You are on page 1of 8

3rd International Symposium on Cone Penetration Testing, Las Vegas, Nevada, USA - 2014

CPT-based compressibility assessment of soils

F. Ciloglu
GAI Consultants Inc, Pittsburgh, Pennsylvania, USA
K.O. Cetin & A. O. Erol
Middle East Technical University, Ankara, Turkey

ABSTRACT: This study presents a simplified methodology for the estimation of one dimensional field
constrained moduli of soil layers based on CPT data. For this purpose, CPTs were performed at soil
sites before the application of surcharge loads, after the placement of which settlement performance of
the site was monitored through field settlement plates. The resulting database consists of a total of 45
CPT data consisting of cone tip resistance (qc), friction ratio (FR), and 59 settlement plate records where
settlements at the field were recorded under various surcharge conditions. Maximum likelihood frame-
work was utilized to develop a new probabilistic CPT-based soil compressibility model. Fines content
(FC) and plasticity index (PI) along with cone tip resistance were selected as model input parameters. If
soil samples are available, FC and PI values were estimated through laboratory testing, otherwise the
CPT-based estimation procedure recommended by Cetin and Ozan (2009) was used. As the concluding
remark, it is shown that the proposed new probabilistically based assessment scheme can reliably esti-
mate field compressibility response of soils through CPT data.

1 INTRODUCTION

Compressibility characteristics of soils can be expressed by soil modulus: one-dimensional (1-D) con-
strained modulus, M, undrained Young’s modulus in compression loading, Eu and compression index,
Cc. In practice, the most frequently utilized parameter for soils is the 1-D constrained modulus, M de-
fined as M=1/mv. In which, mv is the coefficient of volume compressibility.
In the past few decades, many correlations have been proposed to predict the 1-D constrained modu-
lus, M, of soils based on the CPT results (e.g. Sanglerat, 1972, Lunne and Christophersen, 1983, Sen-
neset et al., 1982, 1989, Kulhawy and Mayne, 1990, Jones and Rust, 1995, Manas et al., 1995, For-
manavicius, 1995, Mahesh et al., 1995, Lunne et al., 1997, Robertson, 2009).
Sanglerat (1972) proposed a relationship for the 1-D constrained modulus expressed in terms of the
cone tip resistance and coefficient, m as given in M = m(qc). The coefficient m in the range of 1 to 8
was proposed for a wide range of the cone tip resistance (qc) and type of the soils.
Another correlation developed by Kulhawy and Mayne (1990) proposed a general relationship of M
= 8.25(qt-0). An alternative correlation for 1-D constrained modulus was proposed by Senneset et al.
(1982, 1989). In this study a linear interpretation model for over consolidated and normally consolidated
clays was proposed, in the form of M = m(qt-0). The coefficient, m ranges from 5 to 15 and 4 to 8 for
over consolidated and normally consolidated clays, respectively.
Most recent correlation by Robertson (2009) suggested following values for m based on normalized
cone resistance (Qtn).
When Ic > 2.2 use

629
m = Qtn when Qtn≤14
m = 14 when Qtn>14
When Ic < 2.2 use
m = 0.03[10(0.55Ic+1.68)]
In this study, a possibly more robust and reliable method covering the different types of the soils, is
proposed by using powerful and flexible probabilistic tools. The proposed correlation to estimate the 1-
D constrained modulus, M is calibrated, by settlement case histories where consolidation settlements in
the field under various surcharge loads were recorded along with CPT-based soil profile data.

2 DATABASE COMPILATION

To develop a robust and defensible correlation for the assessment of the soil compressibility, a database
consisting of CPT soundings and settlement plate readings was compiled. The database is composed of
59 settlement plate records and 45 CPT boring data consisting of raw cone tip resistance (qc), friction
sleeve resistance (fs). CPT data and settlement plate readings were obtained from two different sites in
Turkey, which are (1) Bursa East Waste Water Treatment Plant Project and, (2) Bursa West Waste Wa-
ter Treatment Plant Project in Bursa, Turkey. 30 CPT’s were obtained from East Waste Water Treatment
Plant site and the remaining CPT’s were obtained in the West Waste Water Treatment Plant site.
CPT data used to describe the soil profile and the sublayers of the soil up to 20 meters at the location
of the settlement plates. CPT based soil classification indicated that the database mostly consists of lay-
ers of silts, clays, silt and clay mixtures. Non-plastic silty sands and well graded sand were also present
in CPT soundings. Roughly, 30% of database is composed of nonplastic soils and remaining 70% are
distributed within the plasticity index (PI) range of 1.0 and 57.7. Overconsolidation ratios (OCR) for the
soils in the database were estimated by CPT-based estimation by Kulhawy and Mayne (1990). Estimates
of OCR showed that soils are generally normally consolidated and slightly overconsolidated with typical
value of 1.5.
Figure 1 shows the typical CPT profile and the characteristics of CPT database with the typical val-
ues of, qc, fs, fines content (FC), and plasticity index (PI) are summarized in Table 1. If soil samples are
available, FC and PI values were estimated through laboratory testing, otherwise the CPT-based estima-
tion procedure recommended by Cetin and Ozan (2009) was used.

Table 1. Characteristics values of CPT Database.


qc (MPa) fs (KPa) FC (%) PI (%)

Minimum 0.5 1.0 1.2 1.0

Maximum 35.9 378.0 99.0 57.7

Average 4.1 55.0 67.1 37.7

630
Cone Tip Resistance Friction Sleeve Resistance
qc (Mpa) fs (Kpa)
0 5 10 0 100 200 300 400
0 0

2 2

4 4

6 6

8 8

Depth (m)
Depth (m)

10 10

12 12

14 14

16 16

18 18

20 20

22 22
Figure 1. Typical CPT Profile

3 DATA PROCESSING

In this study, the general form of the 1-D constrained modulus M = m qc was adopted and a correlation
between coefficient m and fines content (FC) and plasticity index (PI) of soils is proposed. Fines con-
tent (FC) and plasticity index (PI) are selected as representative physical parameters defining the com-
pressibility behavior. In the process of establishing a correlation, regression analysis between recorded
settlement plate readings and the predicted settlement values based on the proposed correlation for M
was performed. To perform this regression analysis and develop a relation form, three main steps listed
as below were completed.
 CPT Data Processing: Normalization of CPT data and determination of soil index parameters FC
and PI.
 Calculation of increase in the vertical stress under the sustained loading.
 Prediction of ultimate settlement.

3.1 CPT data processing


Consistent with the available studies (Lunne et al., 1997), raw cone tip resistance (qc) was corrected for
unequal end area effects and net corrected tip resistance, (qt - v), was normalized to an effective vertical
stress, v’ of 1 atm, qt,1,net to eliminate the effects of the variability in vertical stress conditions as given
in Equations (Eqs.) 1 through 3.
Corrected tip cone tip resistance
qt  qc  u  (1  a) (1)

631
Normalized net cone tip resistance
qt ,1,net  qt   v   C N = Qt
c
P 
CN   a  (2)
  v 
Normalized friction ratio
fs
FR   100 % (3)
qt   v
Power law stress normalization exponent (Cetin & Isik, 2007)
R  272.38
c  0.085 ; 272.38  R  275.19 (4)
275.19  272.38
q 
R  (log( F R )  243 . 91 ) 2  (log  t ,1 , net   126 . 24 ) 2 (5)
 Pa 
where u is pore pressure behind the cone shoulder; Pa is atmospheric pressure in the units of v’; and c
is power law stress normalization exponent as recommended by Cetin and Isik (2007).
After a couple of iteration rounds, compatible qt,1,net and c pairs were estimated as defined in Eqs 1
through 5. Friction ratio, FR, was calculated as in Eq. 3.
After normalization, soil index parameters FC and PI based on qt,1,net and FR were calculated as rec-
ommended by Cetin and Ozan (2009).
Mean Values of FC
R  238.50
 FC   100  20.93
240.25  238.50
R  (log(FR )  55.42) 2  (logqt ,1,net   233.52) 2 (6)
Mean values of PI
 PI  102.371.33log( F
R )  log( qt ,1, net ) / 2.25 
 9.83 (7)

3.2 Prediction of ultimate settlement


Each settlement plate’s records were used to predict the ultimate settlement at the end of primary consol-
idation under the applied loading. The settlements in the field were recorded for a period of 120-140
days. Required time to observe the final settlement is too long (theoretically infinite), and it is not prac-
tical to wait for this ultimate settlement to occur. Hence, the available methods proposed by Asaoka
(1978) and Horn (1983) were adopted to estimate the ultimate settlement using available field readings.

3.3 Calculation of stress distribution under applied surcharge


The project sites of Bursa East Waste Water Treatment Plant and Bursa West Waste Water Treatment
Plant were preloaded by 4-meter high embankments. The change in the vertical stresses produced in the
subsoil due to the applied external loads was calculated by Flac 3-D finite-difference program.

4 DEVELOPMENT OF CORRELATION

Using the database consisting of normalized CPT data, estimated soil index parameters and ultimate set-
tlements based on settlement plate readings, a correlation to predict the compressibility of the soils was
developed.

632
4.1 Bayesian limit-state models
The Bayesian model assessment technique was used to develop limit-state models incorporating the im-
portant descriptive variables for the CPT-based soil compressibility problem. The basic framework is the
well-known Bayesian updating rule for the estimation of model parameters, which is given by
f (Θ )  cL (Θ ) p (Θ ) (8)
where Θ is the set of model parameters to be estimated, p(Θ) is the prior distribution representing our
knowledge about Θ before making observations, L(Θ) is the likelihood function representing the new
knowledge contained in the set of observations, c=[∫L(Θ)p(Θ)d(Θ)]-1 is a normalization factor, and f(Θ)
is the posterior distribution representing our updated state of knowledge about Θ (Box and Tiao 1973,
Der Kiureghian 1999). The likelihood function is proportional to the conditional probability of making
the observations for a given value of Θ and requires the selection of a problem-specific probabilistic
model.

4.2 Limit-state models for CPT-based soil compressibility


The first step in developing a limit-state model for the CPT-based soil compressibility is to select a
mathematical model which captures the essential of the problem. Motivated by previous studies, we
adopted the following model for prediction of 1-D constrained modulus, M.
M   m qc
 m  1 Ln ( FC  )   3 Ln ( PI  )   5
2 4
(9)
To evaluate the compressibility behavior of soils, specifically the 1-D constrained modulus coeffi-
cient, αm, the values of ultimate observed settlements at the fields were used in the regression analysis
together with the predicted settlement by developed formulation of αm.
i n
S p   S i (q ci , H i ,  i ;  mi )
i 0
                        (10) 
S pi (q ci , H i ,  i ; mi )   1   H  
 ( mi  qci ) i i

where Sp is the total predicted settlement; Spi is the predicted settlement at the ith sublayer; i=0, n is the
number of the sublayer for each soil profile; ∆Hi is the thickness of ith sublayer; ∆σi is the increase in
vertical stress for ith sublayer; qci is the average cone tip resistance for the ith sublayer.
By using the proposed correlations for αm in Eqs. 9 and 10, formulation given in Eq. 11 was devel-
oped to predict the final settlement of the soil. A representative functional form that was used to develop
the relationship for αm.
i n
  H   i 
g ( S o , q ci , FC i , PI i , H i ,  i ,  )  S o    i (11)
i 0 
(( 1 Ln ( FC )   3 Ln( PI )   5 )  q ci )
2 4

where So is the observed settlement in the field; FCi and PIi the soil index parameters for ith sublayer;
and θ=(θ1,…,θ5) is the set of unknown model parameters.
The limit state function given in equation 11 assumes that the compressibility behavior of soils can be
completely explained by the three descriptive variables qc, FC and PI. Obviously, there may exist addi-
tional parameters (such as stress history) that influence the compressibility behavior of soils. To account
for the influences of the missing variables and the possible deficiencies in the functional form, a random
model correction term, ε was introduced and limit state function was modified as
i n
 H i   i 
g ( S o , q ci , FC i , PI i , H i ,  i , )  S o  
   (12)
i 0  (( 1 Ln( FC 2 )   3 Ln( PI 4 )   5 )  q ci )
 

633
With the aim of producing an unbiased model, the mean of ε is set to zero. The ε was decided be best
represented by normal distribution, a symmetrical distribution with zero mean. The standard deviation of
ε, denoted σε, however is unknown and must be determined. The set of unknown parameters of the pro-
posed model is presented as Θ=(θ,ε).

5 FORMULATION OF THE LIKELIHOOD FUNCTION

Soi and Spi are the observed and predicted settlement values at the ith location in the site, respectively and
εi is the corresponding realization of the model correction term. Assuming the observations compiled
from different locations at the field to be statistically independent, the likelihood function can be written
as the product of the probabilities of the observations:

 Pg (S 
55
L( ,  z )  oj , q ci , FC i , PI i , H i ,  i ,  j ,  )  0 (13)
SPno  J 1

Here it is assumed that the measured values each observation Soj, qci, FCi, PIi, Hi, i are exact, i.e.,
no measurement or estimation error  is present. Then, noting that
g(Soj , qci , FCi , PIi , Hi , i , j ,)  g(Soj , qci , FCi , PIi , Hi , i ,)   j (14)
has the normal distribution with a mean of g (Soj, qci, FCi, PIi, Hi, I, θ) and standard deviation of σz,
the likelihood function can be written as
55
 g ( S oj , qci , FCi , PI i , H i ,  i ,  j , 
L( ,   )      (15)
j 1   
where φ[ ] is the standard normal probability density function. Note that the above expression is a func-
tion of the unknown parameters and θ and σz.

6 PARAMETER ESTIMATION

Having formulated the likelihood function, the values of the unknown model parameters θ = (θ1,…,θ5)
for proposed correlation and standard deviation σz of the model error term ε, which maximize the likeli-
hood function, were computed. Table 2 summarizes model parameters of the CPT-based compressibility
assessment of soil limit state models.

Table 2. Maximum likelihood estimates of proposed model parameters


1 0.30
2 -9.32
3 0.15
4 -2.89
5 3.15
Ln (Max. Likelihood value) 45.99

When the model parameters given in Table 1 are put into the Equation 11, the proposed correlation
for αm, becomes:
 m  0.3Ln ( FC 9.32 )  0.15 Ln ( PI 2.89 )  3.15 (16a)
M  q c (0.3Ln ( FC 9.32 )  0.15 Ln ( PI 2.89 )  3.15) (16b)

634
The recorded settlement value in the site and the predicted settlement using the proposed correlation
values are presented in Figure 2. The accuracy of estimations was also comparatively illustrated in Fig-
ure 2 along with 1V:1H, 1V:2H, and 2V:1H lines. As the figure illustrates, the proposed CPT-based re-
lation for 1-D constrained modulus, M can be judged acceptable under a certain loading. The resulting
correlation is summarized in Table 3 indicating that αm varies with PI and FC. αm values decreases with
increasing PI and FC values and varies from 3 to 13, where the low values apply to clays with high plas-
ticity. The predicted m values in this study are in good agreement with the m values proposed by Sen-
neset et al. (1982, 1989) ranging from 5 to 15 and 4 to 8 for over consolidated and normally consolidat-
ed clays, respectively. Good correlation is observed with Mayne (2007), which showed that αm values
are in the range of 1 to 10 where low values apply to soft clay.

0.35

0.30

0.25

0.20
Predicted Settlement (m)

0.15

0.10

0.05

0.00
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
Observed Settlement (m)
 
Figure 2. Distribution of observed settlement at field and predicted settlement based on the proposed correlation.

7 SUMMARY AND CONCLUSION

A brief summary is provided that describes the development of a new relationship for the prediction of
soil compressibility M (for clays and silts) in the form of soil index parameters FC and PI values of
soils. For this purpose, a database composed of 45 CPT’s consisting of cone tip resistance (qc), friction
sleeve resistance (fs) along with 59 settlement plate records was compiled from two sites in Turkey. The
resulting database was assessed probabilistically by applying the maximum likelihood (ML) method on
select representative limit state models. A probabilistic CPT-based new correlation was developed and
presented in equation 16 which is a function of plasticity index (PI), fine content (FC) of soils. The pro-
posed correlation is also summarized in Table 3 indicating that αm varies with PI and FC and varies from
3 to 13, where the low values apply to clays with high plasticity.

635
Table 3. Proposed CPT-based soil compressibility relationship.
Soil Classification Soil Definition αm
Silty or clayey sands
15 < FC ≤ 50 PI ≤ 4.0 Silty sands 13.5 > αm > 7.0
PI > 4.0 Clayey sands 10.8 > αm > 6.3
Silts
50 < FC ≤ 65 PI ≤ 12.0 Low plasticity silts 10.1 > αm > 5.9
PI > 12.0 High plasticity silts 6.6 > αm > 5.1
Silt-Clay Mixtures
65 < FC ≤ 80 PI ≤ 20.0 Clayey silt 9.3 > αm > 4.9
PI > 20.0 Silty clay 5.5 > αm > 3.8
Clays
FC > 80 PI ≤ 50.0 Low plasticity clays (CL) 4.9 > αm > 3.6
PI > 50.0 High plasticity clays (CH) 4.2 > αm > 3.1

8 REFERENCES

Asaoka, A. (1978), Observational Procedure of Settlement Prediction, Soils and Foundations, Vol.18, pp. 87-100.
Cetin, K. O. and Ozan, C. (2009), CPT-Based Probabilistic Soil Characterization, ASCE Journal of Geotech and
Geoenvironmental Engineering, 135 (1), page 84-107.
Cetin, K. O. and Isik, N. (2007), Probabilistic Assessment of Stress Normalization for CPT Data, ASCE Journal
of Geotech and Geoenvironmental Engineering, 133 (7), page 887-897.
Furmonavicius L., Dagys A. (1995), Cone Penetration Testing In Lithuania, International Symposium on Cone
Penetration Testing, Vol.1, pp.125- 131, Linköping, Sweden.
Horn, A. (1983), Determination of Properties of Weak Soils by Test Embankments, International Symposium on
Soil and Rock Investigations by Insitu Testing, Paris, Vol.2, pp.61-66.
Jones, G.A. and Rust, E. (1995), Piezocone Settlement Prediction Parameters for Embankments on Alluvium,
Proceedings of the International Symposium on Cone Penetration Testing, CPT’95, Linköping, Sweden,
Vol.2, pp.501-508, Swedish Geotechnical Society.
Kulhawy, F.H. and Mayne, P.H. (1990), Manual on Estimating Soil Properties for Foundation Design, Electric
Power Research Institute, EPRI, Report EL-6800.
Lunne, T. and Christophersen, H.P. (1983), Interpretation of Cone Penetrometer Data for Offshore Sands, Pro-
ceedings of the Offshore Technology Conference, Richardson, Texas, Paper No. 4464.
Mahesh, D. and Vikash, J. (1995), State of Art of CPT in India, International Symposium on Cone Penetration
Testing, Vol.1, pp.87-95 Linköping, Sweden.
Manas, S. and Linaresh, H. (1995), The Use of Cone Penetration Testing in Spain (NR4), International Symposi-
um on Cone Penetration Testing, Vol.1, pp.213-220, Linköping, Sweden.
Mayne, P. W. (2007), Cone Penetration Testing: A synthesis of highway practice. Project 20-5. Transportation
Research Board, NCHRP synthesis 368, Washington D.C.
Robertson, P. K. (2009), Interpretation of Cone Penetration Tests – a Unified Approach, Canadian Geotechnical
Journal, Vol. 26, pp.1337-1355.
Sanglerat, G. (1972), The Penetrometer and Soil Exploration, Elsevier, Amsterdam, pp. 464.
Senneset, K., Janbu, N. and Svano, G. (1982), Strength and Deformation Parameters from Cone Penetration Tests,
Proceedings of the 2nd Europen Symposium on Penetration Testing, ESOPT-II, Amsterdam, 2, 863-70,
Balkema Pub., Rotterdam.
Senneset, K., Sandven, R. and Janbu, N. (1989), The Evaluation of Soil Parameters from Piezocone Tests, Trans-
portation Research Record, No. 1235, pp.24-37.

636

You might also like