Fabian Bonilla - ESG4 Conference @ UCSB

Analysis of borehole data
Luis Fabian Bonilla
Universite Paris-Est, IFSTTAR, France

Outline
• Advantages of borehole data
• Difficulties of working with these data
• Understanding linear and nonlinear
modeling
• Working proposition?

1. Advantages of borehole data
Garner Valley - USA
(Borehole Obs.)
Wave propagation from bedrock to surface

PGA distribution
(KiK-net)
Field data observation of soil
nonlinearity onset?
Statistical analysis with
respect to magnitude and
Vs30

Calibration of soil models
Stress computation from
deformation data
Waveform modeling

Revealing nonlinear response
after Bonilla et al. (2011)
• 2011 Tohoku earthquake data
• Predominant frequency more affected than fundamental
• Affected frequency increases as Vs30 increases

Port Island, Kobe / Kushiro Port
Loose sand => liquefaction
-Lowpass filtering
-Deamplification
Dense sand => cyclic mobility
- High frequency peaks
- Amplification
Velocity model is not always enough!

2. Difficulties of borehole data
Downgoing
wavefield
Site response (outcrop
response) is not the
same as borehole
response

Vs30 uncertainty
(lack of knowledge of
the medium)
• Variability within each
soil class is important
• This variability is even
larger at depths greater
than 30 m
• Is Vs30 enough?
• Not always core
sampling, thus no
dynamic soil parameters

Analysis of
KiK-net
boreholes
• Similar Vs30
(between 350 and
450 m/s)
• Different velocity
distribution at depth
• Different site
response
• Is Vs30 enough?
After Regnier et al. (2010)

Vs30 = 400 +/- 5 m/s
-"+(.,-/(0,1"2"'-(3455(
'"2,$(=+(>2&?#"+()&2(=+*5(@(A55BC+(D3&E&'("-(7#F(G55HI(
0
Vs (m/s)
500 1000 1500 2000 2500 3000
IWTH07 C 800
= 0.46
After Regnier et al. (2010)
20
HRSH13 C 800
= 0.49
NGSH01 C 800
= 0.65
40
NGNH03 C 800
= 0.66
NARH05 C = 0.69
800
YMGH02 C 800
= 0.71
60
HRSH18 C 800
= 0.78
KMMH09 C 800
= 0.82
Depth(m)
80
100
SZOH31 C = 0.86
800
KKWH07 C 800
= 0.90
HRSH01 C 800
= 0.95
AICH22 C 800
= 1.11
YMGH13 C 800
= 1.20
120
KMMH13 C 800
= 1.22
YMTH10 C 800
= 1.28
(
140
160
IKRH01 C 800
= 1.36
FKIH06 C 800
= 1.37
MYGH02 C 800
= 1.49
generic profil
No comments!
The data speak alone

3. We need to know well the linear response
(example of the CORSSA array, Greece)
1. H/V spectral ratio (noise data)
2. H/V spectral ratio (earthquake data)
3. Standard spectral ratio (borehole response)
4. Borehole response inversion (velocity, thickness,
and Q profiles)

8 -- 800.0*
Inverting for nonlinear soil properties
In the first and second layers, the values in parentheses are identified from the part just after the main parts of the mainshock records, and the values
without parentheses are identified from the main parts. Asterisks indicate S-wave velocities based on the logging results, i.e., not identified here.
170 F. De Martin, H. Kawase, and A. Modaressi-Farahmand Raz
in the main part and the part just after the main part of the
strong motion, vertically propagating S waves are dominant
(a)
Sand and gravel (b)
Clay
in the period range from 0.1 to 2.0 sec, while in the later
Identified
Equivalent linear part, horizontally 1 propagating waves 11 are 3, 5, 9dominant in the pe- 1
10 4
1 2
~t the main part of the foreshock record
A riod range longer than 0.7 sec, and vertically propagating S
0.8
0.8
8
the main part of the mainshock record O waves are still dominant in the shorter period range. For the
just after the main part of the mainshock record m
weak motion, 0.6 only the main part can be examined because 0.6
of the small signal-to-noise ratios in other 7 parts. In the main
1.0
20
0.4
0.4
part of the weak motion, vertically propagating S waves are
0.8
16
6
also dominant. 0.2 Inversion - sand
0.2 Inversion - clay
..q 0.6 I lllllll IIIIIIII
12
Inversion - gravel
Vucetic and
0.4 IIIlllll IIIitll
~7
Based on these Seed and results, Idriss (1970b) we decided to analyze these three Dorby (1991)
8 > S-wave dominant 0 time segments, that is, the main part of the 0
0.2
IIIIIIll LtlI_'. -
10 -6 10 -5 10
4
strong motion, the part just after -4 10 -3 10 -2 10 -6 10 -5 10
the main part of the strong
-4 10 -3 10 -2
γ
γ
'--FI-fl-flff- I IIIIIll
0.0 motion, and the main part of the weak motions, based on 1D
10-4 10- 3 10- 2 10- 1 100 0
Nonlinear Soil Behavior in Kashiwazaki-Kariwa Nuclear Power Plant during 2007 Earthquake 773
Figure '-n
o
(a) First layer
>.
24. wave (a) Comparison propagation oftheory strain-dependent for vertically De characteristics propagating Martin inverted S et waves. al. for sand (2010) and gravel layers with laboratory test results for s
1.0
(Seed and Idriss,
20 ,q
The 1970b) observed whose spectral effective ratio vertical between stress KD2 rangesand fromKS2 0.05for to 0.5 the MPa. (b) Comparison of the strain-dependent characteris
inverted for clay (a)
< 0.8
o main
layers
part
with
of the
laboratory Surface
strong
test
motion
results Layer
shows
for clay
a longer
(Vucetic
peak
and Dobry, (b)
period
1991) whose plasticity Intermediate index Layer (PI) ranges from 0% to 200
16 For both panel, numbers indicate layers of Table 3.
0,6 t ll,,Jl,,I ,,,lll,Jt I
with lower 1.4 amplitude at the peak around 0.5 sec compared 1.4
12
Empirical
to the corresponding peak for the weak motions. The shift
Relationship
Earth Science and Disaster 1.2 Prevention (NIED). This work was performed based 1.2 on Lotung downhole ground motion data, in Proc. of
041j r
8
120 Empirical
while one of the
02 I IIIIIJl lJ l-- llrllll I
of authors the peak (F. period D. M.) was can abe visiting clearly scholar seen at in the the Relationship Disaster Fourier spectra 2nd Int. Conf. on Recent Advances in Geotechnical Earthqu
140
4 Prevention Research of the Institute, KS2 1.0 records. Kyoto, onTo leave quantify from Bureau this deperiod Recherches and amplitude Engineering 1.0 and Soil Dynamics, St. Louis, Missouri, 11–15 Ma
o.of-FFIll-Itff-Tlflllltl I IIIIIIII I IIIIIIII Géologiques et Minières, Orléans. This work was supported by the CAR- 1991, 111–118.
8070
0
shift, the S-wave velocities and the damping factors are iden-
60
60
10 -4 10-3 10- 2 10-1 100 NOT Institute and the 21st 0.8 Century COE Program of Kyushu 100 University Chávez-García, 0.8 F. J., and D. Raptakis (2008). Inversion
50
of soil structure
SHEAR STRAIN (%)
(H-14). Drs. S. tified Parolai, by T. minimizing Satoh, and an anonymous the residual reviewer between madethe valuable observed spec- analysis of the seismic wavefield from a vertical array, in Proc. of
80 50
(b) Second layer
suggestions totral improve ratio this 0.6and article. the theoretical amplification 90 70 factor calculated 14th 0.6 World Conf. on Earthquake Engineering, Beijing, 40 China, 12
140
from the 1D wave propagation theory. The S-wave velocity October 2008.
0.4
Figure 15. Relationships between the effective
and the damping References
Chin, B.-H., 0.4
Fujisawa factor Sand
and K. Aki (1991). Simultaneous study of the source, path,
Disturbed Samples
in the surface alluvial layer identified
40
site effects on strong ground motion during the 1989 Loma Pr
shear strain and the shear modulus reduction ratio or for the main 0.2 part Undisturbed of the strong motion are about 10% smaller 0.2
the Pioneering damping factor for (a) work the first layer by and T. (b) the Aguirre, J., and K. Irikura (1997). Samples Nonlinearity, liquefaction, and velocity earthquake: A preliminary result on pervasive nonlinear site effe
and 50% greater, respectively, than those : void ratioidentified for the
second layer are estimated by two methods. Solid variation of soft soil layers in Port Island, Kobe, during the Hyogo-ken Bull. Seismol. Soc. Am. 81, 1859–1884.
0
0
symbols Satoh shows the since values based the on linear 90’s 1D theory Nanbu earthquake, main part Bull. 10 of -8 Seismol. the 10 -7 weak Soc. 10 motions. Am. -6 87, 101244–1258.
-5 The 10relationships -4 -3 10between
-2 Cramer, C., 10 -8 and 10 C. -7 Real 10 (1992). -6 A10 -5 statistical 10 -4 analysis 10 -3 of submitted 10 -2 s
with the S-wave velocities and the damping factors Aki, K. (1993). the Local effective site effects shear onstrain weak and strong the shear groundmodulus motion, reduction effects predictions for the weak-motion blind prediction test condu
Strain
Strain
estimated by the identification method. Open symbols Tectonophysics 218, 93–111.
at the Turkey Flat, USA, site effects test area near Parkfield, Califor
ratios or damping factors estimated by the identification
show the values based on the equivalent linear 1D Aki, K. (2003). A perspective on the history of strong motion seismology, in Proc. of the Int. Symp. on the Effects of Surface Geology on Seis
method agrees with the laboratory test results. We corrobtheory
in which the S-wave velocities and the damp- Phys. Earth Planet. In. 137, 5–11. (c) Mogi et Base al. Motion, Layer (2010)
Odawara City, Japan, Volume 2, 15–20.
ing factor for the main part of the foreshock are used Archuleta, R. J., orate S. H. that Seale, the P. main V. Sangas, part L. of M. the Baker, strong andmotion, S. T. Swain whose Cultrera, maxi-
1.4
G. C., D. M. Boore, W. B. Joyner, and C. M. Dietel (19
as initial values for iteration. The shear modulus re- (1992). Garner mum Valley acceleration downholeat array the ofsurface accelerometers: station Instrumentation
and and preliminary whose data duration analysis, is Bull. 3 sec, Seismol. has Soc. the 1.2 potential Am. 82, 1592– of making the
KS2 is 220 cm/sec Nonlinear 2 soil response in the vicinity of the Van Norman Comp
duction
• Use of vertical arrays
ratio for the foreshock are assumed to be unity
following the 1994 Northridge, California, earthquake, Bull. Seism
and the damping factor to be 3.4% at the effective 1621.
Soc. Am. 89, 1214–1231.
surface soil nonlinear at an effective shear strain on the order
50
shear strain of 10-4%. Solid and dashed curves rep- Archuleta, R. J., S. H. Seale, P. V. Sangas, L. M. Baker, and 1.0 S. T. Swain Darragh, R. B., and A. 40F. Shakal (1991). The site response of two rock
• Inversion of G/Gmax
resent the shear modulus reduction ratios and the (1993). Garner of 0.1%. only
7060
valley downhole array of accelerometers: Instrumentation
and preliminary The S-wave data analysis, velocities Bull. Seismol. in the Soc. alluvial 0.8Am. 83, layers 2039. identified Am. 81, 1885–1899.
soil station pairs to strong and weak ground motion, Bull. Seismol. S
damping factors as a function of the effective shear
strain given by JESG (1991) from laboratory tests. Asano, K., and from T. Iwata the (2006). part just Source after process the main and near-source part of the ground strong motion Field, E. H., P. A. Johnson, I. A. Beresnev, and Y. Zeng (1997). Nonlin
motions of the 2005 West Off Fukuoka Prefecture earthquake, 0.6 Earth ground-motion amplification by sediments during the 1994 Northri
G/G max
G/G max
PI = 0%
PI = 200%

Inverting for nonlinear soil properties
D. Assimaki et al. Pure Appl. Geophys.
A Wavelet-based Seismogram Inversion Algorithm for the In Situ Characterization of Nonlinear Soil Behavior
Figure 18
Best fit nonlinear cyclic soil response predicted at convergence of the proposed inversion using the six-parameter model at the midpoint
between consecutive receivers of the DHB downhole array in Lotung (LSST): 0–6, 6–11, and 11–17 m. Comparison of results with previously
Assimaki et al. (2010)
Figure 17
rray instrumentation of Lotung LSST experiment site (modified from ANDERSON, 1993), b shear wave velocity profile (V s ) at
adopted from ELGAMAL et al. (2001). Density used in the nonlinear soil response inversion was q = 2 tn/m 3 and attenuation
factor Q = 10
motions ranging from medium to high intensity were
shown to converge to consistent nonlinear properties.
The target best fit model was defined as the average
modulus reduction and damping curves across the
Table 5 ensemble of ground motions used. The inversion was
next employed on field data at a downhole array site,
corded events at the LSST downhole array used in the inversion of seismogram and results were recordings shown to (from compare GLASER, very1995)
well with
published data on generic soil conditions. Results
ate Magnitude (Ml) Epicenter from Focal thisdepth work (km) advance thePeak stateAcceleration of the art in in[gals]
situ
strong motion site response inversion, where the soil
Longitude (E) Latitude (N)
is for the most part approximated
E–W
by an
N–S
equivalent
V
obtained inversion data at the same site (ZEGHAL et al., 1996; ELGAMAL, 1995) and nonlinear dynamic soil properties proposed by EPRI (1993)
for the corresponding profile depths
Inverting for G/Gmax and damping ratio
nonlinear site response during future events. The
authors are currently extending the formulation to
multilayered 1D profiles with a downhole and surface
receiver, and in the future will be working to implement
the proposed inversion for 3D near-surface soil
formations and non-destructively extract inelastic soil
characteristics from active surface vibration sources
(i.e., Spectral Analysis of Surface Waves).
Acknowledgments

25.4 5634 0.56 30
10 72 3.51 8
21 180 2.32 18
For each case study, the soil response and the soil-structure interaction are studied. The soil
response of the three different soil profiles are presented in §5.3. The soil-structure interaction is
An
studied
insight §5.4. Different indexes
of
and
nonlinear
control points are defined to simplify
soil
the way
response
the results
are displayed. Figure 5.8 shows these indexes and the control points in the model.
able 5.1: Properties of the three selected buildings ts
perties
ff
P o.A
sb
a
a/2
P o.B
P o.C
a
Foundation length
files are considered to represent different level of nonlinearities. 110 The Chapter soil 5. Effect of soil nonlinearity
different layers. The geometry and elastic properties of these soil profiles
Figure 5.8: Definition of the control points of the soil-structure models
.
g/m 3
m/s
10m
Soil-structure interaction model
ff represents the free field, ts the top of the structure, out the outcropping and sb the base
of the soil profile. The control points P o.A, P o.B and P o.C represents the points where we
ρ
will study the 1 = 1930kg/m
soil and structure 3
!5
ρ
responses. The parameter 1 = 1930kg/m
a is the 3
length of the foundation.
V s1 = 220m/s
10m
1
2
3
0
V s1 = !10 185m/s
out
Gandomzadeh (2011)
Soil profile 1
Soil profile 2
Soil profile 3
!15
g/m 3
m/s
20m
ρ 2 = 1980kg/m 3
V s2 = 400m/s
20m
ρ 2 = 1940kg/m !20
3
Depth (m)
V s2 =
!25
335m/s
!30
Depth (m)
g/m 3
m/s
20m
ρ 3 = 2040kg/m 3
V s3 = 550m/s
20m
ρ 3 = 2000kg/m !35 3
V s3 = !40 460m/s
!45
g/m 3
m/s
ρ = 2100kg/m 3
V s = 800m/s
ρ = 2100kg/m !50
3
0 200 400 600 800 1000
V s = 800m/s
Shear Modulus (MPa)
1)
(b) Soil profile (#2)
(c) Soil profile (a) (#3) Low-strain shear moduli of the profiles
(b)
Confining pressure dependency
ometric and material properties of the three selected soilFigure profiles 5.3: Low-strain shear moduli and elastic shear

shift of the fundamental frequency of the soil-structureshear system components. when soilFor example, Figure 5.32 displays the iso-values of the cumulative dissipated
is acting is quite low. Numerical simulations show this is not energy a good in the index soil, tofor three soil-structure systems (Case studies 2, 5 and 8) at the end of the
computation. The dissipated energy presented here is only due to the shear terms. These
structures are based on the second soil profile, and the input motion of 0.25g outcropping PGA
is applied at the base of the model after its division by two.
.2.5 Case studies
hree different structures and three different soil profiles are considered. Consequently, we have
from nonlinear
An
behavior.
insight
However, the amplitude
of
may be
nonlinear
more discriminant.
soil response
terial behavior is assumed linear, consequently, the aforementioned remark
different case studies. For each case, one linear and four nonlinear analyses (with the input
otions presented in §5.2.3) are performed. Therefore, our parametric study consists of 45
ifferent analyses. Table 5.5 displays the properties of each case study.
ering For eacha case nonlinear study, the soil structure.
response and the soil-structure interaction are studied. The soil
esponse of the three different soil profiles are presented in §5.3. The soil-structure interaction is
tudied in §5.4. Different indexes and control points are defined to simplify the way the results
re displayed. Figure 5.8 shows these indexes and the control points in the model.
dissipation in the soil and maximum strain during
out
pagation
ts
dissipation
ff
P o.A
a
P o.B
a/2
P o.C
a
2
y the role of the different energy dissipation mechanisms in the problem and
Foundation length
f soil nonlinearity on SSI, the dissipated3
energy due to the hysteresis behavior
sb
1
computed during the wave propagation. For the soil, an energy dissipation
0.
2.5
uted by
.5
3.
136 Chapter 5. Effect of soil nonlinearity on dynamic
1.
SSI : Parametric3.5
study
Figure 5.8: Definition of the control points of the soil-structure models
I soil = 1 ∫ ∫
1.5
4.
σ(x,t):dɛ(x,t)dV (5.3)
2.
4.5
Ω
2.5
5.
bottom boundary of the soil layers and also close to the surface was described before in §5.5.1.
(d) Legend: (J/m 3 )
We observe that at these areas the maximum shear strain is in accordance with the dissipated
energy Figureand 5.32: is larger Cumulative than other dissipated parts. energy at the end of the wave propagation due to the shear
components of the stress 0.2g and strains for case studies 2, 5 and 8 obtained 0.7gfor
the 0.25g outcropping
PGA
ff represents the free field, ts the top of the structure, out the outcropping and sb the base
f the soil profile. The control points
Ω
P o.A,
t
P o.B and P o.C represents the points where we
ill study the soil and structure responses. The parameter a is the length of the foundation. (c) Dissipated energy (b03 and soil profile #2)
Dissipated energy is
higher at interfaces
and close to the free
surface
(a) Dissipated energy (b01 and soil profile #2) (b) Dissipated energy (b02 and soil profile #2)
0
!5
!5
!5
!5
These graphs help us to find the parts of the soil mediumPo.C
that dissipate more energy during
the wave propagation. The parts that are white in this figure dissipate more than 5J/m 3
!10
!10
!10
!10
energy. We observe that the soil energy is strongly dissipated at the bottom boundary of each
!15
!15
!15
!15
layer and also close to the surface where the up and downgoing waves are combined. Firstly,
!20
!20
!20
!20
the shear modulus reduction curve is related to the confinement stress (equation 5.1), therefore
the!25soil becomes softer by!25
getting closer to the surface. !25 Secondly, the properties !25 of each layer is
weaker than the one below it. Consequently, more energy dissipation is observed at the bottom
!30
!30
!30
!30
boundary of each soil layer. Thirdly, the impedance contrast is higher at the boundary between
layers, !35 thus amplification!35
is also produced, which!35competes with nonlinear !35 effects.
The effect of the structure weight on the low-strain shear modulus is taken into account (see
!40
!40
!40
!40
1.6.4). This may change the nonlinear behavior of the soil particularly around the foundation.
free field
free field
free field
!45
Po.A
!45
Po.A
!45
!45
Po.A
In this work, we neglected the effect of initial static condition. This effect changes the behavior
Po.B
Po.B
Po.B
Po.C
Po.C
Po.C
of !50 the soil before applying!50
the dynamic loading. In !50 addition, the weight !50 of the structure and soil
0 0.2 0.4
I
produces soil
(J/m
the 3 0 2 4 6
0 50 100
)
settlement in theShear media strain (%) that x 10 we did not study in our work. Therefore, the effect
!3
I soil
(J/m 3 0 0.02 0.04 0.06
)
Shear strain (%)
Depth
Depth
0
Depth
0
free field
Po.A
Po.B
Depth
0
Gandomzadeh (2011)
(a) Dissipated energy
(b) Maximum shear strain
(c) Dissipated energy
(d) Maximum shear strain
Figure 5.36: Maximum shear strain reached during the calculation and the dissipated energy due
to the stress and strain shear components for the case study 2 with two 0.1g (two left figures

What do we observe?
• Energy is strongly dissipated at the bottom of each
layer and close to the free surface
• Since shear strength increases with depth, the
energy is dissipated in the weaker part (transition
between layers)
• Furthermore, the impedance contrast increases at
each layer interface
• Thus, nonlinear response has a cumulative effect
(number of cycles) and competition between
impedance contrast (linear part) and material
strength (nonlinear part)
• It is therefore necessary to instrument not only the
middle of the layers but near their interfaces

Conclusions
Sources of uncertainty (variability) in site response
• Input ground motion (e.g. near- and far-field)
• Low strain properties (linear site response)
• Dynamic soil properties (nonlinear site response)
• Methods of computing site response
What do we need?
• Understanding linear site response
• Inverting earthquake data to obtain dynamical soil
properties (up to bedrock?)
• Core sampling and laboratory tests (material
strength, granulometry, pore pressure effects, etc.)
• Instrumenting middle of layers and near their
interfaces