You are on page 1of 19

1

Liquefaction potential of silts from CPTu


Dawn A. Shuttle and John Cunning

Abstract: Silt tailings (slimes) are difficult materials to test in that, like sands, it is extremely difficult to obtain undis-
turbed samples and subsequently re-establish them in a triaxial cell for element testing in a laboratory in anything ap-
proaching their in situ condition. Evaluation of silt tailing behaviour has to depend on in situ tests, and the piezocone
(CPTu) in particular. However, CPTs in silt generate substantial excess pore pressure and there is no established meth-
odology to evaluate the measured responses in terms of soil properties, as drained sand-based CPT interpretation is in-
applicable. A case history of particularly loose silt tailings is reported in which the National Center for Earthquake
Engineering Research (NCEER) liquefaction assessment method would lead to uncertainty in the liquefaction potential.
However, the extremely high CPTu excess pore pressure ratio, Bq, and low dimensionless CPT resistance, Qp, at this
site indicates liquefaction is likely occurring during pushing of the CPT. Detailed finite element simulations of the CPT
using a critical state model provided an effective stress framework to evaluate the in situ state parameter of the silt
from the measured CPT data. This framework shows that the group of dimensionless CPT variables Q(1 – Bq) + 1 is
fundamental for the evaluation of undrained response during CPT sounding. And, despite the high silt content, the in-
terpretation indicates that the tailings are indeed liquefiable.
Key words: liquefaction, CPT, silt, finite element, critical state.
Résumé : Les stériles de silt (limons) sont des matériaux difficiles à tester en ce que, comme les sables, il est extrê-
mement difficile d’obtenir des échantillons non remaniés et de les reconstituer subséquemment dans une cellule
triaxiale pour tester les éléments en laboratoire de la manière qui s’approche de leurs conditions in situ. L’évaluation
du comportement du stérile de silt doit dépendre des essais in situ, et en particulier du piézocône « CPTu ». Cepen-
dant, les essais de CPT exécutés dans le silt génèrent de substantiels excédents de pressions interstitielles, et il n’y a
pas de méthodologies reconnues pour évaluer les réactions mesurées en fonction des propriétés des sols, puisque
l’interprétation du CPT sur le sable n’est pas applicable. On fait état d’une histoire de cas de stériles de silt particuliè-
rement lâches dans lesquels la méthode d’évaluation de la liquéfaction « National Center for Earthquake Engineering
Research (NCEER) » conduirait à une incertitude dans le potentiel de liquéfaction. Cependant, le rapport de l’excédent
de pression interstitielle extrêmement élevé dans le CPTu, Bq, et la faible résistance CPT sans dimension, Qp, sur ce
site indiquent que la liquéfaction se produit semblablement au cours du fonçage du CPT. Des simulations détaillées en
éléments finis du CPT au moyen d’un modèle d’état critique a fourni un cadre de référence de contrainte effective pour
évaluer le paramètre d’état in situ du silt en partant des données mesurées de CPT. Ce cadre de référence montre que
le groupe de variables CPT sans dimension Q(1 – Bq) + 1 est fondamental pour évaluer la réponse non drainée durant
le sondage au CPT. Et, en dépit de la teneur élevée en silt, l’interprétation indique que les stériles sont en effet liqué-
fiables.
Mots-clés : liquéfaction, CPT, silt, éléments finis, état critique.

[Traduit par la Rédaction] Shuttle and Cunning 19

Introduction mine, thereby creating additional storage for the tailings. In


some cases, the tailings can be used as a material for the
In the mining industry a byproduct of the ore milling pro- dam raise construction, or the raises can be constructed on a
cess is large quantities of sand- and silt-sized particles called foundation of the deposited tailings. Today tailings im-
tailings. Typically tailings are hydraulically transported and poundments are designed and built to high engineering stan-
deposited into an impoundment. This impoundment is usu- dards and require safe containment of the tailings during,
ally formed using engineered dam structures (not always and after, the mine operating life. However, designing for
around the perimeter), constructed throughout the life of the closure was not necessarily the case in the past. Closure de-
sign of older tailings impoundments requires applying up-
Received 6 January 2005. Accepted 19 June 2006. Published dated design criteria under which the behaviour of the silt
on the NRC Research Press Web site at http://cgj.nrc.ca on tailings (colloquially called slimes) can become an issue, es-
16 February 2007. pecially with regard to liquefaction.
D.A. Shuttle.1 Department of Civil Engineering, The Assessing the liquefaction potential of silt tailings using
University of British Columbia, 6250 Applied Science Lane, standard methods is problematic. The National Center for
Vancouver, BC V6T 1Z4, Canada. Earthquake Engineering Research (NCEER) summary report
J. Cunning. Golder Associates Ltd., 500 – 4260 Still Creek
(Youd et al. 2001) presents a liquefaction evaluation ap-
Drive, Burnaby, BC V5C 6C6, Canada.
proach based on a review of case histories. The correlations
1
Corresponding author (e-mail: dshuttle@civil.ubc.ca). presented focus on “clean sand” (i.e., sand with less than 5%
Can. Geotech. J. 44: 1–19 (2007) doi:10.1139/T06-086 © 2007 NRC Canada
2 Can. Geotech. J. Vol. 44, 2007

fines), with higher fines contents soil behaviour being “cor- ner to use of this state measure in sand (Been and Jefferies
rected” relative to a “clean sand” reference. The highest 1985, 1986). The NorSand model (which is based on ψ) is
fines content explicitly considered is 35%, but tailings then shown to be a good representation of general silt behav-
slimes often have a far higher silt-sized fraction and have iour by being fitted to measured triaxial data. This constitu-
liquefied, for example at Merriespruit (Fourie et al. 2001). tive model is then used in large strain finite element
A feature of silts is that their behaviour is partially like simulations of spherical cavity expansion, which is a usual
sand and partially like clay. Similar to sands, loose silts are analogue for the CPT. The non critical state soil parameters
essentially impossible to sample and test undisturbed. Even are varied parametrically, and the computed results are pre-
using freezing and setting-up at the in situ stress in the labo- sented in a dimensionless framework for evaluation of the
ratory before thawing (e.g., Sego et al. 1994), the problems state parameter in situ and are used to infer ψ from the CPT
are that the in situ stress must be known, and further, it is data.
unclear whether the samples can truly be preserved with in- Parameters and variables are described as they are intro-
tact grain contact arrangements using freezing. Silts also ex- duced, and are listed in a “List of symbols” at the end of the
ist over a range of densities, with loose silt behaving very paper.
differently from dense silt. These factors lead to the modern
electronic piezocone penetration test (CPTu) becoming a ba-
sic reference test for assessing tailings. However, this test is Example of silt tailings behaviour
no panacea as there is no accepted methodology for evaluat-
In situ state
ing silt properties from the CPT data.
By way of background, data are presented from the Rose
CPT data in silt present two fundamental issues that must Creek Tailings Impoundment, which is located at the Anvil
be addressed for data evaluation. First, CPT soundings in Range Mining complex near Faro, Yukon. Between about
silts have substantial excess pore pressure so that standard 1969 and 1992 tailings produced from the adjacent lead and
sand-like evaluations of the data, which assume drained pen- zinc mine processes were deposited in the Rose Creek tail-
etration, are inapplicable. Second, silt density substantially ings impoundment. Figure 1 presents an aerial view photo of
affects its behaviour so evaluation of CPT data using total the impoundment taken in 1997. Mining operations ended at
stress methods to indicate an su value leaves open questions the site in 1998 following bankruptcy of the owner. The site
about what such strengths mean in terms of liquefaction or is now in the care of a Receiver, under whose direction final
other behaviours (su from the CPT is a concept associated closure is being carried out. This closure involves detailed
with Tresca-like constitutive behaviour, which is far re- engineering assessments of the tailings impoundment to al-
moved from the undrained behaviour of a liquefiable soil). low for the final closure design and the possible remediation
Use of the CPT in siltier soils to interpret soil type is not of the Rose Creek tailings impoundment.
uncommon in the literature (e.g., East et al. 1988; The Rose Creek tailings impoundment lies, as its name
Dasenbrock 2005; Yafrate and DeJong 2005), but such work suggests, in a section of the now infilled Rose Creek Valley.
has concentrated on silty sands (typically less than 20% silt). The flow of Rose Creek has been diverted upstream of the
Even at 0%–15% silt, problems have been noted with the impoundment area into a diversion canal that runs along the
standard liquefaction assessment approach (e.g., Carraro et south side of the impoundment. The impoundment is about
al. 2005). Quantitative evaluation of CPT data in true silt is 3700 m in length by 700 m in width, with a maximum tail-
rarer, typically relying on the use of site specific correlations ings thickness of about 30 m. However, the natural ground
(e.g., East et al. 1988; Lou et al. 1991), and an appropriate slope is somewhat reflected in the slope of the tailings, and
framework for such correlations is unclear. Moreover, to there is an elevation change of about 20 m in the tailings
date, CPT chamber calibration tests have not been under- surface from the highest upstream end to the crest of the
taken using silt, and correspondingly, there are no directly downstream retention dam.
measured empirical behaviour trends to underpin a method- Tailings deposition patterns evolved over the life of the
ology. mine’s operation and as the tailings impoundment was ex-
Some initial steps for evaluating the CPT in silt using an panded. Hydraulic tailings deposition resulted in separation
effective stress approach have been presented by Been et al. of the tailings to a range of gradations distributed through
(1988), Plewes et al. (1992), and Been and Jefferies (1992), the impoundment. The coarser tailings fraction was gener-
but these methods are very much first approximations and ally deposited close to the spigot discharge points, with the
indeed might be argued as speculative. On a theoretical very fine (and soft) tailings (slimes) fraction generally trav-
level, effective stress solutions for undrained CPT soundings elling to backwater pond areas furthest from the discharge
have been provided by Cao et al. (2001) and Chang et al. points. Meandering of the tailings stream on the surface pro-
(2001) using the Modified Cam Clay model. However, Mod- duces an interlayering of coarser and finer soils.
ified Cam Clay is not an appropriate model for silt as it can- The site is in a seismically active zone. A site specific
not dilate properly with dense soils nor display liquefaction seismic hazards assessment was carried out to assess various
behaviour in loose soils. closeout measures for the tailings impoundment. The analy-
This paper presents numerical simulations of undrained sis determined that a maximum credible earthquake (MCE)
CPT penetration from a variable density silt tailings deposit event of Magnitude 7, corresponding to a 10–4 per annum
to develop a mechanics-based methodology for determining probability of occurrence, would result in a median peak
the in situ state parameter of the deposit. The silt’s state is ground acceleration of 0.3g and a mean peak ground accel-
characterized by the state parameter ψ in an identical man- eration of 0.5g. One of the concerns is whether the silt tail-

© 2007 NRC Canada


Shuttle and Cunning 3

Fig. 1. Rose Creek tailings impoundment at the Anvil Range Mining Complex near Faro, Yukon.

ings (slimes) might liquefy and flowslide over the retention referred to subsequently as Ic,RW, that is used in the NCEER
dams because of the elevation of these slimes in their pres- method for liquefaction resistance of soils. The Ic,RW classi-
ent location. The soil of interest in the present context is the fies this silt on the boundary between soft clay and organic
very fine deposits found in the backwater areas. soil. Both versions of Ic are compared in Fig. 4.
An example of a CPT sounding in the backwater area is The CPT used at the site was a seismic cone, and the
shown in Fig. 2 (SCPT03–21). The basic measured CPT data shear wave velocity was measured at 1 m depth intervals.
are plotted as well as the two dimensionless standard ratios for These data, converted to shear modulus, are also shown in
soil type identification; the friction ratio, F(= fs /(qt − σv0)) and Fig. 3. The shear modulus is surprisingly high for such a
excess pore pressure ratio, Bq (= (uc − u0)/(qt − σv0)). The silt soft soil, ranging from about 55 MPa at the top of the silt to
of particular interest lies between 4 and 12 m below the sur- some 130 MPa at the base of the unit. The increase in shear
face, and it can be readily identified by the extremely high modulus with depth is close to linear and corresponds to a
Bq values induced during CPT sounding. The tip resistance constant dimensionless rigidity Ir ≈ 550 for the estimated in
during sounding was very low, less than 1 MPa, for the en- situ geostatic stress ratio K0 = 0.7. This value of rigidity, and
tire silt layer. The friction ratio was typical of silt and fluctu- its lack of depth dependence, is not atypical of the finer
ated quite tightly around F = 2%. grained tailings at the site, as shown in Fig. 5.
Close inspection of Fig. 2 indicates Bq values above unity, The investigation of the site also included a limited num-
which would suggest an increase in pore pressure greater ber of borings. One boring was put down close to the CPT,
than the increase in tip stress. Even allowing that uc and qt and the water contents of the retrieved samples were care-
are measured at different locations, values of Bq much above fully logged. These have been converted to void ratio, as-
1.0 are unlikely. It can be seen that Bq is certainly close to suming full saturation, and are also plotted in Fig. 3. The
1.0, but the numeric value is susceptible to transducer accu- void ratio averages about e = 1.1, and any trend for decrease
racy and digital discretization. with depth is obscured by local variability in the data. It
The silt has a downward seepage regime, being drained by should be noted that the samples were disturbed and may
the underlying tailings sands and alluvial foundation. In situ have increased or decreased in water content because of this
pore pressures were established through a review of pore disturbance. Correspondingly, these estimates of void ratio
pressure dissipation tests during the CPT sounding and from are somewhat approximate (which is one of the reasons why
a number of multilevel standpipe monitoring wells in the im- the CPT was required).
poundment installed to investigate water quality. The silts re- The silt soils were nonplastic. A particle size distribution
main saturated as the surface water resupply rate exceeds the representative of the finer tailings is shown in Fig. 6, and it
underdrainage. This is further confirmed by the extremely can be seen that the soil is about 66% silt-sized particles
high Bq, which could only occur with a saturated soil. with the remaining being fine sand.
The CPT data were processed to give the soil type index,
Ic(= f(Q, Bq, F); Been and Jefferies 1992), and this index is Triaxial compression behaviour
plotted against depth in Fig. 3. Note that the Ic used here in- It is well known that the stress–strain response of sands
cludes a “+1” in the log term and hence differs slightly from and silts is fabric dependent prior to critical state (e.g.,
that defined in Jefferies and Davies (1993). As might be an- Vaid et al. 1999; Høeg et al. 2000). However, this does not
ticipated from the very high Bq, the silt classifies on a soil preclude the use of critical state models to represent sand
behaviour type basis as very soft clay to organic soil. Rob- and silt behaviour. The effect of fabric may be represented
ertson and Wride (1998) adopted an alternate definition of in a plasticity based constitutive model by changing the
Ic, which neglects the pore pressure. It is this simplified Ic, plastic hardening modulus (e.g., Been et al. 1991). This is

© 2007 NRC Canada


4 Can. Geotech. J. Vol. 44, 2007

Fig. 2. Example of CPT data in silt tailings deposit (silt layer of interest lies between 4 and 12 m depth).

Fig. 3. Soil type index from CPT, in situ void ratio from borehole sample water contents, and in situ shear modulus.

© 2007 NRC Canada


Shuttle and Cunning 5

Fig. 4. Comparison of the Soil Classification Index, Ic, from Been and Jefferies (1992), and Ic,RW from Robertson and Wride (1998).

Fig. 5. In situ shear rigidity versus mean effective stress. Fig. 6. Particle size distribution data for tailings slimes in
4–12 m depth zone of Fig. 1.

the fits in Fig. 7, with the effect of fabric on the stress–strain


response being well captured by plastic hardening only.
In the following analyses laboratory testing was carried
illustrated in Fig. 7 for Erksak sand. In this example, the out to determine the critical state parameters Γ, λ10, and Mtc
critical state (CS) parameters, Γ, λ10, and Mtc are obtained of the fine tailings. The tested soil was prepared from a mix-
from the moist tamped sample. The water pluviated triaxial ture of several samples that were estimated to be representa-
sample is then fitted keeping all three critical state parame- tive of the in situ slimes. The use of representative samples
ters constant (i.e., with a unique critical state locus, CSL). to obtain tailings material properties is not uncommon; tail-
Typically both plastic hardening and shear modulus could be ings are typically highly stratified, making it difficult to ob-
expected to vary with fabric, but this was not necessary for tain homogeneous samples that are large enough for testing.

© 2007 NRC Canada


6 Can. Geotech. J. Vol. 44, 2007

Fig. 7. Example of soil fabric represented by hardening modulus, H.

The samples were dried and then combined to create a bulk tamped samples has been questioned (e.g., Vaid et al. 1999;
sample for the testing program. Frost and Park 2003), Been et al. (1991, 1992) have shown
The specific gravity (Gs) was determined using proce- that the CSL obtained using average sample void ratios from
dures following ASTM Standard D854 (ASTM 2006a), to more uniform wet pluviation and moist tamping is indistin-
give Gs = 3.97. This rather high value (relative to the more guishable within experimental resolution. Moist tamped
usual experience of natural soils) is explained by the heavy samples are unavoidable primarily because laboratory
nature of lead–zinc ore tailings. Minimum and maximum pluviation procedures do not produce samples as loose as
void ratios of 0.837 and 2.017 were determined following those found from pluviation at field scale, and at least some
ASTM Standards D4254 (Method A; 2006c) and D4253 very loose samples are needed to fully define the CSL.
(Method 2A; 2006b), respectively. However, the high fines Samples were prepared very loose and were saturated, fol-
contents of the silt tailings are greater than the 15% by dry lowing assembly of the triaxial cell, while maintaining a
mass limit recommended in these standards. Hence, these positive effective confining stress of at least 20 kPa.
limit values have only been determined as an estimate of the Changes in sample volume during this saturation phase were
expected range in void ratio and could not be expected to be tracked by monitoring the change in volume of fluid confin-
used in a relative density approach for these silts. ing the sample in the triaxial cell. The procedure to saturate
A total of four isotropically consolidated triaxial tests were the sample included carbon dioxide (CO2) treatment, initial
undertaken on reconstituted samples prepared by the moist saturation with de-aired water, and final back pressure satu-
tamping method; two being undrained and two drained. ration. The CO2, which is many times more soluble in water
Moist tamping was required to produce contractive samples than air, is first percolated under low pressure through the
that will reach a clear critical state within the strain limits of voids in the moist sample. This is followed by an initial sat-
the triaxial equipment. Although the homogeneity of moist uration stage in which de-aired water under very low pres-

© 2007 NRC Canada


Shuttle and Cunning 7

sure head is allowed to flow into the base of the sample and tuted samples, the in situ critical void ratio for the depth
continued until the flow returns through the top of the sam- range of 5–11 m would be between about 0.76 and 0.71.
ple and pore pressure lines. The last step is back pressure Thus the in situ state parameter is about +0.3 (depending on
saturation and B value testing, carried out with cell pressure stress). Such very positive states are normally catastrophi-
increments of between 30 and 35 kPa. During each cell pres- cally liquefiable under minor static triggering, and they are
sure increase the pore pressure response is monitored to cal- certainly exceedingly liquefiable in any credible design
culate the B value. Samples were considered saturated with a earthquake scenario.
B value greater than 0.96. Following this saturation stage the A different conclusion about liquefaction follows from the
samples were then isotropically consolidated (in several NCEER method, based on Robertson and Wride’s (1998)
steps) to the desired test consolidation stress. During consol- CPT-specific liquefaction methodology. This prescribes the
idation volume changes were recorded to continue to track soil as likely nonliquefiable if Ic,RW > 2.6 and F > 1%. As
void ratio. can be seen from Fig. 4, the CPT results indicate that the
In determining the critical state line, the main concern for Robertson and Wride based inference of liquefaction is
specimen preparation is that the void ratio of the samples be “likely no liquefaction potential”.
known. Following displacement controlled shearing to reach Which one of these conclusions is correct? Clearly it must
the critical state, the final sample void ratio was determined be the conclusion that the soil is highly liquefiable. The
using the total sample freezing method. This method allows denser samples in the laboratory show very contractive be-
for accurate final void ratio determination at the critical state. haviour even though they only exhibit a small post-peak re-
Determining the final void ratio by freezing allows for a duction in undrained strength. Also Bq > 1 is consistent with
check on the void ratio calculated by monitoring the sample an essentially complete loss of strength from shearing
volume change from preparation, saturation, consolidation, around the CPT tip (i.e., local static liquefaction induced by
and shearing. Even though there were large volume changes CPT penetration alone). There is no need for recourse to the
during the initial water saturation stage on these tests, there “Chinese criteria” (Seed and Idriss 1982), this conclusion
was good agreement with the void ratio determined by accu- can be developed from mechanics.
rate tracking of sample volume and that determined by the
sample freezing method (Sladen and Handford 1987). Numerical modelling of CPT
No membrane penetration corrections have been applied
to the calculated void ratio data. These procedures complied Framework
in all respects with the methodology presented by Been et al. Although it may seem appealing to model the actual geom-
(1991) for accurate determination of a CSL. etry of a CPT in finite element simulations, and a few work-
Figure 8 presents a summary of the test program, showing ers have attempted this (e.g., van den Berg 1994; Yu et al.
the stress and state paths followed for all four samples, to- 2000; Lu et al. 2004), there are still difficulties with large dis-
gether with the stress–strain behaviour measured during placement formulations. To date most of these simulations
shear. The CSL has been estimated by fitting a semi- have focused on very simple soil constitutive models with un-
logarithmic trend line through the end points of the void ra- realistic dilatancies. Understanding is therefore based on the
tio versus mean effective stress and on the mean effective classical spherical cavity analogue. This analogue reduces a
stress versus deviator stress as shown in Figs. 8 and 9. The 3D analysis to one space dimension and allows elegant and
CSL is well defined, suggesting that any nonuniformity of simple large displacement analysis within fast numerical
the moist tamped samples had little effect on the final, criti- codes. There is also no difficulty with using advanced soil
cal state, void ratios. Based on this fitting, the critical state models in large strain spherical cavity expansion, and un-
parameters determined were Γ = 1.076 (at 1 kPa), λ10 = drained liquefaction induced by cavity expansion is readily
0.159, and Mtc = 1.25. modelled.
The layer of principal interest, 4–12 m depth, has an in The extent to which spherical cavity expansion is an ana-
situ mean effective stress range of approximately 70 kPa < logue for CPT penetration was tested experimentally by
p′ < 210 kPa. The mean stress for critical conditions deter- Ladanyi and Roy (1987). Their key result is presented in
mined in the laboratory triaxial tests lie within the range Fig. 10, wherein the resistance of a blunt penetrator (i.e.,
100 kPa < p′ < 1350 kPa. Hence, the semilogarithmic ideal- pile or CPT) is compared to the limiting cavity expansion
ization for the CSL shown in Fig. 9 was assumed reasonable pressure across a range of sand strengths. Penetrometer re-
for modelling the in situ tailings behaviour. sistance and cavity pressure are similar functions of friction
angle, but the two are not equal; there is approximately a
Liquefaction potential factor of two between the two situations. Similarly, in the
It is noteworthy that it was not possible to test these re- case of a total stress approach using the Tresca model, it is
constituted silt samples at anywhere near their in situ void widely appreciated that the theoretical Nk ≈ 9 (the actual
ratio as estimated from in situ water contents. There was value depends on Ir) derived from spherical cavity expansion
substantial contraction of the samples during sample satura- is rarely found in practice relating undrained strength to
tion, with additional contraction occurring during consolida- CPT penetration resistance. A common range is 10 < Nk <
tion as shown in Fig. 9. The samples were therefore tested in 25, which is about the same divergence between cavity ex-
shear starting from void ratios in the range 0.7 < e < 0.8. pansion theory and reality as found with the drained experi-
These void ratios can be compared with in situ void ratios ments of Ladanyi and Roy (1987). In summary, cavity
shown in Fig. 3, which lie in the range 0.9 < e < 1.3. expansion is only an analogue for the CPT; calibration is re-
Broadly, using the CSL determined on the blended reconsti- quired to obtain the appropriate parameter values. Neverthe-

© 2007 NRC Canada


8 Can. Geotech. J. Vol. 44, 2007

Fig. 8. Summary of triaxial compression tests on silt tailings.

Fig. 9. Summary of consolidation curves and CSL determination Fig. 10. Comparison of experimental spherical cavity limit pres-
from triaxial compression tests on silt tailing. sure with penetration resistance of blunt indenter (after Ladanyi
and Roy 1987).

less, cavity expansion is an excellent basis to investigate the associated Mohr Coulomb materials (Carter et al. 1986;
effects of parameters and to develop the framework in which Houlsby and Yu 1991). Drained spherical cavity expansion
CPT data should be evaluated and normalized. solutions in realistic soil models have been provided by Col-
The reduction of the problem from a true 3D to spherical lins et al. (1992) and Shuttle and Jefferies (1998). The latter
symmetry necessarily removes any role for geostatic stress work provided substantial support to the trend inferred by
ratio K0 in the computed results. This may not be a defi- Been et al. (1986, 1987a, 1987b) from chamber testing of
ciency however, as the data obtained with sands in calibra- sands. The chamber data of the CPT penetration of sand in-
tion chamber studies show that there is negligible influence dicates that dimensionless CPT resistance Qp(= (qt − p0)/ p0′)
of K0 in itself, provided that mean stress rather than vertical is related to the soil’s in situ state parameter ψ 0 by the sim-
stress is used in normalizing the data (Clayton et al. 1985; ple equation
Been et al. 1986). It is assumed that this experimental result [1] Qp = k exp (− m ψ 0)
continues to be true in the undrained situation.
Semiclosed form large strain spherical cavity solutions in which k and m are soil-specific coefficients. Equation [1]
have been developed for drained cavity expansion in non- only applies to drained soundings. Although this equation
© 2007 NRC Canada
Shuttle and Cunning 9

relating CPT resistance and ψ0 is simple and consistent, and Fig. 11. Definition of state parameter ψ and overconsolidation
has a proper dimensionless approach, it has been criticized ratio R (after Jefferies and Shuttle 2002).
on the grounds that the reference chamber test data indicate
significant bias with stress level. At least k, and possibly m,
are functions of stress level (Sladen 1989a, 1989b). Detailed
numerical simulations by Shuttle and Jefferies (1998)
showed that eq. [1] provides an accurate representation of
the relationship between drained CPT penetration resistance
and state. The observed stress level bias was caused by treat-
ing k and m as constants. These parameters should be func-
tions of G/p′; this was neglected in the work of Been et al.
(1986, 1987a, 1987b).
The question of how to estimate ψ in undrained CPT
soundings was addressed independently by Been et al.
(1988) and Houlsby (1988). Houlsby2 (1988) noted that the
group of dimensionless CPT variables Q(1 – Bq) + 1, where
Q = (qt − σv0)/ σv0
′ , corresponded to a simplification of the
methodology for measuring overconsolidation of clay devel-
oped by Konrad and Law (1987), and that it is equivalent to
normalizing CPT resistance by the vertical effective stress
established during CPT penetration as a consequence of
CPT-induced excess pore pressure. Been et al. (1988) sug-
gested that Q(1 – Bq) + 1 was related to the soils state pa-
rameter analogously to eq. [1] as changing state parameter. Second, the critical state does not
usually intersect the yield surface. This divergence of yield
[2] Qp(1 − Bq ) + 1 = k exp (− m ψ) surface from critical state is used as the basis of the harden-
ing law, and the hardening law acts to move the yield sur-
It is noted here that the soil type index Ic introduced ear- face towards the critical state under the action of plastic
lier is also based on the same Q(1 – Bq) + 1 grouping. shear strain, which directly captures the essence of critical
state principles. Whether the yield surface hardens or softens
NorSand depends on two things: the current state parameter and the
In choosing a constitutive model to represent sands and direction of loading. Loading past the internal cap always
silts, the goal is to predict the spectrum of behaviours caused softens the yield surface, as does principal stress rotation.
by changes in the soil’s void ratio and confining stress level. NorSand is a sparse model with just seven soil model pa-
There are now several such models in the literature, all rameters, all of which are dimensionless. Three parameters
of which are based on the state parameter, ψ. NorSand (Mtc, Γ, λ10) are used to define the critical state. Two param-
(Jefferies 1993; Jefferies and Shuttle 2002, 2005) was the eters are associated with plastic hardening, χ determining
first of such models and is used here. the influence of ψ on maximum dilatancy and H being the
NorSand is a critical state model and is based on some plastic hardening modulus. Finally, two properties, Ir and ν,
simple idealizations about the dissipation of plastic work; define elasticity.
the widely known Cam Clay model (Schofield and Wroth The version of NorSand used in the present work corre-
1968) is a special case of the more general NorSand model. sponds to the modified version given in Jefferies and Shuttle
Perhaps the most striking feature of NorSand is its recogni- (2002). Further details, including procedures used to deter-
tion of an infinity of normal compression loci (NCLs). An mine the properties and validation examples over a range of
infinity of NCL forces two parameters to characterize the stress paths, can be found in Jefferies and Shuttle (2005).
state of a soil: ψ and R. The state parameter, ψ, is a measure
of the location of an individual NCL in e–p state space. The Finite element formulation
overconsolidation ratio R represents the proximity of a state The CPT is idealized as a cavity in an infinite uniform
point to its yield surface. The difference between ψ and R is medium under an isotropic stress state, p0, with the internal
illustrated in Fig. 11. effective pressure of the cavity, p0, initially equal to the iso-
NorSand is a plasticity model for soil. As such, and in tropic stress. The cavity is expanded by a monotonically
common with other plasticity models, it consists of: (i) a increasing radius until a limiting (constant) pressure is ob-
yield surface; (ii) a flow rule; and (iii) a hardening law. tained, this being the pressure of interest. This idealization
Yield surfaces in NorSand have the familiar bullet-like shape greatly simplifies the analysis because the spherical symme-
of the classical Cam Clay model, and examples for both try allows only radial displacements and in turn this permits
loose and dense soils are illustrated in Fig. 12. There are two a 1D description of the problem. The corresponding stresses
aspects where NorSand differs from standard critical state are a radial and two equal hoop stresses.
models. First, there is an internal cap, which is how The 1D problem was analyzed using 140 elements, with a
NorSand gives realistic dilatancy despite being an associated single stress sampling location at the centre of each element.
model. The position of the internal cap evolves with the The problem has no intrinsic measure of scale and so, for con-
2
The paper gave Q(1 – Bq) – 1; this is apparently a typographical error as the derivation given leads to the “+1” term.
© 2007 NRC Canada
10 Can. Geotech. J. Vol. 44, 2007

Fig. 12. Illustration of NorSand yield surfaces and limiting stress ratios (after Jefferies and Shuttle 2005). (a) very loose soil; (b) very
dense soil.

venience, the original cavity radius was set equal to unity. The model time-dependent material behaviour and was then
outer boundary was set as a zero displacement node at a dis- adapted to represent plasticity. In this context viscoplasticity
tance of 550, chosen to avoid artefacts of boundary conditions. refers to a numerical technique and does not involve any time
To capture the rapid variation of stress close to the CPT, dependence in the strength properties of the soils. The visco-
the element spacing was set to logarithmically increase with plastic solution technique involves a conventional elastic
distance from the cavity. Similarly, a second order numerical analysis with material nonlinearity (i.e., plasticity) introduced
difference was used to extrapolate the element stresses to the by iteratively modifying the loading vector. The loads vector
cavity wall to accurately capture the high stress gradients at each load step consists of the externally applied loads and a
present at low strains. set of self-equilibrating body loads. Body loads are used to
The spherical analyses were carried out using a large strain redistribute stresses within the system, without altering the
viscoplastic finite element research code (Shuttle and Jefferies external loading. In the finite element code developed for
1998). The viscoplastic algorithm was originally developed to cavity expansion analysis, the incremental viscoplastic formu-

© 2007 NRC Canada


Shuttle and Cunning 11

lation (Zienkiewicz and Cormeau 1974) was implemented us- Fig. 13. Comparison of NorSand simulation with drained triaxial
ing the general approach described in Smith and Griffiths tests on silt (H = 25, Ir = 150, ν = 0.15, ψ 0 = +0.055).
(1998). The accuracy and stability of this viscoplastic for-
mulation for critical state models is excellent with verifica-
tion examples showing deviations of less than 0.1% between
the finite element solution and the reference case (Shuttle
2004). Undrained behaviour was obtained in the effective
stress analysis using the standard numerical approach of as-
signing no volume change through the relationship ε vp = − ε ve
(which follows from setting ε v = 0 within an elastic–plastic
strain decomposition).

Numerical results for cavity expansion


Calibration of NorSand to tested silt
The purpose of the triaxial testing of the silt tailings was
twofold: (i) to obtain critical state parameters (Mtc, Γ, and
λ10 ), which are independent of state and fabric; and (ii) to
validate NorSand’s ability to represent general silt behav-
iour; it is not a calibration to perceived field conditions. It is
well known that non critical state parameters can be fabric
dependent, and hence the laboratory derived values are
likely not applicable to in situ conditions. This parameter
uncertainty is addressed later through parametric variation of
soil properties.
Calibration of NorSand to laboratory test data starts with
determination of the CSL. The relevant data were discussed
earlier and are presented in Fig. 9, giving the properties Γ =
1.076, λ10 = 0.159 from the best-fit trend through the end
points of state paths in triaxial compression tests on very
loose samples. Similarly, the critical friction ratio is obtained ple’s void ratio prior to the shearing stage of the test and the
from the end points of the stress paths in the same tests, giv- idealized CSL. An excellent fit is obtained in both cases; the
ing Mtc = 1.25 (see Fig. 8). parameters for these fits being shown in the figures. Hence,
The dilatancy property χ is ideally determined from one NorSand is an appropriate constitutive model for represent-
or more very dense triaxial tests and is done by calculating ing silt. Note that the rigidity values for both tests are lower
the ratio of maximum dilatancy to the test’s state parameter than those measured in situ. As discussed previously, this is
(i.e., χ = Dmin / ψ). Two or more tests are used to provide re- to be expected because of aging and fabric effects.
dundancy in the parameter estimate. However, no tests were
available on dense drained silt samples. Therefore, χ was de- Cavity expansion of tested silt
termined as part of the procedure to determine H, using iter- Simulations were carried out for a range of soil properties
ative modelling. to both model the anticipated in situ conditions as well as to
The value of H is determined from modelling drained step out from these conditions to more clearly define the
triaxial tests. The calibration procedure used is to guess H, trends. These simulations then provide a relationship be-
compute the entire stress–strain behaviour, and compare the tween the dimensionless CPT parameters Q, Bq and the ini-
computed behaviour to that measured. The value of H is tial state (which is of course the parameter of interest in
then revised and the process repeated until a best fit is ob- liquefaction assessment). Seismic velocity was measured at
tained. The χ parameter was also varied; the best fit was ob- 1 m depth intervals during the CPT sounding to give Ir, us-
tained with a typical value for sands of χ = 4.0. ing the mean effective stress estimated from the bulk unit
The laboratory tests did not include measurements of weight of the soil (22 kN/m3) and the depth.
shear modulus using bender elements, and so shear modulus In these simulations five soil properties in the NorSand
was determined from fitting the stress path for undrained model were kept constant: Γ = 1.076, λ10 = 0.159, Mtc = 1.25
tests and assuming a Poisson’s ratio ν = 0.15 (an uncontro- (equivalent to ϕcv ≈ 31° in triaxial compression), χ = 4.0,
versial choice for an effective stress model). For the CPT in and ν = 0.15. Three of the five parameters (Γ, λ10, and Mtc)
situ, the shear modulus was measured during the soundings. are critical state properties that are independent of fabric ef-
The modulus measured in situ should not be used for cali- fects, and therefore can be inferred from the triaxial testing
bration to laboratory samples because these samples are on reconstituted samples. Both χ and ν vary over a very nar-
unaged, as well as likely having a different fabric because of row range and therefore were neglected in the sensitivity
the differing deposition conditions. analysis. Two properties were varied, Ir and H; Ir was varied
The fit of NorSand to the tested silt is shown in Fig. 13 because the in situ rigidity from the seismic measurements
for a drained test and in Fig. 14 for an undrained test. The differed from the value needed to model the laboratory tests
initial state parameter values were determined from the sam- on reconstituted samples. Such variance is not surprising,

© 2007 NRC Canada


12 Can. Geotech. J. Vol. 44, 2007

Fig. 14. Comparison of NorSand simulation with undrained triaxial tests on silt (H = 25, Ir = 110, ν = 0.15, ψ 0 = +0.075).

but does require consideration. In the case of the plastic sure during undrained shear, can be accounted for by simply
hardening modulus, H, it is clearly speculative to assume using the measured pore pressure itself. This plot provides a
that the soil structure in the reconstituted samples (which is basis to evaluate the in situ test data from the site.
what a plastic hardening modulus characterizes in any plas- As noted earlier, Houlsby (1988) suggested that Qp(1 –
ticity model) is the same structure as found in situ. There- Bq) + 1 was fundamental to the evaluation of undrained CPT
fore, H was taken as a parameter to be varied. Finally, data using an effective stress framework. Houlsby’s sugges-
different stress levels were simulated to verify that there was tion was based on arguments for appropriate combinations
no stress-level bias. for dimensionless groups, and the results reported here are
Spherical cavity expansion needs to continue to about believed to be the first complete numerical substantiation us-
double the initial cavity radius before the total and effective ing a realistic constitutive model. Figure 17, anticipated by
cavity expansion pressures stabilize at their limiting value Houlsby, is a remarkable result. Of course, k and m in
for the parameter–property choice being simulated. In devel- eq. [2] will vary with soil specific coefficients such as Mtc
oping the relationship between soil state, soil properties, and and λ10. But Fig. 17 provides a practical, mechanics-based
CPT resistance it is this large displacement limiting value methodology with which to determine soil state from the
during cavity expansion that corresponds to the CPT data. CPT in silts.
Thus, although a large numerical effort is needed, only two
results are relevant from each simulation. In situ tailings state and liquefaction potential
Table 1 summarizes the property–parameter cases simu- Before Fig. 17 can be used to estimate ψ 0 in situ from
lated and the computed soil response to cavity expansion. CPT data, allowance must be made for the limitation of the
The initial state parameter was stepped over six nominal val- spherical cavity idealization and the effect of nonisotropic
ues but with slight variations in each case so that the result- conditions in situ.
ing data points would never plot on top of each other when As discussed previously, there is evidence that the spheri-
comparing trends against ψ 0. cal cavity idealization underestimates CPT resistance by a
Figure 15 shows the trend in Qp values with ψ 0. The ef- factor of about 2. As a first estimate, it would not be unrea-
fect of Ir at constant H is highlighted in Fig. 15a, while sonable to double the normalized k, giving k = 3.7, but leave
Fig. 15b presents the opposite effect of variable H at con- m unchanged. The inferred trend between ψ 0 and eq. [2], in-
stant Ir. Figures 16a and 16b show the same trends but now cluding the factor of two correction, is also shown in
for the computed Bq values. In the case of Ir variation, there Fig. 17. A tacit assumption here is that the excess pore pres-
are unique trends in Qp and Bq for each value of Ir, regard- sure ratio computed by the spherical cavity is the same as
less of stress level. Similarly, there are unique trends in Qp the Bq ratio measured with the CPT for a “shoulder”
and Bq for any choice of H at constant Ir. mounted piezometric element.
The numerical results can be normalized using the Calibration chamber data in sands, also discussed previ-
dimensionless form of eq. [2]. Figure 17 plots all of the results ously, shows that K0 has no effect on the CPT resistance pro-
of the numerical simulations as Qp(1 – Bq) + 1 versus ψ 0. This vided that the CPT data is normalized by the mean rather
plot brings results with differing Ir and H values onto a single than the vertical effective stress. That is Qp = 3Q/(1 + 2K0).
trendline. Further, this trendline has the anticipated Figure 18 shows the measured CPTu data previously pre-
semiexponential form within a rather small tolerance, the sented in Fig. 2 and now transformed to the normalized
trendline drawn in Fig. 17 being that from eq. [2] and using dimensionless variable Qp(1 – Bq) + 1. The scales have been
k = 1.86 and m = 14.6. This figure indicates that the change in expanded on the plot so as to window in to the silt stratum
elastic and plastic properties, which affect the excess pore pres- of interest. As can be seen, the normalized resistance is

© 2007 NRC Canada


Shuttle and Cunning 13

Table 1. Summary of simulation parameters and computed spherical cavity results.

ψ0 p′ (MPa) H Ir qt (MPa) uc (MPa) Qp Bq Qp(1 – Bq) + 1


0.079 0.050 25.0 100 0.164 0.135 2.285 1.180 0.589
0.052 0.050 25.0 100 0.195 0.151 2.898 1.045 0.870
0.026 0.050 25.0 100 0.211 0.147 3.213 0.917 1.267
0.001 0.050 25.0 100 0.227 0.137 3.545 0.774 1.802
–0.020 0.050 25.0 100 0.246 0.123 3.926 0.626 2.468
–0.050 0.050 25.0 100 0.269 0.080 4.386 0.365 3.784
0.076 0.050 25.0 250 0.173 0.142 2.456 1.160 0.608
0.054 0.050 25.0 250 0.195 0.153 2.902 1.057 0.834
0.029 0.050 25.0 250 0.209 0.149 3.181 0.936 1.204
0.004 0.050 25.0 250 0.226 0.139 3.517 0.789 1.741
–0.025 0.050 25.0 250 0.259 0.127 4.176 0.610 2.627
–0.047 0.050 25.0 250 0.279 0.098 4.573 0.428 3.616
0.076 0.050 62.5 250 0.178 0.148 2.568 1.153 0.606
0.051 0.050 62.5 250 0.214 0.170 3.281 1.037 0.878
0.025 0.050 62.5 250 0.243 0.179 3.853 0.929 1.273
0.003 0.050 62.5 250 0.274 0.186 4.479 0.831 1.757
–0.022 0.050 62.5 250 0.321 0.195 5.422 0.720 2.519
–0.045 0.050 62.5 250 0.373 0.196 6.468 0.605 3.557
0.080 0.500 25.0 100 1.642 1.345 2.284 1.178 0.593
0.050 0.500 25.0 100 1.963 1.505 2.926 1.029 0.916
0.026 0.500 25.0 100 2.109 1.461 3.218 0.908 1.297
0.003 0.500 25.0 100 2.267 1.361 3.535 0.770 1.812
–0.020 0.500 25.0 100 2.447 1.186 3.893 0.609 2.521
–0.048 0.500 25.0 100 2.681 0.814 4.361 0.373 3.732
0.079 0.500 25.0 250 1.755 1.478 2.510 1.178 0.554
0.054 0.500 25.0 250 1.975 1.556 2.950 1.055 0.838
0.028 0.500 25.0 250 2.113 1.485 3.225 0.921 1.255
0.003 0.500 25.0 250 2.278 1.372 3.557 0.771 1.813
–0.024 0.500 25.0 250 2.548 1.211 4.096 0.591 2.673
–0.046 0.500 25.0 250 2.761 0.932 4.522 0.412 3.658
0.079 0.500 62.5 250 1.813 1.530 2.626 1.165 0.567
0.052 0.500 62.5 250 2.155 1.720 3.311 1.039 0.870
0.025 0.500 62.5 250 2.462 1.807 3.925 0.921 1.310
0.003 0.500 62.5 250 2.783 1.868 4.565 0.819 1.828
–0.024 0.500 62.5 250 3.283 1.923 5.565 0.691 2.720
–0.048 0.500 62.5 250 3.828 1.909 6.657 0.574 3.838
0.077 0.500 137.5 550 1.981 1.695 2.961 1.145 0.571
0.050 0.500 137.5 550 2.358 1.912 3.717 1.029 0.892
0.023 0.500 137.5 550 2.794 2.116 4.587 0.923 1.355
0.001 0.500 137.5 550 3.271 2.321 5.542 0.838 1.900
–0.026 0.500 137.5 550 4.051 2.627 7.101 0.740 2.847
–0.050 0.500 137.5 550 4.937 2.914 8.875 0.657 4.047
0.080 0.050 137.5 550 0.184 0.155 2.677 1.160 0.571
0.050 0.050 137.5 550 0.231 0.187 3.627 1.031 0.888
0.020 0.050 137.5 550 0.279 0.210 4.585 0.918 1.377
0.000 0.050 137.5 550 0.322 0.230 5.435 0.845 1.843
–0.020 0.050 137.5 550 0.379 0.256 6.589 0.778 2.465
–0.055 0.050 137.5 550 0.507 0.302 9.139 0.661 4.100
Note: ψ0, initial state parameter; p′, mean effective stress; H, dimensionless plastic hardening modulus for loading; Ir,
dimensionless elastic shear rigidity; qt, total tip resistance; uc, spherical limit pore pressure assumed proportional to the pore
pressure measured by CPT during sounding (at shoulder location); Qp, dimensionless CPT resistance based on mean stress as
required by mechanics; Bq, CPTu excess pore pressure ratio.

close to unity over much of the weak silt zone. Application The in situ void ratio of the silt was estimated from log-
of the inverted form of eq. [2], and using the estimated coef- ging the water content of samples in the adjacent boring and
ficients m and k, gives the profile of in situ state parameter then converting this water content to void ratio assuming full
also shown in Fig. 18. Broadly, the in situ state of the weak saturation. These void ratios can be combined with esti-
silt is estimated to lie in the region +0.08 < ψ < +0.12. mated critical void ratios from the laboratory testing of re-

© 2007 NRC Canada


14 Can. Geotech. J. Vol. 44, 2007

Fig. 15. Results of numerical simulations for normalized CPT resistance. (a) effect of soil rigidity; (b) effect of plastic hardening.

constituted samples to give alternative estimates of ψ in situ. pected for ψ ≈ +0.1 based on Fig. 16. There is consistency
Values calculated in this way are also shown in Fig. 18 and in the ψ estimated from the spherical cavity analogue. On
they indicate that +0.15 < ψ < +0.38. the other hand it is possible that the spherical cavity ana-
The state estimated from the CPT is not quite as loose as logue has underestimated the CPT resistance by more than
that inferred from water content measurements, although the estimated CPT:spherical factor (which could arise while
both methods indicate a soil markedly looser than its critical still preserving consistency in the computed behaviour).
state. In looking to resolve this ambiguity, it can be noted However, the large strain formulation used in the present
that the Bq measured with the CPTu (about unity for the ho- work is conventional and widely used by others working on
rizon in question) is actually a little less than the Bq ex- cavity expansion idealizations. The undrained implementa-

© 2007 NRC Canada


Shuttle and Cunning 15

Fig. 16. Results of numerical simulations for normalized CPT pore pressure. (a) effect of soil rigidity; (b) effect of plastic hardening.

tion of NorSand was also verified by comparing NorSand Another potential error could be in the cavity expansion
parameters chosen to mimic the Tresca soil model (ψ = 0, idealization itself. Ladanyi and Foriero (1998) suggest that
R = 2.7), with the analytical Tresca solution. The solutions there are inconsistent aspects with such formulations in that
limiting cavity expansion pressure differed by approximately large strain measures (ψ, M, H) are combined with Ir, which
20%; this difference being caused by drift in the NorSand is a ratio based on small strains (rigidity). Certainly using a
pore pressures. The difference between solutions is small spherical:CPT factor of three rather than two would bring
and does not sufficiently account for the difference in in- the value of ψ estimated from CPT much closer to that esti-
ferred ψ. mated from water contents.

© 2007 NRC Canada


16 Can. Geotech. J. Vol. 44, 2007

Fig. 17. Normalized results of numerical simulations (all 48 simulation results plotted).

Fig. 18. Comparison of ψ inferred from CPTu with ψ inferred from water content of disturbed samples combined with CSL of recon-
stituted silt.

© 2007 NRC Canada


Shuttle and Cunning 17

Fig. 19. Example of undrained triaxial behaviour that is representative of the inferred in situ conditions from CPTu (ψ 0 = +0.1,
Γ = 1.076, λ 10 = 0.159, Mtc = 1.25, χ = 4.0, Ir = 550, ν = 0.15).

A more likely reason for a mismatch, as discussed above, astrophically brittle. Interestingly, this is exactly the soil
is that in calculating Bq it is tacitly assumed that the excess behaviour type indicated from the classification index Ic
pore pressure ratio computed by the spherical cavity is the (Fig. 3).
same as the Bq ratio measured with the CPT for a “uc” or
“shoulder” mounted piezometric element. This is unproven,
and adding a factor to Bq to account for geometry effects on Conclusion
the pore pressure would change the inferred CPT trend line
A framework for the evaluation of in situ state for silt tail-
in Fig. 17.
ings has been presented that combines in situ CPT measure-
Finally, the water-content based ψ estimates are not neces- ments, laboratory testing of the silt tailings, and numerical
sarily correct. A maximum inferred ψ > +0.4 is on the outer modelling.
bounds of credulity. Samples this loose could not be pre- A case history of silt tailings has been considered in
pared in the laboratory even using moist tamping methods. which there are adjacent data for a CPTu sounding and sam-
And it is unclear whether the low strengths corresponding to ples from a borehole. The NCEER approach (Youd et al.
such a loose state could support the current impoundment 2001), which is based on Robertson and Wride (1998), indi-
geometry. cates that the soil is likely not liquefiable. However, compar-
Regardless of the actual ψ, either approach indicates the silt ison of the logged water contents from the borehole to the
is liquefiable. Figure 19 shows the computed undrained triaxial critical state determined from tests on reconstituted samples
behaviour of a typical in situ silt having Ir = 550 and ψ = +0.1 indicates that the soil was catastrophically liquefiable. This
for three values of hardening parameter, H. The lower value of conclusion is corroborated by the high Bq values measured
H = 25 corresponds to that interpreted from the laboratory tests during CPT testing; the near equivalence of qt and uc indi-
and would be appropriate if H was independent of Ir. An alter- cate a soil strength approaching zero.
native approach would be to assume that the H to Ir ratio stays Finite element simulations of undrained cavity expansion
approximately constant, giving H ≈ 100. Of course, there is no were undertaken using NorSand. This soil model closely cal-
reason to assume that the value of H in situ is related to that ibrates to the measured silt behaviour of reconstituted sam-
from the moist tamped samples. Therefore one additional ples in the laboratory. These finite element simulations were
value of H = 300 is also shown, to span the likely range of found to closely support evaluation of undrained CPT data
hardening values. All of the simulations indicate that the silt using the dimensionless framework suggested by Houlsby
is indeed liquefiable. The chosen value of ψ = +0.1 is the best (1988). This approach provides a mechanics-based method-
estimate of the range of state values in Fig. 18; higher ψ ology with which to determine soil state from the CPT.
would give more collapsible behaviour. The numerical simulations further confirmed that the soil
The residual or postliquefaction strength for soils with stress was liquefiable. However, the extent to which the silt might
paths such as those computed would usually localize at the exhibit brittle strength reduction was much less than that in-
minimum post-peak strength. This strength ranges between ferred from the in situ water contents.
about 37% and 60% of the peak (depending on H). It may Simulation of the expected in situ behaviour based on the
seem surprising that such high in situ ψ values lead to only CPT inferred in situ state showed that the soil was predomi-
moderate post-peak strength drops, but this is an effect caused nantly very contractive and soft, rather than brittle. Interest-
by the rather large slope of the CSL. In effect, the soils are ingly this is the same behaviour identified from the CPT soil
weak, compressible, and somewhat liquefiable, rather than cat- type behaviour chart.

© 2007 NRC Canada


18 Can. Geotech. J. Vol. 44, 2007

References Penetration Testing 1988 – ISOPT-1, Edited by Jacob de Ruiter.


A.A. Balkema, Rotterdam, The Netherlands. pp. 745–750.
ASTM. 2006a. Standard Test Methods for Specific Gravity of Soil Fourie, A.B., Blight, G.E., and Papageorgiou, G. 2001. Static liq-
Solids by Water Pycnometer, ASTM D 854-06. Philadelphia, uefaction as a possible explanation for the Merriespruit tailings
PA. American Society for Testing and Materials. dam failure. Canadian Geotechnical Journal, 38: 707–719.
ASTM. 2006b. Standard test methods for maximum index density Frost, J.D., and Park, J.-Y. 2003. A critical assessment of the moist
and unit weight of soils using a vibratory table ASTM D 4253- tamping technique. Geotechnical Testing Journal, 26(1): 1–14.
00 (2006). Philadelphia, PA. American Society for Testing and Høeg, K., Dyvik, R., and Sandbækken, G. 2000. Strength of undis-
Materials. turbed versus reconstituted silt and silty sand specimens. Journal of
ASTM. 2006c. Standard test method for minimum index density Geotechnical and Geoenvironmental Engineering, 126: 606–617.
and unit weight of soils and calculation of relative density, D Houlsby, G.T. 1988. Introduction to papers 14–19. Penetration test-
4254-00 (2006). Philadelphia, PA. American Society for Testing ing in the UK, Thomas Telford, London. pp. 141–146.
and Materials. Houlsby, G.T., and Yu, H.S. 1991. Finite cavity expansion in
Been, K., and Jefferies, M.G. 1985. A state parameter for sands. dilatant soils: loading analysis. Géotechnique, 41: 173–183.
Géotechnique, 35: 99–112. Jefferies, M.G. 1993. NorSand: a simple critical state model for
Been, K., and Jefferies, M.G. 1986. Reply to discussion. Géotechnique, sand, Géotechnique, 43: 91–103.
36: 123–132. Jefferies, M.G., and Davies, M.P. 1993. Use of the CPTu to esti-
Been, K., and Jefferies, M.G. 1992. Towards systematic CPT inter- mate equivalent SPT N60. Geotechnical Testing Journal, 16:
pretation. In Proceedings of the Wroth Memorial Symposium, 458–468.
Thomas Telford, London. pp. 121–134. Jefferies, M.G., and Shuttle, D.A. 2002. Dilatancy in general
Been, K., Crooks, J.H.A., Becker, D.E., and Jefferies, M.G. 1986. Cambridge-type models. Géotechnique, 52: 625–638.
The cone penetration test in sands: Part I, state parameter inter- Jefferies, M.G., and Shuttle, D.A. 2005. NorSand: Features, cali-
pretation. Géotechnique, 36: 239–249. bration and use. Geotechnical Special Publication No. 128, Soil
Been, K., Jefferies, M.G., Crooks, J.H.A., and Rothenberg, L. Constitutive Models: Evaluation, Selection, and Calibration.
1987a. The cone penetration test in sands: Part II, general infer- Edited by J.A. Yamamuro and V.N. Kaliakin. ASCE, Reston, Va.
ence of state. Géotechnique, 37: 285–299. pp. 204–236.
Been, K., Lingnau, B.E., Crooks, J.H.A., and Leach, B. 1987b. Konrad, J.-M., and Law, K.T. 1987. Preconsolidation pressure from
Cone penetration test calibration for Erksak (Beaufort Sea) sand. piezocone tests in marine clays. Géotechnique, 37: 177–190.
Canadian Geotechnical Journal, 24: 601–610. Ladanyi, B., and Roy, M. 1987. Point resistance of piles in sand. In
Proceedings of the 9th Southeast-Asian Geotechnical Confer-
Been, K., Crooks, J.H.A., and Jefferies, M.G. 1988. Interpretation
ence, Bangkok. Southeast Asian Geotechnical Society, Asian In-
of material state from the CPT in sands and clays. In Penetration
stitute of Technology, and Engineering Institute of Thailand.
testing in the UK. Thomas Telford, London. pp. 215–218.
pp. 6-29–6-42.
Been, K., Jefferies, M.G., and Hachey, J. 1991. The critical state of
Ladanyi, B., and Foriero, A. 1998. A numerical solution of cavity
sand. Géotechnique, 41: 365–381.
expansion problem in sand based directly on experimental stress–
Been, K., Jefferies, M.G., and Hachey, J. 1992. Reply to discus- strain curves. Canadian Geotechnical Journal, 35: 541–559.
sion. Géotechnique, 42: 660–663.
Lou, J.K., Byrne, P.M., Garner, S.J., and Marcuson, W.F. 1991. As-
Cao, L.F., Teh, C.I., and Chang, M.F. 2001. Undrained cavity ex- sessment of seismic stability of Dolphin Pool Slope of John
pansion in modified Cam clay I: Theoretical analysis. Géo- Hart Dam. In Proceedings of the 2nd International Conference
technique, 51: 323–334. on Recent Advances in Geotechnical Earthquake Engineering
Carraro, J.A.H., Bandini, P., and Salgado, R. 2005. Liquefaction and Soil Dynamics, St. Louis, Missouri, 11–15 March 1991.
resistance of clean and non-plastic silty sands from cone pene- Edited by Shamsher Prakash. University of Missouri-Rolla,
tration resistance. In Proceedings of Geo-Frontiers 2005, ASCE Rolla, Mo. Paper no. 7.11, pp. 1013–1020.
Geotechnical Special Publication No. 133 [available on CD- Lu, Q., Randolph, M.F., Hu, Y., and Bugarski, I.C. 2004. A numeri-
ROM]. cal study of cone penetration in clay. Géotechnique, 54: 257–267.
Carter, J.P., Booker, J.R., and Yeung, S.K. 1986. Cavity expansion Plewes, H.D., Davies, M.P., and Jefferies, M.G. 1992. CPT based
in cohesive frictional soils. Géotechnique, 36: 349–358. screening procedure for evaluating liquefaction susceptibility. In
Chang, M.F., Teh, C.I., and Cao, L.F. 2001. Undrained cavity ex- Proceedings of the 45th Canadian Geotechnical Conference, To-
pansion in modified Cam clay II: Application to the interpreta- ronto, 26–28 October 1992. Canadian Geotechnical Society.
tion of the piezocone test. Géotechnique, 51: 335–350. pp. 4-1–4-9.
Clayton, C.R.I., Hababa, M.B., and Simons, N.E. 1985. Dynamic Robertson, P.K., and Wride (Fear), C.E. 1998. Evaluating cyclic
penetration resistance and the prediction of the compressibility liquefaction potential using the cone penetration test. Canadian
of a fine-grained sand – a laboratory study. Géotechnique, 35: Geotechnical Journal, 35: 442–459.
19–31. Schofield, A., and Wroth, C.P. 1968. Critical state soil mechanics.
Collins, I.F., Pender, M.J., and Yan, W. 1992. Cavity expansion in McGraw-Hill, London.
sands under drained loading conditions. International Journal for Seed, H.B., and Idriss, I.M. 1982. Ground motions and soil lique-
Numerical and Analytical Methods in Geomechanics, 16: 3–23. faction during earthquakes. In Earthquake Engineering Research
Dasenbrock, D. 2005. Improved site stratigraphy and layer charac- Institute Monograph, Oakland, Calif.
terization using cone penetration testing methods on Minnesota Sego, D.C., Robertson, P.K., Sasitharan, S., Kilpatrick, B.L., and
DOT projects. In Proceedings of Geo-Frontiers 2005, Site Char- Pillai, V.S. 1994. Ground freezing and sampling of foundation
acterization and Modeling. ASCE Geotechnical Special Publica- soils at Duncan Dam. Canadian Geotechnical Journal, 31: 927–
tion No. 133 [available on CD-ROM]. 938.
East, D.R., Cincilla, W.A., Hughes, J.M.O., and Benoit, J. 1988. Shuttle, D.A. 2004. Implementation of a viscoplastic algorithm for
The use of the electric piezocone for mine tailings deposits. critical state soil models. In Proceedings of the 9th International

© 2007 NRC Canada


Shuttle and Cunning 19

Symposium on Numerical Models in Geomechanics (NUMOG Ic,RW soil classification index from Robertson and Wride (1998),
IX), Ottawa, Ont., 25–27 August 2004. Edited by G.N. Pande I c, RW = [ 3.47 − log (Q)]2 + [(1.22 + ( log F )]2
and S. Pietruszczak. A.A. Balkema Publishers, London.
pp. 243–247. Ir dimensionless elastic shear rigidity (= G/p′)
Shuttle, D.A., and Jefferies, M.G. 1998. Dimensionless and unbi- k drained soil-specific coefficient in eq. [1]
ased CPT interpretation in sand. International Journal for Nu- k undrained soil-specific coefficient in eq. [2]
merical and Analytical Methods in Geomechanics, 22: 351–391. K0 in situ lateral effective stress ratio, σh′ / σv′
Sladen, J.A. 1989a. Cone penetration test calibration for Erksak m drained soil-specific coefficient in eq. [1]
(Beaufort Sea) sand: Discussion. Canadian Geotechnical Jour- m undrained soil-specific coefficient in eq. [2]
nal, 26: 173–177. M Value of ratio η at the critical state (varies with Lode angle)
Sladen, J.A. 1989b. Problems with interpretation of sand state from Mi value of M used in the flow rule (Mi – η), Mi = M – |ψ|
cone penetration test. Géotechnique, 39: 323–332. Mtc value of ratio η at the critical state under triaxial com-
Sladen, J.A., and Handford, G. 1987. A potential systematic error pression conditions
in laboratory testing of very loose sands. Canadian Geotechnical Nk cone strength factor, N k = ( qt − σv0 )/ s u
Journal, 24: 462–466. p mean total stress, p = (σ1 + σ2 + σ3 )/3
Smith, I.M., and Griffiths, D.V. 1998. Programming the finite ele- p′ mean effective stress, p′ = ( σ1′ + σ2′ + σ3′ )/ 3
ment method. 3rd ed. John Wiley & Sons, New York. p i′ mean effective stress at the image state
Vaid, Y.P., Sivathayalan, S., and Stedman, D. 1999. Influence of p 0′ in situ mean effective stress, p 0′ = ( σx0 ′ + σy0
′ + σz0′ )/3
specimen-reconstituting method on the undrained response of p0 in situ mean total stress, p0 = ( σx0 + σy0 + σz0 ) / 3
sand. Geotechnical Testing Journal, 22: 187–195. q deviatoric stress invariant, q = (1/ 2( σ1 − σ2 )2 + 1/ 2( σ2 −
van den Berg, P. 1994. Analysis of soil penetration. Ph.D. thesis, σ3 )2 + 1/ 2( σ3 − σ1 )2 )1 / 2
Technische Universiteit Delft, Delft, The Netherlands. qt total tip resistance
Yafrate, N., and DeJong, J.T. 2005. Detection of stratigraphic inter- Q dimensionless CPT resistance based on vertical stress
faces and thin layering using a miniature piezoprobe. In Proceed- corresponding to standard usage within the in situ test-
ings of Geo-Frontiers 2005, Site Characterization and Modeling. ing community, Q = ( qt − σv0 )/ σv0 ′
ASCE Geotechnical Special Publication No. 133 [available on
Qp dimensionless CPT resistance based on mean stress re-
CD-ROM].
quired by mechanics, Qp = ( qt − p0 )/ p 0′
Youd, T.L., Idriss, I.M., Andrus, R., Arango, I., Castro, G., Chris-
R overconsolidation ratio, R = pmax ′ / p current
′ (see Fig. 11)
tian, J., et al. 2001. Liquefaction resistance of soils: summary re-
port from the 1996 NCEER and 1998 NCEER/NSF workshops su undrained shear strength
on evaluation of liquefaction resistance of soils. ASCE Journal of u current pore pressure
Geotechnical and Geoenvironmental Engineering, 127: 817–833. uc pore pressure measured by CPT during sounding (at
Yu, H.S., Herrmann, L.R., and Boulanger, R.W. 2000. Analysis of shoulder location)
steady cone penetration in clay. ASCE Journal of Geotechnical u0 in situ pore pressure
and Geoenvironmental Engineering, 126: 594–605. χ proportionality coefficient, relating dilation at peak
Zienkiewicz, O.C., and Cormeau, I.C. 1974. Viscoplasticity, plas- stress ratio, Dmin, to state parameter, ψ
ticity and creep in elastic solids. A unified numerical approach. Γ “altitude” of CSL, defined at 1 kPa
International Journal for Numerical Methods in Engineering, 8: εv total volumetric strain, εv = ε ev + ε vp
821–845. ε ev volumetric strain, ε ev = ε1e + ε 2e + ε 3e , e superscript indi-
cates elastic
ε vp volumetric strain, ε vp = ε1p + ε 2p + ε 3p , p superscript in-
List of symbols dicates plastic
B Skempton’s pore pressure parameter, B = ∆u/ ∆σ3 ε& q shear strain measure work conjugate with σ& q , dot
Bq CPTu excess pore pressure ratio, Bq = ( uc − u0 )/( qt − σv0 ) superscript denotes rate ε& q = 1/ 3 [(sin θ + 3 cos θ) ε& 1 −
Dmin maximum dilatancy; equal to the dilation at peak stress 2 sin θ ε& 2 + (sin θ − 3 cos θ) ε& 3 ]
ratio (the minimum arising due to the compression posi- η dimensionless shear measure as ratio of stress
tive sign convention) invariants, η = q/p′
e void ratio ηL limiting stress ratio, corresponding to the location of the
ec void ratio at the critical state for the current mean stress inner cap yield surface
F stress normalized CPTu friction ratio, F = fs /( qt − σv0 ) θ lode angle, sin ( 3 θ) = −13.5σ1′ σ 2′ σ 3′ / q 3
fs friction sleeve stress measurement λ 10 slope of CSL, defined on base 10
G elastic shear modulus ν Poisson’s ratio
Gs specific gravity σv0 in situ vertical total stress

σv0 in situ vertical effective stress
H dimensionless plastic hardening modulus for loading
Ic soil classification index from Been and Jefferies (1992), φ′ effective friction angle
ψ state parameter, e – ec
I c = 3 − log [Q (1 − Bq ) + 1]2 + [1.5 + 1.3 (log F )]2
ψ0 in situ (or initial) state parameter

© 2007 NRC Canada

You might also like