Sie sind auf Seite 1von 6

Physical Modelling in Geotechnics – McNamara et al.

(Eds)
© 2018 Taylor & Francis Group, London, ISBN 978-1-138-34419-8

Shear wave velocity: Comparison between centrifuge


and triaxial based measurements

G. Cui, C.M. Heron & A.M. Marshall


Nottingham Centre for Geomechanics, Department of Civil Engineering,
University of Nottingham, Nottinghamshire, UK

ABSTRACT: As part of a project investigating ground-borne vibration using a geotechnical centrifuge, which
involves very small strain shear waves, the need for an accurate measurement of small strain shear modulus (Gmax )
arose. Gmax can be back-calculated from measured shear wave velocities (vs ). Previous researchers have measured
vs in-flight by using an air-hammer device and sensitive accelerometers to detect the speed of the generated shear
wave. However, issues relating to inaccurate position measurement of the accelerometers, background noise,
and the recording of very small signals at sufficiently high frequency can cause errors in the measurement of
vs . This paper compares centrifuge air-hammer based measurements of vs against those obtained from a bender
element system integrated into a triaxial apparatus. The paper details the testing methodologies adopted and
provides results which indicate that the two methods agree well with each other and with predicted values from
literature. The paper highlights how the precise processing methodology adopted can affect the accuracy of the
obtained values of vs .

1 INTRODUCTION involve the vertical oscillation of a single pile embed-


ded within a dry silica sand. An accurate evaluation of
With ever increasing urban population density and the soil properties (e.g. dynamic shear modulus) over the
corresponding need for efficient and expedient means pile depth is therefore required.
of transport, rail-based public transportation systems Dynamic shear modulus (Gmax ) is one of the most
are a popular choice (such as Crossrail and HS2 in the important parameters in soil dynamics and can be
UK and the high-speed rail network in development back-calculated by measuring shear wave velocity (vs ).
in China). However, the noise and vibrations gener- Previous researchers have measured vs in centrifuge
ated by such railways can cause problems in an urban models by using an air-hammer device and sensitive
setting. The need for accurate evaluations of the envi- accelerometers to detect the speed of the generated
ronmental impact of noise and vibrations from these shear waves. However, issues relating to inaccurate
transport systems has motivated recent research in this position measurement of the accelerometers, back-
area (Kuo 2010, Nugent et al. 2012). ground noise, and the recording of very small signals
Yang et al. (2013) conducted centrifuge tests and at sufficiently high frequency can cause errors in the
numerical modelling to study the effect of the varia- measurement of vs .
tion of soil parameters with depth on ground-borne The purpose of this paper is to demonstrate how air-
vibrations. This work highlighted the sensitivity of hammer based measurements in a centrifuge compare
ground-borne vibration behaviour to the precise selec- against shear wave velocities measured from a bender
tion of key soil parameters, such as shear modulus element system integrated within a triaxial cell. Com-
and damping. However, obtaining accurate values for parisons are also made against two empirical equations
these parameters, and how they vary with stress level, available from the literature. The paper provides an
is challenging. This is particularly the case at low overview of the testing methodologies as well as details
stress levels, near the ground surface, where parameter of data processing, which had an important impact on
values tend to vary significantly. outcomes.
A research project being conducted at the Uni-
versity of Nottingham is investigating ground-borne
vibrations using a geotechnical centrifuge.This project 2 SHEAR MODULUS
focuses on the transmission of vibrations to, or from,
a pile. Vibrations generated by a railway can propa- Railway-induced vibration is small (0.1 to 1 cm/s2 )
gate through a soil-pile-building pathway and affect (Tsuno et al. 2005) and induce soil shear strains within
the operation of sensitive equipment or cause distur- the very small range (10−6 to 10−5 ) (BS 2005). Within
bance to occupants (Kuo 2010). The centrifuge tests this range, shear modulus essentially remains constant

137
(i.e. at Gmax ) (Rollins et al. 1998), which can be Table 1. Physical properties of Congleton sand (Lauder
calculated from measured values of vs using 2010).

Parameter Value

Specific gravity (Gs ) 2.63


where ρ is soil density. There are various factors that D10 0.090 mm
affect the magnitude of Gmax , including confining D30 0.120 mm
stress, void ratio, over-consolidation ratio and geologic D60 0.170 mm
age (Dobry & Vucetic 1987); this study focuses solely D90 0.195 mm
on the effect of confining stress. Gmax is commonly Minimum dry unit weight (γd min ) 14.6 kN/m3
measured in the laboratory at the element scale using Maximum dry unit weight (γd max ) 17.6 kN/m3
resonance column, torsion, and bender element tests Maximum void ratio (emax ) 0.769
Minimum void ratio (emin ) 0.467
(Shantz 2012). Several methods have been developed
by centrifuge modellers to measure in-flight values
of vs , including bender elements (Brandenberg et al.
2006, Lee et al. 2012, Kim 2010, Lee et al. 2014) and a vibration absorbing material, Duxseal (Steedman et
air-hammers (Ghosh 2002, Arulnathan 2000). al. 1990, Cheney et al. 1990), was used to absorb the
Several empirical equations have been proposed vibrations. The steel container was lined with Duxseal
to estimate Gmax of dry sand, including Hardin and to a thickness of 19 mm along the sides and 34 mm on
Richart (1963): the bottom. A thicker layer was placed on the base to
absorb vertically propagating waves from the end of
the pile.

4.2 Air-hammer setup


and Hardin and Black (1968):
A miniature air-hammer (AHA) was embedded in the
sand to generate shear waves. As shown in Figure 1,
the air-hammer is a 10 cm long, 6 mm diameter tube
with a 20 mm long pellet inside. The pellet is made
of 2 PTFE hollow cylinders, an steel cylinder and 2
where e is the void ratio of the soil model, K0 is
bolts. To generate the shear waves within the soil, air is
the at rest earth pressure coefficient, and σm is the
supplied through a tube to force the pellet to impact an
mean effective stress (kPa). These expressions are both
end of the tube. Consequently, the tube moves slightly
applicable to the round-grained sand employed in the
in the soil resulting in a small shear wave. A thin layer
current study. Predictions obtained from these meth-
of sand was bonded to the surface of the tube to ensure
ods are compared against experimental data obtained
shear waves are transferred to the soil effectively.
from triaxial tests later in the paper.
The progression of the shear wave is detected
by a vertical array of four MEMS (Micro-Electro-
Mechanical Systems) (type: ADXL001-70BEZ)
3 MATERIAL PROPERTIES
accelerometers labelled AHA-M1 to M4 in Figure 1.
The MEMS have a range of ±70 g and were mounted
HST95 Congleton sand was used for all experiments.
on a continual flexible plastic strip to reduce uncer-
It is a very fine and uniformly rounded silica sand.
tainty in the inter-MEMS distances. The range of
The basic parameters of Congleton sand are shown in
the MEMS is considerably greater than the measured
Table 1 (Lauder 2010). Samples were prepared by air
accelerations generated from the shear wave, however
pluviation using flow rate and drop height calibrated to
MEMS with a smaller measurement range and a high
obtain a relative density of approximately 85% (± 2%)
frequency response are not commercially available.
for both the centrifuge and bender element tests.
The sensitive axis of the MEMS was aligned with the
horizontal particle motion direction of the shear wave.
A high sampling rate of 100 kHz was used to record
4 CENTRIFUGE MODELLING AND
the acceleration time histories from the MEMS to cre-
AIR-HAMMER TESTING
ate a sufficiently small sampling interval relative to
the shear wave travel time, thereby minimising errors
4.1 Centrifuge modelling
(Lee et al. 2014).
A cylindrical steel container with a diameter and depth MEMS were installed at 60 mm (3.6 m proto-
of 500 mm was used for the centrifuge tests, which type), 180 mm (10.8 m), 300 mm (18 m), and 420 mm
were performed at an acceleration of 60 g. The sand (25.2 m) from the ground surface (shown in Figure 1)
was poured to a depth of 440 mm (26.4 m at proto- to measure the average shear wave velocity (vs,AHA,1 ,
type scale).An undesirable problem caused by the rigid vs,AHA,2 and vs,AHA,3 ) of each soil layer (L1 , L2 and
boundaries of the container is that vibration waves are L3 ). All dimensions regarding depths of MEMS and
reflected back into the model. To reduce this effect, distances between soil layers, shown in Figure 1, are

138
with a sampling rate of 2000 kHz was used to obtain
readings.
Two linear variable differential transformers
(LVDTs) with a range of ±2.5 mm were installed to
measure vertical deformation of the specimen. The
measurement resolution using a 16 bit data acquisi-
tion is 0.1 µm, allowing the detection of very small
axial strains (10−6 ).
It has been found that σv and σh contribute equally
to the value of the shear wave velocity, which implies
that σm in equations 2 and 3 can be taken as (Stokoe
et al. 1985):

σm = (σv )0.5 (σh )0.5 (4)

For normally consolidated dry sand, the relationship


between σv and σh satifies:
Figure 1. Air-hammer and bender element (BE) test setup σh = K0 σv (5)
(not to scale). Prototype depths/lengths shown in round
brackets ( ).
where K0 = 1 − sin (φ  ) (Jaky 1944), φ  is the angle of
internal friction for dry sand. The ratio of horizontal
specified in both model scale and prototype scale. The to vertical stresses applied in the BE tests was 0.5 to
average shear wave velocity in a soil layer can be calcu- match the anisotropic stress state in the centrifuge with
lated by measuring the elapsed time for a shear wave to an assumed K0 = 0.5.
travel between two adjacent accelerometers (methods Wave velocities were determined at 11 levels of
to do this are presented later in this paper). confining stress, corresponding to 11 depths within
the prototype soil model, as illustrated in Figure 1.
The tested confining stresses include those evaluated
in the centrifuge to provide a basis for comparing
5 TRIAXIAL BENDER ELEMENT TEST
air-hammer and BE data.
In this study, a GDS Bender Element System (BES)
was used to conduct the element scale shear wave
6 SIGNAL PROCESSING
velocity tests. A pair of bender elements, capable
of producing separate shear and pressure waves, are
The peak to peak and cross correlation methods are
incorporated into the top cap and base pedestal of a tri-
commonly used to determine wave travel time. The
axial cell (Figure 1). The soil samples measured 50 mm
peak to peak method uses the time difference between
in diameter and 100 mm in length. The wave velocity
the first two peaks as the travel time. Whereas, the
is measured in the BES as the time for a wave to the
cross-correlation method works by performing a point-
travel the distance LTT between the source and receiver
wise multiplication to determine where two signals
elements.
are best aligned. Two correlated discrete signals are
During the generation of shear waves in the ben-
denoted as s[k] and r[k] and the cross correlation
der element (BE) tests, three components appear. Of
function Tsr [n] is defined as (Rabiner and Gold 1975):
them, two propagate at shear wave velocity and the
third travels at pressure wave velocity. The polarisa-
tion of the pressure wave is opposite to the input shear
wave. If the source is too close to the receiver, the
first arrival of the shear wave is masked by the pres-
sure wave. This is referred to as the near field effect. where the parameter n is an integer between positive
The pressure wave attenuates quicker than the shear and negative infinity.The time shift between two corre-
waves and its effect on readings can be minimised lated signals is obtained at the maximum (or minimum
by increasing the ratio between the tip-to-tip distance if the signals are negatively correlated) of the cross cor-
LTT and the wavelength λ of the faster shear waves relation function. The cross correlation method avoids
(Camacho-Tauta et al. 2012). The length of the tri- complications with picking characteristic points (i.e.
axial sample should therefore satisfy 2 < LTT /λ < 9, first peak) from discrete signals, however the data
where λ = vs T ; the period T is an input parameter length selected may affect results if more distorted data
for the GDS BES. Using values of vs obtained from is involved, as discussed below.
air-hammer tests, T for the BE tests was estimated as Typical data from air-hammer and bender element
0.05 × 10−3 s to 0.1 × 10−3 s. However, it can be seen tests are shown in Figure 2 and 3, respectively. The
from Figure 3 that the near field effect still existed data from a depth of 3.6 m was chosen here as an
in the BE tests. A high speed data acquisition system example; data from triaxial bender element tests at

139
Figure 2. Typical data from a centrifuge air-hammer test. Figure 3. Typical data from a triaxial bender element test
(depth = 3.6 m).
all depths had a similar pattern. background noise and
Table 2. Shear wave velocity from air-hammer testing using
reflected waves. Figure 2 shows that the determina- different data lengths.
tion of a first peak in each of the centrifuge MEMS
data is not straightforward. The MEMS signals in Data length Strikes Mean vs,AHA,1
the centrifuge are affected by several factors, includ-
ing a relatively low resolution of data (9 points over 1 cycle 8 293.7 m/s
one cycle, shown in Figure 2), reflecting waves from 1.5 cycles 8 281.1 m/s
boundaries or elements buried in the soil (pile, piezo- 2 cycles 8 264.8 m/s
electric accelerometers), decay of the wave signals,
and electrical and mechanical noise. These issues are
less pronounced for the first detected cycle (i.e. high-
est signal-to-noise ratio data). Despite this, the low
resolution of data still poses a challenge for imple-
menting the peak to peak method. The output of the
cross correlation method is also be affected by these
same signal quality issues. By including low signal-
to-noise ratio data (i.e. the later detected cycles), the
determined time difference between two signals can be
impacted. It is therefore important to consider remov-
ing the distorted data in order to obtain a higher degree
of correlation and hence a more accurate value for the
travel time. This data length aspect will be discussed
in the following section.
In contrast, the high resolution signals obtained Figure 4. Shear wave velocity comparison using three
from the BE tests (Figure 3) suffer to a lesser degree approaches to analyse BE data.
from these issues, and both the peak to peak and
cross correlation methods prove to be reliable. Fur- using the entire data length (normal xcorr), cross cor-
ther results from these two methods are compared in relation using the first cycle of the shear wave (first
the next section. cycle xcorr), and the first peak to peak method (peak
to peak), as shown in Figure 4. Results indicate that
the peak to peak method agrees well with the first
cycle xcorr method, but that using the full length of
7 TEST RESULTS the data set results in a lower estimate of vs as well as
an unrealistic scattered pattern of vs with depth.
7.1 Cross correlation method
To investigate the effect of data length on cross corre- 7.2 Empirical curves versus BES data
lation results, the mean value of vs,AHA,1 for the bottom
soil layer (L1 ) from 8 air-hammer tests using differ- Equations 2, 3, 4 and 5 can be manipulated to obtain
ent data lengths was calculated. Table 2 shows results expressions for vs :
and indicates that a longer data length resulted in
lower estimations of vs,AHA,1 . This outcome highlights
the importance of considering data with the lowest
possible signal-to-noise ratio.
In addition, three approaches were used to calculate
vs profile from the triaxial BE data: cross correlation

140
Table 3. Shear wave velocities from air-hammer and BE.

Soil Soil Air-hammer BE


layer depth (m) (m/s) (m/s) Difference

L3 (vs,3 ) 3.6-10.8 229.3 258.37 11.4%


L2 (v2,3 ) 10.8-18.0 304.13 310.64 2%
L1 (vs,1 ) 18.0-25.2 347.13 344.43 0.7%

Shear wave velocities of three soil layers from air-


hammer and BE tests are given in Table 3. Shear wave
velocities in Column 3 are measured from air hammer
tests. Shear wave velocities in Column 4 are calcu-
Figure 5. Shear wave velocities from BE tests. lated using Equation 10. The data indicate that the
middle (2% difference) and lower (0.7% difference)
regions show good agreement. The larger difference
of 11.4% for vs,3 in the upper region is due to sev-
eral factors: (1) the weak signal received by the upper
MEMS, as shown in Figure 2; (2) the low stress lev-
where ρ is sand density (1739 kg/m3 ), e is 0.512, g els in this shallow region, causing a relatively poor
is gravitational acceleration, and h is soil depth (see coupling between the soil and the accelerometer; (3)
Figure 1). It can be seen from Figure 5 that Equation 8 the influence of reflected pressure waves generated
agrees very well with the bender element test results, by the air-hammer at the air-hammer end; and (4) the
with differences less than 8 m/s. This indicates that low signal to noise ratio, making it hard to accurately
Equation 3 can be used to calculate the shear modulus distinguish the arrival of the shear wave.
of dry Congleton sand. The difference between Equa- Despite these issues, it can be said that the air-
tion 7 and BE data was significant, reaching more than hammer is a good method to measure the in-flight
100 m/s at 20 m depth. shear wave velocity in the centrifuge in a simple,
The trend of vs with depth can be fitted with a power cheap, and efficient way. However, a greater level of
function in the form vs = ah0.25 , where a is a fitting scepticism should be applied to air-hammer results
parameter. Figure 5 shows the result of fitting this obtained near the ground surface.
curve to the BE data (BE-curve fitting); a value of
a = 159.6 provided the best fit to the data. This curve
is used for the analysis in the following section.
8 CONCLUSIONS
7.3 BE data versus air-hammer data
An air-hammer device in the centrifuge and a ben-
The measured shear wave velocities from air-hammer der element system (BES) integrated into a triaxial
tests represent mean values across the depth between apparatus have been used to determine the shear wave
MEMS accelerometers (i.e. span lengths L1 , L2 and velocity within a dry sand. Details of experimental
L3 from Figure 1). However, the BE test provides the methods, the data processing techniques, an evalua-
shear wave velocity at a specific stress level, i.e. depth. tion of two empirical methods based on the BE data,
Therefore, in order to make a direct comparison, the and a comparison between the aim-hammer and BE
BE data spanning each of the depths (L1 , L2 , L3 ) needed shear wave velocity data have been presented. The
to be averaged. To do this, the vs = ah0.25 curve fitted main conclusions are as follows:
to the BE data was used. The elapsed time t can be
calculated by integrating dh/(ah0.25 ) from h1 to h2 with • The cross correlation method is significantly
respect to h: affected if more distorted data is included in the
analysis; the data for the first detected shear wave
cycle should be selected from the recorded signals
and used to conduct the cross correlation analysis;
• The empirical relationship for shear wave velocity
proposed by Hardin and Black (1968) fitted well to
Then the average velocity va from BE data can be the BE data for the sand and test conditions applied.
obtained using: • The air-hammer and BE data compared well, indi-
cating that the air-hammer can be used as a reliable
method for measuring shear wave velocity in the
centrifuge. However, a higher level of scepticism
should be applied to air-hammer measurements with
where h1 and h2 correspond to the limits of the regions distorted or weak signals, or those obtained near the
of L1 , L2 and L3 . ground surface.

141
REFERENCES Lauder, K. (2010). The performance of pipeline ploughs.
Ph. D. thesis, Dundee University.
Arulnathan, R. (2000). New Tool for Shear Wave Velocity Lee, C.-J., W.-Y. Hung, C.-H. Tsai, T. Chen, Y. Tu, &
Measurements in Model Tests. C.-C. Huang (2014). Shear wave velocity measurements
Brandenberg, S., S. Choi, B. Kutter, D. Wilson, & and soilpile system identifications in dynamic centrifuge
J. Santamarina (2006). A bender element system for mea- tests. Bulletin of earthquake engineering 12(2), 717–734.
suring shear wave velocities in centrifuge models. In Lee, C.-J., C.-R. Wang,Y.-C. Wei, & W.-Y. Hung (2012). Evo-
6th International Conference on Physical Modeling in lution of the shear wave velocity during shaking modeled
Geotechnics, pp. 165–170. in centrifuge shaking table tests. Bulletin of earthquake
BS (2005). Mechanical vibration Ground-borne noise and engineering 10(2), 401–420.
vibration arising from rail systems Part 1: General guid- Nugent, R. E., W. Village, & J. A. Zapfe (2012). Designing
ance. BS ISO 148, 54. Vibration-Sensitive Facilities Near Rail Lines. Sound and
Camacho-Tauta, J., J. D. JimenezAlvarez, & O. J. Reyez-Ortiz Vibration (November), 13–16.
(2012). A procedure to calibrate and perform the bender Rabiner, L. R. & B. Gold (1975). Theory and application of
element test. Dyna 79(176), 10–18. digital signal processing. Englewood Cliffs, NJ, Prentice-
Cheney, J. A., R. K. Brown, N. R. Dhat, & O. Y. Hor (1990). Hall, Inc., 1975. 777 p. 1.
Modeling free-field conditions in centrifuge models. Rollins, K. M., M. D. Evans, N. B. Diehl, & W. D. Daily
Journal of Geotechnical Engineering 116(9), 1347–1367. III (1998). Shear modulus and damping relationships for
Dobry, R. & M. Vucetic (1987). Dynamic properties and gravels. Journal of Geotechnical and Geoenvironmental
seismic response of soft clay deposits. Engineering 124(5), 396–405.
Ghosh, B ; Madabhushi, S. P. G. (2002). An efficient tool Shantz, B. R. W. J. T. D. T. (2012). Guidelines for Estimation
for measuring shear wave velocity in the centrifuge. In of Shear Wave Velocity Profiles. Technical report.
Proceedings of the international conference on physical Steedman, R. S., S. P. G. Madabhushi, & U. of Cam-
modelling in geotechnics, pp. 119–124. AA Balkema. bridge. Engineering Department (1990). Wave Propaga-
Hardin, B. O. & W. L. Black (1968).Vibration modulus of nor- tion in Sand Medium. CUED/D – soils TR. University of
mally consolidated clay. ASCE – Proceedings – Journal Cambridge, Department of Engineering.
of the Soil Mechanics and Foundations Division 94(SM2, Stokoe, K., S. Lee, & D. Knox (1985). Shear moduli mea-
Part 1), 353–369. surements under true triaxial stresses. In Advances in the
Hardin, B. O. & J. F. E. Richart (1963). Elastic wave velocities art of testing soils under cyclic conditions, pp. 166–185.
in granular soils. ASCE – Proceedings – Journal of the Soil ASCE.
Mechanics and Foundations Division 89(SM1, Part 1), Tsuno, K., W. Morimoto, K. Itoh, O. Murata, & O. Kusak-
33–65. abe (2005). Centrifugal modelling of subway-induced
Jaky, J. (1944). The Coefficient of Earth Pressure at Rest. vibration. International Journal of Physical Modelling in
Journal for Society of Hungarian Architects and Engi- Geotechnics 5(4), 15–26.
neers (October), 355–358. Yang, W., M. F. M. Hussein, A. M. Marshall, & C. Cox
Kim, N R ; Kim, D. S. (2010). Development of VS tomog- (2013). Centrifuge and numerical modelling of ground-
raphy testing system for geotechnical centrifuge experi- borne vibration from surface sources. Soil Dynamics and
ments. 7th Physical modeling in geotechnics, 7th ICPMG, Earthquake Engineering 44, 78–89.
Springman 50(1), 349–354.
Kuo, K. A. (2010). Vibration from Underground Railways:
Considering Piled Foundations and Twin Tunnels. Ph. D.
thesis, University of Cambridge.

142

Das könnte Ihnen auch gefallen