osipov et a1 (1984) examined the structural arrangements ofvarious clays under different loading regimes using an electronmicroscope and confirmed Mitchell's hypothesis about thestructural changes caused by externally applied forces. Theymeasured the torque of a knurled cylinder continuously rotatingin wet clays with and without an externally applied vibration tothe clay. The torque required was considerably less when thesoil was vibrated, but soon regained the original level when theexternal vibration was removed. The clay structure was examinedby freezing the samples under various loading conditions usingliquid nitrogen, and was found to become much more uniform underthe vibrating load, but to revert to its more disorganised statewhen the vibration was removed.
19
2.11 END RESTRAINT
Perloff and Pambo (1969) used a finite element method to analysethe triaxial test to determine the effect of end restraint. Theyconsidered the case with total fixity at the ends and comparedit with the results obtained using smooth end platens. Theyfound considerable non-uniformities in the stresses and hence thepore pressures and strains at the ends of the samples. However,the middle section was relatively uniform even with total fixityat the ends. Carter (1982) used a finite element programme toexamine constant strain rate triaxial strength tests, whichshowed non uniformity in stress at the edges of the sample aswell as at the ends. Kimura and Saitoh (1983) carried out aseries of undrained shear tests on saturated clay samples withpore pressure transducers in the centre, at the end and on theperiphery of the samples and demonstrated that each section ofthe sample followed a different effective stress path. The pathsbecame more similar at low strain rates of approximately 4%/houror less. Ba1asubramanian (1976) consolidated lead shot intriaxial samples of kaolin. X ray photographs were then taken ofthe samples during shearing. The axial strain was found to bereasonably uniform up to 2% for fixed ends, and up to 8% for freeends. The material tended to form rigid cones on the platens atapprOXimately 01 to the platen
Lee (1976) reported on the various methods to reduce end frictionand concluded that the most satisfactory way was to use siliconegrease on polished steel end platens with a rubber membraneseparating the soil and grease. Brown (1974) reported theresults obtained from an investigation into end restraint duringthe research programme on Drammen clay, and reached the sameconclusion as Lee provided high vacuum grease was used. Leereported that if the greased end platens were left under load foras little as 10 hours then some effectiveness was lost and a peakfriction angle developed which reduced once radial movementcommenced and resmeared the grease. Overy (1982) reported asimilar effect and stated that a modified system with a greasereservoir consisting of two membranes stuck together at the
20
edges, with bleed holes to allow the grease onto the platen,proved to be effective for at least seven days. The technique wasoriginally developed by Austin (1975). However, Overy reportedpractical difficulties in assembling and using the reservoir, andfound that the extra grease could clog the side filter paperdrains.
Lee raised another point concerned with the effect of theviscosity of the grease. He demonstrated that, assuming freeends, the shear stress developed at the circumference of thesample under a 1 Hz loading was quite significant. He statedthat this effect could be reduced by using more than one greasedmembrane, but this would cause more uncertainty in axialmeasurement for those researchers still measuring axial strainsbetween the end platens.
Most triaxial testing equipment measures pore pressures at thebase of the sample, and because some end restraint will always bepresent, the measured pore pressure will not be representative.Some researchers, such as Sangrey et al (1969), used very lowfrequencies of loading to allow time for pore pressureequalisation, while others, Koutsoftas (1978), allowed time afterfaster cyclic loading for the pore pressure to equalise. Ineither case the recorded pore pressure will still be distorted bythe end effects but the second method has the advantage oftesting at representative frequencies or rates of loading. Itseems likely that the most accurate pore pressure measurementswill be from a centre probe in the relatively uniform centralsection of the sample before pore pressure equalisation has takenplace (Hight 1983), assuming that the transducer itself does notcause any significant effects.
2.12 SUCTION
Water is held in the soil matrix by absorption and surfacetension forces at a pressure less than atmospheric (Croney 1977).Soil mOisture suction, or suction, is a measure of therequired to abstract water from a sample of soil which
pressureis free
21
from external loading. Suction is expressed on the pF scale,which is the common logarithm of the height in centimetres of anequivalent column of water.
Krahn and Fredlund (1972) define the total suction as thenegative gauge pressure relative to the external gas pressure onthe soil water to which a pool of pore water must be subjected inorder to be in equilibrium through a semi permeable membrane withthe soil water. The total suction was proposed as the sum of theosmotic suction and the matrix suction and the relationship wasproved by measurements of all three parameters on a variety ofsoils. The osmotic suction is due to the different salts whichmay be dissolved in the pore water and which affect the surfacetension. The matrix suction is due to the size and shape of poresin the soil skeleton and the clay particle surface absorption.
The total suction of a soil sample is measured using apsychrometer which measures the humidity of the air surrounding,and in equilibrium with, the soil sample by using thermocouples.The osmotic suction can be computed from electrical conductivitymeasurements. All types of suction apparatus which involveallowing the sample to reach moisture equilibrium across a porousstone or membrane measure the matrix suction as the stones arepermeable to salts dissolved in the pore water.
Krahn and Fredlund found that the variation in total suction withmoisture content of the sample was almost entirely due tovariations in the matrix suction as the osmotic suction remainedvirtually constant. Krahn and Fredlund concluded that forremoulded compacted soils the matrix suction and hence the totalsuction, was dependent on the moisture content of the sample butessentially independent of the dry density. Shackel (1973)concluded that the suction was slightly influenced by the drydensity but primarily depended on the degree of saturation of thesample. Croney et al (1958) found that the dry density haslittle effect on the suction moisture content relationship forheavy clays, but that dry density affects the suction moisturecontent relationship for sandy or silty clays below a pF of
22
approximately 2.
Snethen and Johnson (1977) state that the matrix suction is bothwater content and surcharge dependent, and that the osmoticsuction is independent of water content and surcharge pressure.
Russell and Mickle (1970) demonstrated that the suction moisturecontent relationship for clays has three distinct regions withpoints of inflection at approximately the liquid limit and theplastic limit. The suction at the plastic limit was found to beapproximately pF 3.7 for heavier clays reducing to approximatelypF 3.2 for silty clays. A vacuum is equivalent to approximatelypF 3.0.
Croney et al (1958), using a pressure plate apparatus,demonstrated that the suction moisture content relationshipexhibits some hysteresis in that the measured suction at aparticular moisture content depends on whether the sample hasbeen wetter and allowed to dry or is being wetted up from a drierstate. This was put down to the fact that pores may empty atdifferent suctions from that at which they fill. Croney alsofound that there was a unique suction moisture content curvefor continuously disturbed soil.
However, Snethen and Johnson (1977) determined the suctionmOisture content relationship for four soils with plasticityindices varying between 13 and 68 using a psychrometer and foundthat there was no hysteresis between the wetting and dryingcurves. They were of the opinion that the pressure plate typeapparatus caused the hysteresis found by other researchers.
Shackel (1973) proposed that suction (in pF) was linearlyproportional to the logarithm of the moisture content, for therange pF 0.6 to pF 1.8, although the constants depended on thedegree of saturation and the dry density. Croney (1977)demonstrated that a plot of moisture content against plasticityindex gave approximate linear contours of suction for the rangepF2 to pF3. Snethen and Johnson (1977) found that there was a
23
linear relationship between suction (in pF) and the moisturecontent over the range pF2 to pF4.7.
Mou and Cha (1981) working with a clay with a plasticity index of19 demonstrated that static and kneading compaction resulted indifferent suction moisture content curves, with a greater suctionfor the statically compacted samples in each case.
2.13 RELATIONSHIP BETWEEN SUCTION AND RESILIENT MODULUS
Croney (1977) states that soil moisture suction is an overridingfactor in determining the elastic modulus for saturated clays.An analysis of the results presented by Brown et a1 (1975) interms of suction show a linear relationship between resilientmodulus and suction at constant deviator stress. A similarrelationship is reported by Finn et al (1972) determined withdata from a test road subgrade. Deh1en (1969), working withsilty clay, showed a measured suction decreasing with increasingdepth below a road surface and a corresponding linear decrease inresilient modulus with suction.
Fredlund et al (1975) measured the resilient modulus and suctionof a till and a clay and found that the modulus increased withincreasing suction, but at a decreasing rate. Most of their workconcentrated on samples much drier than the plastic limit withquite high suctions.
Richards and Gordon (1972), working with a clay subgrade,determined the resilient modulus in repeated load triaxial testsand subsequently measured the suction of each sample. Theyconcluded that the resilient modulus is very sensitive to changesin soil moisture suction and that there is a linear relationshipif the logarithms of each are plotted. However, Edris and Lytton(1977) found that resilient modulus increased with suction untila mOisture content about 2% less than the optimum, and that therewas little change in resilient modulus as the suction increasedfUrther. Sauer and Monismith (1968), working on a glacial till,determined resilient modulus from repeated load triaxial tests
24
and also measured the soil suction of the specimens. Theyreported that the resilient modulus increased rapidly withincreasing soil suction until a moisture content equivalent tooptimum, thereafter the curve reached a steady value of resilientmodulus as the suction increased. Edil et al (1980) found thattheir was little change in the resilient modulus with suction upto 100 kPa from their test results.
2.14 PREDICTION OF RESILIENT MODULUS FROM SOIL PROPERTIES
Kirwan and Snaith (1976) published a chart showing contours ofresilient modulus on a graph of relative compaction againstrelative moisture content. This chart was based on experimentalresults on Irish glacial tills (P.I. range 14-20%) and showedthat there was an optimum moisture level for maximum resilientmOdulus at any particular degree of compaction.
Later work by Kirwan et al (1982) showed that the resilientmodulus increased gradually with increasing load applications andreached a steady value after 10,000 cycles. At higher values ofdeviator stress the modulus decreases with increasing number ofcycles and did not reach a steady value after 80,000 cycles.
The variations in steady state resilient modulus and permanentdeformation (after 40,000 cycles) with dry density and moisturecontent showed these parameters to be more sensitive to changesin moisture content than dry density. Kirwan et a1 concludedthat for their soil a reasonable estimate of resilient moduluscould be obtained from the dry density and moisture content.
Edris and Lytton (1977) tested three soils in a repeated loadtriaxial rig and monitored the elastic and plastic deformation,the vertical load and the soil suction. The soils had a widerange of clay content and plasticity indices. Expressions weredeveloped for resilient modulus and plastic strain from theirresults showing that it is possible to predict these parametersfor any soil with a clay content between 20% and 70% using theirequation. They found that the resilient modulus was dependent on
25
temperature, increasing as the temperature decreased, and providedtemperature correction factors for their expressions.
26
CHAPTER THREE
THE TEST PROGRAMME
3.1 INTRODUCTION
The main body of the research concentrated on repeated load testson a reconstituted anisotropically consolidated clay (KeuperMarl) under triaxial conditions. The initial stress conditionsand the loading magnitude and frequencies were chosen to simulateactual road subgrade conditions as closely as possible. A servohydraulic testing rig which had been developed at Nottingham overa number of years was modified and used for the research.Chapter 4 gives details of the equipment and modifications. Inview of the importance attached to the California Bearing Ratio
Res. Lab (1970)(CBR) in the design of roads, Road· , and Powell et al (1984)a subsidiary test programme on consolidated CBR samples wascarried out.
A smaller parallel investigation into the behaviour of compacted(i.e. not necessarily saturated) clay samples was carried outusing a much simpler repeated loading triaxial rig. A parallelseries of tests on compacted CBR samples was also carried out.Three different clays were used in this part of the research andthe suction moisture content relationship of each clay wasdetermined over the range of moisture contents used in thetriaxial and CBR test programmes. The following sections detailthe range of stress conditions chosen for each of the testprogrammes.
3.2 CONSOLIDATED TRIAXIAL SAMPLES
3.2.1 Initial stress conditions
There is little data available on the insitu stress conditions inroad subgrades under roads in serviceable condition in thiscountry, as only roads that have failed are investigated indetail. The range of stress conditions used in the test
27
programme was based on a survey of the literature and someassumptions and calculations which are detailed below.
Croney (1977) states that it is reasonable to assume that themajority of road subgrades in this country are saturated ornearly saturated, and therefore the use of fully saturatedsamples, which allow analysis in effective stress terms, can bejustified.
Roads in Britain are generally provided with a drainage systemdesigned to maintain the water table a metre or so below theformation level. It is recognised though, Russam (1967), thatthis is not always feasible for plastic clays where thepermeability may be less than that of the surfacing material.However, in practice the water table can be found as high as thetop of the subbase. Early research at TRRL into the seasonalvariation of moisture content in the ground beneath an extensivepaved area showed that, provided there were no cracks, there waslittle variation in moisture content with change in season (Blacket al 1958). Therefore it can be assumed that there will belittle water flow in the subgrade under a road in good conditionand consequently there will be an approximately linear increasein pore pressure with depth. Water held above the water table bycapillary action will cause a linear increase in negative porepressure with height above the water table until desaturationOCcurs. Croney (1977) states that this is unlikely to occurunless the water table is greater than 5m deep. A range of porepressures at the top of the subgrade of between 0 and -70 kPa waschosen to give a reasonable range of effective stresses for thetest programme.
The total vertical stress on the subgrade can be calculated fromthe overburden pressure due to the road construction. Thehorizontal stress though cannot be calculated without assuming avalue for Ko, which in turn depends on the overconsolidationratio. It is reasonable to assume that most natural clays inthis country are overconsolidated to some degree, and there issome evidence to suggest that compacted clays also behave as if
28
they were overconsolidated (Croney, 1977). Naturallyoverconsolidated clays would be anisotropic as they wereconsolidated one dimensionally. As there were no values oftypical overconsolidation ratios available it was decided to usevalues of 6, 12 and 18 as defined by the mean normal effectivestress, as this covered a reasonable range. Some undisturbed U4sample tubes of road subgrades were obtained part way through theresearch project and the preconsolidation pressures determined.The results are presented in Chapter 9 and show that the range ofoverconsolidation ratios chosen was reasonable.
Assuming an overburden pressure of approximately 20 kPa fromabout 1 m thickness of road construction, and a value of Ko ofbetween 1.5 and 2.0 at an overconsolidation ratio of 12 for clayswith a Plasticity Index in the range 10 to 40 approximately,lambe and Whitman (1979), gives, for a water table at the top ofthe subgrade, a mean normal effective pressure of about 30 kPa.lowering the water table to give a negative pore pressure ofabout -70kPa increases the mean normal effective stress toapproximately 100 kPa.
It was decided therefore to test samples at mean normal effectivestresses of 33 kPa, 65 kPa and 100 kPa at overconsolidationratios of 6, 12 and 18. Figure 3.1 shows the sample numberingsystem.
3.2.2 loading conditions
The test programme was designed to investigate the elasticresponse of the samples to low level stress pulses, and theresilient and permanent responses of the samples to loadingsequences typically undergone by road subgrades.
The range of test frequencies and pulse amplitudes was limited toSome extent by what it was feasible to measure and control. Thiswas especially true for any analysis in terms of effective stresswhich required accurate pore pressure measurement. The mainlimitation in the equipment was the X-V plotter which was unable
29
AGURE 3·' NUMBERING SYSTEM FORCONSOLIDATED SAMPLES OFKEUPER MARL
30
to follow an input signal accurately above approximately 0.3Hz.The introduction of the computer data recording system part ofthe way through the test programme allowed the stress and strainparameters to be plotted against each other at higher frequencies.
In order to investigate the elastic response of the samples thestress paths had to be recorded to allow a full effective stressanalysis. Therefore the samples were tested at as low a frequencyas possible (approximately 0.05Hz) to allow accurate recording ofthe pore pressure and low stress levels generally in the range~2.5kPa to ±10 kPa were used to limit any permanent deformation.
To establish a realistic range of test frequencies and stresspulse magnitudes to simulate traffic loading two differentpavements were analysed using a finite element computer programcalled SENOL. Full details of the program and the results ofthe analyses are presented in Chapter 10. The results suggestedthat a deviator stress pulse magnitude of between 10 and 20 kPawith a period of 0.15 seconds would be suitable. This wasequivalent to a vehicle speed of 50 km/hr. Croney (1977) andBarksdale (1971) suggest that periods of about 0.1 second would beapplicable, and Croney (1977) suggests deviator stress pulses ofbetween 20 and 50 kPa would be typical.
To simulate traffic loading more accurately the load was appliedas a number of individual pulses of constant magnitude ratherthan continuously.
Hyde (1974) suggested that rest periods between pulses or groupsof pulses could be important as it allowed for what he termeddelayed elastic recovery. However too many long rest periodswould make it impractical to apply the large numbers of stresspulses required. It was decided that these tests would becarried out with a deviator stress pulse of 0.1 second durationfollowed by a rest period of approximately 0.25 seconds. Thestress pulses were derived from a single cycle of a sine wave asthis is the closest easily derivable waveform to the pulse shapeproduced by the computations using SENOl. The arrangement of the
31
triaxial cell does not allow the direction of the principal axesof stress to rotate, except by 90°, and consequently it isimpossible to simulate fully the stress conditions caused by awheel load rolling over the subgrade.
Single pulses of 0.1 second duration can be considered undrainedwhen applied to clays, but there will be some dissipation ofexcess pore pressure caused by repeated loading in roadsubgrades. It was decided to run the tests undrained and tomonitor the development of pore pressure, and if necessary toallow some drainage if required.
The initial deviator stress pulse magnitude for these tests waschosen as 20kPa and the test was continued until the sampleeither reached equilibrium or failed. If the sample reachedequilibrium the test was terminated and the sample left for atleast 24 hours to recover with the drainage tap open. The testwas then repeated with a deviator stress pulse magnitude ofbetween 15kPa to 20kPa greater than the previous repeated stresslevel. This process was continued generally until the repeateddeviator stress level was high enough to cause failure.
In order to determine the effect that this type of test had on theresilient properties of the sample a simple test consisting of afew cycles at each of a range of pulsed deviator stresses up toabout 75% of the test level was applied to the sample just priorto and immediately following the main test. These tests werealso undrained but used a pulse length of 1 second.
Figure 3.2 shows diagrammatically the type and sequence of thetests applied to each sample.
Cyclic load data is presented on nine anisotropicallyoverconsolidated samples. There was only a one in three successrate at consolidating the samples due to the length of timerequired (about 8 weeks on average) and the number of equipmentfailures which occurred and caused excessive sample disturbance or
32
~~ ~ 1·0s.pulses at various stress levels
..J D 0·1s.pulses at fixed stress levelUJ>~
~UJfl'::ti;fl':: Drained rest periods~
~ l ~ 6~<t(U
etc.0~
Time
FIGURE 3·2 SEQUENCE CF REPEATED LOAD TESTS ONSATURATED TRIAXIAL SAMPLES
33
failure. The results of all tests on the consolidated triaxialsamples are presented in Chapter 6.
3.3 COMPACTED TRIAXIAL SAMPLES
A number of much simpler repeated load triaxial tests wereperformed on compacted samples of three different clays. Theaim was solely to determine the variation of resilient moduluswith moisture content and no attempt was made to measure thepore pressure or the development of permanent strain with thenumber of cycles applied. The tests on the Keuper Marl sampleswere designed to provide a comparison between compacted andreconstituted samples, and those on the Gault clay and Londonclay to determine if different clays showed similar behaviourunder similar loading conditions. The test programme on eachclay included suction measurements which would give anindication of the effective stress state of the compactedsamples and enable the results to be compared more easily withthose from the consolidated samples.
The range of moisture contents for the compacted Keuper Marlsamples was chosen to cover that of the reconstituted samples atthe end of consolidation. The samples were compacted by hand toas dense a state as possible at each moisture content. It wasconsidered to be too difficult to produce uniform samples withlower dry densities using this method of compaction. The rangeof moisture contents of the test samples of the London and Gaultclays was chosen by ensuring that the same liquidity index wasachieved as that used for Keuper Marl tests. This method waschosen to try and take into account the different plasticityindices, and ensure that the samples were tested under similarconditions.
The frequency of loading could not be varied and was set at 1Hz,which was similar in frequency to some of the tests performed onthe reconstituted samples. The samples were tested for 100pulses at each of a series of increasing deviator stressmagnitudes. The axial and radial deformations were monitored
34
during the first and one hundredth pulse. The deviator stresswas not increased beyond that causing a resilient axial strain ofabout 3500 microstrain, and the resilient response was thenmonitored for decreasing deviator stress pulse magnitudes. Allsamples were pulsed with a positive deviator stress pulse fromthe initial isotropic state.
The Keuper Marl tests were divided into three sets at differentcell pressures, namely OkPa, 15kPa and 30kPa. The London clayand Gault clay samples were tested with a cell pressure of OkPaonly, as this allowed the results to be interpreted in terms ofthe measured suction, and because further development of theapparatus would be required to allow accurate control of thecell pressure.
The test details for the Keuper Marl, Gault clay and London claysand the test results are presented in Chapter 7. The equipment isdescribed in Chapter 4 and the sample preparation procedure inChapter 5.
3.4 CALIFORNIA BEARING RATIO TESTS
As the elastic stiffness of subgrades for use indesign is at present obtained from the CBR value of theeither by using the empirical expression E = 10 x CBR,and Klomp (1962), or a more complicated relationshipE = 17.6 x CBRO.64, Powell et al (1984), it was decided
analyticalsubgrade,Heukelom
~ tryand measure CBR values for the clays used in this research at themoisture and stress conditions used for triaxial samples.
A CBR mould was adapted to enabJe overconsolidated samples ofKeuper Marl to be prepared from a slurry. The same range ofoverconsolidation ratios as the triaxial samples was used.However, it was found to be too difficult to control the finalmean normal effective stress precisely enough so all sampleswere allowed to swell back to the same final vertical effectivestress.
35
The compacted CBR samples of London clay, Gault clay, and KeuperMarl were manufactured in the same manner as the triaxialsamples and covered approximately the same range of moisturecontents.
The tests were carried out at the standard rate of loading ofImm/minute, although some tests were unloaded and reloaded atthe same rate after Imm, 2.5mm and 5mm deflection. This was inan attempt to obtain more information from the standard CBR test.
The results and test details of all the CBR tests are presentedin Chapter 8.
36
CHAPTER FOUR
THE EQUIPMENT
4.1 THE SERVO HYDRAULIC TRIAXIAL TEST FACILITY
The servo controlled hydraulic triaxial testing facility atNottingham was first developed by Lashine (1971). The originaldevelopment is described in detail by Lashine (1971) andmodifications to the system are described by Cullingford et al(1972), Parr (1972), Hyde (1974), Austin (1979) and Overy (1982).The equipment as it existed at the start of this project isdescribed below, followed by a description of the changes andimprovements made by the author.
4.2 EXISTING TRIAXIAL EQUIPMENT
4.2.1 The test equipment
The equipment was contained in an air conditioned laboratory andconsisted of a loading frame with two hydraulic actuators, theelectronic control system and the data monitoring and recordingsystems. Hydraulic power was supplied by a pump at a normaloperating pressure of 14MPa. Figure 4.1 shows a schematic layoutof the load and control systems. The control systems comparedthe output of tbe relevant transducer with the input commandsignal, and the difference was amplified and fed to the servovalve which adjusted the oil flow to the actuator to reduce thedifference between the input and output signals. The summingjunction allowed a signal from a signal generator to besuperimposed on the command signa] which caused the load on thesample to cycle. A high frequency (400Hz) dither signal wasapplied to the servo valve to prevent the shuttle sticking.
Axial load was applied to the sample by connecting the load ramin the triaxial cell directly to the hydraulic actuator. Theload ram was designed by Austin (1979) and incorporated a straingauge bridge located within the triaxial cell to 'avoid errors
L..
~
Q.I
ts:-=0
~a.g.U)
._0
.~"50L.."0>..s:
37
Q.IU"'0oo
CQ.I
EuQ.Ia.Ul
-Uo
gz8u_.~a:::o>-::cI
~a:::~
-
38
caused by friction in the cell bearing. The load cell providedthe feedback signal required for the control system. There wasan option of position control instead of load control which useda large linear variable differential transformer (LVDT) connectedto the load ram to provide the feedback signal.
The confining fluid used in the triaxial cells was silicone oil(Dow Corning 200/20 cs) and pressure was applied by connectingthe hydraulic actuator to a piston which acted on the siliconeoil. The feedback transducer was a strain gauged diaphragmpressure transducer located in the cell top.
The triaxial cells were designed by Austin (1979) and consistedof an epoxy resin base, perspex side walls and an aluminium celltop. The specimen size was 1S0mm high by 78mm in diameter. Thesamples were encased in latex rubber membranes. The bases weremade of resin to eliminate any bi-meta11ic corrosion caused bydifferent metals on contact under water, which was the originalcell fluid. Pore pressure was measured at the base of the samplethrough a porous stone secured to the cell base, with a tappingconnecting to a pressure transducer which was screwed into thebase.
Overy (1982) modified the triaxial cells to allow additionalinstrumentation to be added to the sample. The top platen andcell base were modified to allow the pore pressure to be measuredin the centre of the sample using a miniature pore pressuretransducer which had been consolidated into position in theslurry moulds. Overy (1982) added a metal support system to thecell base which carried four linear variable differentialtransformers (LVDTs) and two proximity transducers (PTs). Theinstrument layout is shown on Figure 4.2 and provided a measureof the axial deformation over the centre third of the sample andthe radial deformation across a diameter.
Back pressure could be applied to the samples from an air pressureregulator via an air water interface. Volume change of thesample could be measured using either a 100ml or a Sml burette
39
~~
1 to 4 LVDT's5,6 A-oximity
t ransdJcers
1
Ys---t Do 05
bIE---- ---
I. rI I
Section A-A Section B-8
Radial 6Ys+6Ys Axial6Z, +6Z2 - 6Z3 +6Z4
strain = strain = 2 2DOZo - ZE
FIGURE 4·2 TRIAXIAL APPARATUS INSTRUMENTATION
40
plumbed into the back pressure system. There were two completecells with the full instrumentation.
Overy (1982) incorporated a bellofram piston unit into the celltops which could be clamped to the load ram and was used to applya deviator stress to the samples during consolidation.
4.2.2 The consolidation equipment
Overy (1982) developed a system to allow anisotropicconsolidation of the samples in the triaxial cell~t· A schematiclayout of the system is shown in Figure 4.3. The systemconsisted of a servo control unit which detected a change in thesample diameter measured using the proximity transducers andadjusted the deviator stress to correct the change by adjustingthe air pressure in the bellofram unit connected to the load ram.The air pressure was adjusted by a motor driven air regulator.The system was used in conjunction with another motor driven airregulator which was connected to the cell pressure system and wasused to increase the cell pressure at a constant rate during theconsolidation phase.
The equipment was arranged to allow control over one cell onlyand only allowed consolidation without extensive rearrangement ofthe drive system.
4.3 MODIFICATIONS TO THE TRIAXIAL TEST FACILITY
4.3.1 !odifications to the control system and loading frame
Plate 4.1 shows loading frame~ control unit and signalconditioning equipment used in this project. A Wykeham Farranceconstant strain rate machine was incorporated in the loadingframe to allow undrained strain controlled strength tests to becarried out on samples after cyclic loading without relieving thestress on the sample by disconnecting the hydraulic loadingsYstems. The triaxial cells were fitted with castors and an oilproof preparation area which connected directly to the load frame
: fI Q"\"I" "I.. .. ..,I ,!"--,.I
~. .. ..Q~~ ...
e ~. , .. I .: •
_j
oIXI-ZoU
oZ<!
w:L<!IXLL
(9Zo<!9WIf-
LLo
_j<!IXWZW(9
w~_j0...
41
u·cc;OL..L.. _
I -----7-1- c~ c u 0._ L..~ Ue QI ....L..u ...,
U):J_"0eU)._ c"00CL..L..
- o
"C·5b-a.-a.
:JQ,I(/)
U
~UJt-V>(;;...J
~t-Z8zo~ac5V>
8ua.oet:6(/)
z«
42
was constructed. This eliminated the need to carry the triaxialcells and limited sample disturbance.
As the equipment was going to be used for long term tests a cutout switch was fitted which limited the travel of the ram intothe cell and therefore prevented damage to the instrumentation inthe event of the sample failing. A safety cut out switch was notrequired for the cell pressure system as the travel of the pistonunit exceeded that of the hydraulic actuator.
The original piston unit on the cell pressure system was sealedby 0 rings and was rigidly attached to the hydraulic actuator. Acombination of friction and binding caused by these two factorsprevented low magnitude ( t 5 KPa) and low frequency ( t 0.05 Hz)stress pulses from being applied to the sample. A new pistonunit was constructed which incorporated a bellofram seal andwhich was not rigidly attached to the hydraulic actuator, butrelied on the cell pressure for the return stroke. This systemwas found to work well.
The control system was found to be very susceptible tointerference which was caused mainly by the air conditioningsystem in the laboratory switching on and off. Rewiring theearth system of the control unit and installing supressors onthe air conditioning system unit removed some but not all of theinterference.
4.3.2 Modifications to the consolidation control unit
The consolidation phase of the test programme required fairlyheavilyanisotropically overconsolidated samples to be produced.This was time consuming and required independent control of eachtriaxial cell to eliminate any time delays in the samplepreparation procedure.
The control system was rebuilt with new motors to allow threecompletely independent consolidation control systems. Plate 4.2shows the layout of the consolidation system and two samples
I-Zw2:0...=>owzo~o_J
o(j)zouw:r:I-
N-..:t
W~_J0...
43
consolidating. A high pressure air line (2.5MPa) was installedin the laboratory to supply one of the consolidation systems.This allowed the high consolidation pressures required to beachieved. The anisotropic consolidation control box had to beextensively rewired to eliminate crosstalk between the channels,and to provide additional power to supply the extra motorscontrolling the regulators.
The clamping arrangement between the bellofram piston on the celltop and the load ram was improved to cater for the higherpressure required from the standard airline by changing fromclamp type A to clamp type B, as shown in Figure 4.4. To caterfor the high pressure airline another bellofram piston unit wasbuilt which acted directly on the load ram, rather than through aclamp, as shown in Figure 4.5. To produce anisotropicallyoverconsolidated samples requires negative deviator stresses asKo exceeds 1.0. Some uplift was supplied to the load ram by thecell pressure acting on the lower bellofram in the case of the lowpressure cells, and this was supplemented by small additional aircylinders which were clamped to the load ram. These actedagainst the cell top. A special bellofram piston unit wasincorporated in the high pressure cell to provide the requirednegative deviator stresses. All the triaxial cells were equippedwith a vacuum connection between the load ram and the top platen.This was installed by Overy (1982) although the system was firstdeveloped by Boyce (1976).
4.3.3 Modifications to the measuring and recording systems
The Transport and Road Research Laboratory generously providedsufficient funds to fully instrument a third cell. Identicalinstrumentation was chosen to ensure compatibility with theexisting triaxial cells.
4.3.3.1 Axial load
The existing load rams, which were designed by Austin (1979),were found to be unsuitable for the relatively high loads
44
PHOSPHOR BRONZEBUSH
BELLOFRAMSEALS
CLAMP A CLAMP B
FIGURE 1..1. ANISOTROPIC CONSOLIDATION CELL TOP
45
~-TIE RODS
LOCATING PIN
.,..._-LOAD RAM
FIGURE '·5 BELLOFRAM LOAD UNIT FOR ANISOTROPICOONSQLIDATION AT HIGH CELL PRESSURES
46
required during consolidation. It is thought that this was dueto their age and past overstressing. The Transport and RoadResearch Laboratory generously provided additional funds andthree small commercial tension/compression load cells werepurchased and installed in the triaxial cells. These were foundto suffer from zero drift in the long term and were replaced byload cells designed in the Civil Engineering Department atNottingham. These were disc load cells and consisted of aloaded web, which was shaped to give an approximately linearstrain profile with radius. This allowed load cells with similaroutputs to be constructed. They were found to work very well.
4.3.3.2 Pressure transducers
The cell pressure, base pore pressure and centre pore pressuretransducers were from a range of strain gauged siliconetransducers made by Druck Ltd and worked well. The transducerlead for the centre pore pressure transducers had to be sealed asit passed through the top platen and the cell base andconsequently could not be fitted with a proper electricalconnection. It was found that clothes pegs with metal faced jawsprovided a good electrical connection with the bared transducerleads.
4.3.3.3 Defonmation transducers
The arrangement of LVDTs and PTs was found to work satisfactorilyand could only be improved upon by adding additionalinstrumentation measuring across diameters at right angles to theoriginal instrumentation. This would be very expensive and theimprovement was thought not to be large enough to justify theadditional cost.
4.3.3.4 Volume change
The original system used by Overy (1982) consisted of fourseparate back pressure/drainage lines, two with 100ml capacityburettes and two with Sml capacity burettes. The back pressure
47
was applied to the sample using air pressure, controlled by anair regulator, acting on an air water interface. The direction offlow through the burettes could be reversed by operating twovalves simultaneously. The lOOml burette was used duringconsolidation while the 5ml burette was found to be sufficient fordrainage periods during cyclic load testing. However to changeburettes required changing back pressure lines.
The system was rebuilt and incorporated both lOOml and 5mlburettes in the same back pressure line either of which could beselected by a valve. The direction of flow could be reversed byusing a single four way ball valve which proved much moresatisfactory than the original system. The original system usingair pressure was retained for two of the cells but a system usingmercury in two pots, one fixed and one adjustable, wasfor the third cell. The aim was to be able to applyback pressures or suctions to the samples but the timethe test programme did not allow any experiments withback pressures to be carried out.
providednegative
scale ofnegative
4.3.3.5 Data Processing Equipment
The outputs from all the transducers were taken through a signalconditioning unit with variable gain amplifiers as described byOvery (1982). The outputs were recorded on ultra violetsensitive paper by an oscillograph. Each circuit could bemonitored by a voltmeter or oscilloscope for instantaneousreadings. A further summing unit was provided which acceptedoutputs for all the transducers, scaled them and output theresults to an X - Y plotter. This allowed total and effectivestress strain paths to be plotted.
The deformation and pore pressure measuring circuits were fittedwith offset generators with outputs to the oscillograph. Thesecircuits enabled the small cyclic component of a large inputsignal to be separated from the d.c. level and amplified further.
There was a lot of electrical noise on the traces and the
48
equipment was very susceptible to electrical interference fromvarious other pieces of equipment in the laboratory. This wascured by rewiring the earthing system within the signalconditioning system.
The amplification of the deformation measuring transducers wasincreased and d.c. offset controls fitted to all the amplifiers.This allowed the maximum gain of the amplifier to be usedwhatever the output signal of the transducer. Appendix D givesthe calibrations for all the transducers.
The original system incorporated only one bridge balance controlfor the load, cell and pore pressure systems, which madeoperating more than one cell at a time rather difficult. Thesystem was extended to give separate bridge balance controls foreach circuit for each cell. These were all switchable throughthe main amplifiers and allowed all three cells to be operatedsimultaneously.
4.3.3.6 Digital Data Recording System
An analogue to digital converter, computer, disk drive andprinter were purchased. The A-D converter was made by CllMicrosystems ltd and the computer system by Commodore BusinessMachines. A Hewlett Packard X-V plotter was used for directplotting of the results. The A-D converter had its own memorysufficient for 16000 readings and could be programmed to take xscans with a y millisecond time delay and store the data forretrieval later by the computer. The machine was 16 bit andaccepted inputs in the range !10 volts giving an effectiveresolution of 15 bits, the 16th bit being used for the sign.Sufficient resolution from the load, cell and pore pressuretransducers could be obtained using the minimum gain availablefrom the signal conditioning unit. Therefore inputs for the A-Oconverter were taken before the gain switch, enabling the gainsto be adjusted without affecting the computer calibrations.However the signals from the deformation transducers had to beamplified further to get sufficient resolution and involved extra
49
software to keep track of the permanent strain as the gains andd.c. offsets were set at the start of each test.
At present the system is limited by the time taken to retrieve,process and store the data using Commodore basic, rather than bythe amount of available memory. This could be overcome byprogramming, at least in part, in machine code which can speed uprepetitive processes by a considerable amount.
4.4 PNEUMATIC TRIAXIAL RIG
4.4.1 Existing Equipment
This rig was originally constructed by Pappin (1979) fordetermining the resilient modulus of aggregate samples. Itconsisted of a triaxial cell similar to those used in the mainservo hydraulic rig with a bellofram loading piston acting on theload ram. This was supplied with air under pressure from an airregulator acting through a solenoid valve, which was switched bya motor driven cam. This provided an approximation to a squarewave with a frequency of about 1Hz. The load was measured insidethe triaxial cell by using a strain gauged load cell. The cellpressure was supplied directly from an air regulator and measuredusing a pressure gauge. Axial deformation only was measuredbetween the end caps using a wire wound linear potentiometer.
4.4.2 Modifications to the equipment
The deformation measuring system was improved by adding two LVOTsworking as a pair over a gauge length in the centre third of thesample. The signals were summed electronically, amplified andoutput to an ultra violet oscillograph through an offsetgenerator. Appendix 0 gives the calibrations.
An attempt was made to measure the radial deformation, initiallyby using strain coils, then subsequently by a strain gaugedaraldite hoop. The strain coils were made by Bison InstrumentsInc and consisted of two 2 inch diameter coils, placed one each
50 ..
side of the sample across a diameter. One coil acted as theemitter and the other as the receiver. The output from thebalancing box was taken directly to an ultra violet recorder. Thestrain hoop was a scaled down version of those developed by Boyce(1976) and consisted of a ring attached concentrically to thesample across a diameter as shown in Figure 4.6. The straingauges were mounted 90° around from the attachment points and theoutput taken to the U.V. oscillograph via an amplifier and offsetgenerator.
It was found thatcontrolling the load
the solenoidinterfered
operating the air valvewith the offset generator
signals. The load system was modified to use mechanicallyoperated air valves running directly from a cam. Air bleeds wereincorporated into the system to provide some adjustment of theshape of the waveform and the facility for supplying an ambientdeviator stress and pulsing from that level was added.
4.5 SLURRY MOULD FOR eBR SAMPLES
A mould was constructed to allow overconsolidated eBR samples tobe produced in a similar manner to the consolidated triaxialsamples. The mould, shown in Figure 4.7, was constructed as atube which was split into three sections. The middle section wasthe same height as a standard CBR mould. The pistons were facedwith porous stones which allowed end drainage. The weight of themould was counterbalanced by two weights connected to the mouldby a pulley arrangement.by a bellofram piston
The consolidation pressure was appliedunit which was connected to the high
pressure air line through an air regulator.
There was appreciable friction between the pistons and the mouldwhich made it impossible to swell back to very low pressures inthe complete mould. To avoid this the mould was dismantledafter consolidation to the maximum pressure and the sampletrimmed to size in the centre section. The bottom plate wasattached to the centre section of the mould and the sample loadedat the swell back pressure through the top piston. The piston
51 ..
SAMPLE
PLAN
STRAIN GAUGES
SIDE ELEVATION
FIGURE 4.6 STRAIN HOOP FOR MEASURINGRADIAL lEFORMATION OFCOMPACTED SAMPLES
52
LOAD
FILTER PAPER
COUNTERBALANCEWEIGHTATIACHMENTPOINTSOLTS HOLDNGMOULD TOGETHER
RESERVOIR
----.i--'-WAT ER
FIGURE 1.·7 CBR SLURRY MOULD
53
was placed free on top of the sample and therefore ensured thatthe correct pressure was applied to the top of the sample. Thesample was kept under water to ensure saturation.
4.6 SUCTION MEASURING APPARATUS
Two types of apparatus were used to measure the suctions of thesamples. The first measured the suctions of the sample at itsoriginal moisture content using a piece of apparatus called theRapid Suction Apparatus which was developed at the Transport andRoad Research Laboratory and is described by Dumbleton and West(1968). Figure 4.8 shows a diagrammatic layout of the apparatus.The apparatus consists of a high air entry porous stone 12mm indiameter connected to a hand operated suction pump via acapilliary tube. The sample was placed on the stone and the pumpoperated until the suction was sufficient to balance that in thesample. This was indicated by no movement of the meniscus inthe capillary tube. The range of the apparatus could beincreased by applying an external pressure to the sample. Bothpressures were measured by pressure gauges.
The second method allowed samples to reach an equilibriummoisture content under a set value of suction. The layout of theapparatus is shown on Figure 4.8 and consisted of 40mm diameterhigh air entry porous stones. These were saturated and a constantvalue of suction was applied to them from a vacuum pump via aregulator. The samples were placed on the stones and allowed toreach the equilibrium moisture content for that suction. Thevacuum regulators worked on the controlled air bleed principle,and once set were found to be very stable. The porous stonestended to dry out after three or four days and then leak air,but this time scale was found to be sufficient for the KeuperMarl samples to reach equilibrium provided that they were mixedto a moisture content reasonably close to the equilibrium valuefor that suction.
54
TO GAUGE
CAALLIARY TUBE
TO PRESSUREPUMP
rTO GAUGEHAND OPERATEDSUC1l0N PUMP
RAPID SUCTION APPARATUS
POROUSSTONE
VACUUM REGULATORS
TO VACUUMPUMP
EQUILIBRIUM SUCTION APPARATUS
FIGURE 4·8 DIAGRAM OF SUCTION MEASURINGEQUIPMENT
55
4.7 DETERMINATION OF PRE-CONSOLIDATION PRESSURE
Some undisturbed site samples were obtained and thepreconsolidation pressure determined, using Rowe cells. Theseare described more fully by Rowe and Barden (1966). By assumingvalues of in situ stresses and the level of the water table anestimation of the natural overconsolidation ratio could be made.The samples were supplied in U4 tubes and had to be trimmed tofit in the Rowe cells. A cutter was manufactured which fittedover the end of the U4 tube and cut the sample as it wasextruded. The tip of the cutter was positioned level with the endof the tube to try and limit sample disturbance, and the cutterbody was the correct length for the Rowe cell samples.
The Rowe cell is shown in Figure 4.9 and consisted of a base witha tapping, cell walls, and a cell top with a rubber diaphragmand two further tappings. The axial load was applied to thesample by hydraulic pressure acting on the sample through therubber diaphragm. Various drainage options and pore pressuremeasurement points are available using the tapping in the cellbase and the second tapping in the cell top. Top and bottomdrainage was used in this research to minimise the drainage pathand speed up testing. No attempt was made to measure porepressure.
The back pressure and load pressure were controlled by airregulators acting through air water interfaces. The pressureswere monitored using pressure cells which consisted of a straingauged diaphragm rigidly clamped at its edges. These weredeveloped by Parr (1972), and were capable of resolving tol~Pa. The axial deformation was recorded by a dial gauge.
The Rowe cells were paired up to operate off one loading system toincrease the rate of testing. The pressure to the diaphragm, theaxial deformation and the volume change were recorded. Theanalysis was based on the axial effective pressure anddeformation results as the volume change readings were found tobe unsatisfactory.
56
DIAL GAUGE
AXIAL LOADPRESSURE INPUT
BEARING AND SEAL
7=7=~¥7=#=~t= ORAl NAGELINE
RUBBER BELLOWSPOROUS DISCCELL WALL
POROUS STONE BASE
FIGURE 4·9 ROWE CELL
57
CHAPTER FIVE
EXPERIMENTAL PROCEDURE
5.1 THE MATERIALS
The main research programme was carried out on Keuper Marl, andsome further simpler tests were performed on London clay and onGault clay. Standard soil classification tests were carried outon all three clays and the results are presented in Appendix A.
Extensive use has been made of Keuper Marl at Nottingham inearlier research projects on repeated load triaxial testing(Lashine 1971, Hyde 1974, Austin 1979 and Overy 1982) and as thesubgrade in the pavement test facility (Bell 1978, Brown et al1980). The Keuper Marl used in this research was taken fromeXisting stocks in the laboratory. The material was originallyobtained from a local brickworks as tailings from unfired bricks.These were dried and then broken down in a mixer. The materialpassing through a BS 52 (150 micron) sieve was retained and usedin the research programme.
The Gault and London clays were provided by the Transport andRoad Research Laboratory from stocks held by them for theirresearch projects.
5.2 CONSOLIDATED TRIAXIAL SPECIMENS
5.2.1 Initial Soil Preparation
Dry Keuper Marl was placed in the bowl of a Hobart mixer togetherwith sufficient de-aired water to give a moisture content ofabout 50%. The Marl was left to slurry for at least three dayswith any water loss through evaporation made good by addingde-aired distilled water to maintain a constant slurry level.
58
5.2.2 First Stage Consolidation - Slurry Moulds
The moulds were designed by Overy (1982) to produce 78mm diametersamples and are shown in Figure 5.1. The porous stones in thepistons were de-aired under water in a vacuum vessel for at least24 hours before use. The tips to the miniature pore pressuretransducers were sintered bronze to ensure rapid response of thetransducer and hence were easier to de-air. Applying a vacuum tothe transducer and then releasing it under de-aired water threetimes just before use was found to be sufficient provided thatthe transducer was then kept immersed in water.
The bottom piston was placed in a container of de-aired water andthe mould placed over it. The mould rested on a spacer whichprevented it from completely covering the piston. The spacer wasremoved just prior to applying the consolidation load. Thebottom piston was covered with a disc of filter paper, which,when wet, swelled sufficiently to close off the clearance gapbetween the piston and the tube. It also prevented the porousstone from clogging. The container and mould was placed on avibrating table and the tube slowly filled with slurry using along handled pot. The slurry was stirred to aid removal of anyair bubbles which may have been trapped when the tube was filled.
The top cap, and filter paper, complete with pore pressuretransducer, was then placed in position and the complete mouldinstalled in the loading frame. A vertical load of 70kPa wasapplied and the sample left to consolidate. It was found thathandleable samples could be produced in approximately 10 days.
5.2.3 Cell Base Preparation
The most important aspect of preparing the cell base was ensuringthat the drainage system and base pore pressure probe weresaturated. To achieve this, a metal collar was clamped to thecell base, see Figure 5.2, and carefully filled with de-airedwater. The tappings were flushed through with a hypodermicsyringe, which was found to be very effective at removing bubbles
59
LOAD
FILTER PAPERDISCS
SOIL
PORE PRESSURE,-+-1--
TRANSDUCER
PERSPEX TUBE
RESERVOIR
FIGURE 5·1 SLURRY CONSOLIDATION MOULD
~60
0 UJ0:: UJ V)a. N «UJ zc) CD0:: Oz ....J:::> 0:_V) COo:V)UJ OW UJ0: Wc) (!)~a. 0:« «WlLJ UJz Z.......... -0: z« og~0 _0:a. lIlO ..... III
~I ,.: II I'I ,,'II 'I, I," I, I', : I ''I,
'I I
I
II ,
II' ,I I
• I I I . :.,I I i I,", I,
0:....J0
Er lLJW
UJ tnz (J
til III C)
tn~ is z0: 0:UJ ~a. (J o(/) tn« z-0UJ «z «ffi tnUJ ~
....J
....JW(J
LLo_.~UJo
61
adhering to the base.
The drainage ring and porous stone had been de-aired by boiling indistilled water for some time and then storing under distilledwater in a vacuum vessel for at least 24 hours. The drainagering was placed in position in the cell base and then coveredby a thin disc of PTFE sheet. The porous stone was then screwedinto place clamping the plastic in position. The PTFE sheetseparated the sintered bronze drainage ring from the stainlesssteel bottom platen and prevented any possibility of bimetallicaction.
The top and bottom platens were prepared by first greasing thestainless steel surfaces with high vacuum silicone grease. Theywere then covered by a disc of polythene sheet. This was cut ina series of concentric overlapping arcs which would allow radialexpansion of the sample to take place. A further layer of greasewas applied and then finally covered with an offcut of latexmembrane. The sides of the pedestal were greased to ensure agood seal with the membrane, and then the bottom platen wasplaced in position.
5.2.4 Sample Installation
The consolidation load was removed from the sample and the sampleextruded from the slurry mould. The sample was trimmed to 150mmlength, weighed and placed on the pedestal. The trimmings wereused for moisture content determinations.soft and could only be handled using
The samples were verytwo 1200 longitudinal
sections of tube of the same diameter as the sample which actedas load spreaders.
A piece of latex rubber membrane, 25mm long, was positioned atthe base of the sample such that it just covered the bottomplaten. This was to ensure that there was no possibility of anygrease from the bottom platen contaminating the filter paperdrains, and to improve the response of the base pore pressureprobe during consolidation by removing any possibility of a
62
drainage path forming between the sample and the bottom platen.
The studs which provided the attachment points for the LVDT coreswere then carefully embedded in the sample using a jig. Thispositioned the bottom studs 50mm up from the base of the sampleand positioned the top studs to give a gauge length of 50mm overthe centre section of the sample. The studs were designed byOvery (1982) and are shown in Figure 5.3.
The radial drainage system was then fitted. This consisted oftwo sets of filter paper drains cut to form a solid band aroundthe base of the sample with eight vertical fingers of paper alongthe length of the sample. Sandwiched between the two sets offilter paper were individual wicks, eight in all, consisting oflengths of unravelled string. The radial drains were cut in theform of fingers to allow clearance for the instrumentationmounting pOints and to eliminate any restriction to radialdeformation of the sample. A rubber band was used to secure thebottom end of the filter paper to the drainage ring and, once thepaper was wet, it was found to stick to the sample and requiredno additional support.
The targets for the proximity transducers were cut from aluminiumfoil and placed up against the sample across a diameter 90°
around from the LVDT studs. It was found that they also adheredwell to the sample and remained in place while the membrane waspositioned.
The membrane, supported by a membrane stretcher, was placed overthe sample and the bottom O-ring rolled off. The top cap wasthen threaded over the pore pressure lead and placed in position.The membrane was then finally rolled off the stretcher and sealedat the top and bottom by three additional O-rings, after ensuringthat there were no air bubbles trapped by the top cap. Thetransducer lead exit from the top cap was sealed using two smallO-rings, one inside the other, and clamped up tightly with ametal collar.
63
SCALEI 10mm
~---LVDT BODY
I I
LOCK NUT
WASHER
FIGURE 5·3 LOCATION STUD FOR LVOTS
64
At this stage the metal collar on the cell base was removed, andthe cell base and specimen wiped clear of excess water.
5.2.5 Installing the Instrumentation
The instrumentation was supported by a metal framework which hadbeen developed by Overy (1982). The framework was placed inposition and bolted to the cell base. The membrane was carefullypunctured over the LVOT mounting studs using a soldering iron andthe supports for the LVOT cores screwed in. The holes in themembrane were sealed using two small 0 rings, see Figure 5.3.The proximity transducers were placed in position. Plate 5.1shows the sample complete with the instrumentation just prior tofitting the cell top. The cell top, complete with load ram, wasthen fitted and filled with silicone oil. The back pressure linewas attached and the sample was allowed to consolidate overnightunder approximately the same pressures as in the slurry mould.This allowed the sample to reconsolidate and expel any excesswater trapped behind the membrane before the LVOTs and PTs werezeroed.
The sample was then checked for saturation by increasing the cellpressure and monitoring the change in pore pressure with thesample drainage off. Generally the pore pressure incrementsequalled the cell pressure increment giving the B value of 1. Ifthis was not the case the back pressure Was increased and thesample left to stand for twenty four hours before rechecking thesaturation. Any sample with a B value of less than 0.97 wasrejected.
The cell was then drained of oil and the cell top removed. Theinstrumentation was then set up accurately. The LVOTs were setat the end of their range to allow for the movement of the sampleduring consolidation and the proximity transducers were set nearthe outermost limit of their range to allow for sample expansionduring testing. The cell top was then replaced and refilled withoil and pressurised to the original pressure.
_J_Juiu
WII-
<.!)Zl-I-LL
oI-
0::o0::0....
zo~I-ZW~:::J0::l-(/)Z
II-
~
WI-W_J0....~ouw_J0....~«(/)
wt:t_J0....
65
5.2.6 Consolidation Procedure
The servo unit controlling the anisotropic consolidation was setat the null position and the system switched on. The motorcontrolling the cell pressure was switched on and the sampleallowed to consolidate. A minimum of two readings a day weretaken from all transducers during the consolidation phase.
The sample was allowed to remain under the maximum cell pressureand deviator stress until the pore pressure had stabilised and nofurther drainage took place. This usually took about four daysafter which the cell motor was reversed and the sample allowed toswell back. When the sample reached the required final meannormal effective stress it was left to reach equilibrium beforetesting. This again took four to five days.
5.2.7 Test Procedure
Before the hydraulic actuators were connected to the sample thedrainage taps were closed and the signal conditioning amplifierswere set at the required level. The cell pressure was connectedto the hydraulic loading system first, and then the load ram wascarefully connected to the hydraulic actuator. With care thesample could be transferred to hydraulic control with a maximumchange in the deviator stress or cell pressure of approximately±2kPa.
Following a period of cyclic load testing the sample was returnedto pneumatic loading at the pressures achieved at the end ofconsolidation by careful ly disconnecting the hydraulic actuators.
The results from the overconsolidated triaxial samples arepresented in Chapter 6.
5.3 CONSOLIDATED CBR SAMPLES
The mould for these samples was a much larger version of themould for the triaxial samples described in section 5.2.2 and was
66
prepared in the same manner. The mould was placed on thevibrating table, with the bottom piston under water in acontainer, and filled with slurry. The spacers supporting themould above the container base could not be removed until thesample was installed in the loading frame and mould counterbalance weights attached. This prevented the mould from fallingunder its own weight and allowed the consolidation load to beapplied directly to the sample through the bottom piston as wellas the top piston.
The sample was left for approximately three weeks to consolidateunder the maximum load. The sample was then removed from theloading frame and the mould split into the three sections. Thesample was cut with a cheese wire. The bottom plate was clampedin place to the centre section forming the CBR sized mould whichwas then returned to the loading frame. The final load wasapplied to the sample through the top piston. The sample waskept under water by a rubber sheath sealing the gap between thepiston and the mould walls as shown in Figure 5.4. The samplewas left for two to three weeks to swell back before testing.
When equilibrium was judged to have been attained the sample wasremoved from the loading frame and the top trimmed off level. Astandard constant strain rate machine was used for the CBR test.The sample was tested with no surcharge weights.
After the CBR test the bottom plate of the mould was removed andtwo standard oedometer samples were carefully cut as the CBRsample was extruded. These were used to determine the maximumpressure applied to the sample during consolidation. Theundrained shear strength was measured at the top and bottom ofthe sample using a 19mm shear vane and also a pocketpenetrometer. The sample was sectioned and the moisture contentat the top and bottom of the sample was determined. The resultsfrom the consolidated CBR samples are presented in Chapter 8.
67
LOAD
WATER LEVEL
o RING
POROUS STONE
r::T=====:::::i::::::::i===========+=======~=====:i:::i::!:::I~\ FI LTER PAPER
BOLTS HOLDINGCBR MOULDTOGETHER
FIGURE 5·4 ARRANGEMENT FOR ALLOWING SAMPLESTO OVERCONSOUDATE IN THE CBRMOULD
68
5.4 COMPACTED TRIAXIAL SAMPLES
The compacted triaxial specimens were 78mm in diameter and 148mmlong. The dry powdered clay was weighed out into the bowl of aHobart Mixer and sufficient distilled water added to give therequired moisture content. It was found that approximatelyfifteen minutes mixing time was sufficient to give a consistentuniform mix.
The mould in which the samples were compacted consisted of threeequal longitudinal sections which were clamped together withjubilee clips to form a cylinder. The inside of the mould wasprepared by inserting a liner made of polythene sheet. This waseasy to separate from the sample and prevented the sample fromadhering to the mould. The mould was clamped to a base plate andthe soil placed in thin layers and compacted by hand using atamping rod approximately 12mm in diameter.
The mould was dismantled, the sample weighed and placed on thecell base. The top and bottom platens were prepared as for theconsolidated samples, see section 5.2.3. A jig was used to placethe LVOT and strain hoop studs in the correct locations and themembrane was then positioned using a standard membrane stretcherand sealed in place with O-rings. Neoprene membranes were usedas they are impermeable to air which was used as the confiningmedium. No provision was made for drainage or for themeasurement of pore water pressure.
The membrane was punctured over the mounting studs using asoldering iron and the instrumentation fitted and zeroed. TheLVOT details were the same as for the consolidated samples, asshown in Figure 5.3, and the radial deformation transducers wereattached to similar studs and sealed in the same manner.
The cell top was fitted and the sample was left for twenty fourhours before testing. This was to allow for any variation inmoistureequalise.
content or suction caused during sample assembly toThe test programme is detailed in Chapter 3. After
69
testing the cell was dismantled, the sample weighed, and thecentre third of the sample used for moisture contentdeterminations. The results of the tests on the compactedtriaxial samples are presented in Chapter 7.
5.5 COMPACTED CBR SAMPLES
The soil was prepared in the same manner as for the compactedtriaxial specimens, section 5.4, and compacted into a standard CBRmould (BS 1377). The soil was placed in thin layers and compactedto refusal using the same hand tamper. The weight of the soilwas determined by subtracting the weight of the mould from thetotal weight after compaction. The top of the mould was screwedinto position and the sample left for twenty four hours beforetesting, again to allow any pore pressure or moisture contentvariations to equalise.
The CBR was determined at the top and bottom of the sample using astandard CBR testing machine. The undrained strength of the soilwas determined using a 19mm shear vane and also a pocketpenetrometer. The moisture content profile with thickness wasalso determined.
The results are presented in Chapter 8.
5.6 MEASUREMENT OF SOIL SUCTION
5.6.1 Rapid Suction Apparatus
The apparatus is described in Chapter 4, section 4.6 and shownin Figure 4.8. The apparatus was prepared by applying a suctionto the porous stone while it was immersed in de-aired distilledwater. This saturated the porous stone and allowed the meniscusto be set at a suitable place in the capillary tube.
The porous stone was approximately 12mm diameter and a smallmould and tamper were manufactured to make samples of a suitablesize. The mould produced samples 12mm in diameter and 10mm long.
70
Sufficient dryhighest water
soil for a number of tests was mixed to thecontent desired for the initial suction
measurement. A sample was compacted in the mould, weighed, andwas recordedA little dry
then extruded on to the porous stone. The suctionand the moisture content of the sample determined.soil was added to the remaining mixture! to decrease the moisturecontent and the process was repeated.
5.6.2 Equilibrium Method
The apparatus is described in Chapter 4, section 4.6 and shown inFigure 4.8. It consists of a number of high air entry porousstones 40mm in diameter on to which each sample was placed. Aknown constant value of suction was applied to the stone and thesample allowed to reach equilibrium. The moisture content of thesample was then determined.
The porous stones were de-aired by leaving them under water in avacuum vessel. It was found that the stones de-aired much morequickly if the vacuum vessel was placed on a vibrating table. Acylindrical mould was manufactured and used to produce suitablysized samples. The soil was mixed to a moisture content whichwas estimated as being close to the equilibrium moisture contentat the value of suction chosen and then compacted in the mould.The sample was extruded and placed on the porous stone and left toreach equilibrium. After a few days the moisture content of thesample was determined and the suction, as measured by a mercurymanometer, recorded.
The results from both types of test are detailed in Chapter 7.
71 ..
CHAPTER SIX
BEHAVIOUR OF OVERCONSOLIDATED SATURATED SAMPLES OF KEUPER MARL
6.1 INTRODUCTION
The objective of this section of the work was to produce and thentest triaxial samples with a known, carefully controlled stresshistory which resulted in final stress conditions similar tothose found in road subgrades in the U.K. The test programme wasdesigned to investigate the resilient and permanent response ofthe samples subjected to loading conditions similar to thoseapplied to a road subgrade by traffic loading. The range ofinitial stress conditions and the test programme is described inChapter 3. The test equipment is detailed in Chapter 4 and thesample preparation procedure in Chapter 5.
6.2 TRIAXIAL CONSOLIDATION RESULTS
6.2.1 Consolidation Paths
The consolidation phase of each sample was monitored regularly toensure that it attained the correct consolidation history.Figure 6.1 shows the results for the samples in mean normaleffective stress-specific volume space and shows quite clearlythat there is a unique virgin consolidation curve for thismaterial. In this case theas the samples are all
curve is the Ko consolidation curveanisotropically consolidated. The
individual swell back lines for each sample are shown and areall of a similar shape. These curves have a very shallow initialgradient which tends to get progressively steeper as the meannormal effective stress decreases further and theoverconsolidation ratio increases. This result was as expectedand is typical behaviour for· overconsolidated samples. Thecritical state parameters determined from this figure are given inAppendix A.
The consolidation results for all samples plotted in mean normal
1·35 '--_--'-_,,_____.....___-'----'-.I .......... _
20 50 100 200 SOO1000 2000Mean Normal Effective Stress,p'Ik Po}
15
-~o-c~6oL.OJ-o3:
25
20
1·65
1·55
1·45
72
-
Pe p'Definition of equivalent
pressure
>- v
-gu....uOJ
(J)
FIGURE 6·1 COMPRESSION AND SWELL - BACKDATA FROM TESTS ON KEUPER MARL
73
effective stress - deviator stress space again show a scatteraround a unique curve for the consolidation phase, the gradientof which gives the value of Ko' However, the overconsolidatedsections of the curves tend to oscillate about a mean curve andFigure 6.2 shows a typical test result only for clarity. Thiseffect was noticed by Overy (1982) and is due to the way in whichthe servo system maintains anisotropic conditions. The systemuses the proximity transducers to detect any change in the samplediameter during consolidation and adjusts the deviator stressaccordingly to maintain a constant sample diameter. Duringvirgin consolidation there is an appreciable volume change as theeffective stress changes, i.e. the gradient of the consolidationcurve, A ; in lnp'-v space is relatively large. This means thatthe servo system is activated frequently and consequentlyproduces small increases in deviator stress which results in asmooth curve in p'-q space.
The swell back curves in lnp'-v space though, have a much lowergradient than the virgin consolidation curve and consequentlythere is a large change in stress for a small change in volume.The samples are effectively a lot stiffer. This means that theservo system operates relatively infrequently and hence causeslarge deviator stress changes, which result in theoverconsolidation path in p'-q space oscillating around the meancurve.
Mayne and Kilhaney (1982) suggested that the value of Ko was afunction of the overconsolidation ratio such that
K = (1 - sin ~') x OCR sin~'o 6.1
This relationship is shown plotted in Figure 6.2 and fits themean curve drawn through the data quite accurately.
6.2.2 Response of the radial drainage system
Overy (1982) tested normally consolidated samples of Keuper Marlprepared using this test equipment. The samples were
74
QIe ~a ....."0Qj ____- Z>. /~ 0
'0 (/)
6 /' - N ....f ..... (/)- /
UJ.~ ....
s:~[ L ~ 0:::
"0 0 LL
Vc .....0 ~Q,I 23c
c>.
I00 ~:z:~ u.~ L- eo
~"OQ,I dQ,I-
IL--Q.~ :::jQ,I
Q_-:--.... UJ0..:
~\ 0
\z«
c Z 0:::0 0 -c
\+: en ~0:g ffl 0:::~
0:::~-, Q_
c
~8 ~-. ~
<, U)
UJ"-..
~LL
co U) ....t N 0 N0 d 0 ad;. d d
, b J
75
consolidated from a slurry initially, and then transferred to thetriaxial cell for the main consolidation phase. In order tominimise the effects of sample disturbance when the sample wasunloaded and transferred from the slurry moulds to the triaxialcell Overy used the lowest consolidation pressure in the slurrymould that would give handleable samples.
As this research concentrated on producing overconsolidatedsamples the first few samples were consolidated to higherstresses in the slurry mould to reduce the consolidation time inthe triaxial cells, although the samples were all consolidated inthe triaxial cell to at least three times the effective stressachieved in the slurry mould. This factor was suggested byLoudon (1967) to eliminate the effects caused by stress relief.
The preliminary tests highlighted deficiencies in the radialdrainage system with large excess pore pressures developingduring consolidation, which took a week or more to dissipate. Itproved impossible to overconsolidate some samples to any realdegree as they seemed to reach an 'equilibrium' state with anegative excess pore pressure of up to -60 kPa without absorbingany more water.
This was thought to be caused by constrictionsin the filter paperdrain either by pinching between the bottom platen and themembrane or general restrictions caused by the cell pressure asthis was higher than had been used previously. Anotherpossibility was that the sample surface had been 'smeared' whenit had been extruded from the mould. Smear is defined as thelocalised remoulding and 'sealing' of a clay surface causing abarrier of lower permeability than that of undisturbed clay.
The permeability of the filter paper drains in the triaxial cellwas measured as a function of the cell pressure, and was found tobe higher than that quoted for Keuper Marl by Overy (1982) by afactor of approximately 6000. Appendix B contains the testdetails.
76 ..
The possibility that the filter paper drains had become cloggedduring the consolidation phase was also considered. However thepermeability of a used filter paper drain could not be measuredin the same manner as they were impossible to remove intact fromthe samples.
The flow path of the drainage system was improved by reducing thediameter of the bottom platens to approximately 3mm greater thanthe sample diameter, and rounding off the edges. Two filter paperdrains were used with a small wick sandwiched between them, asdiscussed in Chapter 5. Samples consolidated to lower effectivestresses in the slurry mould seemed to suffer less drainageproblems, and the author considers that the 'smear' effect ismore apparent on the more heavily consolidated samples. Thesamples tested in the main test programme were all consolidatedto low effective stresses in the slurry mould before transfer tothe triaxial cell.
One sample was consolidated under end drainage by modifying thetop and bottom platens to allow porous stones and drainage linesto be fitted. This system proved to be more efficient than theoriginal radial drainage arrangement, but less efficient than the
•modified radial drainage system used in this research.
Excess pore pressures generated during consolidation weregenerally dissipated within twenty four hours. The timecalculated for 95% consolidation using values for thepermeability and for the coefficient of volume compressibilityfrom Overy (1982) (k = 2.5 x 1O-7mm/sec., m,v = 3.53 x 10-4
l
m2/kN), and the consolidation curve for radial drainage fromLambe and Whitman (1979), is of the order of 6 hrs. The radialdrains could therefore be improved further, but were judged to besufficiently good for this research and no further modificationwas attempted.
Atkinson et al (1985) carried out a series of tests on triaxialsamples to investigate the distribution of moisture in samples ofkaolin consolidated with radial drainage under various rates of
77
consolidation, and compared these with samples consolidated withend drainage under the same conditions. The samples consolidatedwith radial drainage in one step were found to be as much as 1.5%wetter in the centre than at the periphery. As the samples weresaturated this led to quite considerable non uniformities ineffective stress across the sample. Atkinson et al reported thatthe outer part of the sample started to consolidate first, and asit consolidated it became stiffer. This relieved some of the totalstress on the inner part of the sample which led to the finaluneven distribution of water content at the end of consolidation.Atkinson et al demonstrated that at rates of loading as low aspi = 10 kPa/hr then a difference in moisture content of about0.4% between the centre and the periphery would be expected fortriaxial samples of Kaolin consolidated with radial drainage.
At the end of the test series on each sample the centre third ofthe sample was sectioned and the moisture content determined atthe centre, at an average radius of approximately 27mm, and at anaverag~ radius of approximately 35.5mm. The overall radius ofthe samples varied but was generally about 40mm. Table 6.1 givesthe moisture contents determined for each sample.
There is little significant variation in moisture content acrossthe radius for each sample although the centre sections do appearto be slightly wetter in most cases. The loading rate pi in allcases was constant at approximately 3kPa/hr, which is consideredto be sufficiently low to minimise the non uniformitiesdemonstrated by Atkinson et al. It may be that the problem isless marked for anisotropically consolidated specimens as thesample is being sheared during the consolidation process. Howeverfurther work would be required to investigate any difference inbehaviour between isotropic and anisotropic consolidation.
6.3 EFFECTIVE STRESS RESPONSE
The effective stress response of the samples to a variety oftotal stress paths over the range of frequencies used isdiscussed below. The main series of such tests were the low
78
Table 6.1 Moisture content of consolidated samples
Sample Outer Inner CentreNumber Annulus Annulus
r = 35.5mm r = 27mm% % %
33/6 22.29 22.45 22.4933/12 21.47 21.37 21.3633/18 19.14 19.33 19.4965/6 20.29 20.21 20.1865/12 17.90 17.95 17.9865/18 17.51 17.68 17.70100/6 19.51 19.74 19.84100/12 16.92 17.09 16.97100/18 17.07 17.01 17.17
frequency stress path tests performed initially on each sample,see Figure 3.2, but the effective stress response has beendetermined from the results of the 1 second and 0.1 seconddeviator stress pulse tests.
6.3.1 Total stress path tests
A series of undrained total stress paths of low magnitude (in therange :t2.5 kPa to :tID kPa) and low frequency (0.05Hz and below)was applied to each sample to allow investigation of the elasticresponse at sufficiently low loading rates to allow the porepressure to be monitored accurately. The stress paths are showndiagrammatically on Figure 6.3. The stress level and frequencywere at the limits of the control system in its present state.Five samples were tested in this manner and all showed that thechange in total stress caused the pore pressure to alter suchthat the mean normal effective stress remained constant. Theresilient shear strain-deviator stress plots were similar foreach sample at any particular stress level, and showed a slighthysteresis although the area enclosed by the loop and hence theenergy absorbed by the sample, was small in all cases. Theresilient volumetric strain recorded in all cases was negligible,and indicates that Poisson's ratio was 0.5. The calibrations ofthe proximity transducers and LVOTs were shown to be valid oversuch small strains by a comparative test detailed in Appendix C.Overy (1982) demonstrated similar behaviour for normallyanisotropically consolidated samples of Keuper Marl at repeatedstress levels of the order of 50kPa or more. Overy carried outthese stress path tests after cyclic loading tests and thereforethe yield locus had been extended from its position at the end ofconsolidation by the cyclic loading, and his stress paths werewithin the new yield locus.
Figure 6.4 shows the effective stress and strainsamples 100/18 and 33/6 to a variety of total stressboth cases the cyclic total stress level applied was
response ofpaths. In
± 2.5kPa·.The fact that the mean normal effective stress remained constantduring the loading sequences and that Poisson's ratio was 0.5
80
If)If)wa::tna::~>wo
TOTAL MEAN NORMAL STRESS
NOTE
1 All stress paths are centred on the finalstresses at the end of consoIida t ion.
2 Six different stress paths were appliedto each sample.
FIGURE 6·3 DIAGRAMMATIC REPRESENTATIONOF TOTALSTRESS PATHS FORUNDRAINED rESrS AT -O·OSHz
fq
jpi1m
:> :;. > :;..p p' Es Ev
p
>p
5kPo
81
>
p I>
Ev 50 JJ.E
FIGURE 6·40 SOME STRESS PATH TEST RESULTS FROMSAMPLE 100/18
qj qi qjpIt 0
> ,> > >p P ES Ev
p
5kPa
> >ES
>Ev
Ev 50 I.I.E
FIGURE 6·4b SOME STRESS PATH TEST RESULTS FROMSAMPLE 100/18
>p
>pi
:>p
5 kPa
83 ..
>
>pi
ES 50J.1E
FIGURE 6·4c SOME STRESS PATH TEST RESULTS FROMSAMPLE 33/6
.'
84
>p
:>pi >
p
>
>p~ D
Ey :>
p
SkPa
Ey SOJ.LE
pi :>
>
>
I >P
FIGURE 6·4d SOME STRESS PATH TEST RESULTS FROMSAMPLE 33/6
.'
85
indicates that the sample is behaving in an isotropic elasticmanner.
Graham and Houlsby (1983) working on a naturally anisotropicallyconsolidated clay demonstrated that changes in shear stressaffected the volumetric strain and that changes in the meannormal effective stress affected the shear strain. This led toeffective stress paths where p' did not remain constant althoughthe effective stress path in q-p' space remained linear. Thesign of the gradient of the stress path in q-p' space indicatedwhether the sample was stiffer axially or radially. Thisbehaviour is not apparent from the results of this research orthat of Overy (1982) although the samples were carefullyanisotropically consolidated. Further work should includeexamining samples of clay consolidated in this equipment under anelectron microscope as it is possible that the samples are notconsolidated from a wet enough slurry to induce anisotropy.
6.3.2 One Second Stress Pulse Tests
These tests were performed just prior to and immediately after thepermanent strain tests, see Figure 3.2. The tests consisted of arange of undrained deviator stress pulses of 1 second duration.The peaks only were recorded, so the complete stress paths cannotbe plotted. Figure 6.5 shows the peak resilient axial strainplotted against the peak resilient radial strain for a number oftests and shows that Poisson's Ratio is approximately 0.5 withinthe experimental scatter. Figure 6.6 shows that on average thechange in pore pressure is approximately one third of the changein the deviator stress, although the value appears to increase asthe deviator stress pulse increases. This may be due to thetransducer affecting the stress distribution at the higher strainlevels. The ratio of a third between the deviator stress pulseand the change in pore pressure indicated that p' remainsconstant during this type of loading as:-
6.2
500
-400w;:j_
zce300l-ll)
_J«0200~
I-a] 100_J
lI)uiCl::
86
SAMPLE NUMBER
x 33/6 + 65/6 • 100/6
• 33/12 • 65/12 et 100 112
.33/18 .65/18 iii 100/18
v =0·5
O~~--~--~----~--~~----~----o 100 200 300 400 500
RESILIENT AXIAL STRAIN (J..lE)
FIGURE 6·5 RESILlENT AXIAL STRAIN vs RESILIENTRADIAL STRAIN FOR CONSOLIDATEDSAMPLES OF· KEUPER MARL.1 SECOND PULSE DURATION
87
SAMPLE NUMBER
W0:::~ 1.0VlW0:::Q..
~30~lJ..J~20zwuzw 10(!)z<l::r:u O~----~--~----~----~----~----o
• 100/6
.65/12 et 100/12
.33/18 11100/18
)( 33/6
10 20 30 1.0 50DEVIATOR STRESS PULSE MAGNITUDE(kPa)
FIGURE 6·6 PORE PRESSURE RESPONSETO DEVIAlORSTRESS PULSE OF 1 SECOND OJRATION(CENTRE PROBE)
88
Therefore as the cell pressure remains constant
6.3
The total mean normal stress p = i (Ov + 2or) 6.4
therefore !::'°v!::,p= -3 6.5
substituting 6q!::'p = -3- 6.6
However as !::,u= from Figure 6.6 it follows
that !::,p= !::,u 6.7
as pi = P _ U 6.8
therefore as !::,pl= !::,p- !::,u 6.9
it follows that !::,pl= 0
6.3.3 One Tenth Second Stress Pulse Tests
These tests were carried out at a variety of undrained pulseddeviator stress magnitudes of 0.1 second duration with a restperiod of approximately 0.25 seconds. The early tests wererecorded on an ultra violet chart recorder and consequently only
89
the peak valves were available. Following the introduction of theanalogue to digital converter and the computer data acquisitionsystem it was possible to record the full pulse and consequentlythe effective stress path resulting from the applied loading.
Figure 6.7 shows some of the effective stress paths recorded anddemonstrated that pi remained essentially constant during thesetests. The centre pore pressure transducer was shown to becapable of following stress pulses of this duration bysupplementary tests detailed in Appendix C.
Overy (1982) reported a loss in response from the base porepressure transducer at a maximum frequency of 0.005 Hz, and atapproximately O.lHz for the centre pore pressure transducer inmost cases. Samples exhibiting this sort of response would beunlikely to show a B value of 1 in the initial saturation testsdescribed in Chapter 5 and would have been rejected under theconditions applied during this research.
The pulsed deviator stress resilient axial strain paths for anumber of deviator stress pulse magnitudes for samples 33/6 and100/12 are shown in Figures 6.8, and 6.9. Figure 6.10a shows thestress softening behaviour of the material evident at the startof the tests, and Figure 6.10b shows the effect of 10,000 pulseson the shape of the stress strain paths. The resilient modulusseems to reduce with increasing number of stress applications.Figures 6.8 and 6.9 shows the effect of numbers of cycles on thestress strain paths in more detail for samples 33/6 and 100/12respectively.
It was found that between 5 and 15 deviator stress pulses wereapplied to the samples before the correct magnitude was achieved.This depended to a large extent on the magnitude of the deviatorstress pulse. Therefore the first full pulse recorded by thecomputer was not the first pulse applied to the sample and somechange in the shape of the hysteresis loop may have alreadyoccurred.
90
100
20
100
80
CT 40<]
_60oo,~
~ 40
-~
20
20~p' (kPo)
Sample 100/12qr = 100 kPaN = ...10
0-t-"5'----,---o 20
~p' (kPa)Sample 100112qr = 100 kPaN = 100 000
60
CT<]20
O-f-llt-----,r-----o al
~p'( kPa)Sample 100/12q, = 65 kPaN = 100 000
20boP' (kPa)
Sample 100/12qr = 65 kPaN =-1)
FIGURE 6·7 EFFECTIVE STRESS PATHS RESULTINGFROM PULSED IEVIATOR STRESSES AT0·1 SECOND OORATION
~l .,
40N= ...10 100(0)
20-&.:::;t.-I-0;0
-er20
0 00 3 0 600 0 1000 r ( 2000E6 (J..LE) EO J..LE)
80'
_50 N=...10 & N=...10200 000 1000000 .x
o,.::t:. er- 40I-er25
O~----~----~~-o 2(0) r 4000e:a(J..LE)
100
I-er50
-
O~----~----II---o :000 r( )10000
EO J..LE
EJGUREse QEVlATOR STRESS PULSE RESlUENT _AXIALSTRAIN PATHS FOR SAMPLE 33/6
92 ..
30 60- N=..10 100000o -a.. N=..10 50000 o.Y. a..- .Y.
L-sr L-
er15 30
O~----~------~--o 1000 2000e:~ (ue)
Of-------~----~--o 2500e:6( us)
5000
90
(XX)_180oa...Y.-
o 0 3000 6CXX)e:~(Jle:)
RGURE 6·9 DEVIATOR STRESS PULSE RESILIENT AXIALSTRAIN PATHS FOR SAMPLE 100/12
"93
N= 10,00080 80
oo,.::L.
cn,.::L.
~er 40
~er 40
O~----~----~-----o 5000 10000
E6 (~E)
O+-----~--------o 5000 10000E~ (~E)
FIGURE 6·100 EFFECT OF CEVIATOR STRESS PULSEMAGNITUDE AND NUMBER OF CYCLESON THE STRESSSTRAIN RESPONSEOF SAMPLE 33/6
N = lQOOO200 200
~
er 100~
er 100
oo,~
-oo,~
o 3000 6000£~ (~E)
o 3cxx) 6000E~ (~E)
FIGURE6·10b EFFECT OF DEVIATOR STRESS PULSEMAGNITUDE AND NUMBER OF CYCLESON THE STRESS STRAIN RESPONSEOF SAMPLE 100/12
..94
There is an indication however from Figures 6.8 and 6.9 that theloops do degenerate with increasing number of cycles, with thetop of the loop becoming similar in shape to the end of the Sshape hysteresis loops generally found by other researchers, anddiscussed more fully in Chapter 2. The S shaped loops referredto are the result of two way loading whereas the above testswere all one way.
At low repeated deviator stress levels the hysteresis loops inFigures 6.8 and 6.9 are closed and do not appear to change withfrequency, indicating that the loading is having little effect onthe sample, which is therefore behaving elastically. As therepeated deviator stress increases then the loops tend to getwider. The area enclosed by the hysteresis loop is a measure ofthe energy absorbed by the sample during each stress pulse, andit follows that as the magnitude of the stress pulse increases,and therefore the magnitude of resilient strain, then the workdone on the sample must also increase.
6.4 RESILIENT RESPONSE
The main bulk of the data comes from the tests carried out with apulsed loading time of 1 second as the stress levels weresufficiently high to produce reasonable resilient strainswithout causing any permanent deformation. The data from thesetests is presented first and the data from the remaining testswith shorter and longer stress pulse durations is then presentedand compared with the main results.
6.4.1 One Second Stress Pulse Tests
All the results from the tests showed a stress softening effectwith increasing deviator stress pulse magnitude. Figure 6.11shows a typical result for a single test. Each point representsthe resilient axial strain measured at that value of pulseddeviator stress and was determined over 3 or 4 cycles. Figure6.11 demonstrates that if the sample was unloaded then thematerial was softer than it was on the initial loading sequence,
95'
_60~ x Loading~ 50 e Unloadingw~~ 40(/)(/)
~ 30tna:: 20~«> 10wo O'~--~--~~--~--~--~--~--~~--~--~---
o 100 200 300 400 500 600 700 BOO 900 1000 1100 1200RESILIENT AXIAL STRAIN (~E)
FIGURE 6·11 TYPICAL DEVIATOR STRESS PULSE vs RESILIENT SHEARSTRAIN PLOT FOR SATURATED KEUPER MARL.1 SECONDLOADING
)( A-ior to permanent strain test• Just after permanent strain test
~ 50 + After rest period....J
~~40w~30~~20>~ 10
O~--~~--~--.--.--'---r--r--~-.o 100 200 300 400 500 600 700 800 900 1000RESILIENT AXIAL STRAIN(~E)
AGURE 6.12 THE EFFECT OF LOADING AND REST PERIODS ON THESTRESS STRAIN RESPONSE OF KElPER MARL.1 SECOND LOADING.SAMPLE 100/18.
96
although there was no measurable change in pore pressure or anydevelopment of permanent strain.
Figure 6.12 shows a typical comparison between the results ofthese tests immediately' prior to and immediately following apermanent strain test. The samples were found to be a lot softerafter the permanent strain test even if there was negligible porepressure development, as was the case for most of the permanentstrain tests. The permanent strain tests are discussed morefully in Section 6.5. A comparison between typical stress-straincurves measured just prior to each permanent strain test on aparticular sample is shown in Figure 6.13 for sample 100/18 anddemonstrates that the sample recovers its original stiffnessduring the rest period between the permanent strain tests. Thiswas found to be the case for all the samples. As there was verylittle excess pore pressure developed during the permanent straintests there was consequently very little volume change during thedrained rest periods and therefore no significant change in thesample's stress state. It is thought that this softening andstiffening is due to a thixotropic type effect and the samplerecovered its original stiffness because it was allowed to restwithout being loaded rather than because it was allowed to drain.
The stress strain curves from the initial 1 second deviatorstress pulse tests on all the samples were used to infer therequired deviator stress pulse magnitudes to give resilient axialstrains of 100, 200, 300, 400, and 500 microstrain in each case.Figure 6.14 shows these values of pulsed deviator stress plottedagainst the mean normal effective stress at the end ofconsolidation. The curves for the strain contours arefit lines drawn through the points and the origin.
the bestTable 6.2
gives the intercept for the curves if the origin is not specifiedand demonstrates that drawing the strain contours through theorlgln is a reasonable assumption. The contours get closertogether as the magnitude of the resilient shear strain increaseswhich is the result of the non-linearity in the stress strainresponse evident in Figure 6.13.
••
oz(V)
ozx + • • • 0
gi'
•
tooZ
0:::WQ.
~co-0-w8w t- -
rJ~~wz :::::> _.- t- Q.
~ ~ ~tn u, U')
~ «_J 0 r:~U')
~ z ~~zU')-w«0:::0:::
t-ZU')«t-~zU')WU')zU')«w~~ffiU')Q.
§.-
ooO'l
8N
oo.-
oto
r- -r------r-----,------,----------~~Ooo.-o 0 0 0U'> .-...3 (V) N
(Od>t) 3S1nd SS3~lS ~Ol'tIA3a
::r:u«woI-
0:::o0:::Q.
t-U'):::::>-,
.,98
Lt)
0 00
0 0 8 .Wo 0
~0
::l...- N ~ Lt)
LV)W • + X • •
0~..-
~0Lt) .
"7 a~ UJ0:::, . J ::>b (!)u:::
99
Table 6.2 Intercept of strain contours for consolidated KeuperMarl under 1 second pulsed loading
Res il ient Intercept on p~ lp'shear strain at qr/p' = 0 e(microstrain) e
100 0200 -.004300 -.007400 -.007500 -.004
ion
Figure 6.15 shows the gradient of the strain contours plottedagainst the magnitude of the resilient shear strain. The curveis reasonably linear and leads to the relationship
6.10
where A and B are material constants and are equal to 1110 and1.552 respectively with the stresses in kPa and the strains inmicrostrain.
Equation 6.10 does not include the ambient deviator stress. Theambient deviator stress varied considerably for these samples andwas positive in some cases. This was caused by the method ofconsolidation as discussed in section 6.2. However the stresspaths followed by these tests are within the yield locus forthese samples, and have been shown to be elastic. The ambientdeviator stress is also a small proportion of the yield stressfor the range of stress conditions chosen for the test programme,and therefore the ambient deviator stress is unlikely to havemuch effect.
An indication of the accuracy of the model, or the scatter of theresults, is shown in Figure 6.16 where calculated values of theresilient strain are shown against the measured values.Approximately 90% of the results are accurate to ~ 20%, whichconsidering the type of test and the magnitude of the strains, isconsidered to be quite reasonable.
6.4.2 One Tenth Second Stress Pulse Tests
These tests were carried out with a constant deviator stress pulsemagnitude of 0.1 second duration and were intended to investigatethe development of permanent deformation and effective stresschange with number of cycles .. The resilient response was alsomonitored during these tests which covered a range of differentstress levels and the stress softening behaviour was again
lor
w:::1.
·3 ·4 ·5 ·6 .7·8·9 10qr IPo'
..._Zl1J...J
tD200a::
10~1
FIGURE 6·15 MODEL DEVELOPMENT FOR SATURATED KEUPERMARL SAMPLES UNDER 1 SECOND LOADING
_ 400w:::1.-olJ.J
~ 300(f)
ti5~
~,$ 200
..102
" 33/6• 33/128 33/18
+ ffi/6A 65/12o 65118
* 1)0/6o 100/12x 100/18
500:!:200/o/
x " ••
/
100
/~ (j,l •
/
O~--~----~----.----.----,--o 100 200 400 500
EQUALITY
:!:200/0r·
FIGURE 6.16 RESILIENT SHEAR STRAIN CALCULATEDvs RESILIENT SHEAR STRAIN MEASURED
300q CALCULATED (!lE)
..103
apparent. Figure 6.17 shows the pulsed deviator stressresilient axial strain response for sample 100/12, which had thelargest number of different stress levels applied to it. Therewas insufficient data from the different samples to draw anystrain contours but the results are shown in Figure 6.18 for allthe tests as resilient axial strain against the deviator stresspulse magnitude divided by the mean normal effective stress. Theshear strain has not been plotted as there was some doubt as tothe accuracy of the radial strain measurements for this type oftest. Figure 6.18 indicates that there is a linear relationship,although there is quite a scatter of results, of the form
ca" = c [~~ r 6.11
where C and 0 are material constants and are equal to 2000 and1.62 respectively. The strain is given in microstrain.
The model is of the same form as that derived from the testswith a 1 second pulse length, but the material constants aredifferent. One of the main factors which could influence thematerial constants is likely to be the stress pulse durationwhich controls the rate of loading.
6.4.3 Total stress path tests
These tests were carried out over a small range of low stresspulse magnitudes and low frequencies. The samples did not showany sign of stress softening, which consequently gave a constantresilient modulus for each sample tested as shown in Figure6.19. It is unclear whether this linearity is due to the lowstress pulse magnitude, or the low test frequency, or acombination of both. It is felt that the effect is due to acombination of low frequency and low stress level, and that if thestress levels had been increased whilst maintaining thatfrequency or the frequency increased whilst maintaining thosestress levels then the stress softening effect would have been
104.
______~--------~~----------~~Ooo 0o 0N ..-
(od>1) 3S1nd SS3~lS ~Ol'1IA3a
oR§le>
~ N..--ZWO(/')0
~
e::{ ....J ..-et: :::> uitnCl.~
~~
~~tno:::o,_o:::lL.
8zozUJ~-0--«N~>CX:~ ~ ti1et: e-
g~OxMe::{
§..-
105
Sample10000 o 33/6
)( 33/12+ 33/18
5000 o 65/6.100/6
- • 100112 •w::l.
• 100/18-z2000<Cl::t-V)
_J1000<x<t- 500zUJ_JV)UJer
200 •100+-----.------.-----.-----.------,---~
.1 ·2 ·5 1q,.lp~
2 5 10
RGURE 6·18 RESILIENT SHEAR STRAIN RESULTS FROM0·1 SECOND STRESSPULSE TESTS
106
200
• • •- •oa.~-150(/) - +::>_J
::>00~.....100zUJ-_J(/) ~ X )(UJCl:: 50
OT-------.-------.-------~------T_-o 2·5 5 7·5DEVIATOR STRESS PULSE (kPa)
10
SAMPLE• 33/12
X 33/6
+100/18
.100/12
FIGURE 6·19 VARIATION OF RESILIENT MODULUSWITH STRESS PULSE MAGNITUDE-LOW FREQUENCY PULSES
107
apparent.
Figure 6.20 shows the results from the low frequency testsplotted as the resilient shear strain against the deviator stresspulse magnitude divided by the mean normal effective pressure.The results show good correlation and lead to an equation of theform
6.12
where E and F are material constants and equal 690 and 1.06respectively. The strain is given in microstrain.
6.4.4 Comparison with previous research at Nottingham
The earlier soil model was of the form
6.13
It was developed from the results of the repeated load triaxialtests of Hyde (1974) and is described by Brown et al (1975).Hyde (1974) also presents some results attributed to Lashine(1971) and Figure 6.21 shows those results and those of Hydereplotted in the form used in this thesis. Hyde's results aremeasured after 105 cycles. It is:apparent from Figure 6.21 thatthere is a reasonable line through all the points for each cellpressure, but that dividing by the mean normal effective stressdoes not result in a unique relationshi~ . Lashine and Hydeboth used an external pore pressure transducer and an externalLVDT, and Lashine also used an external load cell. The externalLVDT may have caused some errors in the measurement of theresilient strain, especially after 105 cycles when the sample maybe behaving in a non uniform manner, due to the development of
w::i-200z~ex:t-V) 1000::-cUJ:::c(/)
t- 50zUJ_J
(/)tU0:: 20
1000
500
IDS'
Sample• 33/6)( 33/12o 100/6A 100/188133/12
10+-----r------.-----r----1I------~----·01 ·02 ·05 ·1 ·2 ·5 1
FIGURE 6·20 RESILIENT SHEAR STRAIN RESULTS FROM20 SEOOND STRESS AJLSE TESTS
109
AFTER LASHINE (1971)10000 0; (KPo)
w )( 302- 5000z • 41
« + 79~ 2000 -191V')
_J
« 1000~
~ 500zLU_J
V')LU 200Cl::
1000·1 0.2 0·5 1 2 5 10
qr,p'01CXX>0
w::l.5000-z«Cl:: 2000In_J
« 1000~~m 500_J
V')
~ 200
AFTER HYDE (1974)0: (KPo).220+ 120)( 40G 33
100+-~~---'--1I--'---~--~0·1 0·2' 0·5 1 2 5 10
qrlt 'PoFIGURE 6·21 RESILIENT BEHAVIClJR OF KEUPER
MARL AFTER LASHINE· (1971) ANDHYDE (1914),
110
permanent strain and the end effects caused by friction on theplatens. There will be some errors introduced into Lashine'sresults from friction on the bearing through the cell top.
It is considered by the author that the difference in behaviour isunlikely to be due to the fact that Hyde's and Lashine's sampleswere isotropically consolidated, as the results indicate that theanisotropically consolidated samples were behaving istropicallyand that the resilient behaviour is independent of moisturecontent. Overy (1982) also found that the stress paths tofailure for isotropically overconsolidated and anisotropicallyoverconsolidated samples were similar. The differences thereforeare more likely to be due to the different, and more accurate,measuring systems used in this research.
6.4.5 Discussion of resilient model
A simple model of the form
€ rs ; A [~]
B6.14
has been shown to fit the experimental results from the differenttests on the overconsolidated saturated Keuper Marl samples.Stress pulses of three different durations were used with arange of different stress pulse magnitudes for each pulse length.The values of A and B were found to differ for each of the threestress pulse lengths. Table 6.3 gives the values of A and Btogether with the length of the stress pulse and the variation indeviator stress pulse magnitude divided by the mean normaleffective stress. There is some overlap in the stress pulsemagnitudes for each test as expressed in the form qr/P~ and thevalues of A and B were found to increase as the duration of thestress pulse decreases. The results show that as the duration ofa particular stress pulse decreases i.e. the rate of loadingincreases, then the stiffness of the sample increases. This wasas expected and is in agreement with the findings of other
"
111
Table 6.3 Model details from tests on overeonsolidated KeuperMarlModel form
B(mierostrain)
Duration of qr/p~ A Bstress pulse min max
sees value value
20 0.04 0.36 690 1.0611 0.2 0.57 1110 1.522
0.1 0.2 2.5 2000 1.62
112
researchers. The soil has been shown to be behaving elastically,certainly in the range of stress pulse magnitudes that arecovered for all three frequencies, and therefore the variation inA and B is likely to be due to the change in frequency ratherthan the change in stress pulse magnitude. However a plot of Aor B against frequency shows no apparent simple relationship.
6.5 SAMPLE BEHAVIOUR UNDER REPEATED LOADING
6.5.1 Introduction
The development of permanent strain and change in pore pressureunder large numbers of constant amplitude pulsed loads of 0.1second duration was investigated in this section of the testprogramme. Generally each test was allowed to run until it wasjudged that equilibrium had been attained or the sample hadfailed. An average of approximately 400,000 stress pulses wereapplied to each sample at each pulsed stress level.
The response of the sample to the loading sequence was difficultto define accurately as the equipment was not capable ofmaintaining the constant ambient deviator stress required duringthe test period, although the cyclic component was in factreasonably constant except at high amplitudes. It is thoughtthat the variation in deviator stress was caused by two separateeffects. The first was a drift in the control system over aperiod of time which caused slow changes in the deviator stresslevel. The second was sudden pulsed loads caused by electricalinterference which affected the control system. Theinterference could have been airborne or in the mains supply andwould have been caused by other items of electrical equipment inthe laboratory switching on and off. This had the effect ofapplying an unknown and uncontrolled stress pulse to the sample,after 'which the system would stabilise close to the originalstress conditions, although it may have caused significantstrains and pore pressures to develop.
113
6.5.2 Development of Permanent Strain
As described previously the change in ambient deviator stressaffected the strains recorded during each test and typicalresults of axial strain against the logarithm of the number ofcycles is shown on Figure 6.22. Also shown is the change inambient deviator stress and it is apparent that there is goodcorrelation between the peaks and troughs in the stress andstrain curves.
Figures 6.23 and 6.24 show the results of all the permanentstrain tests on samples 100/6 and 100/18 respectively, andindicate that the lower repeated stress levels have little effecton the sample, but that above a certain level the sample beginsto deform. It is interesting to note that sample 100/6 showed acontinuous build up of permanent strain which then accelerated tofailure, while sample 100/18 showed very little development ofpermanent strain initially, and then failed suddenly. Theresults generally tend to support the proposition of a thresholdstress level, loading above which causes the sample to fail by abuild up of permanent deformation.
Figure 6.25 shows the permanent strain against the logarithm ofthe number of cycles for the highest stress level applied to eachof the samples tested under these conditions. Apart from sample100/18 there is a gradual build up of strain with numbers ofcycles, although there is evidence of drift and interference inthe load control system.
Figure 6.26 shows the effective stress paths for the tes~s shownon Figure 6.25. The paths are based on the peak pore pressurerecorded by the centre probe from each of the stress levelsapplied to that sample at that frequency. The base pore pressureprobe was unable to follow the stress pulse at these frequencies,see Appendix C. Samples 33/18, 100/6 and 33/6 failed under therepeated loading with a reasonably steady build up of strain, therate of which then increased to failure. Sample 100/12 did notfail and sample 100/18 failed quite suddenly after a relatively
114
(Od>1) SS3~lS ~Ol'VIA3a :>11'V15 NI 3DN'VH:>oC"l
oN
o .0 o-I-~ ~ ~ -+ -L~ W::I:....
IJ)
WIII -..3' _J
III U UC11 tL- ~1;) ~L-
e.!) Vl0 9- l1J0
~III ::I:> .....
'0 C"lZZ ~U e.!)« Z- 00:: 00
c 1A-Jti; fi
0 c .... a:L- Z
~1;)~
LU
"8 c N ~ LU
~0 ~ J:.J:; .....U
0::
X • ~ u..0
_J
G«u UJCC H:.-~ UJ
o8N
ooLf).-
o8
ooIJ)-(311) N 1'V~lS 1'V1X'V IN3N'VW~3d
115
~~
fil~:::>..Jo~I-o
LLMo IIL-
er
u.oer:~2:::>z
("t)NcD
0 UJLO 0 LO If) 0
N N.- .- er:
(311) OOOP' NIV1:I1S TvIX'V lN3N'tW~3d ~u:
KJ 0::Ja._.:::!~o(Y')15-
0-~o~
o-..t.9 It.
CT
& LL
.:::!IJ') et:r-.... UJ-II ~
L-CT =>Z
~3:
lfb
N~S 1\f1X't lN3~3d
(Y')z
Cl)UJ..Ja..~
~
0LU
~0::::::>~Cl)
0:::0U.
0:::UJCD~:::>zUJ..JUt>
z <!)
9V)
~ ~ & & e >
~ ~ .x .x .x Z
0 CO 0 -..t s «CO lO
(1) ~0:::
~ ~ ~ li;" II II II IIL- L- L- L- L-
0" 0" C1' C1' C1' ..J«CO N ex>
X
lO ...- ...- so ...-«
..J- - - - - ~0:::
('t') ('t') 0 0 0 «('t') ('t') 0 0 0 ~
...- ...- ...- ~
01 .Qlw 0:::
01 .Ql .Ql ~tu
a.. a. Cl. a. a. a..E E E E E 9 :::>
~ ~0
~ ~tu
Cl)UJ ~
+ <l>
• X 0 ~lJ..0
IJ')NCDW0:::s
0 u:0 0 0 0 0 0IJ') ~ (Y) N ...-
(311 )OOO~~ 3
IH3
eo N eocD ..- ..- co ..-- - - - -M M 0 0 0M M 0 0 0 ..(II
..- ..- ..- ._a.a.
~ Q.I ~ ~ ~a. 15.. a. a. a. cD
E E E E E~ ~
0~ ~(j)
• )( + • 0Lf')
........................ ...,t............ -......
q..-
Cl::0LL
(j)_,
I Cl::
~<t
a..: ~
(j) Cl::(j) UJUJ a..Cl:: ::::>t;; UJ
~
WLL0>
I- (j)
b:J ~LL a..LL ~UJ
~
0 0llJ(j) UJ_,
~<t Cl::~ ::::>Cl::
~0Z (/)
~~cD
UJa::::::>(!)
LL..-I
119
small build up of strain. This may have been caused byinterference. Figure 6.26 suggests that there is a thresholdstress which can be expressed as a single line on the normalisedplot, and that effective stress paths crossing this line arelikely to cause excessive deformation and eventually failure, aswas the case for samples 33/18, 100/6 and 33/6. Effective stresspaths approaching this line are likely to cause significantdeformation, as sample 100/12.
The mean normal effective stress paths tend to become curved asthey approach the failure level, although it is not clear if thisis a true measurement, or if it is due to inaccuracies ofmeasurement when the sample deformations become large and thepulsing of the load ram affects the cell pressure. At largedeformations it is also unlikely that the sample is behavinguniformly.
6.5.3 Development of pore pressure
The pore pressure was monitored during the tests using the baseand centre pore pressure probes. The lack of any significantpore pressure development during any of the tests except those atthe highest stress level corresponds to the lack of any permanentstrain development and supports the supposition that the samplesare generally behaving elastically. There was no significantvolume change recorded in the drained rest periods between testsat the lower stress levels, which is as expected as there wasnegligible pore pressure development. The volume changemeasuring system could resolve to 0.00 cm3• There was noevidence that the sample states progressed to equilibriumpositions determined from the previous stress history as reportedby Sangrey et al (1969) and discussed in Chapter 2.
Figure 6.27 shows the change in ambient pore pressure asmeasured by the centre probe for the samples tested with thehighest repeated deviator stress. The tests are the same asthose shown in Figure 6.25. The general trend is for a negativeexcess pore pressure to develop, although the magnitudes recorded
1io
SAMPLE• 33/6x 33/186100/6+100/12
FIGURE 6·27 PORE PRESSURE DEVELOPMENT vsNUMBER OF LOAD APPLICATIONS FORCONSOLIDATED KEUPER MARL SAMPLES.CENTRE PROBE
-d!. SAMPLE.x 100- 33/18ur xer 80 6 100/6:::> + 100/12l/)l/)
60 o 100/18wera..ur 40er0 20a..z 0 2urC!) LOG Nz -20<t:::r:u -40
FIGURE 6·28 PORE PRESSt,RE DEVELOPMENT vsNUMBER OF LOAD APPLICATIONS FORCONSOLIDATED KEUPER MARL SAMPLESBASE PROBE
121
and the rate of development are all very variable. The porepressure also changed as the ambient deviator stress changedwhich masked the behaviour under cyclic loading. The change inboth axial strain and pore pressure for sample 33/6 shown inFigures 6.25 and 6.27 respectively are examples of this.Unfortunately the centre pore pressure probe had failed in sample100/18 and the change in the centre pore pressure just prior tothe sudden failure, see Figure 6.25, was not recorded.
Figure 6.28 shows the change in ambient pore pressure as measuredby the base probe for the same tests as shown in Figure 6.27.All samples showed an initial positive excess pressure, which isopposite to that recorded by the centre probe, which eithercontinued to build up or decreased and became negative, dependingon the sample. All the samples were overconsolidatedsufficiently to be on the dry side of critical where a negativeexcess pore pressure would be expected on shearing. Figure 6.28shows that one of the samples with the highest overconsolidationratio, sample 100/18, in fact recorded the highest positiveexcess pore pressure. Unfortunately the centre pore pressuretransducer in this sample was not working. It is the opinion ofthe author that the base probe is affected considerably by samplenon uniformity caused by the end restraint when the sample beginsto deform, and that the centre probe gives a more accuratemeasure of the pore pressure. However overconsolidated samplesdevelop negative excess pore pressure when sheared which tends tocause non uniformity in the sample generally and leads todiscrete failure planes, in which case a centre pore pressureprobe is unlikely to provide a true measure of the pore pressureas the sample fails.
6.5.4 Variation in Resilient Modulus with numbers of Cycles
Generally the resilient modulus was found to remain fairlyconstant with number of stress applications for low leveldeviator stress pulses, and to decrease with increasing number ofcycles as the deviator stress pulse magnitude increased as shownon Figure 6.29. It is interesting to note that the greatest
501'22
00 1 2 3LOGN4 5 6
FIGURE 6·290 VARIATION OF RESIL.IE~I MOOUL.USWITH NUMBER OF CYCLES FffiSAMPLE 33/6
qr= 40 kPa60
• •qr= 50qr= 70
(/)::::>_J
§40z~30UJ_J -80(/)
r-UJ 20er
r= 100
10
q, =170.· ..6 6 •00 1 2 3 4 5 6 7
LOGNFIGURE 6.29b- VARIATION Of RESILIENT MQ()JWS
WITH NLMBER Of CYCLES FORSt\MPLE 100/18
reduction in resilient modulus occurred for the midrange deviatorstress pulse magnitudes, and that the resilient modulus remainedfairly constant during the tests at the deviator stress pulsemagnitude which caused failure for samples 33/6 and 100/18.
The rate of change of deviator stress with number of cyclesappears to be reasonably constant for a particular value ofdeviator stress pulse magnitude, and these results do not suggestany relationship between this rate and the magnitude of thedeviator stress pulse. Hyde (1974) reported some results fromsimilar tests on isotropically overconsolidated samples of KeuperMarl. Hyde used a smaller range of repeated deviator stresses,which were generally higher than those used in this research.Hyde demonstrated that the modulus decreased with increasingstress pulse magnitude, and that those samples with the smallerrepeated deviator stress showed the greatest reduction inresilient modulus with increasing number of cycles.
The effects described above may be due to the thixotropy of thematerial, as it is the energy input from the external appliedloading which breaks down the thixotropic bonds. Under lowexternal loads the thixotropic bonds remain essentially intactand the modulus therefore remains constant. As the externalloading increases in magnitude the thixotropic bonds begin to bedisrupted and the sample becomes softer. As the magnitude of theloading increases the number of stress applications required tobreak the majority of the bonds reduces until at high externalloads, i.e. those approaching the failure level, the bonds areall broken in the first few cycles.
The first few cycles for each repeated stress levelshown in Figure 6.29 as the applied stress leveladjusted and any change in thixotropic behaviour wasthe nonlinearity of the stress strain response.
are notwas beingmasked by
124
6.6 UNDRAINED STRENGTH
Three samples were loaded to failure at a constant rate ofdeformation. The tests were undrained and a rate of deformationof O.lmm/min was used. Figure 6.30 shows the deviatorstress-axial strain plot and the excess pore pressure axialstrain plot. The deviator stress plots do not show the peak andresidual strength normally associated with overconsolidatedmaterial. The curves show a general shape of an initial stiffsection followed by an approximately linear, much less stiffsection, followed by a flat section at failure where the deviatorstress no longer increases with increasing strain. It is thoughtthat the initial stiff section is where the stress path is withinthe yield surface defined by the Hvorslev surface, and that thefirst change in behaviour is when the effective stress pathreaches the Hvorslev surface and curves to follow that surface.The second change in behaviour, then, would be when the samplereaches the critical state at the intersection of the Hvorslevand Roscoe surfaces.
Figure 6.31 shows the stress paths plotted in normalised stressspace with the stresses indicated at which the changes inbehaviour occurred. The curves show a similar shape but there issome variation in the failure points and the curves do not followa single line up to failure. It is thought that this is due tothe non uniformity present in overconsolidated samples once yieldhas occurred. The initial yield points marked on the curve seema little low to coincide with the Hvorslev surface. AlthoughHyde (1974), Overy (1982) and Hyde and Ward (1985) were allworking with Keuper Marl the various material parameters quoteddiffer and so does the location of the critical state point innomalised q-p' stress pace. The material used by Hyde and Wardis closest in terms of the Atterberg limits and the clay content,and Figure 6.31 shows the Hvorslev surface drawn from theirfailure tests. It is apparent that the results agree quitereasonably though all the results tend to show a scatter andsample failure before the critical state is attained. Hyde andWard plot the critical state at a value of the normalised mean
.,125
• Sample 65/6• Sample 65/126 Sample 100/12
500
(/)(/)
~300(/)
~200~
~ 100
O·~------'-------'------'-------.-------r~=2CXXXl
AXIAL STRAIN (~E)
-20-tua..== -40
• Sample 6516• Sample 651126 Sample 100/12
40
000 60000AXIAL STRAIN (~E)
000 1 00
UJS -60~~ -80
~-100~-120z
~-14zet:I:u-16
FIGURE 6·30 DEVIATOR STRESS AND EXCESS PORE PRESSUREvs AXIAL STRAIN FOR UNDRAINED STRENGTH TESTS
·,126
·7
HVORSLEV SURFACE AFTERAND WARD (1985)
·6
·5
·4
~.3er
·2
)( •)( A
" A •)(
)C AI A)( A •
If)( A
·1
A Sample 100/12"CHANGE IN SLOPE OF q-Ea PLOT.SEE FIG 6·30
FIGURE 6·31 NORMALISED STRESS PATHS FORUNDRAINED STRENGTH PLOTS
127
normal stress of 0.77, and a value of 0.83 for the normaliseddeviator stress. None of their failure tests on overconsolidatedsamples attained the critical state. This is due to the formationof discrete failure planes as the sample dilates on shearing andtherefore the transducers are unlikely to measure the truefailure conditions on the shear plane. Figure 6.32 shows the Mohrcircles for the samples which show good agreement with aneffective angle of friction of 28.5°, and no effective cohesion.
6.7 EQUIPMENT PERFORMANCE
6.7.1 The Consolidation Control System
A number of samples were failed during the consolidation phase bybreakdowns in the consolidation control unit. As discussed inChapter 4 the control unit and servo system were modified andimproved as far as possible without completely redesigning thecontrol system, and the bellofram piston system supplying theaxial consolidation load. This was originally designed by Overy(1982), and was modified to supply the higher axial loadsrequired to produce heavily overconsolidated samples.
One major problem with the system was that there was stillinterference between the circuits for each cell. The sixproximity transducers were wired with one acting as a masteroscillator driving five slave oscillators. It was found that ifone transducer was moved over its full range then it affected theoutputs from the other five. This was only of the order of a fewmillivolts in a full range output of lv, but it was sufficient tosignificantly disrupt the servo systems which were operating.Unplugging a sensor from a proximity transducer to install it on atriaxial cell for example also affected the outputs of the others.
The bellofram units incorporated into the cell top by Overy(1982) to apply axial load to the sample during the consolidationphase also proved troublesome. The modified clamping systembetween the piston and the load ram was not easy to operate andstill proved troublesome at high consolidation loads. There was
128
0(/)
LO W
CO.....J
NCl.
II::E
&<.((/)
8 .....J
0 co Cl::
II<.(
... ::Eu Cl::
0 wR Cl.
::>w~
0 0
~w~0.....J
~
0(/)z8- Cl::010 Er00--...t=... (/)
0 W.....J
N 0 U~ N ~
Cl::- ~ ~0- - UOLll LI'\~~ -o (/)-<II <II <II Cl::- - -a. a. a. :r:e e e 0tU tU
~::E
V') V')
4 X • Ntr:'<D
WCl::{5u:
oo~
ooN
ooLO
ooM
oo-(ect)!) SS3~lS MV3HS
d
129
insufficient travel in the bellofram unit to accommodate thesample movement during consolidation, which necessitatedunclamping and repositioning the piston during consolidation. Theunit also contained a small bellofram seal against the cellpressure, and consequently any significant movement of the loadram caused a change in cell pressure.
The concept of the anisotropic consolidation control system hasbeen shown to work well for the consolidation phase but requiressome modification to provide accurately controlled anisotropicoverconsolidation. Significant shortcomings in the consolidationcontrol equipment have been found.
6.7.2 Data Monitoring System
The digital recording system fitted to the equipment part of theway through the research project proved to be accurate andrepeatable. There were almost no spurious signals recorded andno trouble was experienced from electrical interference. Thesystem extended the range of the test facility considerably asstress paths could be plotted over a much higher frequency rangethan with conventional X-V plotters. The signal conditioningunit and ultra violet chart recorder also worked satisfactorilyfollowing the modifications to reduce the electrical noise onthe signals. The system is limited though as only peak valuescan be determined from the traces.
Both systems could resolve axial and radial strains to fivemicrostrain.
6.7.3 Instrumentation
Generally the electrical transducers worked satisfactorily. Thelatest disc load cells, the pressure transducers and the LVDTswere all found to be sufficiently stable for this type of test.The crosstalk between the proximity transducers has beendiscussed above, and in addition they were found to needrecali~ating. over their linear range at the end of each test.
130
The volume change system using burettes with dyed water paraffininterfaces were found to be temperamental and required frequentcleaning. It was difficult to achieve a water tight seal eachtime the system was dismantled for cleaning. The meniscus wasnot always even and this led to inaccuracies in the volume changemeasurements.
6.7.4 The Control System
Once the hydraulic system and servo valves were overhauled thesystem proved adequate for short term tests where the load wascontinually monitored and could be frequently adjusted. Asdiscussed earlier in Chapter 6 both the static and cycliccomponents of the axial load drifted in the long term, which madeanalysis of the long term tests rather difficult.
The control system was capable of following a haversine loadsignal of 0.1 second duration where the sample was stiff andconsequently the deflection was small. With higher load signalsand deflections the unload part of the pulse degenerated as shownin Figure 6.33 for samples 33/6 and 100/12. The pulse shape onlydegenerates so far and further increases in load do not thencause any more degeneration of the load pulse shape, as shown bycomparing the pulse shapes for repeated loads of 100kPa and abovefor sample 100/12, and 50kPa and above for sample 33/6. Thisdegeneration was not noticeable when the stress pulse duration was1 second or lo~gel partly because the rate of loading was slowerand also because the magnitudes of the load pulses were smaller.
The lack of response at the higher amplitudes and frequencies ismaybe a consequence of using a low flow rate servo valve. Lowflow rate servo valves were originally fitted to try and improvethe stability of the system, and therefore a balance must bestruck between stability and response at the higher frequenciesand amplitudes which may be required during a test programme.
O+JL--r---..JL., O+S----r--~.2 ·4 0 .2 ·4 0TIME (SECS)SAMPLE 33/6
50
131..
35 50 60
35
·2 ·4
65
180
o O~--~.2 ·4 0 ·2 ·4 0 .2TIME (SECS)
SAMPLE 100/12 (CONTIt-IJED)
FIClJRE 6·33 IYPICAl DEGENERATIONOF STRESS BJl.SESHAPE WITH INCREASINGSTRESS BJl.SEMAGNITUDE
·4
132
CHAPTER SEVEN
BEHAVIOUR OF COMPACTED TRIAXIAL SAMPLES
7.1 INTRODUCTION
This chapter presents the results of the suction measurements andrepeated load triaxial tests on compacted samples of the threeclays used in the research project (Keuper Marl, Gault clay andLondon clay). The suction moisture content curves are presentedfirst, and then the results from the pneumatic repeated loadtriaxial rig. The test programme is detailed in Chapter 3, andthe equipment is detailed in Chapter 4.
7.2 SUCTION MEASUREMENTS
The suction/moisture content curves measured for the Keuper Marl,Gault clay and London clay are shown in Figures 7.1, 7.2 and 7.3respectively. Measurements from both types 'of apparatus describedin Chapter 4 are shown for comparison. Two sets of measurementsusing the equilibrium method were made for the Gault and Londonclays. For the first set the clay was mixed to approximately theequilibrium moisture content for the suction applied to thesample, and for the second set the three samples were all mixedto the same moisture content as that estimated as the equilibriumvalue for the middle value of suction used in the experiment.
The two types of test on the Keuper Marl, Figure 7.1, show generalagreement. A small air bleed was discovered at higher suctionsin the 'araldite' sealing the porous stone to the glass tube. Itwould seem likely that this gave rise to the points scattered tothe left of the line shown in Figure 7.1. The leak was sealedand the remaining tests gave satisfactory results.
The agreement between the two types of test on the Gault andLondon clays, Figures 7.2 and 7.3, is not very good. It wasfound that the porous stones used in the equilibrium apparatustended to dry out and admit air after three or four days under
10-~------~10--------2~0------~30~--MOISTURE CONTENT (0/0)
-LLa.~2.Q-u::JV)
LLa.-c 2o-o~
3
u,a.-52gV)'
3
3
~
)( Rapid Suction Method• Equilibrium Test 1
Ke}!)( Rq:>idSuction Method• Equilibrium Test 10Equilibrium Test 2
10 20 30 40 50 60MOISTURE CON1ENT (%)
Ke>!x Rapid SuctionMethod• Equili brium Test 10Equilibrium Test 21L---~--~--~~--'_--~--~--~
o 10 20 30 40 50 60 70MOISTURE CONTENT (%)
FIGURES 7-1,7·2 & 7·3 RESULTS FROM SLCTION-MOISTURE CONTENT TESTS
7·1 KeuperMarl
7·2GaultClay
70
7·3 londonClay
134
vacuum. This would seem to be insufficient time for the lowpermeable clays, like the Gault and London, to reach equilibriumbefore the tests had to be terminated. This is indicated by thedifference between the gradients of the test results from thefirst and second tests using the equilibrium method. The secondtest, when all three samples started at the middle moisturecontent, produced a steeper line than the first test when thesamples started nearer the equilibrium moisture content for eachsuction. The fact that the tests were terminated after the porousstones started leaking throws some doubt on the accuracy of themoisture contents and the curves derived from the rapid suctionmeasurements are used to give the values of suction quoted inthis thesis.
The three clays seem to have similar suctions at their plasticlimits and therefore an attempt was made to 'normalise' themoisture content using the plasticity index. Figure 7.4 showsthe suction plotted against the liquidity index for all threeclays and demonstrates that the suction moisture content curve isapproximately linear for each clay over this range of suctions.The plastic limit is difficult to determine accurately using thestandard test (BS 1377) and this will cause some variation in theresults. The minearology of the three clays is also differentand this may also influence the suction moisture contentrelationship. The factors affecting suction are discussed morefully in Chapter 2, section 12. However, the relationship betweenliquidity index and moisture content shown in Figure 7.4 wouldseem reasonable enough to enable suctions to be estimated from aknowledge of the Atterberg limits and moisture content for thethree clays tested. Also shown on Figure 7.4 are some results,presented by Croney (1977) for a variety of heavy clays which donot fit the pattern of results found in this research. Croneymeasured the suction on the drying part of the curve, if thesuction was measured on the wetting part then the suctions wouldbe lower in each case, though the agreement between the resultswould not improve.
'135
r--r-,Q')
"""""
>- ~eN 0W 0Z -0a::o
L:a::
~ w0 ~ _g ~~ 0 U LLL-
e...> -c<11 Ca. - 0 <{::J ::J "0
~QI 0 § •~ (!) 0
X • + 8
/ ~
• -
I e.Yo
ZQf-/.+ e...>
9~
+
xlI
0N
-..t... • <It e. e-UJa:::~
_ 0 LLLO -..t ("') N - 0.
X3ONIAlIOlnOIl
136
7.3 COMPACTED TRIAXIAL SAMPLES
7.3.1 Introduction
The sample details for the Keuper Marl, Gault clay and Londonclay are given in Tables 7.1, 7.2 and 7.3 respectively.
The test programme on the compacted samples was designed toinvestigate the initial resilient response of the material whensubjected to various deviator stress pulse magnitudes. However,some samples were tested for up to 4000 stress pulses at eachstress level and some samples were subjected to load-reloadtests. The onset of permanent strain was determined but therewere insufficient tests to correlate the development of permanentstrain with number of cycles.
7.3.2 General behaviour
All samples from all three clays were subjected to at least onetest which consisted of pulsing the deviator stress at a certainvalue and monitoring the resilient strains at 1 and 100 cycles.The deviator stress pulse magnitude was then increased and themeasurements repeated. The load sequence was reversed when theresilient strains were approximately 3500 microstrain.
Figure 7.5 shows a typical deviator stress pulse magnitude versusresilient axial strain plot. The samples all exhibit typicalstress softening behaviour, and also above a certain pulseddeviator stress level the resilient strain starts to increasefrom 0 to 100 cycles. The unload line appears nearly linear atthe higher repeated deviator stress, but joins the loading curveat the lower de~iator stress pulse magnitudes. There is also nodifference between the axial resilient strain measured after 1and 100 cycles when the sample was being unloaded.
Six of the tests on the Keuper Marl were repeated, four of themthree times. In all cases the samples were stiffer with eachreloading. Figure 7.6 shows a typical plot. These tests were
.....><its.....s,+'Eos,4-
In+'r-:::sInQJ0::
137
In
s::oIn 0In .......... +'o its0.. 0::
s::4- 0o .....
+'QJ itsQJ ~ ~~ :::sC"I+'QJ itsCl In
.....
s::o+'U:::sV'>
QJs;
,..... :::s,..... In itsQJ In 0..U QJ ~~
0..
,.,E-C"I~
L.()r-....-iL.()~NL.()L.()"'Or-..~r-......~O"'("t')~"'..-iO~OOON("t')","'00",,,,00,,,,,,,,,00"'''''''
L.()OOOOOOL.()L.()OOL.()OO..-i..-iOO..-i"''''~L.()(''t')c::t~Nr-..r-..~OOOOL.()~r-..OOL.()~r-..oo..-i..-i..-i..-i..-i..-i..-i..-i..-i..-i..-i..-ir-"'l
.~~"'NL.()O..-ir-..~L.()OON("t')~r-..~N",~~OOON~r-..'"..-i ..-i
oor-.."'~O("t')~NL.()r-.."''''~~..-iN"'NO("t')~OO~("t')~L.(). ., . .O"'N~r-..("t')Or-..L.()~..-iOO~N..-iN..-i..-iNN..-i..-iNN..-i..-l
OOOOOL.()L.()L.()L.()OOOO("t')("t')("t')("t')("t')..-i..-i..-i..-l
O..-lN("t')..-iN("t')~L.()~r-..OO"'..-I..-I-..-I
138
Table 7.2 Results from unconfined triaxial tests on Gault Clay
Sample Moi sture Suction Dry Densi ty Degree of Poisson'sNumber Content saturation Ratio
% kPa kg/m3 %
1 38.96 18 1230 88.7 0.1162 31.93 37.5 1390 92.1 0.0873 27.11 56 1525 95.4 0.0714 22.54 77 .5 1670 98.6 0.121
Table 7.3 Results from unconfined triaxial tests on London clay
Sample Moisture Suction Dry Densi ty Degree of Poisson'sNumber Content saturation Ratio
% kPa kg/m3 %
1 40.33 19 1245 92.3 0.0932 31.81 43 1410 92.7 0.0723 24.31 76 1570 89.5 0.0284 34.73 32.5 1330 90.3 0.0845 24.92 73 1520 - Unreliable
139
oooN
..J«-w ~
0::1. t-o-~Z
~tn wa:::..J« Cl)
>X« lLJt-
III..J
Z eroW0-o~-lIl
CJ)lLJ
~a::
u~
CJ) CJ)U
Q.I Q.I 0U U 0~ ~ -U u 0<50 0 00- 1J 01J 1J 0 0
8 0 0 Lt")
s C..J ~
X .. 0
III.e--Wa:::::>C!)
0 u::0 0 0 0 0 0 0N 0 CO (.0 ...., N- -(Dd>t) 351nd 553~lS ~OlVIA3a
-oo,=:100tU(/)..J
ir80(/)(/)UJ0::: 60tn0:::o 40~
~ 20
140
00 200 400 600 800 1000 1200 1400 1600RESILIENT AXIAL STRAIN (~e:)
FIGURE 7·6 DATA FROM REPEAT TESTING AFTER ATLEAST 24 HOUR DELAY
141
carried out after at least 24 hours had elapsed since the lasttest. The samples were left undrained during that time.
A single load, partial unload, reload and final unload test wasperformed on Keuper Marl sample 7, see Figure 7.7, whichindicated that the previous highest repeated stress levelaffected the modulus at lower repeated stress levels even on thereload section of the curve. This behaviour must be timedependent as the same trend is not shown when the tests arerepeated with a rest period, see Figure 7.6.
Some attempt was made to monitor the effect of a larger number ofcycles at each pulsed deviator stress level. The results areshown in Figure 7.8 and show that the modulus decreases withincreasing numbers of cycles at each stress level. It is notclear whether the samples would reach equilibrium or fail and atwhat stress level this would occur. This type of test wasconsidered too time consuming and outside the original brief forthis section of the research and no further tests of this typewere performed.
7.3.3 Resilient behaviour
7.3.3.1 Axial Strain
All modulus and strain contour values plotted for each samplewere taken from the initial loading curve from the testsdescribed above. The modulus used is a secant resilient modulusdefined as qr/€ar where qr is the repeated deviator stress and€ar the resilient axial strain.
The values of repeated deviator stress for set resilient axialstrains were read off the basic stress strain loading curves andthese were plotted against the suction. The suction of thesample was obtained from the suction moisture content curvesderived using the rapid suction method. The strain contour plotsfor the unconfined samples of Keuper Marl, Gault clay and London
"0
~C:J
'U 'U 0'U
0 0 0 0
0 1:: ..Q .Q L()
_g «£ Q) ca:: =>
• 0 X • r-..e-l.LJa::::::>(!)
0 u:0 0 0 0 0 0 0 0 0CO r-.. (,0 L() ....t (Y') N -
(Dd>t) 3S1nd SS3~lS ~OlVIA3a
142
ooL()N
oooN
w::l
Z«~
olf)oL() _J-«
X«~zW_J
lf)oWoa::o-
oX
s:-X •
oX
oX
143
•••
oooN Wa:::U)>W~
~
~oif]0-o~-(f)
Wa:::
8LO
••x~ ~ ._ ~ __~~~oo 0 0 0(£) ~ N
(Dd>1) 3S1nd SS3~lS so;VIA30oco
~zW
_Ja:::<t:~
a:::wa.::>w~
Z<t:a:::tn...J<t:
~
cor.:.Wa:::::><!)lL.
144
clay are shown in Figures 7.9, 7.10 and 7.11 respectively. Theresults have been plotted in terms of resilient axial strainrather than resilient shear strain as the low values of Poisson'sRatio measured indicate that the resilient radial strainsmeasured may not be correct.
The results for the London clay, Figure 7.11, show goodcorrelation, with strain contours radiating from the origin andgetting closer together as the strains increase, demonstratingthe stress softening behaviour. The Gault clay and Keuper Marlplots follow the same sort of pattern though not as convincingly.There appears to be a change in slope of the contours at highersuctions for both the Keuper Marl and the Gault clay. Thesechanges occur at suctions approximately equivalent to the plasticlimit and perhaps do not appear on the London clay results as allthose tests were conducted at moisture contents above the plasticlimit.
A comparison between results in Figures 7.9, 7.10, and 7.11 andthose in Figure 6.14 indicate that the suction of a compactedsample could be analogous to the mean normal effective stress ofa consolidated saturated sample. The resilient axial strainsfrom the confined Keuper Marl samples are shown in Figure 7.12and indicate that increasing the confining pressure increases thestiffness of the sample. An attempt was made to allow for theeffect of cell pressure on the suction by using the relationship
s = u + ap 7.1
with u as the suction corresponding solely to the moisturecontent of the sample, p as the cell pressure, and a was takenas 0.3 as used by Croney (1977) for this material. Figure 7.13shows the effect of adding ap to the suctions in Figure 7.12.Generally the trend is correct although the magnitudes are not.This may be due to inaccuracies in monitoring the cell pressureas the cell medium was air and the pressure was measured using a
145
8.....
0(Y)
w 8._~8 0 0 0 0S 0 0 0 NW N cD co .....
• + x • •0.....
fD • x
3S1nd SS~lS ~OlVIA30
146
•
oto
o eLOa...x
o('Y)
oN
~ S ~(Od>t) 3S1nd
~
SS3~lS ~lVlA3a
-
147
~~ § ~ ~ ~'-0w • + x ••
o1.0
o 0 0 0-..t M N ~
(Od>t) 3S1fld SS3~lS ~OlVV\3a
~.JU
Z
Z90Wt-
o a:CD 8-
~ ~o oX LL1.0_
z«a::t-V')
a::
~~r-:..wa::sII:
148
z
00s ~
vl?- ~~
a CS're, 0..,. en
0C"')
W
:l.8 ° o§o'-0 ~
o 0WN <.0 ~ 0
- N
~ 0 • 0 <:l D> El.::i-
cY)-b'-
Lt') + • <4 a. • 0~ ~ N~
UJ0 0:::
0 0 0 0 0 0 0~\0 I/) ~ C"') N ~
(Dd>t) 3S1nd SS3~lS ~OlVI"3a CL
149
00
~-
a ~ °%~ 0')
..J~ a::«
° ~co a::wo,::::>
R w~IJ..
8 e- Cl 0
I ILLJ (/')
~ ~o_z Wto ~ IJ.. ..J
I~
n,.::tt.z ~
8_0ZU «
° 0 a::(/')
L() t3 IJ.. 0W
~~Z
- IJ..
~« Z
0 0 0 0 0 0~
0- 0 0 0 0 0U
L.....,.
WC N ....,. (,0 ex> 0 Z- ::::>..J a::
0« e
a, 0 ~.::tt.
M
0 • 0 <n I> G .... (/')
M aJ a::
d-::::>
..J ~0 (/) Z0
N 8o,.xL() + • ~ ~ II ('Y')- -d- o r-:.- LLJ
Cl:::::>o
0 u..0 0 0 0 0 0 0 0r.... (,0 L() ....,. M N -(Dd>t) 3S1nd SS3~lS ~OlVI/\3a
150
relatively insensitive pressure gauge. The system was notimproved as further work concentrated on unconfined specimens.
Figures 7.14, 7.15 and 7.16 show the strain contour gradientplotted against resilient axial strain for the Keuper Marl, Gaultclay and London clay respectively. These figures givereasonable approximations to straight lines and lead torelationships of the form
7.2
where A and B are constants. This can be rearranged to give anequation for the resilient modulus.
7.3
Table 7.4 gives the values of A and B obtained from Figures7.14, 7.15 and 7.16. Plots of calculated resilient axial strainagainst measured resilient axial strain for the Keuper Marl,Gault clay and London clay are shown in Figures 7.17, 7.18 and7.19 respectively. Apart from the test results from samples atmoisture contents below the plastic limit the calculatedresilient axial strains are all within 20% of the measured values.
7.3.3.2 Poisson's Ratio
The resilient radial strain was recorded for each test. Figure7.20 shows a typical plot similar for all three clays anddemonstrates that the Poisson's Ratio was not stress dependent.The values obtained for each test for the Keuper Marl, Gault clayand London clay are given in Tables 7.1, 7.2 and 7.3.
The value of Poisson's Ratio measured for saturated samples insimilar undrained tests was 0.5 for the range of moisturecontents tested. The dry density of the saturated samples varied
.. 151
...J<{
X 300<{ (
~ZUJ...J
t3 200a:::
w~500z~ 400t;;
800
700
600
·2 ·4 ·5 ·6 ·7 ·8 ·9 1·0
FIGURE 7·14 LOG RESILIENT AXIAL STRAIN vs LOG GRADIENTOF STRAIN CONTOUR FOR KEUPER MARL
1000900800700
600
~
~ 200...J(j')lJJ0:
152
·2 .3 .4 ·5 .6 .7 .8 ·9 1·0qr/P~
FIGURE 7·15 LOG RESILIENT AXIAL STRAIN vs LOG STRAINCONTOURGRADIENT FOR GAULT CLAY
153
1000900800
700
600
500w~z 400«Cl::t-V) 300....J« (~
t-~ 200::JV)
~
·2 .4 . 5 . 6 ·7 . 8 ·9 1·01OS.1
RGURE 7·16 LOG RESILIENT AXIAL STRAIN vs LOG STRAINCONTOUR GRADIENT FOR LONDON CLAY
.,154
Table 7.4 Coefficients for equation 7.2
Materials A B
microstrain
Keuper Marl 2740 2.1Gault clay 2640 1.35London clay 1690 1.27
155
~0~\+1
-, 00
\ 00.-
g. -, 0~" 0<0 -W
0:1.-
0 0
-, 0 Wto .....<t...J::::>
0 ~~0 <to..:0 .- N M -..t
L.
U e.- .- .- .-..... ..... ..... U;til til (/)
aw UJ 0
W e--..... ..... ..... ..... .-
0
+N e--
• X 0 UJ0:::r5
0 CL0 0 0 0 0 00 0 0 0 00 co U) .....t N.-
(3r1) °3a3~ns'V3W J
156
a::'if!. fZa (/)~'\ z
0 «a::
-, f-a
(/)
Xa --'0- «-w x
~ • ::1. «0
-a 853 f-~" Z
COf- lJJ
<, «....J
X ::::>
-, 8~ ~<DU-, L- a
w
<, 0a~ u
- N ('t') ~co
a -f- tn tn f- 0 t'.(/) (/) NlJJ lJJ UJ lJJ Wf- f- f- f- a::X • + 0
::::>(!)
0 u,
a 0 a 0 0 0a 0 a a 0a co <D ~ N- °3t3r1) a3~nSV3W J
157
o8-
~ -"00~, -,
<; -,-, -,-,,~<, l'- N M ....r " ~
to- ll; I-- I--U') U') U') <, mW W W
~I-- I-- I--.-:-e--
X • + 0 W0::::::>~u::
0 0 0 0 0
8 0 0 0 0co to ....r N- (3rl) a3~nSV3~ J °3
158
000N
_J
<tx<tt-ZUJ_J
(J')
0 UJ0 Cl::tf)
~-z<tCl::
tU t-::t (J')
_J
Z <t<t 0Cl::
~gti;t-O_J m-<t
x _J N<t (J')
t- UJ~Z Cl::
UJ LL U_J 0 .z(J') 0W t- o
0\ Cl:: 9 z0\
c a. 0c 15 8
_J
15 0 _J00 - tf) <t z0 c U
~_J :J a.+ • ~
.....(J')
~r'UJCl:::JC!)
0 u:0 0 0 0 0«D N co ....,t- -(311) NlVaLS lVlav~ lN3111S~
159
with moisture content though the degree of saturation was 100%.The behaviour of the compacted samples is rather different andgives much lower values for Poisson's Ratio. The London clay,for example, gives a decreasing Poisson's Ratio with decreasingmoisture content at a roughly constant degree of saturation. TheGault clay and Keuper Marl also seem to show a similar trendexcept when the degree of saturation is close to 100% when thevalue of Poisson's Ratio is rather higher than the other resultswould indicate, for example Gault clay sample 4 and to a lesserextent Keuper Marl, sample 13. The value of Poisson's Ratiowould seem to depend on the degree of saturation and also to someextent on the moisture content. Rather more tests would berequired to confirm this conclusion.
7.3.3.3 Permanent Strain
The cyclic deviator stress at which a measurable amount ofpermanent strain had accumulated was determined for each test andwas plotted against the suction of the sample, see Figure 7.21.The figure shows that there is an approximate linear relationshipwhich holds for all three clays. Expressions relating permanentstrain in the subgrade to the repeated vertical stress would bevery useful for the analytical design of road pavements, as oneof the modes of failure of a road is by rutting caused bysubgrade deformation. The spread in the results may be due to theinaccuracies in determining the onset of plastic strain in thetests, or because the results are from three different clays withthree different minearologies and stress strain relationships.
7.4 UNDRAINED SHEAR STRENGTH
The four Gault clay samples and two of the London clay sampleswere subjected to undrained constant strain rate failure tests. AWykeham Farrance constant strain rate machine was used at astrain rate of O.3040mm/minute.
The deviator stress axial strain plots are shown in Figures 7.22and 7.23 for the Gault and london clays respectively. As would
160
00..-
•0en
0
0CO
XX
0e-
oa...xoM GII
L-e
o~tf) 0..-IIL-e
o~o • X +IIL-e
0 0c.D -0a.. z.x 0
~a t;~;::
U~:::>
(J)
9 z-«~
•
~ X&: _g ~o o~ u -UL- e~
0 -"C -::J~
e C~ .9 C>
otf)
000 0....,M N _(Dd>!) NIV~lS IN3NVW~3d
.:10 13SNO 1'1 3S1nd SS3~lS ~01VIA3a
161
00000-
~_,u
0 u, t-8 _,
z ::::>2 «- N M ...J' « C!)
OJ W W w a:::_, _, _, _, t-n, 0.. 0.. 0.. V')
~ ~ z; z _, In« « « « 0 « OJIf) If) V') V') 8 - t-
~+ x 0 0 0 OJto - l/) a:::w > ::::>:1 _,z lf
~~
OJ
~a:::_J a:::«
~zx« «> a:::
0 0 t-
§ V')
NN Ne-Wa::::::>C!)
0 LL0 0 0 0 00 0 0 0...J' M N -(Od>t) SS3~lS ~Ol'1I/\3a
~ Lt) §0
UJ UJ ~_J _JCL. CL.~ ~~ ~(/) (/)
• X
§-OW("t'):::t-Z
~ ~t; _Ju
_J z§~ e> 0O~ Z
N o_J
Cl:: ~tn (/)UJ
Cl:: r-e w8 ~
~. 8 >UJ
~- 0C")Nr.:.wCl:::::>(!)
0 LL0 0 0 0 0 0Lt) 0 Lt) 0 LON N - -( DePt) SS3~lS ~l '111\31
162
o§LO
163
be expected the drier samples sustained the higher values ofdeviator stress. As shown previously the drier samples alsohave the higher suctions.
The type of failure varied with the moisture content. Thewettest Gault clay sample, sample 1, showed no signs of failureplanes and deformed into a uniform barrel shape. Gault claysample 2 developed an appreciable barrelled shape but there wasevidence at the surface of small failure planes. Gault claysamples 3 and 4 failed with excessive deformation on a discretefailure plane, with the rest of the sample showing littledistress. The two samples of London clay also failed alongdiscrete failure planes.
These two types of failure, the uniform barrelling and thediscrete failure planes, would indicate a change inoverconsolidation ratios for saturated samples. Those with lowoverconsolidation ratios would be wet of critical state andconsequently deform uniformly, while those with overconsolidationratios' greater than about 4 would be dry of critical and becomenon uniform with the resultant shear planes. Schofield and Wroth(1968) report that when soil is completely sheared by remouldingthe overconsolidation ratio is often between 4 and 5. This waslimited to saturated soils but it would seem reasonable thatcompacted unsaturated soils would also show some signs ofoverconsolidation. It is likely that compacted soils would alsohave a locked in radial stress caused by the compaction process.The results suggest that there is a change from plastic tobrittle behaviour as the moisture content decreases, but thatthis does not coincide with the conventional plastic limit.
Figure 7.24 shows the peak values of deviator stress plottedagainst suctions at the start of the test on a log-linear scale.The results for the Gault clay give a good linear relationship,showing a direct relationship between the suction and log peakdeviator stress. There are only two results from the London clayand these are shown to fit the curve from the Gault clay quitereasonably.
164
0co
•0e-- V)
I-- m0 tuU) a=
:::>...J
~0U") - 0d!. W
.:;lI- z«
~~a= a=0
I-~ Z~ u- :::>
5 « • ~iU...J UJu o 0 I-U("t') ~ ~z
~ 8 «:::> z z« 9 «o
(!) a=x • l-
V)
O~------~---------o~------~--------------------~o000O .~ ...o 0 .0,...., N ...
(Dd>t) SS3~lS ~OlVIA30 )fV3d DOl
...
165
Saturated samples tested to failure under undrained conditionsfail on the critical state line at the same voids ratio as theyhad initially. The failure point is independent of pi andtherefore independent of the overconsolidation ratio. Theundrained shear strength can be expressed as a function of thespecific volume as
Cu = 1. M exp [ (r - vo)/>' l2 J 7.4
It follows therefore that for saturated samples a plot of log Cuagainst specific volume or moisture content will give a linearrelationship.
Figure 7.24 shows that for unsaturated Gault clay there is alinear relationship between log Cu and suction. It has beenshown previously that there is a linear relationship betweensuction and moisture content, and therefore between log Cu andmoistu~e content for the unsaturated samples.
An equation similar to 7.4 can be seen to describe the behaviourof the unsaturated samples as well as saturated samples, althoughthe various constants would be different.
The two results from the London clay tests do not give sufficientinformation to describe the behaviour of London clay and rathermore tests on Gault clay would be required before a firmconclusion could be drawn.
166
CHAPTER EIGHT
CBR TEST RESULTS
8.1 INTRODUCTION
This chapter presents the results from the tests on theconsolidated CBR samples of Keuper Marl, and on the compacted CBRsamples of Keuper Marl, Gault clay and London clay. The samplepreparation is detailed in Chapter 5, and the test programme isdiscussed in Chapter 3.
8.2 CONSOLIDATED CBR SAMPLES
The sample details from the tests on the consolidated Keuper Marlsamples are given in Table 8.1. The CBR was determined at the topof each sample. The preconsolidation pressure was determined fromoedometer tests carried out on two samples cut from the bottom ofeach CBR sample. The shear strength was determined by a 19mmshear' vane and is the average of at least three readings, whichgenerally showed a scatter of ±2 kPe •
The results from the CBR tests are shown in Table 8.1. Thetrends are generally as expected, with a decreasing moisturecontent leading to an increase in CBR and also in shear strengthas measured by the shear vane. The overconsolidation ratio alsobehaves as expected with· an increase in the moisture contentleading to a decrease in OCR. This is because the samples wereall swelled back to approximately the same vertical effectivestress therefore the overconsolidation ratio depends mainly onthe preconsolidation pressure.
Figure 8.1 shows a typical plot of one of the CBR tests showingthe effects of the load unload cycle. The shape of the curvecompares well with that produced when a sample of annealed copperis loaded and unloaded axially in tension to successively higherstress levels. The curve shows a predominantly plastic behaviourfor the virgin loading section with the recoverable part of the
VI
.,....ro+.JQ)"0
cQ)
EroVI
VIQ)r-
EroVI
.,....r-oVICoU
.CO
Q)r-.0rot-
"167
~+.J t').,.... E 0 0 0 0 0 0 0
~ VI - co L,() c:t L,() C"') ,...... 0s, C 01 ,...... ,...... ,...... ,...... ,...... 1.0 ,......Cl Q) .:::.t. .-4 .-4 .-4 .-4 ...... ...... ......
Cl
CQ)0- ro
~ +.J 0- 0 ~u Q) .:::.t. I c:t I I I .-4 I
.:::.t.U0Q-
Q)Cro>- ro co c:t c:t co co 0 0
~ 0- · · · ·u s, .:::.t. 0 c:t C"') .-4 co 0 ~ro c:t c:t C"') C"') N N .-4
Q).s::.V')
0 0'\ ~ 0'\ co ~ 0 ,......~ ~ · · ·
L,() C"') C"') N N N N .-4
ex:c:l.U
Lt'> 0 0'\ C"') N co co c:t.~ · · · ·
N L,() L,() c:t c:t C"') N N
-0 Lt'>
ex: 0- .-4 co co 1.0 L,() Lt'> coU - · · · · ·0 - .-4 0'\ co ~ co 1.0 Lt'>
E .-40-
+.JC E 0 ,...... co 1.0 N ,...... C"')
Q) 0 c:t .-4 1.0 N c:t c:t co+.J +.J ~ . · · · ·C +.J 0 .-4 .-4 .-4 N C"') C"')
0 0 N N N N N N NU c:l
Q)s,~ Lt'> 1.0 .-4 N ,...... 0 c:t+.J C. ,...... C"') ,...... c:t c:t 1.0 C"')
VI 0 ~ . · · · · · ·.,... I- 0 .-4 .-4 N N C"') c:t0 N N N N N N N~
Q) s,,.... Q)C. .0 N 1.0 c:t C"') .-4 Lt'> ,......E Ero ~V') z:
168
EE-Z
-...to~0:::to-W(Y)z l.U
~W 0:::a.
..J Z
~ §0: l.U~ 25~(/)
UJSC!)
0 LL0 lJ) lJ) Lf) 0r:-.. 0 N~
0 0(N)t) OV01 ~3€)Nn1d
169
deformation becoming apparent on the unloading sections. Thereload part of the curve is of a similar gradient to the unloadsection of the curve (allowing for the hysteresis), until the loadapproaches the previous maximum load when the gradient decreasessharply as the sample starts to yield again.
Figure 8.2 shows the complete load deflection curves plotted forall the consolidated samples with the unload reload sectionsomitted for clarity. Figure 8.3 shows the loads recorded for thefirst millimetre penetration together with the unloading andreloading section of three of the curves. The curves on Figure8.2 are all fairly similar and the relative strengths apparent inthe initial stages of the test continue on to the maximumdeflections. It is interesting to note from Table 8.1 that theratio of the CBR values at 2.5mm penetration to 5mrn penetration isapproximately 1.4 to 1 for all the tests. This may indicate thatthe material under the plunger is already in a general state offailure at 2.5mm penetration and that the stress distributionsare similar at 2.5mm penetration and 5mm penetration although theamount of sample affected will obviously be greater with thelarger penetration. Therefore to determine the stiffness ofsamples before yield it might be prudent to use the results fromthe early part of the CBR test. Hight and Stevens (1982)reported a similar conclusion from their work which consisted ofa finite element analysis of the CBR test on soils with variousload deformation characteristics.
The initial part of the curves shown in Figure 8.3 suggest thatthe stronger the sample the greater the initial gradient.However, the initial part of the curve is very dependent on howwell the plunger is seated on the sample and is thereforedifficult to measure accurately. To some extent this applied tothe unload reload parts of the curves as well as the penetrationat zero load after unloading is difficult to measure accuratelyand was found to be time dependent as the samples continued torecover slowly. The reload parts of the curves shown in Figure8.3 also show that the stronger the sample the greater thegradient of the reload curve.
170
--...,t(",) co
E<D E
zor-
LO~u,llJo
...,to::llJC!>Z::>...J
("')"-
- co
- (N)t)OVOl H38Nnld
o«
0::«~
auWzr-o«_at--u...JUJ~...JZLLoUJuc
N
171
lJJ.....Ja....- N M ...r LO (,0 e-- N~<e. X + 0 • • <l(/)
C"')
eoLU0:::>C)
u:e-- (£) If) -s C"') N .-. .
(N~) OVal ~39Nnld
172
There was too much scatter in the gradients of the reload curvesto suggest any form of relationship with the CBR value other thanthat the gradient increased as the CBR increased. This scatteris probably due to the experimental errors discussed above andalso as there was only one sample tested at each moisturecontent. It would seem reasonable to assume that there would besome form of relationship if the samples were uniform, preparedin the same manner, and of the same material.
Figure 8.4 shows the CBR value plotted against the undrainedshear strength as determined by the shear vane for each sample,and a pocket penetrometer for samples 5 and 6. The penetrometerresults show reasonable agreement with the shear vane results forthe two samples. The CBR and undrained shear strength (in kPa) arerelated by the expression Cu = 7.8 x CBR for this material overthe range of moisture contents tested. Black (1979) quotes therelationship between CBR and undrained shear strength as Cu =11.5 x CBR for undisturbed overconsolidated soil. Black used acone penetrometer to measure Cu. The results quoted by Black showa ra~ge of factors from 8.6 to 11.0, so the results from Figure8.4 are at the low end of this range.
Figure 8.5 shows that there is a linear relationship between themoisture content and the logarithm of the undrained shearstrength. This result is as expected from critical state theoryfor undrained strength tests on saturated samples as the samplefails on the critical state line at a constant specific volume.This is discussed more fully in Chapter 7, section 7.4.
8.3 COMPACTED CBR SAMPLES-Tables 8.2, 8.3 and 8.4 give the sample details and test results
from the tests on the compacted samples of Keuper Marl, Gaultclay and london clay respectively. The CBR, moisture content andundrained shear strength were determined at each end of thesample. The undrained shear strength was determined using a 19mmshear vane and a pocket penetrometer.
6
5
0:3cou
2
1
173x SHEAR VANE• POCKET PENETROMETER
•
10 20 30 40 50 60UNDRAINED SHEAR STRENGTH (kPa)
FIGURE 8·4 CBR AGAINST UNDRAINED SHEARSTRENGTH FOR OVERCONSOLIDATEPSAMPLES OF KEUPER MARL
- (
~ 3·8 XXZ Xx~(/) 3·60:<{l1Jdi 3·4@Z~ 3·2oZ23.05
2.8 '____-L...-_~ "&""-__ __'
20 21 22 23 24 25MOISTURE CONTENT (%)
FIGURE 8·5 LOGARITHM (UNDRAINED SHEAR STRENGTH)AGAINST MOISTURE CONTENT FOR cmSOUDATEDCBR SAMPLE S OF KEUPER MARL
•r-ro.j-)o-0
Cl)r-
VlCl)r-
0::COU
N
00
Cl)r-.cro~
174
e 00 N ,...._ 0"1 ~0 co N ~ ~ ~ 0"1.r- ro.j-) 0-U ~:;, ,...._
N 00 00 0(/) ~ N ~ I.!") r-, 0r-i
I.!") CV) ~e 0"1 0"1 0 0 004- 0 co . . . .0 .r- 0 0 r-i r-i 0.j-)Cl) ro ~Cl) s,s, :;, ~ CV) 00 0"1 ~en .j-) 0"1 0"1 0"1 0"1 00Cl) ro l- .Cl (/) 0 0 0 0 0
a a I.!") a I.!")a ~ r-i 0"1 ,...._~ co ~ ~ ,...._ ,...._ ,...._,., r-i ...... r-i r-i r-i
~.r- EVl <,
Cl e enCl) ~ a a a I.!") I.!")Cl I- a ~ CV) 0 00~ ~ ,...._ 00 ,...._...... - ...... ...... ......eCl)0- co ~ a I.!") 0"1 CV)ro - CV) I.!") 00 I.!"):;, ..., 0- r-i
U Cl) ~~U I- ~ N I.!") 00 a0 ...... CV) I.!") 0"1 0"10- ......
Cl)e CV) I ,...._ ~ 0"1ro co - N ~ 0"1> ro:;, 0-
U s, ~ro CV) I CV) 0"1 0Cl) I- - CV) I.!") .......s: ......(/)
a ~ 0"1 a ~ ~· · ·I.!") a a ...... N ~co
I.!") ~ a a 0 0"1· · · ·N a ...... ...... CV) ~0::co Nu a ~ 0"1 I.!") ...... 0· · · · · ·I.!") a a ...... CV) I.!")
l-I.!") ~ 0 ~ CV) 0"1· · · · · ·N a ...... ...... CV) I.!")
~ I.!") ~ 0"1 CV)a N ~ 00 I.!")Cl) co . · · · ·~ ..., ~ N ...... 00 ~:;, e N N N ...... .......j-) Cl)Vl .j-) M.,.... e0 0 ,...._
I.!") ~ 00 CV)~ u CV) N N 0 0l- . · · · ·~ N 0 00 \0N N N ...... ......
Cl) ~r- Cl)0- .cE E N I.!") ...... CV) ~ro :;,(/) z
"175
E10VI
,..... ~ ~ oo:::te co .-t CV) LO ,.....0.r- 10+-' 0..U .::.t!- ,..... ~ CXl a;:, I- .-t CV) LO CXltil
N CT'I LO .-te CT'I CXl CT'I CT'I~ 0 co . .0 .r- a a a a+-'(1) 10(1) s, ~s, ;:, N CT'I oo:::t CT'ICl +-' CT'I CXl CT'I CXl(1) 10 l- .Cl til a a a a
a a a aLO ~ N a
~ co N CV) LO ~.., .-t .-t .-t .-t>, .r- Es, VI <,Cl e Cl
(1) .::.t!- a a LO LOCl I- LO ~ N .-t
N CV) LO ~.-t .-t .-t .-t
e(1) CXl CV) LO a0.. co .-t oo:::t .-t CV)
.-t CV)+-'
;:, (1) 10U .::.t!- o..
U .::.t!.0 CXl LO N a0.. I- .-t oo:::t a CT'I
.-t oo:::t
(1)e CXl CV)rtI co · CXl:::- CXl ~ LO a;:, rtI N LO .-t
U s, 0..rtI .::.t!.(1) LO N.c: l- ·til CT'I ~ ~ a
N LO CV).-tr-;
a ~ ,..... a a· · · · ·LO Cl .-t oo:::t CXl
coLO co .-t oo:::t ,.....
· · ·c::: N a N oo:::t CT'Ico ~u
a ~ ~ a a· · · · ·LO a f'"""4 oo:::t CXl
I-
LO co CXl oo:::t ~· · · ·N Cl f'"""4 oo:::t N.-t
.. Cl co ..-I CV)oo:::t N ..-I ..-I· · · ·(1) co CT'I N ,..... CV)~ +-' CV) CV) N N;:, e
+-' (1)VI +-' ~.r- e LO CT'I CT'I ~0 0 CV) ..-I LO CT'I:E u · · · ·I- CT'I N ~ ..-I
CV) CV) N N
.(1) ~,..... (1)~ .&lE E f'"""4 N CV) oo:::trtI ;:,til Z
.r-rtI+-'(1)"0
(1),.....
VI(1)r-
E10VI
,.....U
+-',.....;:,10c.!:I
"0(1)+-'U10
Eou
.co(1)r-.&l10I-
...-.,_10....,Q)'0
Q)...-E10III
IIIQ)r-
E10III
0::COW
>.10r-U
eo'0eo....J
'0Q)....,U10
Eow
176
e .-i .-i .-i L.()
0 co M o:::t 1.0 1.0.,_ 10+-' 0-U .:><:.;:, .-i .-i L.() NV'l I- M o:::t 1.0 r-
0 "" M L.()
"" CXl "" ""e CO .4- 0 0 0 0 00 .,_
+-'Q) 10 ~Q) s, 0 "" .-i Ms, ;:, "" CXl "" ""en +-' l- . . .Q) 10 0 0 0 0Cl V'l
CXl 0 0 0.-i r- .-i L.()
~ CO M M L.() L.()
<'l .-i .-i .-i .-i
>. .,_ Es, III -Cl e en
Q) .:><:. CXl 0 0 0Cl .-i r- N r-
I- M M L.() L.()
.-i .-i .-i .-i
e 0 L.() 0 0Q) CO 0 CXl CXl M0- .-i N M
10;:, ...., 0-W Q) .:><:.
.:><:.u00- I- "" 1.0 0 0
"" CXl .-i o:::tM o:::t
Q)e r-IO CO I I CXl I::::- 10
;:, 0-W s, .:><:.
10Q) Ms: l- I I 0 IV'l .-i
0 "" CXl o:::t o:::t· ·L.() .-i N 1.0 ""CO
L.() M r- 0 o:::t· · · · ·N N M CXl .-i
0:: .-i
CO ~W
0 "" M 0 0· · · ·L.() .-i N r- .-i
.-i
I-L.() M .-i 0 N· · · ·N N M "" M
.-i
.-i "" M ""en M M M o:::tQ) · · ·~ ...., L.() N r- 1.0;:, e M M N N...., Q)III
...., ~.,_ e0 0 L.() L.() 0 L.()
:::£ w M .-i o:::t .-i
l- · · · ·L.() N 1.0 L.()
M M N N
Q) ~r- Q)Co .DE E M .-i o:::t N10 ;:,V'l z:
"177
Figures 8.6 and 8.7 show the results for the Keuper Marl, Figures8.8 and 8.9 show the results for the Gault clay, and Figures 8.10and 8.11 show the results for the London clay. The unload andreload sections of the curves have been omitted from Figures 8.6,8.8 and 8.10 for clarity. As for the consolidated samplesdiscussed in section 8.2 the curves are a fairly similar shape,and the relative strengths are apparent in the initial stages ofthe tests. Again as for the consolidated samples the samples withthe higher CBR values showed the larger initial gradients andreload gradients. There was too much variability in the measuredgradients to indicate any relationship with the CBR.
A comparison between Figures 8.2, 8.6, 8.8 and 8.10 show that thecurves for the Gault and London clays are approximately the sameshape, while those of the consolidated Keuper Marl samples arestiffer initially, and those of the compacted Keuper Marl aresomewhat flatter. This is reflected in the ratios between theCBR at 2.5mm and 5.0mm penetration which are very approximately1.25 for the Gault clay and London clay, 1.4 for the consolidatedKeupe~ Marl and approximately 1.1 for the compacted Keuper Marl.
Figure 8.12 shows that there is a good relationship between theCBR value and the undrained shear strength as measured by apocket penetrometer. The relationship can be approximated to astraight line of the form Cu = 3.7xCBR (kPa) The pocketpenetrometer is in reality a small scale version of the CBR testand a linear relationship is not surprising. Figure 8.13 showsthe relationship between the undrained shear strength as measuredby the shear vane and the CBR value. The relationship for theKeuper Marl and Gault clays can be approximated to straightlines, although of different gradients. There are only tworesults for the London clay· and these suggest that the behaviourmight be similar to the Gault clay. The relationships areapproximately Cu = 20 x CBR for the Keuper Marl, and Cu = 12.5 xCBR for the Gaul t clay,) with 'C in kPa.
Black (1979) suggests that C = 23 x CBH (kPa)for remoulded soils ...The results from the Keuper Marl are in reasonable agreement with
178
~ MM- 0 If')
.- .-
~ ....:'-"":\8 r-,N ....:O....:.M• -NN"":.....rN NNNNN
0'1CO
•cooCO •
-If') E
Ezo
....: ~0::t-UJm
M CL
.-
N • • \8 ....:• • •(N)t) 0'101 ~3E)Nnld
N
•.- .-
I-lJ)UJt-o::CDu
ol1Jt>~~ou
~~~ ~ _ N M
«::>. + X(/)Z
..179
. . .(N)t)0\f01 ~38Nn1d
-
(/)
~:::>(/)UJa:::~(/)- UJ
~~
- ~z a:::0 ~I-~ a:::Cl:: CO~UJ UZUJ ...JQ.. Cl::«
~
a:::UJQ..:::>UJ~
0OJ~U
~
~
e-.CO
UJer:::>C!>u:
.·180
0:::l1JCO~ - N M -..t
~ • + X •t-U')l1Jt-
O- .....,- 0')
I:"-- MN -- -'::!!.lO M 0') COO') LO
2-0') - LO N- M
~- COM lO NN 0')
EN N N MM M
UJ0:::
t-(/')UJt-
o:::COU
Z ~0 _J
i= u.....,« t-o::: _Jt- ::>l1J <Z (!)
WMa. Cl
UJt3~~0U
Cl).- CO
UJ0::::>(!)
u:N 0 CO lO ....., N 0 Cl) lO -..t N
N.
N - - - - -(N)f)a\101 ~3E>Nn1d
M
0,::
~
N
N M ...r-We + x • m._J - coCl.~ W« Cl::Cl) ::>
(!)
u:M N - 0 m co e- (£) Lf) ...r M N -. . .- - - - (N)t) 0\'01 M3ElNnld
181
--o-
co
-l:' E, Ezot-
(!)<{'0,::
ttlzUJLf)Cl.
-ELf') E-
Z0I- Z
~<{ 0a:: 0I- ZWZ _.W0.. 0M W
I-
~0..
N ~
U
0-.- CO
WCl:::::>(!)
0 LL~ N 0 CO (0 ~ N 0 CO (0 ....r N 0. .N N N - - - - -
(N>t) (]\fOl ~3DNnld
.J.82
ffim~ - N M ~:::>zl- • + x •(/)UJI-
-~ Lf') en M (j)Lf') -Lf)SL. - ~ ~ M M- cry cry cou Lf') (0 (Or-... NN Lf')Lf)- N N NN MM MM~
'183
0:::UJen~:::>z- N M -..t .-UJ + X •...J.Q..::E-cCl)
-..t N q CO cD ...t N- - - (N)O ovoi ~3E>Nnld
.--o-
M
....0:::«E D-
EcD - 0:::·z mo u~0:::
tn .....·UJZUJQ..
~...JU
Zooz9oUJ....U
~~ou
--eoUJ0::::::>C>u.
185
•x
x ~....JU
....J~0:::
~<! _.~ ....J U
U0::: ZUJ
~0o, 0
~ ~ Z MUJ <! 0 -~ c:> ....J CO
+ • X UJ0:::::>c:>LL
0 0 0 0 0 0N 0 CO 4D -...t N- - (Od>1) 3N'IA ~'13HS AS n~
186
this, but those from the Gault clay are rather lower, and more inagreement with the relationship he proposed for undisturbedoverconso1idated material (Cu = 11.5 x CBR). Black states thatmany compacted fills at low equilibrium suctions can behave asthough they were 'undisturbed overconso1idated soils' as the actof shearing them causes a negative change in pore pressure whichresults in small strain at failure. As shown in Chapter 7 thebehaviour of compacted samples changes in nature from plastic tobrittle as the moisture content decreases. The liquidity indexat which this occurs does not correlate with the plastic limitand may vary with different clays. The difference in behaviourthen may just be the result of the two clays having differentminera10gies and Atterberg limits.
Figures 8.14, 8.15 and 8.16 show that there is an approximatelylinear relationship between the suction of the compacted samplesof Keuper Marl, Gault clay and London clay respectively, and thelogarithm of the CBR value. The CBR value for these samples wasfound to correlate directly with the undrained shear strength asmeasured by the shear vane or pocket penetrometer, andconsequently Figures 8.14, 8.15 and 8.16 show the same type ofbehaviour as the compacted triaxial samples, where there was alinear relationship between the logarithm of the undrained shearstrength as measured in an undrained constant strain ratetriaxial test, and the suction, as shown in Figure 7.25.
....J0 0...0 ~- «
Cl)
....J
0 a::m «
~a::l1J
00...
co~
0l1J
0 t-e- U~
~- 0o 0 ucoo.. a::~
0Z LL0 Z0-Lt')t- 0U t>:::>t/)
~
0 t--..t V)
z-et<!)
0 -etC"'l
et::mu
0 (!)N9-.:t
0..-:-co- l1Ja:::::>(!)
0 LLCD Lt') -..t C"'l N _m
0(0/0) ~8:> 001
187
..188
(/)UJ_Jo,~«(/)
a~en_JU
a !:ico :::>«
(!)
0a UJ~ t
~~
a 8(,Q - er:ro f2c,
~- zaLOZ 00 t;F :::>u (/):::>
alf)-...t t-
V')Z-c<!)
a «M er:
mu
a <!)N 9
in-a .- coUJet::::><!)
a LL-...t N 0 ex> (,Q -...t N - co- - - (0/0) ~8 J €lO,
0to -ro
0..~
0lJ")
z0 _J
t- o,U t- ~
0 :::> U) <{....t (f) Z (f)-<{
(!)~<{
0_J
MU
Z0Cl
0 ZN 0
_J
to-0 CO- UJ
~(!)-
0 LL....t N 0 CO (.0 ....t N- - - (0/0) ~8:> 001
189
oen
oCO
190
CHAPTER NINE
UNDISTURBED SAMPLES
9.1 THE SAMPLES
To try and obtain a measure of the natural overconsolidationratio of road subgrades some undisturbed samples from under tworoads in the south of England were obtained and tested. Thesamples were taken in U4 sample tubes and were kindly supplied bySomerset County Council and Northampton County Council.
The samples from Somerset were taken from under the hard shoulderof the M5 in a cutting approximately 10m deep and consisted of aKeuper Marl. Those from Northamptonshire were taken from asection of the M1 which was being rebuilt and consisted of claywith some sand. The area was generally flat. There were norecords available indicating the ground water level at eithersite.
Some of the samples were tested in standard Rowe cells todetermine the natural overconsolidation ratio, and the remainingsamples were placed in the pneumatic repeated load triaxial rigand tested in the same manner as the compacted samples.
9.2 DETERMINATION OF NATURAL OVERCONSOLIDATION RATIO
The equipment is described in Chapter 4. Some diffic~lty wasexperienced in extruding and cutting good samples, especiallyfrom the Keuper Marl, as the natural material contained stonesand harder natural lumps of Marl.
The results from the tests on the Somerset samples are shown inFigure 9.1, and those on the Northamptonshire soil in Figure9.2. The water table was assumed to be at the top of thesubgrade in each case to allow an estimate of theoverconsolidation ratio to be made. The overconsolidation ratiowas estimated from the vertical effective stresses. The Somerset
·191
0N
0::U 00 ~
0
o ·5 1·0 1·50..~ 0gf5
...-c?w 00:: ~:::>(J)(J)w §0::0..Z ~N0 8~
("I") ~~0 ::J ::J- -_J 0
~0 Q.I Q.IN Q_Q_z E E0u 0 o 0
ui 0(J)(J)
ff ~ • )(
0
LO ·5 1-0 1·5("I")-~0-..... 0zw ("I")
.....z0U LOW N0:::::>
~~.....(J)
0 0~ N
LO~---------.----------~----------~-~O .5 1·0 1·5DEPTH BELOW ROAD SURFACE (m)
FIGURE 9·1 ROWE CELL RESULTS- SOMERSETSAMPLES
192
N
a:::U....to
- 0
C£x 1·0 2·0 3·0.::lo-e-~~ x~~
I
f~~
CLzo00- ~~0_j,0 0If) Lt>Z
§a:::CL o .
1·0 2·0 3·0-~ Lt>0 ('I')
t-ZWt-
OZ0 ('I')
uwa::::::> Lt>tn N
-0~
0N 2·01·0 3·0
DEPTH BELOW ROAD SJRFACE (m)
FIGURE 9·2 ROWE CELL RESULTS ONNORTHAMPTON SAMPLES
193
samples show a range of overconsolidation ratios similar to thatchosen for the test programme, while those from Northamptonshireindicate a rather less overconsolidated material.
Natural Keuper Marl is mudstone which ages and weathers into astiff and overconsolidated clay. The results shown in Figure 9.1indicate that the material is overconsolidated, though not asmuch as might have been expected. There is likely to have beensome sample disturbance at the top of the subgrade during roadconstruction although the samples are natural, not fill, as theywere obtained from the bottom of a cutting. As stated previouslythere was some difficulty in cutting samples to fit accurately inthe Rowe cell, which would result in the measured values ofpreconsolidation pressure erring on the low side. The Northamptonmaterial was fill as various bits of glass and wood werediscovered, and it is interesting to note that it does seem to beslightly overconsolidated which is in agreement with Schofieldand Wroth (1968), who stated that soil completely sheared onremoulding exhibits some degree of overconsolidation.
The moisture content appeared very variable with depth. Themoisture content was determined from trimmings around the edgesand ends of each Rowe cell sample, and generally did not includethe least disturbed material from the centre of U4 tubes.
Both samples indicate that as the moisture content increased, andtherefore the specific volume, the preconsolidation pressuredecreased. There does not appear to be any link betweenoverconsolidation ratio and the depth of the sample. The depthrange is very small though and would need to be increased toinvestigate this aspect more fully.
The variable results are thought to be due to the fact that thenatural material was not uniform and therefore difficult toextrude and cut neatly. The point of steepest curvature on theload deformation curve was difficult to determine accuratelywhich caused variations in the results.
194
9.3 DETERMINATION OF RESILIENT MODULUS
It was impossible to cut 78mm diameter triaxial samples from theKeuper Marl extruded from the U4 tubes without excessive sampledisturbance. Five triaxial samples were successfully cut from theNorthamptonshire material and installed and tested in thepneumatic triaxial repeated load rig in the same manner as thecompacted triaxial sample. Table 9.1 gives the sample details.
There was no provision for drainage and the samples were testedunconfined. Figure 9.3 shows the results plotted as repeateddeviator stress against resilient axial strain. There isreasonably good agreement between four of the tests which wouldindicate that the material is fairly uniform over the depth rangetested. The unload part of the test follows the load sectionmore closely than for the compacted samples and would suggestthat the material is less thixotropic than the Keuper Marl, Gaultclay and London clays tested in the research project. Theplasticity index of the Northamptonshire material lay somewhere
•between that of the laboratory Keuper Marl and Gault clay, andthe difference in thixotropic behaviour may be due to the largerpercentage of sand sized material.
The resilient radial strains were also measured and are shownplotted against the resilient axial strain in Figure 9.4. Thevalue of Poisson's Ratio varies quite considerably for the fivesamples tested as shown in Figure 9.4 and does not appear torelate to either the moisture content or dry density. Thevariation may be due to inaccuracies of measurement, or due tothe natural variability of undisturbed soil samples. The valuesof Poisson's Ratio recorded are much lower than those which aregenerally associated with this type of material.
The estimated suction for each sample given in Table 9.1 wascalculated using the Atterberg limits for this material and theapproximate relationship between suction and liquidity indexdiscussed in Chapter 7. Figure 9.5 shows the resilient axialstrain plotted against the repeated deviator stress divided by
195
VIQ)
c::0 Ll")
-e- 10+-' CL. Ll") ...-I o:::t N Ll")
U .!o: ~ ~ ~ o:::t (V)
::lV')
>,+-'.~ O'l co ...-I (V) (V)
"0 X Ll") ,...._ ~ ...-I ~.~ Q) 0 0 0 N N::l "0c- c::.~ .......__J
N ...-I 0 ,...._co ~ ~ O'l
Q) co N N (V)
C'I N (V) (V) (V)
c::~ VI 10 I I I I
Q) 0 C'I 0::s, c:: Ll") 0 ,...._ N::l +-' .~ ~ I (V) (V) (V) co+-' c::
~,-.. ·VI Q) ,...._ Ll") o:::t ~ ...-I.~ +-' .~ N N N (V)
0 c:: s, ~::IE: 0 +-' 0
U ~ ...-I N N> N (V) co ,...._c:( . ·........ ,...._ co O'l N
N N N (V)
~ ~Q) 0 VI ,-..s, Q) (V) Ll") N o:::t ,...._ N::l +-' r- 0 ,...._ ...-I o:::t N+-' c:: 0- ~ ·VI Q) E 0 Ll") Ll") Ll") 0 N.~ +-' 10 N N N (V) (V)
0 c:: VI >::IE: 0 c:(
U ........
>,+-' M
>, .~ E 0 0 0 0s, VI - (V) ...-I o:::t 0Cl c:: C'I I Ll") Ll") o:::t o:::t
Q) .!o: ...-I ...-I ...-I ...-I
Cl
30r-Q) Q).0 "0 U Lt) Lt)
c:: 10 CO ~ CO N (V)
.c ::l ~ . . . . ·+-' 0 ~ E ...-I ...-I ...-I N N0- ~ ::lQ) C'I VICl .-
Q) ~r- Q)0- .0 ...-I N (V) o:::t Ll")
E E10 ::lV') Z
.cVIc::o+-'
E10VI
-r-X10
.~"0c::::l
c::oVI
.~10+-'Q)"0
Q)
E.~V')
196
z
~~«Ib:
)( 8-...t_
)(
0-OWN::1._-
)(z
+ • 0<0 00::
+ Or-_U')
)( _J«• O~0 )( 2r-
.t> zUJ
x O~'\
0U')~UJ
0 X0::
+0x • 0
0-...t
UJ N (Y) -...t Lf) •_J _0a.. . 0 0 0 0 ~ 0 (T"I
~O Z Z Z Z N c;;..<{Z
e QtUJ
V') • )( + 0:::>\!)
0 tr0 0 0 0 0 0Lf) -...t (Y) N
_(Dd>t)3S1nd SS3~lS ~Ol\fI/\3a
co aa a+ d LO
II
>LO u,....,tN X Za W <tII a:::> :1. ti;a
az ....J 2....,t-
+ <t <t11:a:::
~I- ~
X (/)
~....J I-<t Z 5:x UJ
X -c+ al-
a2MW....Jv;
+ Wet:
X
aaN
197+ •
x
- N ("') ....,t LO
0 0 0 0 0
~ 2 2 2 Z Zo, • X + 0 •~ 8<t(/) -
aa("I")
a aa aN -
(3J1) NIV~lS lVlav~ lN3111S3~
1000
500
w~200
z<{a::r- 100(f)
....I<{
~ 50 ·r-zW....I
~ 20
Sample• 1)(2+30485
ee
)(
• +0
+)C
0
•
10+-----~~~--~--~r---~r-----~------0·01 ·02 ·05 ·1 ·2 1·5
FIGURE 9·5 RESILIENT AXIAL STRAIN vs REPEATEDDEVIATOR STRESS DIVIDED BY SUCTIONFCR THE UNDISTURBED NORTHAMPTONSAMPLES
199
the suction. Samples 1, 2, 3 and 5 show reasonable agreementabove a resilient axial strain of approximately 200 microstrain.The results from sample 4 show a similar gradient to the others,but a different intercept. The spread in the results below 100microstrain is thought to be due to inaccuracies in measuringstrains of this magnitude.
The resilient response of the samples above a resilient qxialstrain of 100 microstrain can be described by the soil modeldeveloped in Chapters 6 and 7, which is of the form
9.1
where A and B are material constants and equal 1870 and 1.61respectively.
Although the stiffness of the samples did not increase with depthwhen they were tested unconfined, they would be stiffer withincreasing depth due to the effect of the surrounding material.A subgrade consisting of this material would appear stiffer withdepth in a non linear manner to a pulsed load as the material isstress softening, see Figure 9.3, and the stress pulse magnitudedecreases with depth as the load is spread over a greater area.'
200
CHAPTER TEN
PAVEMENT ANALYSIS
10.1 INTRODUCTION
Two different pavements were analysed using a finite elementcomputer program which can cater for non linear elasticmaterials. The original non linear soil models developed atNottingham by Brown and Pappin (1982) from data by Brown et al(1975) were used initially to give an indication of stress pulsemagnitude, and pulse shape under a standard wheel load for use inthe experimental programme.
The subgrade model developed in Chapter 6 has been incorporatedinto the program and these results are compared with thoseproduced using the original subgrade model, by re-analysing thesame road structures under the same conditions.
10.2 PAVEMENTS ANALYSED
The two pavements which were analysed are shown in Figure 10.1together with the loading details. The difference between thetwo pavements was in the thickness of the granular layer only,and they are referred to as 'thin' and 'thick' pavements (200mmand 700mm granular layer respectively). The pavements do notcorrespond to any specific design but are two of a number ofstructures previously analysed using the finite elementprogram, and described more fully by Brown and Pappin (1985).
Three different pore pressures were chosen and used in thecomputer program to cover the range of effective stresses used inthe main test programme. These were OkPa, -50kPa and -l00kPa.
10.3 THE COMPUTER PROGRAM
The finite element program, called SENOL (SEcant NOn Linear), hadbeen developed at Nottingham by Pappin (1979) from an original
EE
~
0::oE EE E88N e--
201
500 KPa J60mm dio JA T
ASPHALT LINEAR ELASTICE = 7 GPo , V = 0·4If = 23·5 ~mJ
GRANULAR NON-LINEARr= 22·3 K;%,.
SUBGRAOE NON LINEAR(r = 19·6 K~n?
FIGURE 10·1 PAVEM ENT DETAILS FORANALYSIS
202
program by Barksdale. It can cater for any practical number oflayers and any layer can be specified to behave elastically usingeither a linear or a non linear model. The program isto allow different soil models to be incorporated easilylayer.moduli
Each model must produce stress dependent shearG and K for use in the main program.
arrangedfor eachand bulk
The program output consists of the stresses and strains for eachelement together with the total deformation. Careful account istaken of the overburden pressure. An equivalent modulus andPoisson's ratio is also estimated for each element assumingisotropic behaviour.
10.4 THE SOIL MODELS
10.4.1 The Granular Model
The granular model was developed at Nottingham from a series ofstress path tests on crushed limestone in a repeated loadtriaxial apparatus. The test programmes produced sufficientinformation about the resilient shear and volumetric strains toallow these to be modelled mathematically. Consequently the modelincorporated into SENOL is quite sophisticated in that G and Kare calculated separately and Poisson~s ratio is therefore notconstant. Full details of the test results on the crushedlimestone and the development of the model are given by Shaw(1982) and Pappin (1979).
10.4.2 The Subgrade Model
The earlier subgrade model referred to above was developed byBrown and Pappin (1982) from repeated load triaxial tests onsaturated overconsolidated samples of Keuper Marl in an earlierversion of the triaxial rig used for the research. Thedifferences between the rigs used and experimental programmes aredetailed in the discussion section of this Chapter. The modeldetails are described below.
203
The earlier model relates the resilient axial strain to thedeviator stress and mean normal effective stress as
e: r = ~,,[q~~sa F PbJ 10.1
where e:/ is the resilient axial strain in microstrainqmax is the maximum value of the deviator stress in kPaP I is the initial mean normal effective stress in kPao
and for this materialF = 50 MPaS = 0.7
The new model is of the form
e: ra = A [:~ J
B
10.2
and pi are as beforeois the magnitude of the deviator stress pulse in kPaand for this materialA = 1110B = 1.52
There was no resilient volumetric strain measured during thetests which gave rise to this model as the soil was behavingelastically with a Poissonls ratio of 0.5. The test results arediscussed in Chapter 6.
10.5 USE OF THE MODELS IN THE PROGRAM
The load was applied in ten increments and the total resilientshear strain after each increment was obtained by adding theresilient shear strain calculated for that increment using theexpressions above to the total resilient shear strains calculatedbefore the last increment. In the case of the first increment
204
the initial shear strain was calculated from the overburdenpressures.
The shear modulus G for each increment was obtained from theexpression
10.3
rusing values of qr and £a calculated after the last increment ofdeviator stress. The bulk modulus K was calculated from theexpression
K=2(l+v}G3{1-v}
10.4
However a Poisson's ratio of 0.5 results in an infinite bulkmodulus using this relationship. The earlier work assumed that
K = 3.233 x G 10.5
which gives a Poisson's ratio of 0.36. This value has been usedhere to allow for a direct comparison between the results fromthe new model and those obtained from the earlier model.
10.6 THE FINITE ELEMENT MESH
The finite element mesh consists of a grid of 336 rectangularelements, made up of 26 rows of 14 elements. Figure 10.2 givesthe mesh dimensions. The mesh can be considered as a sectionalong a radius of a vertical cylinder, with the centreline of theload down one edge of the mesh.
The boundary conditions allow for vertical movement only of themesh at the sides and the base is taken as rigid to simulate asection of soil restrained by the surrounding material andfounded on stiffer material.
205
3·4m r_v_A."..0'" "'I"
~ ~
I
~ ~
I EC?<D
~~ ,J'j..~
I
L
?I .,-y
I
Aot- ~-r j h 17 17 "V
AXIS OF SYMMETRY
14 ELEMENTS J-I)RIZONTALLY
24 ELEMENTS VERTICALLY
AGURE 10·2 FINITE ELEMENT GRIDUSED FOR ANALYSIS
206
10.7 RESULTS AND DISCUSSION
Figures 10.3, 10.4, 10.5 and 10.6 show the equivalent modulusplotted against depth on the load centreline for the thin andthick pavements with pore pressures in the subgrade of OkPa and-50kPa respectively. The program did not give a satisfactoryconvergence for the pavements with a pore pressure of -lOOkPa,and therefore the results have not been included. The resultsfrom the new model indicate a very rapid increase in subgradestiffness with depth, while those from the earlier model predicta subgrade stiffness which is approximately linear with depth.Figures 10.3 and 10.5 also show the modulus variation with depthassuming a linear elastic subgrade. The stiffness predicted byboth models increases as the pore pressure becomes more negative,which results in a higher mean normal effective stress. Theincrease in stiffness would be expected as the more confined aparticulate sample the smaller the effect of a stress pulse of acertain magnitude. Equations 10.1 and 10.2 demonstrate that thehigher plO the smaller the resilient axial strain which resultstherefore in a larger value of modulus.
The main difference between the earlier and new models is thebehaviour of stiffness with depth. The earlier model predicts asmall increase in modulus as the depth increases followed by adecrease in modulus, which is more pronounced when the meannormal effective stress is higher. The new model predicts quite arapid increase in the subgrade modulus with depth for all thepore pressures examined. This would seem reasonable as the stresspulse magnitude decreases with depth as the load spreads, andtherefore the material would be expected to increase in stiffnessnon linearly from the general non linear stress softeningbehaviour of the soil.
Figure 10.7 shows the stress pulse shape at the top of thesubgrade for the thin and thick pavements with the water table atthe top of the subgrade ( u = OkPa). The new model gives a lowerpeak value of deviator stress in both cases, although the stresspulse is wider to compensate. Also shown on Figure 10.7 are
-EUJZ_JUJa::t-z2UJuo<tsZo~3Lta:::::>If)
~o_JUJCO
4I5:UJo
207
MOOULUS (M Po )o 100 200 300 1.00O+-----~--~----~----~----~--~
ASPHALT
SOO 600
- -
1
GRANULAR LAYER- - ---SUBGRADE
• MODEL AFTER BROWNAND PAPPIN
• NEW MODEL
+ LI NEAR SUBGRADE
5
FIGURE 10·3 EQUIVALENT MODULUS VARIATIONWITH DEPTH FOR THIN PAVEMENTSUBGRADE PORE PRESSURE = 0 kPg
208
MODULUS (MPa)o 100 200 300 400O+-----~--~----~----~--~~--~ASPHALT=~=..=G~NULAR LAYERSUBGRADE
.E1wz_J
W0::.....ZWu
2oS_J
zowuLt0::3
~
~_J
wm:J:.....fu4o
5
500 600
\11I
x MODEL AFTER BROWNAt-{) PAPPIN
• NEW MODEL
•
FIGURE 10·4 EQUIVALENT MODULUS VARIATIONWITHDEPTH FOR THIN PAVEMENT.SUBGRADEPORE PRESSURE = - 50 kPg
209 .
MODULUS (MPa)o 100 200 300 400 500 600O+-----~--~----~----~--~~--~
ASPHALT
-E 1
uiZ::JW0:::
!zuiu
2o<{
9zowu<{
~3~
~uiCD
::r:t-fbI.o
5
GRANULAR LAYER
------ ---SUBGRADE
)( MODEL AFTER BROWNAND PAPPIN
• NEW MODEL
+ LINEAR SUBGRADE
FIGURE 10·5 EQUIVALENT MODULUS VARIATION WITHDEPTH FOR THICK PAVEMENT. SUBGRAD5PORE PRESSURE = 0 kPa
)(
)(I
210
MODULUS (MPa)o 100 200 300 400O+-----~--~----~----~----~--~
ASPHALT
-ELUZ..JLUa::t-zLUU 2o«szo1.LJu«~3:::>(/)
~s1.LJCDIt-fb 4o
500 600
-------
GRANULAR LAYER
1----- ----
SUBGRADE
5
)(
)(
x MODEL AFTER BROWNAND PAPPIN
• NEW MODEL
rFIGURE 10·6 EQUIVALENT MODULUS VARIATION WITH
DEPTH FOR THICK PAVEMENT. SUBGRADEPOOE PRESSURE = - 50 kPa
z
~co ua:: .....UJZ_J~.....-UJu.R:o UJ« 0 >_J~~a::~UJ :::.U5UJ8~[Uzz~«z:Ju;)( • + I
UJa::::::>o
CV'> LL
211
_JUJoo~
waLL
212
the results from earlier computations using the same pavements,but with a linear elastic soil model. The mesh in this case wasonly 1.5m wide although there were the same number of elements.The stress increment at the edge is not zero. For this reasonthe mesh was doubled in size for the later computations. Howeverthis does not seem sufficient for the new model as the behaviourat the edge of the mesh appears to be inaccurate, and a mesh witha larger number of elements is required to correct this.
A sine wave is shown plotted on Figure 10.7. The period has beenadjusted to give the best fit to the central portion of the pulseshape, but this leads to a rather shorter stress pulse than thatproduced from the calculations. It can be seen though that of thestandard waveforms available from signal generators the sine wavegives the best fit to the pulse shape, and it is easier for aservo system to follow as there are no step changes in the rateof loading.
213
CHAPTER ELEVEN
COMPARISON BETWEEN THE BEHAVIOUR OF CONSOLIDATED AND COMPACTEDSAMPLES
11.1 INTRODUCTION
This chapter contains a comparison and discussion of the resultsfrom the different test series on the consolidated and compactedsamples. It can be divided into two sections: the first containsa comparison between the results obtained from the compacted andconsolidated samples in the same type of test, and the secondpart compares the results of the CBR test series and triaxialtest series from the consolidated and then the compacted samples.
11.2 COMPARISON BETWEEN THE BEHAVIOUR OF COMPACTED ANDCONSOLIDATED TRIAXIAL SAMPLES OF KEUPER MARL
11.2.1 Resilient Modulus
The resilient modulus determined from the consolidated sampleshas been shown to depend on the mean normal effective stress andnot on the moisture content. Overconsolidated samples withdifferent stress histories giving different moisture contents atthe same effective stress will therefore have the same modulus.
The model proposed for calculating the resilient modulus of thecompacted samples is identical to that proposed for theconsolidated samples except that the suction is substituted forthe mean normal effective stress term. The suction referred to isthe matrix suction and evidence from other researchers (Chapter2, section 2.12) indicates that it depends on moisture contentand is relatively insensitive to changes in dry density.Therefore samples with different dry densities at the samemoisture content may be expected to have similiar suctions andtherefore a similar resilient modulus.
The Keuper Marl was the only material to be tested in the
214
consolidated and compacted states and the results from bothseries of tests demonstrate that increasing the confiningpressure increases the resilient modulus. For the saturatedsamples the term plO in equation 6.10 is the actual effectivestress within the sample and naturally includes the confiningpressure. The soil model for the compacted samples was developedfrom the results on unconfined samples and therefore the suctionterm in the equation is the matrix suction of the sample under noexternal load. There were insufficient confined triaxial tests onthe compacted samples to quantify the effect of a cell pressure.
A comparison between the results for the compacted andconsolidated samples cannot be made by using the same moisturecontent in each case as the resilient modulus of the consolidatedsamples does not depend on moisture content. If the suction istaken as the negative pore pressure of the compacted samples thenin the absence of an external load it is equivalent to the meannormal effective stress for the consolidated samples. Equatingthe suction and mean normal effective stress terms in the modelsdemonstrates that the stiffnesses obtained from each model are ofthe same order, see Figure 11.1, with the best agreementoccurring at the lower deviator stress pulse magnitudes. Thesuctions used in the model to plot Figure 11.1 are the suctionsof the samples with no external load. Applying a confiningpressure to the compacted samples will increase the effectivestress even in undrained condition, as the air in the samplecompresses, and will therefore lead to higher resilient moduli.An increase in the confining stress of 20 kPa, representingapproximately 1m of overburden, for the compacted samples willincrease the calculated stiffness by increasing the suction term.According to Croney (1977) the actual value of the increase willdepend on the value of a in the equation
u = s + up 11.1
where u is the pore pressures is the suctionp is the overburden pressurea is the compressibility factor
',\ \ \ \ \'~"" '\ \ \\ \ , \
,\, '\ \ \" \ \ \
'\\, \ \ \ \\", \ \ \ ," \ \ \\ \\ '" \ \ \ ~
" \ \ \ '\ ", \ \ \ \, '\ \'\ ", \" \ 0
, ,,\ \ \, \D
\ \ \ \ '\ \ \ \ ~
\ \~ \ \ , s\ \ \ \ '" \' \ ~\ \ ' , 0\ \ , -..t
\ "~ I\, l,\ \:,\, , 0
\ \\\ ,I ("t)
\ \\,'\ \\ ~\ "~I 0
\\t, N
215
oLf')II'-
CT
o00.-- Lt)II II'- '-
CT tr
IIL..
CT
oo
(J")
~a..~«(J")
(J")
l1J.....Ja..~
~
ClUJ~Cl
ClUJt-U~~8,I
I
.....Jsz8III
oLI\N
'--------~------~------~------~------~o·0(:)
an00 0o '" 0N .... .--(l!dW) srrmos lN3111S3~
oo-
(,/)UJ.....Ja..~
~
ClUJ
~Cl
~
8oen
oco
Clz«ClUJG~
9
- ---
216
There were insufficient tests on confined samples to determine avalue of ~ for the Keuper Marl.
11.2.2 Thixotropy
Both the compacted and consolidated samples exhibited thixotropy,with the measured stiffness at a certain stress pulse smaller ifany of the preceeding stress pulses were larger, within a certaintime period. However the consolidated samples tended to regaintheir original stiffness when they were left to recover, but theresults from the compacted samples suggest that they tended tobecome stiffer after recovering from periods of loading than theywere when loaded for the first time. Drainage was allowed inbetween tests for the consolidated samples, although there was anegligible change in volume in most cases. The compacted sampleswere not provided with any provision for drainage. The gain instiffness of the samples following cyclic loading would then bedue to a slow reorganisation of the internal structure withoutany change in state. The interparticulate forces would besufficiently small that they could be disrupted by quite lowexternal stresses and hence the sample would soften on reloadingeven if the pore pressure remained constant. This is discussedmore fully in Chapter 2, section 2.10.
11.2.3 Development of permanent strain
The test programme on the compacted samples was not designed toinvestigate the development of permanent strain. However, theapproximate repeated deviator stress level which caused the onsetof permanent strain was recorded and was found to be linearlyrelated to the suction of the sample.
The test programme on the consolidated samples demonstrated thatthere was a threshold repeated stress level, above which failurewas likely to occur, but below which equlibrium was attained. Asdiscussed previously the deviator control system was notsufficiently stable to allow the onset of permanent strain to bedetected with any confidence for the consolidated samples.
217
11.3 COMPARISON BETWEEN THE BEHAVIOUR OF COMPACTED ANDCONSOLIDATED CBR SAMPLES OF KEUPER MARL
The CBR values determined for the compacted and consolidatedsamples of Keuper Marl are shown plotted against moisture contentin Figure 11.2, and it is apparent that the consolidated samplesshow a considerably higher CBR than the compacted samples.Neither sample was tested with any surcharge weights, althoughthe consolidated samples were allowed to swell back under avertical effective stress of 20 kPa. This was removed just priorto testing the sample and the consequent change in pore pressurewould maintain the vertical effective stress at 20 kPa. Hightand Stevens (1982) report that initial stresses affect the slopeof the CBR curves at low penetrations but do not affect the CBRvalue for stiff samples (CBR approximately equal to 10%). Forsamples with CBR's of 4 or 5 they found that there was someeffect on the CBR value as well as on the initial stiffness.The compacted samples would have some residual radial stressescaused by the compaction process.
The consolidated samples were consolidated anisotropically andthe sample structure would be expected to be anisotropic with theclay plate like particles arranged horizontally. The structurewould then be stiffer along the axes perpendicular to theplatelets than parallel to them. However, the anisotropica1lyconsolidated triaxial samples did not exhibit any evidence ofanisotropy. This is discussed more fully in Chapter 6. Thecompacted samples were compacted to relatively high degrees ofsaturation by a small tamper. Seed et al (1962) report that thistype of compaction in samples with high degrees of saturation islikely to result in some reorientation of the clay particles intoa parallel structure. Seed et al found that soil with this typeof structure is very much softer than the same soil compacted bystatic compaction which results in a randomly orientatedstructure.
The other difference between the two types of sample is thedegree of saturation. The consolidated samples were all
(/')
218~ (/')o, w~
....Jo,(/') ~
<!0 (/')W~ 0
LO
W
N
0 I-....J U0 <!(/') o,z ~
-s0 0
N
U U
• X
("I")N
/ -•N~
0
/
u-•~ u,
0N
x
-
NN
Cl-
e--N.
CD --- UJet:::>~
If) LL0-
219
saturated, while all the compacted samples, apart from thedriest, varied from 93% to 100% saturation. It seems likely thatsamples compacted to this high a degree of saturation wouldbehave similarly to saturated samples, as the air pockets wouldbe small and not continuous.
The difference in response between the two types of sample wouldseem to be due to a combination of the differences in structureand the initial stresses.
11.4 CONSOLIDATED SAMPLES OF KEUPER MARL - RESILIENT MODULUS ANDCBR
The model developed in Chapter 6 from test results on a varietyof samples with different moisture contents and mean normaleffective stresses demonstrates that the resilient behaviour ofthe overconsolidated samples of Keuper Marl depends on plO and isindependent of the moisture content and hence the consolidationhistory. The model was developed from low stress pulse tests onfairly heavily overconso1idated samples and the stress paths werewell within the yield locus.
The CBR test results covered a similar range of moisture contentsbut the samples were all allowed to swell back to approximatelythe same effective stress. A standard CBR test at 1mm/min iscompleted in 7.5 minutes, and this can be compared withapproximately 150 minutes for 90% consolidation assuming that thecoefficient of consolidation equals 2.28 m2/year (Overy 1982) andassuming a zone of failure 30mm thick under the CBR plunger at2.5mm penetration. This value was taken from the results of thefinite element analysis of the CBR test by Hight and Stevens(1982). The tests can therefore be considered undrained.
The undrained shear strength for saturated soils depends on themoisture content, not the overconsolidation ratio andconsequently not the mean normal effective stress at the start ofthe test. The measured CBR value was found to be linearlyrelated to the undrained shear strength as measured by a shear
220
vane. This may not be the case at other mean normal effectivepressures outside the range of the test programme as Hight andStevens found that the CBR reflected the strength of stiffsamples, and a combination of strength and stiffness for sampleswith lower CBRs. The initial gradients of the load penetrationcurves, and the gradients from the reload sections of the curvesshowed no apparent correlation with any other measured parameter.Hight and Stevens suggest that the initial slope of the loadpenetration curve depends to some extent on the initial stressand does not relate directly to the stiffness. It is the author'sopinion that the CBR tests in this instance reflect the undrainedshear strength rather than the stiffness, although it is possiblethat the initial gradients do correlate with stiffness and thevariability found in the results is due to experimental error.
Figure 11.3 shows a plot of resilient modulus against CBR, wherethe CBR is the measured value, and the resilient modulus iscalculated from the soil model using the mean normal effectivestress estimated for the CBR sample. However a 50% increase inthe mean normal effective stress almost doubles the resilientmodulus and probably has much less of an effect on the CBR value.The calculated values for resilient modulus for a deviator stresspulse of 40kPa then lie on either side of therelationship Mr = 10 x CBR.
11.5 COMPACTED SAMPLES - RESILIENT MODULUS AND CBR
The model developed in Chapter 7 from repeated load triaxialtests on unconfined samples of three different clays demonstratedthat the resilient axial strains were controlled by the suctionof the sample as well as the level of the repeated deviatorstress. The material constants in the model differed for thethree clays. The suction was found to be related to the moisturecontent of the sample, and, as discussed in Chapter 2, wouldappear to be relatively insensitive to changes in dry density.
The CBR test results demonstrated that the CBR was related to thesuction for each clay, although the constants in the equations
If If cf..x..x .x 00 0 0N ~ ....II II II
C'" C'" C'"
X • <l
221
x • <l
o~
oM
o 0N ....(edW) SnlOOOW !N311IS3~
co
~o
et:~CDu
M........UJ
S(!)
u::
222
again differed with the different materials.
Figures 11.4, 11.5 and 11.6 show the relationship between CBR andresilient modulus for different values of pulsed deviator stressfor the Keuper Marl, Gault clay and London clay respectively. Thefigures were plotted using the soil model derived in Chapter 7,rearranged to give the resilient modulus
Mr= ~r [~r 11.2
and the relationship found between the CBR and suction anddescribed in Chapter 8.
log CBR = CS + 0 11.3
where C and 0 are constants and equal 0.019 and -0.417respectively
combining these equations gives
M = ~ [ (109(CBR)-D)] Br A C x qr
The empirical relationships Mr = 10 x CBR, originally derived byHeukelom and Klomp (1962), and Mr = 17.6 (CBR) 0.64 from Powellet al (1984) are also shown on the figures.
11.4
The results indicate that the magnitude of the repeated deviatorstress pulse has a significant effect for all three clays,although the effect is greatest for the Keuper Marl, which is theleast plastic. The marl however exhibits less non linearity inits stiffness - CBR relationship than either of the other twosoils.
The empirical relationships give a reasonable approximation tothe stiffness of the marl, but overestimate the stiffness of theLondon clay and Gault clay at the higher CBR values. Kirwan! etal (1982) reported similar inaccuracies at CBR values of 5% andabove, although they were working with glacial tills with
223Z<{0:::I- ~(/') 1.0
0::: 00:::o:::W m (/')ma.. uuLL x- xO totU
Cl. s;?fij e-,..lI:: -II(/,) II 0LoO 0 c.Zo Lo00
C" N ~ ~OtO~ <0-, 0-
IJJCl:::::>C)
____ _L ~ ~~ ~ ~ ~o LL000-0 LI"I
oLnN
ooN -(edW) SnlnOOW lN3111S3~
Vl0::: Vlen UJu 0:::
I-0 Vl
Z LL« °V) W(/') C)w Zz <{U. 0:::U.
-~<{to '#. (/')_ 0:::
Zfro:::wmUJ_,u~o:::
I.l> I- <{UJ~en
0:::Q..UJa..~:::>z~o
- --
Z 224<{
~
~a:::~
- U.ro 0c,..:.::- 0,-0 000
C'" N c.DCO- 0- V)lJJ0::>f-..JCl.
~
lJJV)..J
-.t::>
CDCl.
0 V)a::: V)CD lJJU a::::)(
f-
coV)-a::: r-, LL-CD II0
U -)( a::: lJJ
0 en (!)- a.nU Z11- ~'-
~
-- -
oLO
o"(!dW)
000("t\ N -SnlnOOW lN3111S3~
Z 225«0::In0::UJa..LL- 0
too,.:.:::- 0,-0 0 000crN ~ cD ex:> - 0- 8 ...Ja..
~0:: «0u,
UJ0::
U')...J
OJ :::>U a..
0 U')
z U')
« lJJa:U') ....U') U')
UJLLZ
LL 0LL
0:: cD - .... UJOJ ?f!. U') (!)
U - Z)( z «0
0:: UJ 0::en
LOU UJ «~'- .... 0::
~ UJ 0OJ LL
~ a..:r: ~U') ...JZ U0
~...JUJ0::
~..a UJC> :r:0::: ....alux to..a .~ --II -'- LUE 0::
:::>(!)-
0 LL
0 0 0 0 0 0 0(,Q LO ...J' ('I') N -
(edW) srmoos lN3111S3~
226
plasticity indices between 14 and 20. These compare more withthe Keuper Marl rather than the London or Gault clays used inthis research.
The relationship between the suction and the magnitude of thedeviator stress pulse required to cause the onset of permanentstrain developed in Chapter 7 is also shown plotted on Figures11.4, 11.5 and 11.6. The relationship is only approximate asdiscussed in Chapter 7, but indicates that there will be somepermanent strain development for deviator stress pulses of 40 kPaon the Gault and London clays even with a CBR of 10%. The marl isshown to be able to support much higher repeated deviatorstresses at particular values of CBR without any permanent straindeveloping than either of the other more plastic clays.
The suction moisture content relationship for the three clays wasshown to be unique in Chapter 7 if the results were plotted interms of liquidity index, and this result was used to plot thevariation in stiffness and CBR with liquidity index. Figure 11.7shows the variations in stiffness with liquidity index for thethree clays, and Figure 11.8 shows the variations in CBR withliquidity index.
The stiffness curves for the London and Gault clays appearcoincident, and the CBR curves for these two clays are parallel,though not coincident. The curves for the Keuper Marl do not fitthose of the other two clays either in gradient or intercept,and therefore if there is a normalising factor to be incorporatedinto a single model then it would appear to depend on otherparameters rather than just the plasticity index.
2001501008060
- 40e 30~- 20~~
10
5-·2 -·1 0 ·1LI
FIGURE 11·7 RELATIONSHIP
227
• KEUPER MARL
x GAULT CLAY
+ LONDON CLAY
·2 ·3 ·4
BETWEEN RESILIENTMODULUS AND LIQUIDITY INDEX FORTHE COMPACTED SAMPLES
• KEUPER MARL
X GAULT CLAY+ LONDON CLAY
.6 LIFIGURE 11·8 RELATIONSHIP BETWEEN eBB AND
LIQUIDITY INDEX FOR THE OOMPACTEDSAMPlES
228
CHAPTER TWELVE
CONCLUSIONS
12.1 CONSOLIDATED TRIAXIAL SAMPLES
12.1.1 Consolidation
(a) The consolidation results show a unique curve, within experimentalscatter, in v-lnp' space with gradient A = 0.1.
(b) The swell-back curves in v-lnp' space were parallel and of similarshape with an average gradient x of 0.035.
(c) The value of Ko was measured as 0.54 for the consolidation phase ofeach sample.
(d) The results from the swell-back phase in q-p' space showed a scatteraround the curve proposed by Mayne and Kilharney (1982) whereKo = (1 sin¢') OCR sin¢'
(e) The rate of loading during consolidation was sufficiently slow toavoid any significant variation in moisture content with radius at the endof consolidation.
12.1.2 Resilient Response
(a) The repeated deviator stress pulse tests of one second and one tenthsecond duration showed a significant stress softening effect. These testswith a repeated stress pulse magnitude of twenty seconds did not show anystress softening, however the stress pulse magnitudes were much smallerand covered a smaller range.
(b) The mean normal effective stress was shown to remain constant for thetests with stress pulses of one second and 20 second duration indicating
229
isotropic elastic behaviour.
(c) Transient pore pressures could be measured under applied loadings ofone tenth of a second duration.
(d) The resilient axial strain resulting from a repeated deviator stresspulse depended on the mean normal stress and the magnitude of the deviatorstress pulse according to the equation:
=
where A and B are material constants, both of which increased as the stresspulse duration decreased, indicating that the material effectively becamestiffer.
(e) For the stress conditions covered by the test programme, the resilientresponse was independent of the preconsolidation pressure and the ambientdeviator stress.
(f) The samples exhibited thixotropic behaviour instiffness at a particular deviator stress was lowerdeviator stress pulses were of greater magnitude. Thetheir original stiffness if they were left to rest.
that the measuredif the preceeding
samples regained
12.1.3 Permanent Strain Response
(a) The test specimens exhibited a threshold stress which was not welldefined but could be represented by a straight line on a normalised qr -Po' stress plot. Stress paths crossing that line are likely to causefailure, and those approaching it to cause large permanent deformationsunder repeated loading.
(b) Stress paths well below this threshold level caused negligiblepermanent deformation or pore pressure development even after 400,000applications of stress.
230
12.1.4 Undrained Strength
(a) The deviator stress-axial strain plots exhibited three distinctregions;region,stress.
an initial stiff section, followed by a linear, less stiff,followed by failure, with large deformations for little change inIt is thought that this shows the initial yield as the stress path
reaches the Hvorslev surface and then failure as the sample reaches theCritical State.
12.2 SUCTION
(a) Plots of suction (in kPa) against Liquidity Index for the Keuper Marl,Gault Clay and London Clay showed a scatter about a unique linearrelationship for the range of suctions, 20kPa to 100kPa, used.
(b) For the soils and conditions of these experiments, suction inpartially saturated soils, was found to be equivalent to effective stressin the influence it had on soil behaviour.
12.3 COMPACTED TRIAXIAL SAMPLES
12.3.1 Resilient Response
(a) The samples exhibited significant stress softening.
(b) The samples exhibited thixotropic behaviour in that the measuredstiffness at a particular deviator stress was lower if the preceedingdeviator stress pulses were of greater magnitude.
(c) The resilient axial strain resulting from a deviator stress pulsedepended on the suction of the sample and the magnitude of the deviatorstress pulse according to the equation:
231
where A and B are material constants which differed for the three clayswhich were tested.
12.3.2 Permanent Response
(a) A unique, approximately linear, relationship was found between themagnitude of the deviator stress pulse required to cause some developmentof permanent strain and soil suction for all three clays which were tested.
12.4 eBR TESTS
(a) The minimum CBR occurred at 2.5mm penetration in all tests.
(b) The consolidated and compacted samples were unloaded and reloadedafter I.Omm, 2.5mm and 5.0mm penetration. The initial slope and the slopeof the reload sections reflected, quantitively, the CBR value, with higherCBR values exhibiting steeper slopes.
(c) For the overconsolidated samples of Keuper Marl, the CBR was linearlyrelated to the undrained shear strength, as measured by a shear vane,according to the equation: = 7.8 CBR (MPa)
(d) For the compacted samples of Keuper Marl, the CBR and undrained shearstrength were related by the expression: Cu = 20 CBR (MPa). For thecompacted samples of Gault clay the expression was: Cu = 12.5 CBR (MPa)
(e) The consolidated and compacted CBR samples exhibited a linearrelationship between the moisture content and the logarithm of theundrained shear strength.
232
12.5 RESILIENT MODULUS AND eBR
12.5.1 Consolidated Samples
(a) For the samples tested, the resilient modulus was found to beindependent of the moisture content, while the CBR varied with moisturecontent.
12.5.2 Compacted Samples
(a) The resilient modulus and the CBR were both found to depend on suctionand are related by a relationship of the form:
= 1109 CBR - DJ BC qr
The empirical relationships E = 10 x CBR (MPa) and E = 17.6 x CBR 0.64(MPa) give a reasonable approximation to the stiffness of Keuper Marl undera deviator stress pulse of 50-60 kPa, but over-predict the stiffness of theGault and London Clays under the same repeated deviator stress.
233
CHAPTER THIRTEEN
RECOMMENDATIONS FOR FURTHER WORK
13.1 CONSOLIDATION TRIAXIAL SAMPLES
The number of samples was not large due to the time consumingnature of sample preparation and therefore a continuation of thetest programme is required to define more accurately theresilient and permanent response to this type of loading. Therange of overconsolidation ratios should be extended toinvestigate whether the resilient strain model is applicable tosamples on the wet side of critical. The range of final meannormal effective stresses should also be increased.
Further tests should be carried out to better define the Hvorslevsurface for the material, and to investigate whether it acts as ayield surface for monotonic tests, and whether it can be modifiedto allow for viscous effects and form a yield surface for repeatedloading.
Attempts should be made to consolidate samples anisotropicallyfrom much wetter slurries of the order of three times the liquidlimit or greater to investigate whether these samples exhibitanisotropic behaviour in effective stress space.
The development of permanent strain under long term repeatedloading is important, and the possibility of relating the onsetof permanent strain to the stress conditions requires furtherinvestigation. The stability of the equipment would have to beimproved before this type of work was undertaken.
The test programme should be extended to cover other clays toinvestigate the possibility of developing a model that applies tomore than one material.
234
13.2 CONSOLIDATED CBR SAMPLES
In order to complete the test programme and investigate theeffect on the CBR of different initial effective stresses CBRsamples should be allowed to swell back to different mean normaleffective stresses. This would require further development ofthe apparatus to allow higher preconso1idation loads to beapplied. The gradient of the load deformation curves should bemeasured more accurately using electrical transducers and furtherattempts made to relate it to stiffness.
13.3 COMPACTED TRIAXIAL SAMPLES
The test programme should be continued to provide a larger poolof results to more accurately characterise the response of thesoils. The programme should also be extended to include samplestested under a confining pressure. The effect of a confiningstress on the suction of the samples could therefore beinvestigated. The development of permanent strain withincreasing numbers of cycles should also be investigated toparallel the test programme on the consolidated samples. Themagnitude of Poisson's Ratio measured in this research alsorequires further investigation.
13.4 COMPACTED CBR SAMPLES
The number of tests should be increasedthe response of the CBR samp1est andshould also be investigated. The
to define morethe additional
gradient of
accuratelymaterialsthe load
deformation curve should be investigated more fully as for theconsolidated CBR samples.
13.5 EQUIPMENT DEVELOPMENT
13.5.1 Consolidation Control System
The proximity transducers must be fitted with an external orseparate oscillator and further alterations to the power supplyand earthing arrangements are required to eliminate crosstalk
235
between the outputs of the individual transducers.
The consolidation control unit requires redesigning to producemore stable switching of the servo motors, and to provide amuch smaller null window.
It is doubtful if the present system will ever producesatisfactory swell back curves and some sort of controller whichadjusts the rate of reduction of deviator stress as the cellpressure reduces may be more suitable.
The bellofram loading pistons on the cell top require redesigningto provide sufficient travel during consolidation, and toeliminate the bellofram seal against the cell pressure whichcaused pulsing of the cell pressure at the higher resilientstrains achieved during testing.
13.5.2 Servo Control System
The system is required to be much more stable than at present andto be capable of applying true haversine waveforms with periodsas short as D.1 second at a reasonable amplitude. The loadsystem must be stable to within ~.DD5kN on both the pulsed andambient components of axial load, and the cell pressure systemmust be stable to within ~lkPa. The systems must be capable ofloading samples with a large range of different stiffness withoutresorting to adjusting the gain while actually under load.
A digital system is most likely to be able to meet the abovespecification, and the author recommends that this type ofcontrol system is investigated with a view to fitting it to boththe axial load and cell pressure systems.
13.5.3 Data Acquisition
The digital data acquisition system fitted during this researchproject proved to be accurate and capable of providi~g far more
236
information from tests over a wider range of frequencies thanother means of data acquisition. The method of measuring volumechange should be digitised to enable the complete stress historyof the sample to be recorded digitally. Combined digital backpressure and volume change devices are available commercially butcould be constructed using a bellofram piston unit with an LVOTor linear potentiometer calibrated to provide a readout of volumechange. The pressure could be controlled by a stepper motorconnected to the bellofram piston.
237
REFERENCESANDERSON, K.H., BROWN, S.F., FOSS,ROSENBRAND, W.F. (1976), 'Effect ofbehaviour', Proc. Conf. on Design andStructures, ICE, London.
I., POOL, J.H., andcyclic loading on clay
Construction of Offshore
ANDERSON K.H., POOL, J.H., BROWN, S.F., ROSENBRAND, W.F. (1980),'Cyclic and static laboratory tests on Drammen clay', Proc. ASCE,Vol 106, GT5.ATKINSON, J.H., EVANS, J.S. and HO, E.W.l. (1985),'Non-uniformity of triaxial samples due to consolidation withradial drainage', Geotechnique Vol 35 No.3.AUSTIN, G.of a siltyUnpublished.AUSTIN, G. (1979), 'The behaviour of Keuper marl under undrainedcreep and repeated loading', Ph.D. thesis, University ofNottingham.
(1975), 'A comparison of creep and repeated loadingclay', Progress report, Nottingham University.
BALASUBRAMANIAN, A.S. (1976), 'Axial strains and displacementpatterns in triaxial specimens of a saturated clay', Soils andFoundations Vol 16, No.1.BANERJEE, P.K. and STIPHO, A.S. (1979), 'An elasto-plastic modelfor undrained behaviour of heavily overconsolidated clays',International Journal for Numerical and Analytical methods inGeomechanics, Vol 3.BARKSDALE, R.D. (1971), 'Compressive stress pulse times inflexible pavements for use in dynamic testing', Highway ResearchBoard Record No. 345.BELL, C.A. (1978), 'The prediction of permanent deformation inflexible pavements', Ph.D. thesis, University of Nottingham.BISHOP, A.W. and GIBSON, R.E. (1963), 'The influence of theprovisions for boundary drainage on strength and consolidationcharacteristics of soil measured in the triaxial apparatus',ASTM Special Technical Publication No. 361.BLACK, W.P.M. (1979), 'The Strength of clay subgrades : itsmeasurement by a penetrometer', Transport and Road ResearchLaboratory, Laboratory Report 901.BLACK, W.P.M., CRONEY, D. and JACOBS, J.C. (1958), 'Field studiesof the movement of soil moisture', Road Research Technical PaperNo. 41.BJERRUM, L. (1973), 'Problems of soil mechanics and constructionon soft clay', Proc. of 8th Int. Conf. SMFE, Vol 3, Moscow.
238
BOYCE, J.R. (1976), 'The behaviour of a granular material underrepeated loading', Ph.D. thesis, University of Nottingham.BROWN, S.F. (1974), 'Sample end restraint and pore pressuremeasurement in cyclic load triaxial tests on Drammen clay',Report to N.G.I.BROWN, S.F., ANDERSON, K.H., and McElVANEY, J. (1977), 'Theeffect of drainage on cyclic loading of clay', Proc. of 9th Int.Conf. on Soil Mechanics and Foundation Engineering, Tokyo.BROWN, S.F., lASHINE, A.K.F. and HYDE, A.F.l. (1975), 'Repeatedload triaxial testing of a silty clay', Geotechnique, Vol 25, No.1.BROWN, S.F. and PAPPIN, J.W. (1981), 'Analysis of Pavements withgranular bases', Transportation Research Record No. 810.BROWN, S.F. and PAPPIN, J.W. (1982), 'Use of a pavement testfacility for the validation of analytical design methods', Proc.of 5th Int. Conf. on structural design of asphalt pavements, Vol1, Delft.BROWN, S.F. and PAPPIN, J.W. (1985), 'The modelling of granularmaterials in pavements', Transportation Research Record 1022.BROWN, S.F. (1980), 'An introduction to the analytical design ofbituminous pavements', University of Nottingham.BROWN, S.F., BRODRICK, B.V. and PAPPIN, J.W. (1980), 'Permanentdeformation of flexible pavements', Final technical reportsubmitted to U.S. Army, University of Nottingham.BS 1377 (1975), 'Methods for testing soils for civil engineeringpurposes', British Standards Institute.CARTER, J.P. (1982), 'Predictions of the non-homogenous behaviourof clays in the triaxial test', Geotechnique Vol 32, No.1.·CRONEY, D. (1977),pavements', HMSO.
'The design and performance of road
CRONEY, D., COLEMAN, J.D. and BLACK, W.P.M. (1958), 'Movement anddistribution of water in soil in relation to highway design andperformance', Highway Research Board, Special Report No. 40.CROOKS, J.H.A. and GRAHAM, J. (1976), 'Geotechnical properties ofthe Belfast estuarine deposits', Geotechnique Vol 26, No.2.CUllINGFORD, G., lASHINE, A.K. and PARR, G.B. (1972), 'Servocontrolled equipment for dynamic testing of soils', GeotechniqueVol 22, No.3.DEHlEN, G.l. (1969), 'The effect of non-linear materialbehaviour of pavements subjected to traffic loads', Ph.D.University of California.
on thethesis,
239
DUMBLETON, M.J. and WEST, G. (1968), 'Soil suction by the rapidmethod: an apparatus with extended range', The Journal of SoilScience Vol 19, No.1.EDIL, T.B., MOTAN, S.E. and TOHA, F.X. (1980), 'Mechanicalbehaviour and testing methods of unsaturated soils', SpecialTechnical Publication No. 740.EDRISS, E.V. and LYTTON, R.L. (1977), 'Dynamic properties of finegrained soils', Proc. of 9th Int. Conf. on Soil Mechanics andFoundation Engineering, Vol 2, Tokyo.FINN, F.N., NAIR, K. and MONISMITH, C.L. (1972), 'Application oftheory in the design of asphalt pavements', Proc. of 3rd Int.Conf. on the Structural design of asphalt pavements. Vall,London.FOLKES, D.J. and CROOKS, J.H.A. (1985), 'Effective stress pathsand yielding in soft clays below embankments', CanadianGeotechnical Journal Vol 22, No.3.FRANCE, J.W. and SANGREY, D.A. (1977), 'Effects of drainge inrepeated loading of clays', ASCE Vol 103, GT7.FREDLUND, D.G., BERGAN, A.T. and SAUER, E.K. (1975), 'Deformationcharacterisation of subgrade soils for highways and runways innorthern environments', Canadian Geotechnical Journal Vol 12, No.2.GRAHAM, J. and HOULSBY, G.T. (1983), 'Elastic anisotropy of anatural clay', Geotechnique 33, No.2.GRAHAM, J., NOONAN, M.L. and LEW, K.V. (1983), 'Yie1d states andstress-strain relationships in natural plastic clay', CanadianGeotechnical Journal, Vol 20, No.3.HEUKELOM, W. and KLOMP, A.J.G. (1962), 'Dynamic testing a~ ameans of controlling pavements during and after construction',Proc. of 1st Int. Conf. on structural design of asphaltpavements, Ann Abor, Michigan.HIGHT, D.W. (1983),'A simple piezometer probe for the routinemeasurement of pore pressure in triaxial tests on saturatedsoils', Draft report for publication.HIGHT, D.W. and STEVENS, M.G.H. (1982), 'An analysis of the CBRtest in saturated clay', Geotechnique, Vol 32, No.4.HOUSTON, W.N. and HERRMAN, H.G. (1980), 'Undrained strength ofmarine soils', Proc. ASCE, Vol 106, GT6.HYDE, A.F.L. (1974), 'Repeated load testing of soils', Ph.D.thesis, University of Nottingham.HYDE, A.F.L. and BROWN, S.F. (1976), 'The plastic deformation ofa silty clay under creep and repeated loading', Geotechnique Vol21, No.1.
240
HYDE, A.F.L. and WARD, S.J. (1985), 'A pore pressure andstability model for a silty clay under repeated loading',Geotechnique Vol 35, No.2.KIM, T.C. and NOVAK, M. (1981), 'Dynamic properties of somecohesive soils of Ontario', Canadian Geotechnical Journal Vol 18,No.3.KIMURA, T. and SAITOH, K. (1983), 'The influence of strain rateon pore pressures in consolidated undrained triaxial tests oncohesive soils', Soils and Foundations Vol 23, No.1.KIRWAN, R.W., FARRELL, E.R., HARTFORD, D.N.D. and ORR,(1982), 'The interpretation of repeated load parametersglacial subgrade from soil properties', Proc. of 5th Int.on structural design of asphalt pavements, Vall, Delft.
T.L.L.for aConf.
KIRWAN, R.W. and SNAITH, M.S. (1976), 'A simple chart for theprediction of resilient modulus', Geotechnique, Vol 26, No.1.KIRWAN, R.W., FARREL, E.R., HARTFORD, D.N.D. and ORR, T.L.L.(1982), 'The interpretation of repeated load parameters for aglacial subgrade from soil properties', Proc. of 5th Int. Conf.on structural design of asphalt pavements, Vall, Delft.
KNIGHT, K. and BLIGHT, G.E. (1965), 'Studies of some effectsresulting from unloading of soils', Proc. of 6th Int. Conf. SoilMechanics and Foundation Engineering, Montreal.KOUTSOFTAS, D.C. (1978), 'Effects of cyclic loads on undrainedstrength of two marine clays', Proc. ASCE, Vol 104, GT5.KRAHN, J. and FREDLUND, D.G. (1972), 'On total, matric andosmotic suction, Soil Science, Vol 114, Pt 5.LADD, C.C., FOOTT, R., ISHIHARA, K.,H.G. (1977), 'Stress-deformati on andProc. of 9th Int. Conf. SoilEngineering, Vol 2, Tokyo.
SCHLOSSER, F. and POULOS,strength characteristics',Mechanics and Foundation
LAMBE, T.W. and WHITMAN, R.V. (1979), 'Soil mechanics S.l.version', John Wiley and Sons.LASHINE, A.K.F. (1971), 'Some aspects of the behaviour of KeuperMarl under repeated loading', Ph.D. thesis, University ofNottingham.LEE, K.L. (1976), 'Influence of end restraint in cyclictests', U.S. Army Engineers Waterways ExperimentalVicksburg, Contract Report S-76-1.LO, K.Y. (1969), 'The pore pressure strain relationship ofnormally consolidated undisturbed clay', Canadian GeotechnicalJournal Vol 6.
triaxialStation,
241
LOUDON, P.A. (1967), 'Some deformation characterics of kaolin',Ph.D. thesis, University of Cambridge.MACKY, T.A. and SAADA, A.S. (1984), 'Dynamics of anisotropicclays under large strains', Proc. ASCE Vol 110,GT4.MATSUI, T. and ABE, N. (1981), 'Behaviour of clay on cyclicstress-strain history', Proc. of 10th Int. Conf. Soil Mechanicsand Foundation Engineering, Vol 3, Stockholm.MATSUI, T., O'HARA H., ITO, T., (1980), 'Cyclic stress-strainhistory and shear characteristics of clay', Proc. ASCE, Vol 106,GHO.
MAYNE, P.W. and KILHANEY, F.H. (1982), 'Ko-OCR relationships insoil', Proc. ASCE Vol 108, GT6.MITCHELL, J.K. (1960), 'Fundamental aspects of thixotropy', Proc.ASCE Vol 86, SM3.MITCHELL, J.K. (1969), 'On the yielding and mechanical strengthof London clays', Canadian Geotechnical Journal Vol 7, No.3.MITCHELL, R.J. AND KING, R.D. (1977), 'Cyclic loading of anOttowa area Champlain sea clay', Canadian Geotechnical JournalVol 14, No.2.MOU, C.H and CHU, T.Y. (1981), 'Soil-suction approach forevaluation of swelling potential'. Transport Research Record 790.OGAWA, S., SHIBAYAMA, T., YAMAGUCHI, H. (1977), 'Dynamic strengthof saturated cohesive soil " Proc. of 9th Int. Conf. SoilMechanics and Foundation Engineering, Vol 2, Tokyo.OSIPOV, V.I., NIKOLAEVA, S.K. and SOKOLOV, V.N. (1984),'Microstructural changes associated with thixotropic phenomena inclay soils', Geotechnique 34, No 3.OVERY, R.F. (1982), 'The behaviour of anisotropicallyconsolidated silty clay under cyclic loading', Ph.D. thesis,University of Nottingham.PAPPIN, J.W. (1979), 'Characteristics of a granular material forpavement analysis', Ph.D. thesis, University of Nottingham.PARR, G.B. (1972), 'Some aspects of the behaviour of London clayunder repeated loading', Ph.D. thesis, University of Nottingham.PERLOFF, W.H. and PAMBO, L.E. (1969), 'End restraint effects inthe triaxial test', Proc. of 7th Int. Conf. Soil Mechanics andFoundation Engineering, Mexico.POWELL, W.D., POTTER, J.F., MAYHEW, H.C. and NUNN, M.E. (1984),'The structural design of bituminous roads', TRRL LaboratoryReport 1132.
242
PUSCH, R. (1982), 'Thixotropic stiffening of clay consolidated inthe laboratory', Canadian Geotechnical Journal Vol 19, No.4.RICHARDS, B.G. and GORDON, R. (1972), 'Prediction and observationof the performance of a flexible pavement on an expansive claysub-grade', Proc. of 3rd Int. Conf. on structural design ofasphalt pavements, Vol 1, London.ROAD RESEARCH LABORATORY (1970), 'A guide to the structuraldesign of pavements for new roads', Road Note 29, HMSO.ROWE, P.W. and BARDEN, l. (1966), 'A new consolidation cell',Geotechnique, Vol 16, No.2.RUSSAM, K. (1967), 'Subsoil drainage and the structural design ofroads', TRRL, Laboratory Report 110.RUSSELL, LR. and MICKLE, J.l. (1970), 'Liquid limit by soilmoisture tension', Proc. ASCE, Vol 96, SM3.SAADA, A.S. and BIANCHINI, G.F. (1975), 'Strength of onedimensionally consolidated soils', Proc. ASCE Vol 101, GT11.SAADA, A.S., BIANCHINI, G.F. and SHOOK, L.P. (1978), 'The dynamicresponse of anisotropic clay', ASCE special Conf. Earthquakeengineering and soil dynamics, Pasadena, California.SANCHEZ, J.M. and SAGASETA, C. (1981), 'Undrained behaviour ofsoft clays in triaxial tests', Proc. of 10th Int. Conf. SoilMechanics and Foundation Engineering, Stockholm.SANGREY, D.A., HENKEL, D.J. and ESRIG, M.I.effective stress response of a saturated clay soilloading', Canadian Geotechnical Journal Vol 6, No.3.
(1969) , 'Theto repeated
SANGREY, D.A., POLLARD, W.S., EGAN, J.A. (1977),associated with rate of undrained cyclic testing of clayASTM special technical publication No 654.SAUER, LK. and MONISMITH, C.l. (1968), 'Influence of soilsuction on the behaviour of a glacial till subjected to repeatedloading', Highway Research Board Record, No.215.
'Errorssoils' ,
SCHOFIELD, A.N. and WROTH, C.P. (1968), 'Critical state soilmechanics', McGraw-Hill.SEED, H.B., CHAN, C.K. and LEE, C.L (1962), 'Resili.encecharacteristics of subgrade soils and their relation to fatiguefailures in asphalt pavements', Proc. of the 1st Int. Conf. onstructural design to asphalt pavements, Ann Arbor, Michigan.SHACKEL, B. (1973), 'Changes in soil suction in a sand-claysubjected to repeated triaxial loading', Highway Research RecordNo. 429.
243
SHAW, P. (1982), 'Stress-strain relationships for granularmaterials under repeated loading', Ph.D. thesis, University ofNottingham.SHERIF, M.A., ISHIBASHI, I. and LING, S.C. (1977), 'Dynamicproperties of a marine sediment', Proc of 9th Int Conf SoilMechanics and Foundation Engineering, Vol 2, Tokyo.SNETHEN, D.R. and JOHNSON, L.D. (1977), 'Characterisation ofexpansive soil subgrades using soil suction data', U.S. ArmyEngineer, Waterways Experiment Station, Vicksburg, Mississippi.Conf of TRB Comm. A2L06.TAVENAS, F., Des-ROSIERS, J-P., LEROUEIL, S., La ROCHELLE, P. andROY, M.(1979), 'The use of strain energy as a yield and creepcriterion for lightly overconsolidated clays', Geotechnique Vol29, No.3.TAYLOR, P.W. and BACCHUS, D.R. (1969), 'Dynamic cyclic straintests on a clay', Proc. of 7th Int. Conf. Soil Mechanics andFoundation Engineering, Vall, Mexico.TOGROL, E., OZUDOGRN, K. and GULER, E. (1979), 'On the dynamicbehaviour of overconsolidated clays', Proc. of 7th Int. Conf.Soil Mechanics and Foundation Engineering, Vol 2, Brighton.WILSON, N.E. and GREENWOOD, J.R. (1974), 'Pore pressures andstrains after repeated loading of saturated clay', CanadianGeotechnical Journal Vol 11, No.2.WOOD, D.M. (1980), Laboratory investigations of the behaviour ofsoils under cyclic loading: a review', Soil Mechanics - Transientand cyclic loads, edited by G.N. Pande and O.Z. Zienkiewicz, J.Wiley and Sons.YASHUHARA, K., YAMANOUCHI, T. and HIRAO, K. (1982), 'Cyclicstrength and deformation of normally consolidated clay', Soilsand Foundations, Vol 22, No.3.
244
APPENDIX A
MATERIAL PROPERTIES
Al SOIL CLASSIFICATION DATA
The grading curves for the Keuper Marl, Gault Clay and LondonClay are shown in Figure AI. The curves were obtained by acombination of wet sieving and the pipette method as decribed inBS 1377.
A series of standard tests were performed on each clay to obtainthe basic soil classification data given in Table AI.
A2 CRITICAL STATE PARAMETERS
The critical state parameters determined from the test results onthe Keuper Marl are given in Table A2.
245
1:::::1:: 220__~~~~r-r-r-r-r-r-~~~----------~g .._.~a::....J~~U....J
Ua::zlJJO~0..0-':;:)Z:;:)
~g<3It ~~~~~~~~~~~~~C,~ot+-+-+-+-+-+-+--+-+--+-+-+-+-t
FtC·'1J-+-~~~~~~~~~-+-4OI~~4-4-4-4-4--+-4--+-4-+-~., - • - - _. __ __ __ - - cf •••
O'~€~§~~~~~~!I~~ES~~~2C·.-fI~4-+-4-+-+-+-+-+-+-+-+-+-+-+-+-~~~~SC'C ·l-HHHI--~I--HI--HHH-t---ll-+-+-+--+--+--t
14-4-+-+-+-+-+-+-+-~+-+-+-+-+-+-~H~~
• x +
'It-~+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-r-I--~~ooc~~+-~~~+-~+-~+-+-+-+-+-+-~~~~III j ,
Oil· ~ - -- -- -- - -- - -- - - --~- ---- - -- - - -
~aRm!EmEOOdc.~r\.
o.. o• o... o.. oII
o•
I
o
-~i•IIII-•
~r-! idIS ~1&1 C2 e-tilZIt
r- I--•~o-! 0IS ~1&1 •2~tilZIt
til.,50()r-2~ iIStil2-1&1ZIt
~'--
5-
~u~
~~__J
0:::«:E0:::WCL:::>w~
WII-
0:::0LL
Z0I-:::>CO0:::I-
~(/)
0 ....J
WN(/)
~U
~0za.: «
~
UJ0::::::>~0:
246
Table Al Basic Material Properties of Soils used in the Project
Property Keuper Marl Gault Clay London Clay
Specific gravity 2.69 2.69 2.73% Clay 33 39 54Plastic limit (%) 18 25 23Liquid limit (%) 37 61 71
Plasticity Index (%) 19 36 48
Table A2 Critical State Parameters for Keuper Marl
Parameter Value
Slope of normal compression line (A) 0.1Intercept for anisotropic consolidation (No) 2.105Mean slope of swell-back lines (k) 0.035Angle of shearing resistance ( ¢ ') 28.5°Friction parameter (M) 1.13Earth pressure coefficient at rest (Ko) 0.54
247
APPENDIX B
Bl PERMEABILITY OF FILTER PAPER SIDE DRAINS
A permeability test on the filter paper side drains was carriedout in one of the triaxial cells used in the test programme. Thedrainage fittings were not modified except that the top cap wasfitted with a porous stone and an additional back pressure line.The soil sample was replaced with an impermeable rubber sample,which was installed in the cell in a similar manner to the soilsamples.
A pressure difference of 15 kPa was maintained across the filterpaper drains and the flow rate was determined for different cellpressures by using the burettes in the back pressure line.
Figure B1 shows the variation in the flow rate through the filterpaper with cell pressure. As expected the flow rate decreases asthe cell pressure increases because the cell pressure pinches thefilter paper against the rubber sample.
A permeability of 1.5 x 10-4 em/sec was estimated for the filterpaper drains at the lowest measured flow rate of 0.8 ml/hr. Thiscompares reasonably well with a permeability of 5.4 x 10-5 em/secquoted for filter paper under an effective stress of 690 kPa byBishop and Gibson (1963). Overy (1982) measured a permeability of4 x 10-6 em/sec for a filter paper drain in contact with asample of Keuper Marl.
z«a::
Lt') Cl..... a::w- o,'-
~tuL:- a::- :::>
E a:: (/')- lJJ (/')
!:J LUtu a::..... u:: o,
0« :r: ...J..... a:: ...JC) tu3: :::> os 0a:: u,u, :r: 0rw z~0ra:: uz
3: :::>9 u,
u, «.....CO
~:::>C)
0 u:0 0 0 0 0 0 0 00 0 0 0 0 0 0e- (0 io ~ (Y) N .....
(Dd~)3~nSS3~d113~
248
249
APPENDIX C
SUPPLEMENTARY TESTS
Cl PORE PRESSURE RESPONSE
Each sample was checked for saturation by comparing the change inpore pressure with a change in cell pressure. This was doneduring the consolidation phase when the cell pressure wascontrolled by an air regulator acting through an air oilinterface, and consequently the rate of increase of cell pressurewas quite low compared with the rates of loading during the testperiods.
The effect of increased rates of loading on the apparent B valuewas investigated by pulsing the cell pressure using the servohydraulic system, and monitoring the base and centre porepressure pulses. Figure Cl shows a typical plot of the measuredB value against the duration of the cell pressure pulse, andshows that centre pore pressure probe showed a value close to 1for a loading period as low as D.1 second. The response of the'base pore pressure transducer started to drop off above a loadingperiod of approximately D.5 second.
C2 COMPARISON BETWEEN AN LVDT AND A PROXIMITY TRANSDUCER
A piece of apparatus was constructed which allowed the output ofan LVOT and a proximity transducer to be compared with each otherover a small range of frequencies at small deflections. Theequipment could not be used to calibrate the transducers directly.
The equipment is shown diagrammatically in Figure C2 and consistsof an eccentric disc with D.D5mm eccentricity on which rests alever arm; the other end of which was pivoted on a knife edge.The LVOT core and proximity transducer target were attached tothe lever arm, and the transducer bodies to the frame. Theeccentric was driven by a variable speed electric motor, which
250
zzu,z0uzV')z0.....e:{
- Cl::V')
~0z0 ~u~ w
V')
..... z~
0 V')
0 wCl::
Cl::
W W wa. N
~
Cl::UJ
::)V')en V')
0 WCl:: Cl::
w a. o,Cl::..... UJ WZ
~Cl::
~ ~• • -ULUCl::::)C)
0 u,-- en co r-:- (£) Lt')
3nl'tA 9
251
0:
Z 5 J:W ~ I-> U ~0::0 0:
~I-~ U 00:
<{ 0:: ...J >I- W ...J
0: Z Z 2tu W <{ <{> Uur trl ~ w_J
0:
~~
0 I-2 W<{ ~ I-0:
~W <{0: I- 0:8 0: <{
0...lJJ 0... ...JI- U ...J0 :::> 0 <{ <{> 0 l- u, ~_J (/') UJ 0 (/')
0: 2 ~ >-e <{ < t- 2 er:0: fE ~ E 0 WI-
~
>l- E~02 >- UJ 0lJJ I-
~ I- _J 0... 0 2 0:::r: ~ CD 0 W W< 0 - Uu til~ ~
I- Z :::>(j') <{ er: 0~
0: :::> b 0... (/')0... --,
W '2~ 0 ?j er: <{w < 0::0: U I-U -(/')
~ >-~ t:
~ ~2 ~
<{~~ CCU (!) CC9 <{ 0...
0 <{
NU
Ww 6 a::LL :::>z > o~ 0... lL
252
allowed the transducer outputs to be compared over a frequencyrange from 0.4Hz to 3Hz.
Table Cl contains the results at one amplitude, equivalent toapproximately 150,..t.on the sample, over the full range offrequencies from the apparatus and demonstrates very goodagreement between the two transducers, with little effect overthe frequency range for the test. The LVOT output recorded variesby approximately 4%.
253
Table Cl Comparison between an LVDT and a Proximity Transducer
Frequency Amplitude recorded (mm)Hz LVDT Proximi ty
Transducer0.4 0.00717 0.00732
1 0.00725 0.007343 0.00748 0.00732
254
APPENDIX 0
CALIBRATIONS
01 SERVO HYDRAULIC RIG
The calibrations given fortransducers are the most recent.
the ~oad cells and pressureThe transducers were calibrated
regularly and found to vary by up to 2% at the most.
Volts/kN or volts~Pa refer to a monitor pOint.Divs/kN or divs/kPa refer to the ultra violet chart recorderwhere 1 div = 2mm
Cell Load Cells Cell Pressure Base Pore PressureTransducer Transducer
No Volts/kN Divs/kN Volts/kPa Divs/kPa Volts/kPa Divs/kPa
1 0.8673 10.915 0.0132 0.130 0.0182 0.1772 0.9017 11.34 0.0120 0.118 0.0178 0.1733 0.8636 10.869 0.00233 0.0227 0.0178 0.173
The above calibrations are given with a gain of Xl. The gain andhence the sensitivity could be increased by X2, X4 or X8.
Centre pore pressure transducer number2006 2015 2286 2402 2405 2475
Volts/kPa 0.0141 0.0139 0.0147 0.0129 0.0139 0.0139Divs/kPa 0.0746 0.0739 0.0780 0.0686 0.0739 0.0736
The above calibrations are given with a gain of Xl. The gaincould be increased by X2, X4, X8 or X16.
255
As the axial and radial deformations recorded were an average ofthe signals from the LVDTs and proximity Transducersrespectively, each type of transducer in each cell was adjustedto give the same calibration.
Transducer CalibrationsMonitor Point U.V. Recorder U.V. Recorder
(Plastic) (Elastic)Volts/mm Div/1lIII Div/mm
LVDT 0.7874 4.697 12.391P.T. 3.937 25.394 67.129
The above calibrations are given for gains of Xl. There were twooutputs to the UV recorder, one was used to record the permanentdeformation, and the second was amplified further through anoffset generator and was used to measure the resilient response.
The gain of the main amplifier could be increased by X2, X4, X8or X16 and increased the sensitivity of all the signals includingthe offset generator. The gain of the offset generator couldalso be increased by X2, X4, X8, X16 or X32.
02 PNEUMATIC RIG
Monitor Point U.V.Recorder U.V.Recorder(Plastic) (Elastic)
Load cell 1.623 mV/kN 251 Div/kN -LVDT 0.157 volts/nun 10.121 Div/mm 97.09 Div/mmStrain Loop 0.251 vol ts/nm 226.3 Div/1lIII226.3 Div/mm
The above calibrations are quoted for a gain of Xl in each case.
The definition of the Plastic and Elastic terms is as given forthe servo hydraulic rig, and the connection between the main gainand offset generator gain is also as for the servo hydraulic rig.
Monitor Point U.V. Recorder U.V. Recorder(Plastic) (Elastic)
Load cell Xl,X2,X5,XlO Xl,X2,X5,XlO -LVOT Xl,X2,X5,XlO Xl,X2,X5,XlO Xl,X2,X5,XlO,X20,X50,XlOOStrain Loop Xl,X2,X4,X8 Xl,X2,X4,X8 Xl,X2,X5,XlO,X20,X50,XlOO
The resolution of the A-D convertor was 0.000305 v/bit.
The drift of the stress and pressure transducers was found to beapproximately ± 2kPa over the longest tests. The calibration of the LVDT'svaried by up to ± 2% over the longest tests. The proximity transducerswere less stable and were found to vary by as much as ± 5%.
he Author considers that the overall accuracy of the stress and strainmeasurements was of the order of 5%.