Вы находитесь на странице: 1из 4

High-frequency shear wave reflections from shallow subsoil layers using a vibrator source:

sweep cross-correlation versus deconvolution with groundforce derivative


Ranajit Ghose*
Delft University of Technology, The Netherlands

Summary To overcome some of the practical limitations of shear-


wave seismic exploration (such as the need to use large
Shear waves are important in geotechnical investigations of sources of limited frequency bandwidth), recently an
soil. We have evaluated on shallow shear-wave reflection electromagnetic vibrator source has been used (Ghose et
data the influence of removing accurately the effective al., 1996). In addition to the possibility of controlling the
source function on the separation and the amplitude of the source spectrum to accomplish a target resolution and/or
reflection events. A small vibrator can produce high penetration, with this shear-wave vibrator one can obtain a
frequencies and the generated groundforce can be estimated reasonably accurate estimate of the source motion. Because
quite accurately. Instead of conventional cross-correlation of small energy, the slipping of the source is considerably
to compress the raw vibrograms, for each sweep we have less, and this helps obtaining a good groundforce estimate.
performed deconvolution using the estimated groundforce
derivative. Our results show that deconvolution of the Since the groundforce is well-estimated for the small shear-
source response improves the separation of the reflections, wave vibrator, we have evaluated on a shallow reflection
correlating better with the actual soil layer boundaries. The section the merits of deconvolving the raw vibrograms with
side-lobe energy of the effective source wavelet is smaller an optimum filter derived from the groundforce derivative,
after deconvolution compared to cross-correlation, and this over the conventional approach of cross-correlation using
improves the separation. As the source function is removed the sweep signal or the groundforce derivative.
separately at each shot location, the true-amplitude
reflection section offers information of the actual lateral Theory
variation of shear-wave reflectivity, which is important to
derive the variation of soil strength at a subsoil boundary. Assuming all operations below in the frequency domain (so
convolution becomes multiplication), the raw vibrogram
Introduction (R) as received by the geophones can be expressed as,

Soil parameters of geotechnical interest are more directly R(ω) = W(ω)V(ω)C(ω)Q(ω) + NR(ω) + NS(ω), (1)
addressed by shear (S) waves than by compressional (P)
waves. The small strain rigidity obtained from the shear where W is the sweep specification given by the input
wave velocity (VS) is important in all dynamic loading current function, V is the vibrator response including
problems. Shear waves are also more sensitive to the subtle coupling, eigenfrequencies and harmonic distortions, C is
changes in the soil properties than the P waves. Recently earth’s impulse response or the reflection coefficient
there are attempts to employ the high-resolution shear- sequence, Q is average attenuation, and NR and NS are,
wave reflection method to geotechnical investigations (e.g. respectively, random noise and source-generated noise.
Ghose et al; 1996, 1998; Brouwer et al., 1997). Shear wave
reflection coefficients are sensitive to the change of soil- Now, the far-field displacement is proportional to the
mechanical parameters at subsoil boundaries (Ghose and groundforce, Fg (Aki and Richards, 1980), and since
Goudswaard, 2000). In addition, the laterally continuous geophones measure velocity rather than displacement, the
and high-resolution image that one obtains makes the source response (S) recorded by the geophones is
shear-wave reflection method an attractive supplement to proportional to the time derivative of Fg:
the conventional point sampling in boreholes.
S(ω) = iωFg(ω). (2)
Because shear waves are more sensitive, and they offer
much higher resolution than P-waves in soft soils (due to The shape of the source wavelet, S is governed by the
much lower velocity for the shear waves), the processing of sweep specification (W) and the vibrator response (V):
shear wave reflections needs greater care. Shear wave
reflection amplitude provides information of the soil S(ω) = W(ω)V(ω). (3)
strength distribution (Ghose and Goudswaard, 2000), hence
recording and preservation of “true” amplitude is crucial in The usual procedure to compress the raw vibrogram is to
data acquisition and processing. cross-correlate the measured record (R) with the sweep (W)
or with the source response (S). Recently, there are also
High-Frequency Shear-Wave Reflection Using a Vibrator: Cross-correlation Versus Deconvolution

suggestions to deconvolve the sweep (Robinson and where ω lies within ωmin – ωmax range. Note that after
Sagaaf, 2001; Brittle et al., 2001). We have attempted the deconvolution the effective source signature is independent
deconvolution of the estimated source response. The theory of vibrator response and the shape of the sweep. The
is briefly outlined here; an overview is given by Brouwer spectral signal-to-noise ratio is identical after cross-
and Helbig (1998). correlation and deconvolution.

After cross-correlation with the sweep (W), the correlated For the small vibrator we have estimated, for each sweep
record (Rcw) and the effective source signature (Hw) are separately, the groundforce, Fg from the measured reaction-
described respectively by the following two equations: mass acceleration, ür and baseplate acceleration, üb:

RcW(ω) = R(ω)W*(ω) Fg = ürmr + übmb, (11)

= |W(ω)|2V(ω)C(ω)Q(ω) + W*(ω)N(ω) , (4) where mr and mb are, respectively, the masses of the
reaction-mass and the baseplate. On shallow, shear-wave
and HW(ω) = |W(ω)|2V(ω)Q(ω), (5) reflection section we have evaluated the cross-correlation
against deconvolution with groundforce derivative.
where * denotes complex conjugation, and N = NR + NS.
Evaluation on shear-wave reflection section
Similarly, cross-correlation with the source response or the
groundforce derivative (S) produces the correlated record A shallow, high-frequency shear-wave reflection survey
(RcS) and the effective source signature (HS) as follows: was carried out at a site where the soil structure is known
from multiple Cone Penetration Tests (CPT). The site is
RcS(ω) = R(ω)S*(ω) covered by a sand layer of 1.5-2.0 m thickness. Fig.1 shows
CPT data at a point located exactly on our reflection line.
= |W(ω)|2|V(ω)|2C(ω)Q(ω) + S*(ω)N(ω), (6) Sand and peat-clay layers alternate at shallow depths.
Accurate mapping of the peat-clay layers is important from
and HS(ω) = |W(ω)|2|V(ω)|2Q(ω) (7) hydrological and environmental reasons. Our goal was to
resolve the distribution of the very shallow soil layers – up
Clearly, after correlation, the effective source signature still to 10-12 m depth, where GPR cannot reach due to the
contains information on vibrator response (V) and sweep presence of conductive clay layers, but a depth range which
specification (W). Thus the effective source wavelet (H) is too shallow for conventional shallow seismic reflections.
may have unwanted properties (e.g., side-lobe amplitude)
or may change considerably from sweep to sweep. The acquisition parameters were as follows: 48 single
horizontal geophones (28 Hz) fixed, and the source moving
In order to deal with this problem we use deconvolution along the line, shear-wave vibrator (cross-line orientation)
instead of cross-correlation. Since the goundforce (Fg) and sweep frequency: 50-300 Hz with end tapers, sweep length:
the source signature (S) are band-limited by nature, we 3.5 s, record length: 4.0 s, vertical stack: 1, sampling: 1
need to define an inverse filter (S’) as the band-limited kHz, geophone orientation: cross-line, source interval: 1m,
inverse of S in an optimum sense taking into consideration Sleeve Resistance, fs (MPa) Friction Ratio, Rf (%)
the usual stabilization factor (ε) to account for the zones of .0 .1 .2 .3 .4 10 8 6 4 2 0

low signal-to-noise ratio: 0 4 8 12 16 20 24 28 Soil


Cone Resistance, qc (MPa) layers
0
sand
S’(ω) = S*(ω) / (|S(ω)|2 + ε). (8) (artificial)
2 A
fs qc peat
After deconvolution with the groundforce derivative (S),
Depth (m)

&
4
the record (RdS) and the effective source signature (H’S) can Rf
clay
B
be described respectively by the following two equations: 6
sand

RdS(ω) = R(ω)S’(ω) 8
peat C
&
10 clay
= C(ω)Q(ω) + S’(ω)N(ω), (9) sand
D

and H’S(ω) = Q(ω), (10) Figure 1: Cone Penetration Test (CPT) data at a point located on the
shear-wave reflection line. Soil layers of varying shear strength can be
identified. A, B, C, D mark the depths to the main interfaces.
High-Frequency Shear-Wave Reflection Using a Vibrator: Cross-correlation Versus Deconvolution

(a) raw data (b) preprocessed data (c) preprocessed data (d) preprocessed data
sweep correlated sweep correlated Fg-derivative correlated Fg-derivative deconvolved
0 0 0 0

100 100 100 100


Time (ms)

200 200 200 200

300 300 300 300


Figure 2: (a) A typical common source gather obtained by traditional cross-correlation of the raw vibrograms, (b) same as (a) but after
preproprocessing, (c) compression via cross-correlation with groundforce derivative and identical preprocessing as in (b), (d) compression
via deconvolution with groundforce derivative and identical processing as in (b) and (c). No AGC has been applied in preprocessing.
geophone interval: 0.5 m, number of shots 36, acquisition of the source wavelet (because of successful removal of the
time: 2 hours by a crew of 3 persons. Penetration to a vibrator response and the sweep from the raw vibrograms).
greater depth was possible by increasing the vertical stack
count, increasing the sweep length, and generating more
(a) sweep: 50-300 Hz, linear
energy at the lower frequencies. However, since our target
was very shallow, high frequencies were emphasized.
Power (dB)

4
Fig.2(a) shows a typical shot gather obtained by
conventional compression of raw vibrograms by sweep 2

cross-correlation. Fig.2(b) shows the same shot gather after 0 100 200 300
amplitude-preserved preprocessing, involving trace editing Frequency (Hz)
and muting, geometrical spreading correction, spectral (b) groundforce: Fg = ür.mr + üb.mb
shaping (10-180 Hz: 100%), fk-filtering to remove the
Power (dB)

surface waves, and bandpass (15-170 Hz) filtering. Fig.2(c) 6

shows, the same shot gather, but obtained by cross- 4


correlation of the raw vibrograms with the groundforce 2
derivative and with identical preprocessing as in Fig.2(b). 0 100 200 300
Fig.2(d) shows the same shot gather, but obtained by Frequency (Hz)
deconvolution of raw vibrograms with the groundforce Figure 3: Power spectrum of (a) the sweep and (b)
derivative and identical preprocessing same as before. the groundforce for the shot gather shown in Fig.2.
Fig.3(a) shows the power spectrum of the sweep which is
Note that cross-correlation with the sweep (Fig.2(b)) results flat between 60-280 Hz, Fig.3(b) illustrates the power
in reflection events of ringing nature, thus hindering the spectrum of the estimated groundforce. It is clear that
separation between two shallow shear-wave reflections. As beyond 100 Hz the groundforce falls gradually and beyond
explained before, cross-correlation does not remove the 170-180 Hz the noise begins to dominate. Deconvolution
vibrator response and the shape of the sweep, causing a with groundforce derivative does compensate correctly for
large side-lobe of the source wavelet - responsible for the this change, but cross-correlation does not.
ringing nature. When the groundforce derivative is used for
cross-correlation (Fig.2(b)), the energy of the reflection Fig.4 illustrates the differences in stack section. A 1-D
events is enhanced, which is due to the multiplication of the velocity function with velocity linearly increasing with
vibrator response spectrum (eqns. (5) and (7)) including the depth (average Vstack = 165 m/s) has been used to stack the
resonances. When the groundforce derivative is used for data. In absence of VSP, tentative depths were inferred
deconvolution (Fig.2(c)), the reflection events are clearly from the stacking velocity, and are shown on the right side.
much better separated, which is due to a smaller side-lobe The use of high-frequency shear waves allows us to
High-Frequency Shear-Wave Reflection Using a Vibrator: Cross-correlation Versus Deconvolution

(a) Time stack: sweep cross-correlation


CPT
0 0.0

Temtative Depth (m)


40 3.0
Time (ms)

80 6.3

120 10.0

(b) Time stack: groundforce derivative cross-correlation


CPT
0 0.0

Temtative Depth (m)


40 3.0
Time (ms)

80 6.3

120 10.0

(c) Time stack: groundforce derivative deconvolution


CPT
0 0.0

Temtative Depth (m)


A
40 3.0
Time (ms)

B
80 6.3
C
120 10.0
D

0 5 10 15 20 25 27
CMP_X (m)
Figure 4: True amplitude stacked time sections; all the processing parameters are identical for the three
sections. (a) Compression by correlation with sweep, (b) compression by cross-correlation with Fg-derivative,
(c) compression by deconvolution with Fg-derivative. Asterisk is the location of the source shown in Fig.(2).
achieve wavelengths less than a meter, and hence correlation. This enables further implementation of the
resolution of a few decimeters. The 4 major soil boundaries high-resolution seismic imaging schemes, using shear
seen in the CPT data (Fig.1) can be interpreted in Fig.4(c). waves and the true reflection amplitudes.
A comparison of the three panels in Fig.(4) shows the merit
of groundforce derivative deconvolution over cross- References
correlation. The ringing nature of the reflection events in Aki K., and Richards, P.G., 1980, Quantitative Seismology.
Fig.4(a) is almost absent in Fig.4(c), thus improving Brittle, K.F., Lines, L.R., and Dey, A.K., 2001, Geophys.
separation. The reflections from the 4 major subsoil Pros., 675-686.
boundaries (A, B, C and D) can be better identified and are Brouwer, J. , Ghose, R., Helbig, K., and Nijhof, V., 1997,
stronger in Fig.4(c). Also, the lateral variation of the Proc. EEGS-ES.
reflection amplitude is more prominent in Fig.4(c), which is Brouwer J., and Helbig, K., 1998, Shallow High-Resolution
geotechnically meaningful (Ghose and Goudswaard, 2000). Reflection Seismics, Elsevier Science Ltd.
Ghose, R., Brouwer, J., and Nijhof, V., 1996, Proc. EAGE.
Conclusions Ghose, R., Nijhof, V., Brouwer, J., 1998, Proc. EEGS-ES.
High-frequency shear-wave reflections can achieve very Ghose, R, and Goudswaard, J.C.M., 2000, Proc. SEG.
high resolution in soft soil. However, for such high- Robinson, E., and Sagaaf, M, 2001, Geoph. Pros., 335-340.
resolution data, the separation between the successive
reflections is often not good, and interpretation is difficult. Acknowledgments
We have found that compression of the raw vibrograms by We thank Gemeentewerken Rotterdam to allow us do the
deconvolution using the time derivative of the estimated field experiment at their site and for the CPT data, D. Ngan
groundforce offers a better separation of shallow reflections Tillard for cooperation, and J. Brouwer for discussions.
and more reliable reflection amplitudes than cross-

Вам также может понравиться