Proceedings ISC-2 on Geotechnical and Geophysical Site Characterization, Viana da Fonseca & Mayne (eds.

)
© 2004 Millpress, Rotterdam, ISBN 90 5966 009 9

CPTu-DMT performance-based correlation for settlement design

M. Arroyo
DIS, Politecnico di Milano, Italy (formerly Iberinsa, Madrid, Spain)
M. Devincenzi
Igeotest, Figueres, Spain
M.T. Mateos & R. Gómez-Escoubès
Iberinsa, Madrid, Spain
J.M. Martínez
UPM, Madrid, Spain

Keywords: CPTu, DMT, settlement, deformability, stiffness, deltaic, embankments

ABSTRACT: A major project in a deltaic area needed careful settlement evaluation. The dilatometric method
was chosen as a basic tool for this purpose. As there were no local precedents of DMT use, extra effort was
needed to prove its predictive abilities. This involved comparison with oedometric tests and, more decisively,
prediction of an independently performed embankment load test. At the test site there was no DMT available
but only CPTu. A site-specific linear relation between cone resistance and dilatometer evaluated constrained
modulus was developed. The relation was very successful in predicting the load test results. To obtain this re-
lation a performance criteria was employed: equal settlement on average under equal load. It is shown how
this criteria offered in this case substantially better results than approaches based on direct averaging or clas-
sical regression. Dilatometric profiles obtained for the area are also commented: they include a sand layer
where high Kd values are associated to high carbonate content.

1 INTRODUCTION 1.2 Geotechnical practice

The Llobregat delta plain is located just south of Past experience in the area clearly indicates that the
Barcelona. It is the major flat extension near the city main foundation problem appears as a consequence
and its natural expansion area. It hosts an expanding of the medium to high compressibility of the inter-
population, a large number of basic infrastructures, mediate silts and clays. The upper sands offer a
important industrial areas as well as several natural fairly good foundation level, but large settlements
reserves and tourist resorts. may ensue when the load extent is such that silts and
clays are also affected.
1.1 Geological setting The depth of the lower aquifer does seriously pe-
nalize any attempt to bridge the problem using piles.
The geological structure of Llobregat delta is similar On the other hand the frequent presence sandy layers
to other Mediterranean deltas. A wedge of low plas- within the silty and clayey levels, does generally re-
ticity silty and clayey deposits, growing up to a sult in relatively fast consolidation. These circum-
thickness of 60 m near the shoreline, overlies a deep stances make preloading a sensible choice in many
sandy & gravelly aquifer and is overlaid by a instances (Alonso et al. 2000, Gens & Lloret, 2003).
roughly 10 m thick, well graded, medium dense Settlement evaluation does require an estimate of
sand. A superficial thin deposit of alluvial and soil stiffness. The critical silty and clayey layers pre-
marshy clays may sometimes appear on top. A de- sent great sampling difficulties, partly due to the
tailed CPT-based stratigraphic and sedimentological presence of finely interbedded sandy layers. There-
analysis of Llobregat delta is presented by Devin- fore intact sample recovery is problematic and labo-
cenzi et al. (2004). ratory measurements of “in situ” stiffness are scarce
The water table is located in the upper sand, gen- and probably biased. For large projects, large in-
erally at 1 to 1.5 m depth. These sands are highly strumented load tests have been employed to over-
permeable with equivalent permeability of 10-2 cm/s. come the ensuing uncertainty.
On several isolated spots they had been quarried, be- The traditional “in situ” measurement in the area
ing replaced by generally uncontrolled fills. These was SPT. Since the early 90’s CPTu testing has be-
locations require special treatment and they are ex- come common practice. Pressuremeter testing is also
cluded from what follows. sometimes performed. There was no previous large-
scale experience of DMT testing.

1605
 % WEIGHT
0 10 20 30 40 50 60 70 80 90 100
0

10

15

Z (m)
20

25

30

35

40

> 0.074 mm >0.002 mm

Figure 2. Granulometric data

(%)
0 10 20 30 40 50 60
Figure 1. Projected enlargement of El Prat airport (Barcelona). 0
A solid line encloses the motorway project area. A rectangle 5
indicates the load test emplacement.
10

15
1.3 Project description
Z (m)
20
The new terminal building of Barcelona Airport is 25
meant to service up to 25 million passengers per 30
year (Fig. 1). The road access to this new terminal is 35
designed as a 8-lane motorway. This motorway flies 40
over a relocated 6-lane motorway, a major flood de-
WL WP
fense waterway, railway access to the airport and
various minor roads. Figure 3. Atterberg limits
The many obstacles present force the motorway
into heights of 12 m and above for more than 2 km. CO3 (%)
A number of large embankments alternate with sev- 0 10 20 30 40 50 60 70 80 90 100
eral bridges and caisson type structures. The ex- 0

pected schedule for work completion is below 3 5

years and the construction sequence may include 10

several successive enlargements. 15
It was clear form the onset that the width and
Z (m)

20
length of the embankment loads may indeed cause 25
important settlements. Estimating the magnitude and
30
rate of these settlements became a critical design is-
35
sue.
The general design approach relied mostly on 40

CPTu data to ascertain rate of settlement (through Figure 4. Percent carbonate

layer definition and dissipation data) and DMT data
to evaluate the settlement magnitude. This commu- The investigation offered a picture that fell well
nication focuses on the role played by DMT tests within expectations. In the project area the mean
and therefore leaves aside the issue of settlement depth of the lower aquifer was 40 m. A roughly 30
rate evaluation. m thick intermediate layer of silts and clay appeared
between the upper sands and the lower aquifer. Fig-
ure 2 presents granulometric data from samples
2 SITE INVESTIGATION where sedimentation results were available, Figure 3
plasticity data and Figure 4 carbonate content.
2.1 General
Specific site investigation for the motorway project 2.2 DMT results
included 9 rotary core soundings, 14 CPTu and 8 Profiling information for the project relied mostly on
DMT. From nearby projects a number of 9 extra CPTu’s and boreholes; but DMT’s may also be em-
soundings, 9 CPTu and 4 DMT were also available. ployed for this purpose. A common criteria
All the CPTu were performed following ASTM (Marchetti, 2001) uses the material index, Id, for
standard (2000); the DMT, according TC-16 report classification. It has clays below 0.6 and sands
(ISSMGE, 2001) and ASTM standard (2001). above 1.8. Figure 5 shows the Id profile for one

1606 © 2004 Millpress, Rotterdam, ISBN 90 5966 009 9

DMT. The fine interbedding of sandy and clay lay- gest an important degree of overconsolidation in the
ers that is broadly classified as silt is well reflected. upper sands. This result, however, is unlikely from a
Averaging several Id profiles (Fig. 6) erases the geological viewpoint. On the other hand, it has been
noisy details and brings forward a clearer picture of observed (Zhu et al. 1995) that in sands bonding
the main deposit characteristics. greatly increases K0 for any given OCR. It may be
Id then speculated that the high Kd value observed in
0.1 0.6 1.0 1.8 10.0 the upper sands (Fig. 7) is mostly due to the substan-
0 tial carbonate content of the profile (Fig. 5).
5 The mean stiffness profile obtained from DMT
10 soundings is presented in Figure 8. An inverted

DEPTH (m)
15 stiffness profile is apparent, with the softest layers
20 just below the higher aquifer. It may also be noticed
25 that a relatively stiffer layer is common between 25
30 and 30 m depth, and that again the stiffness falls just
35 below that layer. Note that here the stiffness mean
40 profile is obtained using compliances, thus giving a
DMT: 2 mean settlement equal to the settlement mean for the
Figure 5. Id profile from a single DMT averaged profiles. This criterion is rather important
as will be shown below.
Id
0.1 0.6 1.0 1.8 10.0 2.3 Comparison with oedometric results
0
5 Several oedometers were performed on borehole
10 Shelby samples. A tangent confined modulus to the
oedometric curve at the in-situ stress state was
DEPTH (m)

15
20 evaluated to compare with MDMT (Pelnik et al.
25 1999). As soundings and DMT probes were not di-
30 rectly paired, a one-to-one comparison of moduli
35 was not feasible and the comparison was made be-
40 tween the average MDMT curve (average obtained on
AVERAGE compliances). The results are again presented in
Figure 8.
Figure 6. Average Id profile for six DMT
There is a generally good agreement, although
Kd oedometer based moduli are somewhat higher. It is
0 2 4 6 8 10 12 likely, however, that oedometric samples had been
0 selected from the more compact zones of the cores.
5

10
3 (COR)RELATING DMT & CPT
DEPTH (m)

15

20
3.1 General
25
It is commonly accepted (e.g. Lunne et al. 1996) that
30
CPTu based methods of settlement evaluation are
35 relatively inaccurate. This is fundamentally related
40 to the limitations of a failure test in recording pre-
DMT1 failure behaviour and therefore obtaining reasonable
stiffness parameters. The possibility of extending the
Figure 7. Typical Kd profile from DMT tests
greater accuracy of the DMT based method to CPT
results based in some local correlation is therefore
Kd, the horizontal stress index, is a good indica-
very attractive.
tor for clays and silts of the lateral stress coefficient,
The DMT settlement evaluation method
K0, and through it, it may be also related to the OCR
(Marchetti, 2001) involves approximating a 1-D in-
(Marchetti, 1997). A typical Kd profile from the site
tegral of deformation using an expression like
is shown in Figure 7. These values correspond to K0
between 0.5 and 0.8. and do not seem to indicate any zi = H
∆σ v ( zi )
H

important overconsolidation in the clay and silty S = ³ ε ( z ) dz ≈ ¦ M (z ) ∆z (1)

zi = E
layers (Mayne & Kates, 1998). E D i

The KD values for sand are harder to relate with The formula uses two depth distributions, that of
OCR (Marchetti, 1997). In our case direct applica- constrained moduli, MD(z), obtained from the DMT
tion of the same relations used for clay would sug-

Proceedings ISCʼ2 on Geotechnical and Geophysical Site Characterization, Viana da Fonseca & Mayne (eds.) 1607
test and an incremental stress distribution ∆σv (z). In from each profile pair (MD(xp, z), qc(xp, z)) it was
our case the latter was obtained from elastic closed- tempting to try a direct average of these ratios
form solutions; this, of course, involves some extra
M (x , z )
α (x j , z ) = ¦ D j
approximations, particularly considering the highly 1 1
α (z ) = ¦ n j qc (x j , z )
contrasted stiffness profile. j : 1...n (3)
n j
3.2 First approach
The resulting α(z) is illustrated in Fig. 9. The values
We aimed for a classical lineal relationship, to ob- are somewhat high, but not completely out of the
tain from every isolated qc profile and equivalent (admittedly wide) range quoted in the literature.
profile of dilatometric moduli MC using However, when this profile was applied to com-
pute settlements from other CPTu’s it resulted al-
M C (x p , z ) = α ( z )qc (x p , z ) (2) ways in lower settlement than what was being ob-
tained were DMT’s were available. This seemed
M (MPa) wrong and a new α(z) profile was obtained by a trial
1 10 100 1000 and error procedure, comparing predicted settle-
0
ments under a standard load (6 x 6 square footing).
5 The α(z) profile employed was manually adjusted
10
until acceptable agreement was obtained between
settlements obtained with isolated CPTu’s and those
15
obtained with isolated DMT’s. The final result of
this process was the α(z) profile presented in Figure
Z (m)

DMT AVERAGE
20
OEDOMETER
25
9 and labelled “DESIGN” in this and subsequent
figures in the paper.
30

35
3.3 An independent check
40 Data from a preload test independently conducted at
a not too distant site was available. The preload site
Figure 8. Average MDMT vs depth and oedometric results (Fig. 1) is closer to the sea, at 2 km from the motor-
way site, and has a thicker layer of silts and clays.
α =MM
/qc/ qc
SETTLEMENT (mm)
0 5 10 15 20 25 30 35 40
0 50 100 150 200 250 300 350 400 450
0 0

5 5
10
10
15
DEPTH (m)

15 20
z (m)

20 25

30
25
35
30 40

35 45

50
40
EXTENSOMETER 2 PREDICTION
DIRECT AVERAGE DESIGN
Figure 10. Measured and predicted settlement at the load test
Figure 9. Estimate of α(z) by direct average of local profiles site

where xp identifies the profile location. The question The preload test involved the construction of a 4
is how to establish the α(z) distribution. m high 160 m by 80 m trial embankment. The exten-
A first step to compare DMT and CPTu profiles sive instrumentation available (Gens & Lloret, 2003)
involves some filtering of the CPT profiles, since included several extensometers which gave a de-
they are sampled at higher rate than the DMT’s. We tailed picture of in depth settlement pattern.
employed the geometric mean suggested by Eslami CPTu logs were also available at the site from
& Fellenius (1995) to obtain filtered qc profiles with tests made before the preload started. The perform-
∆z = 0.20 m. ance-based correlation obtained for the motorway
In this case we had 12 paired DMT/CPTu tests, site was applied to a CPTu located near the em-
that is, tests that have been performed side by side. bankment center, very close to the extensometer
Since it is immediate to obtain a locally exact α(z) number 2.

1608 © 2004 Millpress, Rotterdam, ISBN 90 5966 009 9

 LOCAL VERTICAL DISPLACEMENT (mm) ment, loading and dilatometric profiles may vary
0 2 4 6 8 10 12 from one emplacement to another.
0 The number of available dilatometric profiles is
5 nD, and the number of CPT profiles is nC. A settle-
10 ment evaluation is possible for each profile
15

S D (x j ) = A(x j , z )mD (x j , z )
DEPTH (m)

20
j = 1...nD
S C (x p ) = A(x p , z )mC (x p , z )
25 (5)
30 p = 1...nC
35
40 Where we have introduced mC(xp,z) to represent
45 the CPT based compliance profile at xp. This may be
50 related to the qc log using (2). Note that in (2) α(z),
EXTENSOMETER 2 PREDICTION is independent of x i.e. assumed unique for all the
database.
Figure 11. Measured and predicted local settlement at the load
test site This unique linear relation α(z) may be obtained
now imposing the condition that the average settle-
The measured and CPTu predicted cumulative ment should be equal for both datasets under the
settlement curves are presented in Figure 10. There same load. This is equivalent to assume that the
is an overprediction of 28% of settlement surface. loading A is independent of x which leads to
A more detailed comparison is possible, plotting
j = nD j = nC

¦ A(z ) m (x , z ) = n ¦ A(z ) m (x , z )
measured and predicted layer settlement, and this is 1 1
made in Figure 11. It may be observed how the D j C j (6)
nD j =1 C j =1
comparison is poorer in the deeper levels. The DMT
database employed for the correlation had an aver-
nD 1 § 1 ·
age depth of 35 m; only one log went beyond 40 m. α (z ) = ¦p ¨¨ q (x , z ) ¸¸
It seems therefore reasonable to expect a poorer per- nC m (
¦ D jx , z ) © c p ¹
(7)
formance beyond these limits. If the comparison is j
restricted to the settlements above 35 m depth the
overprediction is just 14%. Applying this more formal approach to our 12
These results are not very different to generally paired tests the α(z) profile obtained in Fig. 12 has
quoted dilatometric method errors (e.g. Marchetti been obtained. It is remarkably similar to the trial
2001). Note, however, that in this case the error cov- and error obtained design one and, indeed, the load
ers that inherent in the dilatometric method, that due test predictions using this new α(z) profile are only
to the DMT-CPTu correlation and that due to site slightly better than those previously obtained.
extrapolation.
The predicted results correspond to an estimated
98% consolidation at the time of measurement. It is 4 DISCUSSION
also interesting to note that some corrections were
applied at a later design stage to take into account The advantage of the formal approach is that it
secondary consolidation; when applied at the pre- makes clear the assumptions involved in the process.
load test site they resulted in greater absolute error In our case the distance between reconnaissance
(circa 40 % overestimation, instead of 28%), al- points was greater than the span of any possible rigid
though they improved the prediction of settlement foundation. It was therefore reasonable to seek an
rate. α(z) profile giving equal average settlement under
equal load. Other cases may require different as-
3.4 Site-specific performance-based relation sumptions (equal average reaction under equal set-
tlement, for instance).
The success of the prediction exercise encouraged us
The formal approach makes also clear that, prop-
to formalize the trial and error procedure by which
erly speaking, we have not obtained α(z) through a
the α(z) design profile was established. The DMT
“correlation”. In fact a linear regression with zero in-
settlement equation (1) is a linear operation that may
tercept may be computed at each z using the pairs
be rewritten as
(MD(xp, z), qc(xp, z)). The α(z) profile is given by
the slopes of such a regression as
1
S (x ) = A(x, z ) = A(x, z ) mD (x, z ) (4)
M D (x, z ) M D (x j , z )qc (x j , z )
α (z ) = ¦
qc2 (x j , z )
j : 1...n (8)
where A(x,z) represents the loading and mD(x,z) the j
inverse stiffness or compliance dilatometric profile.
The explicit dependence on x indicates that settle-

Proceedings ISCʼ2 on Geotechnical and Geophysical Site Characterization, Viana da Fonseca & Mayne (eds.) 1609
 α =M/qc
M / qc This relation was simple: linear and depth de-
0 5 10 15 20 25 30 35 40 pendent. It was established using a performance
0
based criteria, i.e. considering settlements and not
5
directly test results. This approach may prove fruit-
10
ful in other geographical settings. Further research
15
on its statistical bearings would be useful for that
z (m)

20 purpose.
25

30

35 ACKNOWLEDGEMENTS
40

PERFORMANCE-BASED DESIGN
The authors gratefully acknowledge the permission
of AENA to use the data presented in this work. An-
Figure 12. Performance-based α(z) profile. drea Galli, from DIS at Politecnico di Milano, gave
some useful statistical hints.
α =M/qc
M / qc
0 5 10 15 20 25 30 35 40
0
REFERENCES
5

10 Alonso, E., Gens, A. & Lloret, A. 2000. Precompression de-

sign for secondary settlement reduction. Gèotechnique,
15
50,6, 645-656.
z (m)

20 ASTM D5778-95. 2000. Standard Test Method for Performing

25 Electronic Friction Cone and Piezocone Penetration Test-
30
ing of Soils.
ASTM D6635-01. 2002. Standard Test Method for Performing
35 the Flat Plate Dilatometer.
40 Devincenzi, M., Colas, S., Casamor, J.L. Canals, M., Falivene,
DESIGN CLASSICAL REGRESSION
O. & Busquets, P.. 2004. High resolution stratigraphic and
sedimentological analysis of Llobregat delta –Barcelona–
Figure 13. Linear regression α(z) profile. from CPT/CPTu Tests. Proc. 2nd International Conference
on Geotechnical Site Characterization ISC-2, Porto.
Eslami, A. & Fellenius, B.H. 1995. The bearing capacity of
The application of this formula to our database piles from cone penetration test (CPT) data .CPT’95 Inter-
results in the profile shown in Fig. 13. Using this national Symposium on Cone Penetration Testing, Swedish
profile the preload test settlement is underestimated Geotechnical Society.
by 25%. This may result in unconservative design. Gens, A. & Lloret, A. 2003. Monitoring a preload test in soft
Another advantage of the performance based ap- ground, Field measurements in Geomechanics. Myrvoll
proach is that it is not necessary to use only results (ed.) Swets & Zeitlinger, Lisse. 53-59.
ISSMGE. 2001. The Flat Dilatometer Test (DMT) in Soil In-
from paired CPT/DMT. Any number of MDMT and qc vestigations. Report of the ISSMGE, Technical Committee
logs may be included, as long as it makes sense to 16 on Ground Property Characterization from In-situ Test-
search an unique relation between them. Obviously ing. IN SITU 2001, Int. Conf. On In Situ Measurement of
this should be based on some previous appreciation Soil Properties, Bali, Indonesia.
of the area variability. Lunne, T., Robertson, P.K. & Powell, J.J.M. 1997. Cone Pen-
tration Testing in geotechnical practice. Blackie Academic
Finally it may be noticed that by enforcing the & Professiona, London, U.K.
identity of higher order statistical moments (e.g. the Marchetti, S. 2001. The flat dilatometer. Applications to geo-
variance of the settlement distributions) it will be technical design. 18th Conferencia geotecnica Torino
possible to fit as well multiparametric MDMT vs qc re- Marchetti, S. 1997. The flat dilatometer. Keynote lecture. 3rd
lationships. Such an approach may ease comparisons geotechnical engineering conference, Cairo University.
with other conceptual schemes available to relate Mayne, P.W. 2002. Equivalent CPT method for calculating
shallow foundation settlements in the Piedmont residual
these quantities, like the two-step correlation proce- soils based on the DMT constrained modulus approach. At
dure exemplified in Mayne (2002). http://www.ce.gatech.edu/~geosys/Faculty/Mayne/papers/
Mayne, P.W. 1998. Commentary on Marchetti flat dilatometer
correlations in soils. ASTM Geotechnical Testing Journal,
5 CONCLUSIONS 21, 3, 222-239.
Pelnik, T.W., Fromme, C.L., Gibbons, Y.R. & Failmezger,
DMT profiles are very informative about the engi- R.A. 1999. Foundations design applications of CPTu and
DMT tests in atlantic coastal plain. Virginia, TRB, 78th
neering characteristics of the Llobregat delta. The Annual meeting, Washington DC. at www.vgspc.com/
dilatometric method of settlement evaluation has technical/990794.htm.
been successfully employed to predict settlements of Zhu, F., Clark, J.I. & Paulin, M.J. 1995. Factors affecting at-
a test embankment. This prediction hinged on a site rest lateral stress in artificially cemented sands. Can. Geo-
– specific relation between MD and qc. tech. J. 32, 195-203.

1610 © 2004 Millpress, Rotterdam, ISBN 90 5966 009 9