ASCE Geotechnical Special Publication honoring Dr. John H. Schmertmann.

FROM RESEARCH TO PRACTICE IN GEOTECHNICAL ENGINEERING

GSP No. 170, 2008
Geo-Institute Meeting in New Orleans March 9 to 12, 2008

In Situ Tests by Seismic Dilatometer (SDMT)

Silvano Marchetti1, Paola Monaco2, Gianfranco Totani3 and Diego Marchetti4

1
Professor, University of L'Aquila, Faculty of Engineering, DISAT, 67040 Monteluco di Roio,
L'Aquila, Italy; [email protected]
2
Assistant Professor, University of L'Aquila, Faculty of Engineering, DISAT, 67040 Monteluco di
Roio, L'Aquila, Italy; [email protected]
3
Associate Professor, University of L'Aquila, Faculty of Engineering, DISAT, 67040 Monteluco di
Roio, L'Aquila, Italy; [email protected]
4
Studio Prof. Marchetti, Via Bracciano 38, 00189 Roma, Italy; [email protected]

ABSTRACT: The seismic dilatometer (SDMT) is the combination of the standard

flat dilatometer (DMT) with a seismic module for measuring the shear wave velocity
V S . This paper summarizes the experience gained from SDMT tests performed at 34
sites. In particular the paper presents an overview of the SDMT equipment,
comparisons of V S measured by SDMT and by other methods, interrelationships G 0 ,
E D , M DMT based on 2000 data points and a selection of significant SDMT results and
related comments. The paper also illustrates the major issues of present research on
use and applications of the SDMT, mostly focused on the development of methods
for deriving the in situ G-γ decay curves and for evaluating the liquefaction
resistance of sands based on SDMT results.

KEYWORDS: Seismic Dilatometer SDMT, Flat Dilatometer DMT, Shear Wave

Velocity V S , G-γ Curves, Working Strain Modulus, Liquefaction

INTRODUCTION

The seismic dilatometer (SDMT) is the combination of the traditional "mechanical"

Flat Dilatometer (DMT) introduced by Marchetti (1980) with a seismic module
placed above the DMT blade. The SDMT module is a probe outfitted with two
receivers, spaced 0.5 m, for measuring the shear wave velocity V S . From V S the
maximum shear modulus G 0 may be determined using the theory of elasticity.
Motivations of the combined probe:
– V S (and G 0 ) are at the base of any seismic analysis.
– The G-γ decay curves are an increasingly requested input in seismic analyses and
in general in non linear analyses.
– Increasing demand for liquefiability evaluations.
– Seismic site classification using directly V S rather than N SPT or c u .

1
– Availability of the usual DMT results (e.g. constrained modulus M DMT ) for current
design applications (e.g. conventional settlement predictions).
This paper comments on the most significant SDMT results obtained in the period
2004-2007 at 34 sites.
Information on the mechanical DMT, not described in this paper, can be found in
the comprehensive report by the ISSMGE Technical Committee TC16 (2001).

THE SEISMIC DILATOMETER (SDMT)

The seismic dilatometer (SDMT) is the combination of the standard DMT

equipment with a seismic module for measuring the shear wave velocity V S .
The test is conceptually similar to the seismic cone SCPT. First introduced by
Hepton (1988), the SDMT was subsequently improved at Georgia Tech, Atlanta,
USA (Martin and Mayne 1997, 1998; Mayne et al. 1999). A new SDMT system
(Fig. 1) has been recently developed in Italy. The seismic module (Fig. 1a) is a
cylindrical element placed above the DMT blade, equipped with two receivers,
spaced 0.5 m. The signal is amplified and digitized at depth. The true-interval test
configuration with two receivers avoids possible inaccuracy in the determination of
the "zero time" at the hammer impact, sometimes observed in the pseudo-interval
one-receiver configuration. Moreover, the couple of seismograms recorded by the
two receivers at a given test depth corresponds to the same hammer blow and not to
different blows in sequence, which are not necessarily identical. Hence the
repeatability of V S measurements is considerably improved (observed V S
repeatability ≈ 1-2 %). V S is obtained (Fig. 1b) as the ratio between the difference in
distance between the source and the two receivers (S 2 - S 1 ) and the delay of the
arrival of the impulse from the first to the second receiver (∆t). V S measurements are
obtained every 0.5 m of depth.

(a) (b) (c)

FIG. 1. Seismic Dilatometer: (a) DMT blade and seismic module. (b) Schematic
layout of the seismic dilatometer test. (c) Seismic dilatometer equipment.

2
FIG. 2. Shear wave source at the FIG. 3. Example of seismograms obtained
surface by SDMT at the site of Fucino (Italy)

The shear wave source at the surface (Fig. 2) is a pendulum hammer (≈ 10 kg)
which hits horizontally a steel rectangular base pressed vertically against the soil (by
the weight of the truck) and oriented with its long axis parallel to the axis of the
receivers, so that they can offer the highest sensitivity to the generated shear wave.
The determination of the delay from SDMT seismograms, normally carried out
using the cross-correlation algorithm, is generally well conditioned being based on
the two seismograms – in particular the initial waves – rather than being based on the
first arrival time or specific marker points in the seismogram.
Fig. 3 shows an example of seismograms obtained by SDMT at various test depths
at the site of Fucino (it is a good practice to plot side-by-side the seismograms as
recorded and re-phased according to the calculated delay).
Fig. 4 (Fiumicino) is an example of the typical graphical format of the SDMT
output. Such output displays the profile of V S as well as the profiles of four basic

FIG. 4. SDMT profiles at the site of Fiumicino (Italy)

3
TABLE 1. Example of repeatability of V S
measurements by SDMT (Zelazny Most
Tailing Dam, Poland)

Z VS V S values [m/s] corresponding Coefficient

[m] [m/s] to different hammer blows of variation
at each depth Z [%]
7.00 179 178,178,180,180,180,179,179,180,180,180 0.50
7.50 231 234,232,232,230,229,231,232,229,230 0.68
8.00 225 227,225,224,225,225,225,226,226,225,224,224 0.40
8.50 276 276,276,280,273,275,273,271,273,287,281 1.68
9.00 296 291,286,301,292,296,288,301,300,304,303 2.09
9.50 248 244,251,250,247,250,249,250,249,242,248 1.11
10.00 292 292,289,290,293,289,292,289,292,296,295,293 0.79
10.50 320 321,323,320,325,323,325,316,314,308,321 1.61
11.00 291 293,291,293,291,291,290,290,291,290,290 0.38
11.50 321 324,320,320,322,320,322,319,319,320,320 0.48
12.00 309 311,307,311,309,309,311,309,309,307,311 0.50
12.50 286 287,285,285,285,287,285,285,287,287,287 0.35 FIG. 5. Comparison of V S
13.00 265 265,265,265,264,265,265,265,266,265,266,264 0.24 profiles obtained by SDMT
13.50 280 287,276,279,276,276,276,294,275,278,279 2.08
14.00 312 313,312,312,322,310,312,310,310,310,312 1.10 and by SCPT, Cross-Hole
14.50 298 301,298,299,299,298,296,299,298,299,298 0.44
15.00 309 307,309,307,309,309,309,309,309,309,309 0.29
and SASW (AGI 1991) at the
research site of Fucino (Italy)

DMT parameters − the material index I D (soil type), the constrained modulus M, the
undrained shear strength c u and the horizontal stress index K D (related to OCR) −
obtained using current DMT correlations. It may be noted from Fig. 4 that the
repeatability of the V S profile is very high, similar to the repeatability of the other
DMT parameters, if not better.
Table 1 shows an example of another kind of repeatability of V S by SDMT
(Zelazny Most Tailing Dam, Poland). Each V S value at a given test depth
corresponds to a different hammer blow. The coefficient of variation of V S is in the
range 1-2 %.
Such repeatability is more than adequate for normal engineering needs. However,
especially in earthquake geotechnical engineering, where the design is increasingly
based on the acceptability of the permanent displacements, accurate V S (and G 0 )
estimates are a necessary prerequisite for correctly predicting such displacements.
V S measurements by SDMT have been validated by comparison with V S
measurements obtained by other in situ seismic tests at various research sites. As an
example Fig. 5 shows V S comparisons at the research site of Fucino, Italy (NC
cemented clay), extensively investigated at the end of the '80s. The profile of V S
obtained by SDMT in 2004 (Fig. 5) is in quite good agreement with V S profiles
obtained by SCPT, Cross-Hole and SASW in previous investigations (AGI 1991).
Similar favourable comparisons are reported by various Authors, e.g. by Hepton
(1988), McGillivray and Mayne (2004) and Młynarek et al. (2006).

INTERRELATIONSHIPS BETWEEN EXPERIMENTAL G0 , ED , MDMT

The experimental diagrams presented in this section have been constructed using
same-depth G 0 , E D , M DMT values determined by SDMT at 34 different sites, in a

4
variety of soil types. The majority of the sites are in Italy, others are in Spain,
Poland, Belgium and USA.
SDMT generates plentiful data points because each sounding routinely provides
profiles of G 0 , E D , M DMT . Of the over 2000 data points available, only 800 high
quality data points have been considered, relative to "uniform" one-m soil intervals
where log I D , K D , E D , M DMT , V S all differ less than 30 % from their average – used
then to plot the data points – to insure a proper match of the data.
The DMT parameters have been calculated with the usual DMT interpretation
formulae as in Marchetti (1980) or Table 1 in TC16 (2001).

Diagrams of the Ratio G0 / ED

The ratio G 0 /E D are plotted versus K D (Fig. 6a) and I D (Fig. 6b). It can be seen that
data points tend to group according to their I D and K D . From Fig. 6, if I D and K D are
available, rough estimates of G 0 can be made from E D , in absence of direct G 0
measurements.
Recognizable trends in Fig. 6 are: G 0 /E D is mostly in the range 1.5 to 3 in sand, 2.5
to 13 in silt, 3 to 25 in clay. The widest range and the maximum variability of G 0 /E D
are found in clay. For all soils G 0 /E D decreases as K D (related to OCR) increases.

Diagrams of the Ratio G0 / MDMT

The ratio G 0 /M DMT are plotted versus K D (Fig. 7a) and I D (Fig. 7b). The diagrams
indicate a wide range of the ratio G 0 /M DMT (≈ 0.5 to 25 for all soils), hence the
unfeasibility of estimating the operative modulus M from G 0 by dividing G 0 for a
fixed number. Again the data points tend to group according to their I D and K D . It
can be seen that G 0 /M DMT is strongly dependent on (at least) both soil type and stress
history. Hence the use of only one parameter (e.g. c u in cohesive soils) as an estimate
for V S (or G 0 ) for the seismic soil classification appears problematic.
Recognizable trends in Fig. 7 are: G 0 /M DMT is mostly in the range 0.5 to 3 in sand,
1 to 10 in silt, 1 to 20 in clay. The widest range and the maximum variability of
G 0 /M DMT are found in clay. For all soils G 0 /M DMT decreases as K D (related to OCR)
increases.
If I D and K D are available, Fig. 7, where the dispersion is slightly less than in Fig.
6, is to be preferred to Fig. 6 to obtain rough estimates of the ratio G 0 /M, i.e. G 0 from
M or M from G 0 , when only one of them is available.

Diagrams of the Ratio GDMT / G0

Fig. 8 shows the same experimental information as in Figures 6 and 7, but involves
the additional modulus G DMT derived from M DMT using the formula of linear
elasticity:

G = M / [2 (1-ν) / (1-2ν)] (1)

For ν = 0.15, 0.20 and 0.25, the denominator in Eq. 1 would be 2.43, 2.67 and 3

5
respectively. For the conversion, the denominator 2.67 has been retained, i.e.:

G DMT = M DMT / 2.67 (2)

All the G DMT have been derived from M DMT using Eq. 2, then the ratios G DMT /G 0
have been calculated too and plotted versus K D (Fig. 8a) and I D (Fig. 8b).
The reason of constructing Fig. 8 is the following. The ratio G/G 0 is the usual
ordinate of the normalized G-γ decay curve and has the meaning of a strain decay
factor. Since M DMT is a working strain modulus one might hypothesize that G DMT is a
working strain shear modulus too, in which case G DMT /G 0 could be regarded as the
shear modulus decay factor at working strains.
It is emphasized that, at this stage, the legitimacy of using linear elasticity for
deriving G DMT from M DMT (Eq. 2) and the assumption that G DMT is a working strain
shear modulus are only working hypotheses, likely more difficult to investigate than
verifying that M DMT is a working strain constrained modulus (the matter is discussed
later in the paper). The very designation working strain shear modulus (or operative
shear modulus) requires clarification. Anyway, if the above hypotheses were
acceptable, Fig. 8 could provide, if I D and K D are known, rough estimates of the
decay factor at working strains. If complete SDMT are available, then said rough
estimates could be skipped and the decay factor could be obtained directly as the
ratio between G DMT from Eq. 2 and G 0 .
Trends emerging from Fig. 8 are: (a) The G decay in sands is much less than in
silts and clays. (b) The silt and clay decay curves are very similar. (c) For all soils the
decay is maximum in the NC or lightly OC region (low K D ).
Fig. 7a and Fig. 8a are reproduced in a more readable format in Fig. 9 and Fig. 10
respectively, where the data points relative to clay, silt and sand have been plotted
separately. Best fit equations are indicated for each of the six diagrams.

IN SITU G-γ DECAY CURVES BY SDMT

SDMT provides routinely at each depth, besides a small strain modulus (G 0 from
V S ), also a working strain modulus (M DMT ). These two moduli could possibly be of
help when selecting the G-γ curves. Such potentiality is heavily founded on the basic
premise that M DMT is a reasonable estimate of the operative working strain modulus.
It is therefore considered appropriate to recall here the presently available evidence.

M DMT as an Operative or Working Strain Modulus

The terms Operative Modulus and Working Strain Modulus are considered
synonyms in this paper and are used interchangeably. They are defined as those
moduli that, introduced into the linear elasticity formulae, provide reasonable
estimates of the settlements under a shallow foundation (say for a safety factor Fs ≈
2.5 to 3.5).

6
 0.6 1.8
30 30
CLAY I D < 0.6 CLAY SILT SAND
G0 G0
SILT 0.6 < I D < 1.8
ED ED Kd = 1-2
SAND I D > 1.8
Kd = 2-4
20 20 Kd = 4-8
Kd = 8-16
Kd > 16

10 10

0 0
0 10 20 KD 30 0.1 1 ID 10
(a) (b)
FIG. 6. Ratio G 0 /E D vs. K D and I D

0.6 1.8
30 30
G0 CLAY I D < 0.6 CLAY SILT SAND
G0
SILT 0.6 < I D < 1.8
MDMT SAND I D > 1.8 MDMT
Kd = 1-2
20 20
Kd = 2-4
Kd = 4-8
Kd = 8-16
CLAY Kd > 16
10 SILT 10
SAND

0 0
0 10 20 KD 30 0.1 1 ID 10
(a) (b)
FIG. 7. Ratio G 0 /M DMT vs. K D and I D

1 0.6 1.8
1
CLAY I D < 0.6 CLAY SILT SAND
GDMT SILT 0.6 < I D < 1.8 GDMT
G0 SAND I D > 1.8 G0 Kd = 1-2
Kd = 2-4
SAND Kd = 4-8
SILT
Kd = 8-16
0.5 0.5 Kd > 16

CLAY

0 0
0 10 20 KD 30 0.1 1 ID 10
(a) (b)
FIG. 8. Decay ratio G DMT /G 0 vs. K D and I D

7
 30 30 30
G0 CLAY G0 SILT G0 SAND
MDMT I D < 0.6 MDMT 0.6 < I D < 1.8 MDMT I D > 1.8
20 20 20

10 10 10

0 0 0
0 10 20 KD 30 0 10 20 KD 30 0 10 20 KD 30
G 0 /M DMT = 26.177 K D –1.0066 G 0 /M DMT = 15.686 K D –0.921 G 0 /M DMT = 4.5613 K D –
0.7967

R2 = 0.61 R2 = 0.81 R2 = 0.65

FIG. 9. Ratio G 0 /M DMT vs. K D for clay, silt and sand

1 1 1
CLAY SILT
GDMT GDMT GDMT
I D < 0.6 0.6 < I D < 1.8
G0 G0 G0

0.5 0.5 0.5

SAND
I D > 1.8
0 0 0
0 10 20 30 0 10 20 30 0 10 20 KD 30
KD KD
G DMT /G 0 = – 0.0002 K D 2 + 0.022 K D – 0.0173 G DMT /G 0 = 0.0241 K D 0.919 G DMT /G 0 = 0.0826
K D 0.7961
R2 = 0.85 R2 = 0.81 R2 = 0.66
FIG. 10. Decay ratio G DMT /G 0 vs. K D for clay, silt and sand

Comparisons of Surface Settlements

Schmertmann (1986) reported 16 case histories at various locations and for various
soil types, with measured settlements ranging from 3 to 2850 mm. In most cases
settlements from DMT were calculated using the Ordinary 1-D Method. The average
ratio DMT-calculated/observed settlement was 1.18, with the value of the ratio
mostly in the range 0.7 to 1.3 and a standard deviation of 0.38.
Monaco et al. (2006) reviewed numerous other real-life well documented
comparisons of DMT-predicted versus measured settlements. The average ratio
DMT-calculated/observed settlement for all cases reviewed by Monaco et al. (2006)
is ≈ 1.3, with an observed settlement within ± 50 % from the DMT-predicted
settlement.

8
 The above settlements comparisons appear to support the assumption that M DMT is
a reasonable estimate of the constrained working strain modulus.

9
Comparisons of Moduli

Even more direct, but rarely available, are data comparing M DMT with moduli back-
figured from local vertical strain measurements.
In 2002 a major research project, funded by the Italian Ministry of University and
Scientific Research and by Consorzio Venezia Nuova, was undertaken by a
consortium of three Italian Universities (Padova, Bologna and L'Aquila).
A full-scale cylindrical heavily instrumented test embankment (40 m diameter,
6.7 m height, applied load 104 kPa – Fig. 11a) was constructed at the site of Venezia-
Treporti, typical of the highly stratified, predominantly silty deposits of the Venezia
lagoon (Fig. 11b). The loading history, the progression of the settlements and the
drainage conditions – practically fully drained – are shown in Fig. 11c.
A specific aim of the research was to obtain a profile of the observed 1-D operative
modulus M under the center of the embankment. For this purpose a high precision
sliding micrometer was used to accurately measure the local vertical strain ε v at 1 m
depth intervals.
M values were back-calculated from local vertical strains ε v in each 1 m soil layer
as M = ∆σ v / ε v , with vertical stress increments ∆σ v calculated at the mid-height of
each layer by linear elasticity formulae (approximation considered acceptable in
view of the very low ε h ).
Fig. 11d, which is believed to be one of the most important results of the Venezia-
Treporti research, shows an overall satisfactory agreement between M DMT and
moduli back-figured from the test embankment performance, also considering the
marked soil heterogeneity. Fig. 11e compares the observed versus DMT-predicted
settlements at each depth. Again the agreement is rather satisfactory, considering that
the DMT predicted settlements were calculated using the simple linear 1-D
conventional approach s = Σ (∆σ v / M DMT ) ∆H, where ∆σ v is calculated by Boussinesq
linear elasticity formulae.
As to the surface settlements, the total settlement measured under the center of the
embankment at the end of construction (180 days) was ≈ 36 cm (Fig. 11c). The
settlement predicted by M DMT using the 1-D approach (before knowing the results)
was 29 cm. Hence the 29 cm predicted by DMT (which does not include secondary)
are in good agreement with the 36 cm observed settlement (which includes some
secondary during construction).
(Cunha (2004) wondered if it is right to attribute the 7 cm settlement difference
(20 %) to secondary or to other causes. The question is legitimate. But in practical
circles settlement predictions within a factor of 1.5-2 are often considered
satisfactory for normal design. Mayne (2005) reports an SPT-DMT settlement
comparison where the SPT prediction was wrong by a factor of 5 (the DMT
prediction "was in line with the observed performance"). Factors of 3 or 5 are not
unusual. In general, engineering decisions on the foundation type are based on
knowing if the foundation will settle 1 mm or 10 mm or 100 mm. While it is
desirable to reduce the error factor below 5 or 3, once it is reduced to 1.5 to 1.2,
additional efforts may not be worthwhile or even inherently unfruitful due the many
uncertainties involved in the definition of the settlement components and in their
precise measurement).

10
 VS (m/s)
0 100 200 300 400
0
4
8
12
16
20

z (m)
24
28
32
36
40
44

(a) Test embankment. Penetrometer

truck for testing after construction. (b) Superimposed profiles of all SDMT data
M (MPa) S (mm)
End of construction
Settlement (mm) Embankment

6.7 0 20 40 60 80 0 100 200 300 400

(q = 104 kPa)
height (m)

0 0

z (m)

z (m)
10 M back-calculated 10
0 180 Time (days) 720 from vertical strain
M from DMT
DMT predicted 20 20
290
360 Measured Measured at end of
30 30 construction (H=6.7 m)
480
∆u (from piezometers) ≈ 0 throughout embankment
Calculated by DMT
construction and subsequently
40 40

(c) Settlement vs. time at the center of the (d) M DMT vs. M back-calculated (e) Observed vs. DMT-predicted
embankment and comparison of from local ε v measured at 1 m settlement under the center at
measured vs. DMT-predicted settlements depth intervals under the center the end of construction
at the end of construction at the end of construction
FIG. 11. Venezia-Treporti Research Embankment. SDMT profiles. Predicted vs. observed moduli and settlements (Marchetti et al. 2006).

11
 FIG. 12. Tentative method for deriving G-γ curves from SDMT

More details on the Venezia-Treporti research can be found in Marchetti et al.

(2006), also containing numerous additional bibliographic references.
In conclusion also the Venezia-Treporti case-history supported the assumption that
M DMT is a reasonable estimate of the constrained working strain modulus.
A note of caution. In OC clays oedometer moduli, even from good quality samples,
are generally too low to be taken as operative moduli in situ. As noted by Ladd
(1971), quoting Terzaghi-Peck (1964), the compressibility of even good oedometer
samples of OC clay may be 2 to 5 times larger than the in situ compressibility.
Indeed many engineers feel that the Skempton-Bjerrum reduction factor for
settlements in OC clays was prompted by the fact that oedometer-based settlement
predictions in OC London clay were systematically too high. Moduli back-figured
from local strain measurements are far more realistic and are preferable for
calibration or comparison purposes.

Construction of the G-γ Curves

Such curves could tentatively be constructed by fitting "reference typical-shape"

laboratory curves (see Fig. 12, where G is normalized to G 0 ) through two points,
both obtained by SDMT: (1) the initial modulus G 0 from V S , and (2) a working strain
modulus G DMT (Eq. 2). To locate the second point it is also necessary to know, at
least approximately, the shear strain corresponding to G DMT . Indications by Mayne
(2001) locate the DMT moduli at an intermediate level of strain (γ ≈ 0.05-0.1 %)
along the G-γ curve. Similarly Ishihara (2001) classified the DMT within the group
of methods of measurement of soil deformation characteristics involving an
intermediate level of strain (0.01-1 %). The above indications, to be supplemented by
further investigations, could possibly help develop methods for deriving in situ G-γ
curves from SDMT.
Lines of research currently under investigation are:
(a) Enter the G DMT /G 0 ratios of Fig. 8 in the vertical axis of "reference typical-
shape" G-γ curves recommended in the literature for the corresponding soil type. The

12
range of abscissas of the intersection points with the G-γ curves could possibly help
to better define the shear strain corresponding to G DMT .
(b) Develop a procedure for selecting the G-γ curve, among the typical curves
recommended in the literature, making use of I D for choosing the band of curves
recommended for the soil type (sand or silt or clay), and K D (possibly G 0 /M DMT too)
for selecting one curve in the band.
(c) Evaluate for each of the 800 data points in Figures 6-7-8 the settlement under a
simple loading scheme using the simple linear analysis with input M DMT (operation
equivalent to converting a DMT investigation into a "virtual" load test). Then
calculate the settlement by non linear analyses with G-γ curves having variable rates
of decay as input. By trial and error identify the G-γ curve (originating in G 0 )
producing agreement between the two predicted settlements. Consider such G-γ
curve reasonably correct and use it in the development of procedures for selecting the
G-γ curves from SDMT data.

DERIVABILITY OF THE OPERATIVE MODULUS M FROM G0

This section discusses the possibility of deriving the operative modulus M by

dividing G 0 by a constant.
Fig. 13 (Barcelona airport site) shows that, while the modulus M DMT exhibits a
drastic drop at ≈ 12 m depth, at the transition from an upper stiff sand layer to a
lower very soft clay layer, V S shows only a minor decrease. Hence G 0 = ρ V S 2, even
considering the power 2, is far from being proportional to the working strain
modulus M. Similar lack of proportionality, with variations of the ratio G 0 /M DMT
often of one order of magnitude, has been observed at many sites (including Venezia,
Fig. 11d), suggesting that it is next to impossible (at least without local layer-specific
correlations) to derive the working strain modulus by simply reducing the small
strain modulus by a fixed percent factor (e.g. 50 %, Simpson 1999).
On the other hand the poor correlability was expected, since at low strains the soil
tendency to dilate or contract is not active yet. Such tendency substantially affects
the operative modulus M, but does not affect G 0 . Said in a different way, M includes
some stress history information, G 0 does not (Powell and Butcher 2004).
The high variability of the ratio G 0 /M was already noted earlier in Fig. 7. That
figure explains that part of the difference in the amount of decrease in the Barcelona
M and V S profiles is due to the transition from sand (where G 0 /M DMT is in the range
0.5-3) to clay (where G 0 /M DMT is in the range 1-20). Fig. 7 (or the equivalent Fig. 9),
entered with I D and K D , can provide rough estimates of the ratio, hence M from G 0 .
However the direct measurements of both M and G 0 are preferable if accurate
estimates of these parameters are required.

ESTIMATING OCR IN SAND FOR ADVANCED PREDICTIVE MODELS

Computer programs based on advanced soil models are increasingly used not only
by researchers but also by practitioners for everyday design. Hence an important aim
of in situ tests is to possibly provide input parameters for such models.

13
 FIG. 13. SDMT profiles at the site of Barcelona – El Prat Airport (Spain)

An input parameter required by many models is OCR. In clays reasonable OCR

estimates can be obtained by laboratory samples, or in situ sometimes by the DMT
OCR-K D correlations. In sands, however, OCR evaluations are problematic. As
today, to the authors' knowledge, one of the least imprecise methods for evaluating
OCR in sand by in situ tests is the one suggested by TC16 (2001). Such method
consists in evaluating OCR based on the ratio M DMT /q c (q c from CPT) using as a
guide: M DMT /q c = 5-10 in NC sands, M DMT /q c = 12-24 in OC sands. The basis of
such ratios is explained in TC16 (2001). In essence: when imparting stress history to
the sand – for instance by compaction – M DMT increases considerably faster than q c
(see e.g. Schmertmann et al. 1986, Jendeby 1992), hence M DMT /q c increases with
OCR.
The above guide provides very rough estimates of OCR. The uncertainty however
may not be due necessarily to the approximate nature of the correlations. Many sands
could simply not have a well defined break in their insitu e-log p curve (Fig. 14) due
to the complexity of the OCR sources (cementation, desiccation, water level
fluctuations, etc.). Besides, layers of sands are often stratified, as the sand layers in
Fig. 11b, with OCR presumably seesawing, so that the break in the e-log p of the
composite layer will be rounded, and assigning a unique value to OCR of the layer
could be inappropriate.
Uncertainty in the values of OCR represents a problem for many models requiring
a precise value of OCR in sand because their predictions are generally critically
sensitive to OCR. For instance in the case of Venezia-Treporti mentioned before,
using advanced models with OCR = 1.2 or OCR = 1.6 leads to completely different
settlement predictions, while the conventional analysis using the tangent modulus
M DMT produced a stable and satisfactory prediction (Fig. 11c).
The cause of the advanced model sensitivity can be understood even with the
simple oedometer scheme. If a sand has an in situ e-log p curve of the type in Fig.
14b, and is modeled as in Fig. 14a, the predicted settlements will be highly
dependent on the choice of the abscissa of the (in reality non existing) break.

14
FIG. 14. Variation of e and M with p in presence/absence of a well defined break

Therefore sands whose in situ e-log p curve are of the type in Fig. 14b should
probably be treated with models attenuating the stiffness discontinuity across the
preconsolidation stress.
In short, the problem could be not so much the determination of the exact OCR in
sand, but the possibly incorrect assumption of the model that the sand has a sharp
break.
It is generally recommended to run both advanced and traditional analyses to avoid
gross errors that are likely if one is not familiar with the details of the program. This
recommendation appears appropriate in this case, because, at least in the described
Venezia-Treporti case, the simplified analysis using M produced a stable and
satisfactory prediction, while the "advanced" analysis was found to be critically
dependent on the possibly elusive value of OCR.

USE OF SDMT FOR LIQUEFACTION

SDMT routinely provides, among other measurements, pairs of profiles of K D and

V S – both correlated with the liquefaction resistance of sands. Hence SDMT permits
to obtain two parallel independent estimates of liquefaction resistance CRR, one
from K D and one from V S , using CRR-K D and CRR-V S correlations – where CRR is
the cyclic resistance ratio, a basic input in the commonly used Seed and Idriss (1971)
simplified procedure.
The use of V S for evaluating CRR is well known. The most popular CRR-V S
correlation (Fig. 15) is the one proposed by Andrus and Stokoe (2000), modified by
Andrus et al. (2004). CRR is obtained as a function of V S1 = V S (p a /σ' v0 ) 0.25, shear
wave velocity corrected for the overburden stress σ' v0 (p a = atmospheric pressure).
The CRR-V S1 curves in Fig. 15 are for magnitude M w = 7.5 earthquakes (magnitude
scaling factors should be applied for different magnitudes).
Correlations CRR-K D have been developed in the last two decades, stimulated by
the recognized sensitivity of K D to a number of factors which are known to increase
liquefaction resistance, which are difficult to sense by other tests, such as stress
history, prestraining, cementation, structure, and by the relationship of K D to relative
density and state parameter.

15
FIG. 15. Curves for evaluating CRR FIG. 16. Curves for evaluating CRR
from V S for clean uncemented soils from K D (Monaco et al. 2005)
(Andrus and Stokoe 2000)

A key element of the correlation CRR-K D (Monaco and Schmertmann 2007,

Monaco and Marchetti 2007) is the ability of K D to reflect aging in sands.
Calibration chamber data (Jamiolkowski and Lo Presti 1998) suggest that sensitivity
to aging of sand is 3 to 7 times higher for K D than for penetration resistance (see Fig.
6 in Monaco and Marchetti 2007). Aging in sand is a factor having a first order of
magnitude influence on liquefaction behavior, as pointed out e.g. by Leon et al.
(2006).
Fig. 16 summarizes the various correlations developed to estimate CRR from K D
(for magnitude M = 7.5 and clean sand) – to be used according to "simplified
procedure" – including the latest CRR-K D correlation (Monaco et al. 2005), based on
all previous data.
Comparisons based on parallel measurements of K D and V S by SDMT at several
sandy sites (Maugeri and Monaco 2006) have indicated that methods based on K D
and V S often provide, at the same site, substantially different estimates of CRR.
Generally CRR from V S was found to be "more optimistic".
This finding opens the question "which CRR should be given greater weight",
which is further discussed in the next section.

SDMT RESULTS AT VARIOUS TEST SITES

This section presents three examples of SDMT results considered of some interest.

OCR and K D Crusts in Sand

"Crust-like" K D profiles, very similar to the typical K D profiles found in OC

desiccation crusts in clay, have been found at the top of many sand deposits. Various
indications (Maugeri and Monaco 2006) suggest that "K D crusts" in sands reflect
stress history (OCR, cementation, aging and/or other effects), rather than higher
relative density. In the case shown in Fig. 17 (Catania), as in many other cases, the

16
existence of a shallow "stress history crust" (believed by far not liquefiable) is
clearly highlighted by the K D profile, but almost "unfelt" by the V S profile. This
suggests a lesser ability of V S to profile liquefiability.

Role of the Interparticle Bonding

The SDMT profiles in Fig. 18 (Cassino) show relatively high V S values coexisting
with very low values of K D and moduli M. A possible explanation: the shear wave
travels fast due to the interparticle bonding (typical of many volcanic sands in this
area), preserved at small strains. By contrast K D is "low" because it reflects a
different material, where the interparticle bonding has been at least partly destroyed
by the DMT blade penetration. As noted by Andrus and Stokoe (2000), weak
interparticle bonding can increase V S (measured at small strains), while not
necessarily increasing resistance to liquefaction, a phenomenon occurring at medium
to high strains (range of K D measurement). Thus, for liquefiability, the K D
predictions could be possibly more fitting in case of strong earthquakes. Very light
earthquakes, however, may not destroy bonding, then CRR evaluated by V S may be
appropriate in this case.

Limiting "No Liquefaction" Values of V S1 and K D

The vertical asymptotes of the curves CRR-V S1 (Fig. 15) and CRR-K D (Fig. 16)
identify limiting values of V S1 and K D (roughly, for clean sand, V* S1 = 215 m/s and
K* D = 5.5) for which liquefaction can be definitely excluded for any earthquake.
At the site of the Zelazny Most dam (Fig. 19) the indications derivable from such
vertical asymptotes put in evidence a clear contradiction. While the values of
V S1 > 215 m/s suggest "no liquefaction" even in case of strong earthquakes, the
values of K D ≈ 1.5-2 indicate that liquefaction may occur above a certain seismic
stress level (high cyclic stress ratio CSR). Even in this case, as noted before, CRR
from V S is "more optimistic" – one reason being that it partly relies on bonding or
equivalent phenomena.
In the case of the Zelazny Most dam the contradiction has no practical effects,
since the region is non seismic and both Fig. 15 and Fig. 16 indicate no liquefaction.
But for high seismicity cases the question "which CRR should be given greater
weight" remains open.
Figures 17-18-19 are also examples of a commonly noted feature: the much
smoother shape of the V S profiles compared with the M or K D profiles.

SDMT INSIDE BACKFILLED BOREHOLES

In cases where the soil is too hard to penetrate (or even in rock), SDMT can be
carried out inside a borehole backfilled with sand (only V S , no DMT measurements).
The good agreement observed between V S profiles obtained by parallel SDMT
soundings carried out, at the same site, in the natural soil and in a backfilled borehole
(Fig. 20) supports the reliability of V S values obtained by this procedure.

17
FIG. 17. SDMT profiles at the site of Catania – San Giuseppe La Rena (Italy)

FIG. 18. SDMT profiles at the site of Cassino (Italy)

FIG. 19. SDMT profiles at the site of the Zelazny Most Tailing Dam (Poland)

18
 Vs (m/s)
0 200 400 600 800
0

10

15

Z (m)
20

25

30
SDMT in the natural soil
SDMT in a drilled hole filled with sand
35

FIG. 20. Comparison of V S profiles obtained by SDMT in the natural soil and in
a backfilled borehole at the site of Montescaglioso – Ginosa (Matera), Italy

CONCLUSIONS

The seismic dilatometer (SDMT) provides accurate and highly reproducible

measurements of the shear wave velocity V S – a basic input parameter for seismic
analyses. Besides V S , SDMT provides the usual DMT results (e.g. constrained
modulus M DMT ) for current design applications.
Based on a large number of results by SDMT, diagrams showing experimental
interrelationships G 0 , E D , M DMT have been constructed. In particular Fig. 9 and
Fig. 10 illustrate the most significant observed trends.
Recent experience indicates that SDMT investigations can be performed with good
results also in unusual conditions, e.g. offshore or in non penetrable soils (V S - only
measurements in backfilled boreholes).
Current research investigates the possible use of the SDMT for deriving "in situ"
decay curves of soil stiffness with strain level, by fitting "reference G-γ curves"
through two points provided by SDMT at different strain levels: the small strain
shear modulus G 0 (from V S ) and a working strain modulus corresponding to M DMT .
Deriving the operative modulus M for settlement predictions from G 0 appears
arduous. Often to drastic variations in the M profile correspond barely visible
variations in the G 0 profile. The ratio G 0 /M varies in the wide range 0.5 to 20
(Fig. 7), hence it is far from being a constant, especially in clays and silts. Its value is
strongly dependent on multiple information, e.g. soil type and stress history. Hence
the use of only one information (e.g. c u in cohesive soils) as a proxy of V S (or G 0 ) for
the seismic soil classification appears problematic.
If only mechanical DMT results are available rough estimates of G 0 from M can be
obtained from Fig. 9.
The SDMT provides two parallel independent evaluations of the liquefaction
resistance CRR from V S and from K D (horizontal stress index) by means of

19
correlations CRR-V S (Fig. 15) and CRR-K D (Fig. 16), to be used in the framework of
the Seed and Idriss (1971) simplified procedure. Preliminary studies indicate that
methods based on K D and V S often provide substantially different estimates of CRR.
In principle, the authors would propose to give greater weight to CRR by K D for
various reasons – above all the higher sensitivity of K D to stress history and aging,
factors which greatly increase liquefaction resistance. Very light earthquakes,
however, may not destroy bonding, and in that case CRR evaluated by V S may be
more appropriate. The above obviously deserves additional verification, supported by
real-life liquefaction case histories.

REFERENCES

AGI (1991). "Geotechnical Characterization of Fucino Clay." Proc. X ECSMFE,

Firenze, 1, 27-40.
Andrus, R.D. and Stokoe, K.H., II. (2000). "Liquefaction resistance of soils from
shear-wave velocity." J. Geotech. Geoenv. Engrg., ASCE, 126(11), 1015-1025.
Andrus, R.D., Stokoe, K.H., II and Juang, C.H. (2004). "Guide for Shear-Wave-
Based Liquefaction Potential Evaluation." Earthquake Spectra, 20(2), 285-305.
Cunha, R.P. (2004). "Discussion Report: Applications to Geotechnical Structures."
Proc. 2nd Int. Conf. on Site Characterization ISC'2, Porto, 2, 1391-1395.
Hepton, P. (1988). "Shear wave velocity measurements during penetration testing."
Proc. Penetration Testing in the UK, ICE, 275-278.
Ishihara, K. (2001). "Estimate of relative density from in-situ penetration tests."
Proc. Int. Conf. on In Situ Measurement of Soil Properties and Case Histories,
Bali, 17-26.
Jamiolkowski, M. and Lo Presti, D.C.F. (1998). "DMT research in sand. What can be
learned from calibration chamber tests." 1st Int. Conf. on Site Characterization
ISC'98, Atlanta. Oral presentation.
Jendeby, L. (1992). "Deep Compaction by Vibrowing." Proc. Nordic Geotechnical
Meeting NGM-92, 1, 19-24.
Ladd, C.C. (1971). "Settlement analyses for cohesive soils." Special Summer
Program on Soft Ground Construction, Report R71-2, Soils Publication 272.
Leon, E., Gassman, S.L. and Talwani, P. (2006). "Accounting for Soil Aging When
Assessing Liquefaction Potential." J. Geotech. Geoenv. Engrg., ASCE, 132(3),
363-377.
Marchetti, S. (1980). "In Situ Tests by Flat Dilatometer." J. Geotech. Engrg. Div.,
ASCE, 106(GT3), 299-321.
Martin, G.K. and Mayne, P.W. (1997). "Seismic Flat Dilatometer Tests in
Connecticut Valley Varved Clay." ASTM Geotech. Testing J., 20(3), 357-361.
Martin, G.K. and Mayne, P.W. (1998). "Seismic flat dilatometer in Piedmont
residual soils." Proc. 1st Int. Conf. on Site Characterization ISC'98, Atlanta, 2,
837-843.
Maugeri, M. and Monaco, P. (2006). "Liquefaction Potential Evaluation by SDMT."
Proc. 2nd Int. Conf. on the Flat Dilatometer, Washington D.C., 295-305.

20
Mayne, P.W. (2001). "Stress-strain-strength-flow parameters from enhanced in-situ
tests." Proc. Int. Conf. on In Situ Measurement of Soil Properties and Case
Histories, Bali, 27-47.
Mayne, P.W. (2005). "Unexpected but foreseeable mat settlements of Piedmont
residuum." Int. J. of Geoengineering Case Histories, Vol. 1, Issue 1, 5-17.
Mayne, P.W., Schneider, J.A. and Martin, G.K. (1999). "Small- and large-strain soil
properties from seismic flat dilatometer tests." Proc. 2nd Int. Symp. on Pre-
Failure Deformation Characteristics of Geomaterials, Torino, 1, 419-427.
McGillivray, A. and Mayne, P.W. (2004). "Seismic piezocone and seismic flat
dilatometer tests at Treporti." Proc. 2nd Int. Conf. on Site Characterization ISC'2,
Porto, 2, 1695-1700.
Młynarek, Z., Gogolik, S. and Marchetti, D. (2006). "Suitability of the SDMT
method to assess geotechnical parameters of post-flotation sediments." Proc.
2nd Int. Conf. on the Flat Dilatometer, Washington D.C., 148-153.
Monaco, P. and Marchetti, S. (2007). "Evaluating liquefaction potential by seismic
dilatometer (SDMT) accounting for aging/stress history." Proc. 4th Int. Conf. on
Earthquake Geotechnical Engineering ICEGE, Thessaloniki.
Monaco, P., Marchetti, S., Totani, G. and Calabrese, M. (2005). "Sand liquefiability
assessment by Flat Dilatometer Test (DMT)." Proc. XVI ICSMGE, Osaka, 4,
2693-2697.
Monaco, P. and Schmertmann, J.H. (2007). Discussion of "Accounting for Soil
Aging When Assessing Liquefaction Potential" by Leon, E. et al. (in J. Geotech.
Geoenv. Engrg., ASCE, 2006, 132(3), 363-377). J. Geotech. Geoenv. Engrg.,
ASCE, 133(9), 1177-1179.
Monaco, P., Totani, G. and Calabrese, M. (2006). "DMT-predicted vs observed
settlements: a review of the available experience." Proc. 2nd Int. Conf. on the
Flat Dilatometer, Washington D.C., 244-252.
Powell, J.J.M. and Butcher, A.P. (2004). "Small Strain Stiffness Assessments from in
Situ Tests." Proc. 2nd Int. Conf. on Site Characterization ISC'2, Porto, 2, 1717-
1722.
Schmertmann, J.H. (1986). "Dilatometer to compute Foundation Settlement." Proc.
ASCE Spec. Conf. on Use of In Situ Tests in Geotechnical Engineering In Situ
'86, Virginia Tech, Blacksburg. ASCE Geotech. Spec. Publ. No. 6, 303-321.
Schmertmann, J.H., Baker, W., Gupta, R. and Kessler, K. (1986). "CPT/DMT
Quality Control of Ground Modification at a Power Plant." Proc. ASCE Spec.
Conf. on Use of In Situ Tests in Geotechnical Engineering In Situ '86, Virginia
Tech, Blacksburg. ASCE Geotech. Spec. Publ. No. 6, 985-1001.
Seed, H.B. and Idriss, I.M. (1971). "Simplified procedure for evaluating soil
liquefaction potential." J. Geotech. Engrg. Div., ASCE, 97(9), 1249-1273.
Simpson, B. (1999). "Engineering needs." Proc. 2nd Int. Symp. Pre-Failure
Deformation Characteristics of Geomaterials, Torino.
TC16 (2001). "The Flat Dilatometer Test (DMT) in Soil Investigations - A Report by
the ISSMGE Committee TC16." May 2001, 41 pp. Reprinted in Proc. 2nd Int.
Conf. on the Flat Dilatometer, Washington D.C., 7-48.
Terzaghi, K. and Peck, R.B. (1964). "Soil Mechanics in Engineering Practice." John
Wiley & Sons.0

21