Sei sulla pagina 1di 242

Geotechnical Characterisation of Compacted

Ground by Passive Ambient Vibration


Techniques

Pavlick Harutoonian

A thesis submitted in fulfilment for the degree of Doctor of Philosophy

Supervisors
Associate Professor C.J. Leo
Dr J.J. Zou
Dr A. Rahman
Dr D.S. Liyanapathirana
School of Computing, Engineering and Mathematics
University of Western Sydney
December 2012

Abstract

Abstract
Ground improvement works are commonly required to overcome poor underlying soils in
conjunction with infrastructure and housing development. An extensively employed and
popular improvement technique is to impart mechanical compaction to the ground in an
effort to achieve adequate strength and favourable load-deformation behaviour (stiffness)
for the construction of civil infrastructure, including buildings and roads. In order to assess
whether a sufficient level of compaction is achieved to meet future design requirements,
the means to confirm the quality of compaction and to acquire an innate knowledge of the
relationships among void ratio, matric suction, moisture content and unsaturated shear
strength will be imperative. There are a number of methods already available for measuring
soil compaction, namely and generally, the invasive and non-invasive methods. All methods
have their inherent strengths and limitations, and there will always be a trade-off in
choosing one method over another. It has been particularly difficult, however, to find a
cost-effective and time efficient method that can be applied to a deep and extensive
compacted site.
In this thesis, a non-classical characterisation method, the horizontal to vertical spectral
ratio (HVSR) technique is proposed for the purpose of verifying the standard of
compaction at a site. This involves a cable-less low cost, efficient and non-invasive passive
ambient vibration (or microtremor) based method, which will be calibrated against in situ
test data. The simplicity of the HVSR technique is manifested in having a short setup time
(relative to other methods), and precluding the need for any active excitation sources. In
the study a lightweight battery operated sensor was utilised to capture the prevailing
ambient vibrations reflecting the structure in the ground.
This thesis is centred on the following phases of research: (i) refinement and extension of
current ambient vibration HVSR techniques, (ii) ambient vibration measurements of the
compacted fills at Penrith Lakes quarry in Western Sydney, (iii) interpretation, validation,
calibration and application of the measured data to characterise the compacted layers and
the underlying soil profile, and (iv) advancing shear wave velocity (Vs) as an
alternative/complimentary measure of evaluating ground compaction. The work in this
thesis first involved conducting a pilot study of a general geotechnical site investigation to
test the proposed techniques and theories. This pilot study was located at the Penrith
- ii -

Abstract
(Kingswood) campus of the University of Western Sydney (UWS). Once the proposed
techniques proved successful, they were applied to compacted fills, in particular by both
rolling compaction and dynamic compaction methods, at the Penrith Lakes quarry in
Western Sydney. The geotechnical site characterisation and the compaction assessment
involved, in general: (1) interpreting the measured HVSR curves for a preliminary
assessment of the soil layering, and (2) fitting the HVSR curves to a theoretical model to
estimate the Vs profile of the ground.
The results from the HVSR technique were verified against data from classical invasive
methods (e.g. borehole data, CPT, DMT and SPT). Further verification was also made
against the results from other surface wave techniques (e.g. MASW and MSOR
(multichannel simulation using one receiver)). The study suggests that HVSR-based
techniques could be used for characterising a site in combination with a reduced number of
mechanical in situ tests, and especially to fill in the gaps in the soil stratigraphy at the
locations not covered by the mechanical tests. The work in this thesis will facilitate the
development of an ambient vibration technique for assessing soil compaction based on
both raw data and Vs that is currently not available for practicing engineers.

- iii -

Preface

Preface
This thesis is submitted in fulfilment of the requirements for the degree of Doctor of
Philosophy (PhD) at the University of Western Sydney (UWS), Australia. The work
described herein was performed by the candidate in the School of Computing, Engineering
and Mathematics at UWS. The candidate was supervised by Associate Professor Chin Jian
Leo, Dr Jeffrey Zou, Dr Samanthika Liyanapathirana and Dr Ataur Rahman.
This thesis has been supported by publications that have been accepted or published in
peer-reviewed local and international journals and conferences. The list of these
publications is provided below:

Journal Publications
1. Australian Geomechanics Journal
Harutoonian, P., Leo, C.J., Liyanapathirana, D.S. & Wong, H. 2012, 'Site characterisation
by the HVSR technique', Australian Geomechanics Journal, vol. 47, no. 3, pp. 103 - 112.
2. Soil Dynamics and Earthquake Engineering
Harutoonian, P., Leo, C.J., Doanh, T., Castellaro, S., Zou, J.J., Liyanapathirana, D.S.,
Wong, H. & Tokeshi, K. 2012, 'Microtremor measurements of rolling compacted ground',
Soil Dynamics and Earthquake Engineering, vol. 41, pp. 23-31.
3. Journal of Geotechnical and Geoenvironmental Engineering
Harutoonian, P., Leo, C.J., Tokeshi, K., Doanh, T., Castellaro, S., Zou, J.J.,
Liyanapathirana, D.S. & Wong, H. 2012, 'Investigation of composite compacted ground
using microtremors', Journal of Geotechnical and Geoenvironmental Engineering. Ahead of Print DOI: 10.1061/(ASCE)GT.1943-5606.0000881.
4. Geotechnique Letters
Harutoonian, P., Leo, C.J., Castellaro, S., Zou, J.J. & Liyanapathirana, D.S. 2013,
'Compaction evolution observed via the HVSR of microtremors', Geotechnique Letters, vol. 3,
pp. 1 - 4.
- iv -

Preface
5. Soil Dynamics and Earthquake Engineering
Harutoonian, P., Leo, C.J., Tokeshi, K., Doanh, T., Castellaro, S., Zou, J.J.,
Liyanapathirana, D.S. & Wong, H. 2013, 'Investigation of dynamically compacted ground
by HVSR-based approach', Soil Dynamics and Earthquake Engineering, vol. 46, pp. 20 - 29.

Conference Publications
1. ANZ2012
Harutoonian, P., Leo, C.J., Liyanapathirana, S., Golaszewski, R. & Moyle, R. 2012,
'Geotechnical characterisation of compacted ground: Interpretation of the HVSR curve',
ANZ2012, eds G. Narsilio, A. Arulrajah & J. Kodikara, Melbourne, Australia, pp. 77 - 82.
2. ANZ2012
Harutoonian, P., Leo, C.J., Liyanapathirana, S., Golaszewski, R. & Moyle, R. 2012,
'Geotechnical characterisation of compacted ground: Forward modelling of the HVSR
curve', ANZ2012, eds G. Narsilio, A. Arulrajah & J. Kodikara, Melbourne, Australia, pp.
1208 - 1213.
3. ICGI 2012
Harutoonian, P., Leo, C.J., Zou, J.J., Liyanapathirana, S., Golaszewski, R., Moyle, R. &
Tokeshi, K. 2012, 'Site compaction assessment: Comparison of surface wave methods',
ICGI 2012, eds B. Indraratna, C. Rujikiatkamjorn & J.S. Vinod, Wollongong, Australia, pp.
1055 - 1061.
4. AVH8
Harutoonian, P., Tokeshi, K., Leo, C.J. & Liyanapathirana, S. 2012 Use of surface waves
for geotechnical engineering applications in Western Sydney. 8th Alexander von Humbolt
International Conference, AvH8-87, Cusco-Peru, 12-16 November 2012.
5. ACMSM22
Harutoonian, P., Leo, C.J., Liyanapathirana, D.S. & Tokeshi, K. 2013, 'A geotechnical site
investigation by surface waves', ACMSM22, eds Samali, Attard & Song, Taylor and Francis
Group, Sydney, Australia, pp. 619 - 624.

-v-

Preface
6. AVH8
Tokeshi, K., Harutoonian, P. Leo, C.J. & Liyanapathirana, D.S. 2012. HVSR inversion
using synthetic curves from one source and enhanced by Rayleigh dispersion curve from
MASW 8th Alexander von Humbolt International Conference, AvH8-45, Cusco-Peru, 1216 November 2012.
7. ACMSM22
Tokeshi, K., Harutoonian, P., Leo, C.J., Liyanapathirana, D.S. & Golaszewski, R. 2013,
'Horizontal to vertical spectral ratio inversion using Monte Carlo approach and enhanced
by Rayleigh wave dispersion curve', ACMSM22, eds Samali, Attard & Song, Taylor and
Francis Group, Sydney, Australia, pp. 641 - 646.
8. GeoFlorida
Harutoonian, P., Chapman, B., Leo, C.J. & Liyanapathirana, S. 2010, 'Characterisation of an
urban site by ambient noise HVSR method: resonance frequencies and site amplifications',
GeoFlorida, GSP 199, eds D. Fratta, A.J. Puppala & B. Muhunthan, ASCE, West Palm
Beach, Florida, pp. 1152-1161.
9. GeoShanghai
Harutoonian, P., Chapman, B., Young, C.N., Leo, C.J. & Zou, J.J. 2010, 'Near surface soil
characterisation by passive ambient noise HVSR method', GeoShanghai, GSP 201, eds M.
Huang, X. Yu & Y. Huang, ASCE, Shanghai, China, pp. 288-293.
10. UWS Civionics Research Centre Inaugural Annual Conference
Harutoonian, P., Leo, C.J., Liyanapathirana, S., Zou, J.J. & Golaszewski, R. 2010,
'Compaction assessment by the passive ambient noise HVSR technique', University of
Western Sydney's Civionics Research Centre Inaugural Annual Conference, eds B. Uy & Z. Tao,
Sydney, Australia, pp. 50 - 57.
11. UWS Civionics Research Centre Inaugural Annual Conference
Young, C.N., Zou, J.J., Harutoonian, P. & Leo, C.J. 2010, 'Application of vine creeping
optimisation to ambient noise analysis', University of Western Sydney's Civionics Research Centre
Inaugural Annual Conference, eds B. Uy & Z. Tao, Sydney, Australia, pp. 129 - 137.

- vi -

Acknowledgements

Acknowledgments
First and foremost, I would like to take this opportunity to express my sincerest gratitude
and thank my principal supervisor, Associate Professor Chin Jian Leo for giving me the
opportunity to conduct my research under his supervision, and for his encouragement,
guidance and helpful advice throughout the period of my research. Without his patience
and assistance I would not have been able to achieve the work presented in this thesis.
Secondly, I would like to express my sincere thanks to my co-supervisors Dr Jeffrey Zou,
Dr Samanthika Liyanapathirana and Dr Ataur Rahman, as well as Dr Ken Tokeshi
(University of Western Sydney), Dr Silvia Castellaro (Universit di Bologna, Italy), Dr
Thiep Doanh (Ecole Nationale des Travaux Publics de lEtat, France) and Professor Henry
Wong (Ecole Nationale des Travaux Publics de lEtat, France) for sharing with me their
valuable experience, discussions, time and helpful guidance.
I would also like to thank the University of Western Sydney for providing me a higher
degree research PhD scholarship, and also the Australian Research Council (ARC), Penrith
Lakes Development Corporation (PLDC) and Coffey Geotechnics (Coffey) for their
generous support in this study, through the ARC linkage grant LP0989534. In addition,
special thanks to Robert Golaszewski (PLDC), Drew Bilbe (PLDC) and Michael Hughes
(Coffey) for their continuous assistance throughout the study.
Last but not least, I would also like to thank my family, fianc Julie (thanks for drawing
Figures 1.2 and 1.3 for me) and friends for their patience and constant help throughout my
university studies. Without their support it would not have been possible to achieve my
goal of completing my university studies.

- vii -

Declaration

Declaration
To the best of my knowledge and belief, the work submitted in this thesis is original unless
otherwise acknowledged in the text. The material submitted in the thesis as a whole has not
been submitted for a degree in this or any other university.
I also appreciatively acknowledge the supervision from Associate Professor Chin Jian Leo,
Dr Jeffrey Zou, Dr Samanthika Liyanapathirana and Dr Ataur Rahman in the research and
preparation of this thesis.

Pavlick Harutoonian

Date

- viii -

Table of Contents

Table of Contents
Abstract ............................................................................................................................................... ii
Preface ................................................................................................................................................ iv
Acknowledgments ........................................................................................................................... vii
Declaration ...................................................................................................................................... viii
Table of Contents ............................................................................................................................. ix
List of Figures ................................................................................................................................. xiii
List of Tables ................................................................................................................................... xix
Abbreviations ................................................................................................................................... xx
Chapter One: Introduction .......................................................................................... 1
1.1

Background ........................................................................................................................ 1

1.2

Impetus for Research ........................................................................................................ 6

1.3

Objectives of Research ..................................................................................................... 8

1.4

Organisation of Thesis...................................................................................................... 9

Chapter Two: Literature Review ............................................................................... 13


2.1

Dry Density and Moisture Content of Compacted Ground .................................... 14

2.1.1

Measurement of Dry Density (d) and Moisture Content................................. 20

2.2

Resistance of Compacted Ground ................................................................................ 23

2.3

Shear Wave Velocity (Vs) of Compacted Ground...................................................... 27

2.3.1
2.4

Measurement of Shear Wave Velocity (Vs) by Geophysical Methods ........... 32

Horizontal to Vertical Spectral Ratio (HVSR) Technique ........................................ 33

2.4.1

Survey of Methods Used for the Theoretical Modelling of HVSR Curves ... 36

Chapter Three: Methodology .................................................................................... 45


3.1

Measurement of Ambient Vibrations ........................................................................... 47

3.2

Theoretical Modelling of the HVSR Curve ................................................................. 50

3.3

HVSRplus Technique ..................................................................................................... 58

- ix -

Table of Contents
3.3.1

MSOR Technique ................................................................................................... 60

3.4

Normalisation for Confining Pressure ......................................................................... 61

3.5

Resonance Frequencies and Higher Modes................................................................. 64

Chapter Four: Parametric Studies and Analysis of HVSR Curves ............................ 68


4.1

Sensitivity Analysis of the Theoretical HVSR Model ................................................ 68

4.2

Parametric Study of Theoretical HVSR Curves .......................................................... 75

Chapter Five: Trial Geotechnical Site Investigation Using the HVSR Technique .. 83
5.1

Test Site............................................................................................................................. 83

5.2

Measurement Reliability ................................................................................................. 86

5.3

Interpretation of Measured HVSR Curves .................................................................. 88

5.4

Forward Modelling to Infer the Vs Profile .................................................................. 89

5.5

Application of the HVSRplus Technique .................................................................... 96

5.6

Summary ......................................................................................................................... 101

Chapter Six: Investigation of Rolling Compaction Using the HVSR Technique ...102
6.1

Penrith Lakes Quarry .................................................................................................... 103

6.2

Compaction Evolution Observed via the HVSR of Microtremors ....................... 105

6.2.1

Interpretation of Measured HVSR Curves ....................................................... 105

6.2.2

Forward Modelling to Infer the Vs Profiles ..................................................... 106

6.3

Stability of HVSR Curves ............................................................................................. 108

6.4

McCarthys Pit (MCP) Area ......................................................................................... 110

6.4.1

Interpretation of Measured HVSR Curves in MCP Area ............................... 111

6.4.2

Forward Modelling to Infer the Vs Profile in the MCP Area ........................ 112

6.5

Wildlife Lake West (WLW) Area ................................................................................ 116

6.5.1

Interpretation of Measured HVSR Curves in WLW Area ............................. 116

6.5.2

Forward Modelling to Infer the Vs Profile in the WLW Area....................... 117

6.6

Assessment of Compaction Quality ........................................................................... 119

6.7

Summary ......................................................................................................................... 122

-x-

Table of Contents
Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction
Using the HVSR Technique .....................................................................................124
7.1

Dynamic Compaction Prototype Trial Area 9 (DCPT-9) ....................................... 125

7.2

Stability of HVSR Curves ............................................................................................. 127

7.3

Interpretation of Measured HVSR Curves in DCPT-9 Area.................................. 133

7.4

Forward Modelling to Infer the Vs Profile in DCPT-9 Area.................................. 135

7.5

Assessment of Compaction Quality ........................................................................... 141

7.6

Summary ......................................................................................................................... 144

Chapter Eight: Investigation of Dynamic Compaction Using the HVSRplus


Technique .................................................................................................................146
8.1

Dynamic Compaction Prototype Trial Area 7 (DCPT-7) ....................................... 147

8.2

Interpretation of Measured HVSR Curves in DCPT-7 Area.................................. 148

8.3

Forward Modelling to Infer the Vs Profiles in DCPT-7 Area ................................ 150

8.4

Evaluation of Compaction Quality ............................................................................. 158

8.5

Empirical Relationships ................................................................................................ 160

8.6

Summary ......................................................................................................................... 162

Chapter Nine: Inversion of HVSR and Dispersion Curves Using Monte Carlo
Simulation .................................................................................................................164
9.1

Monte Carlo Inversion Procedure .............................................................................. 165

9.1.1

Constraining Information for the Inversion Procedure ................................. 165

9.1.2

Obtaining the Measured Rayleigh Wave Dispersion Curve and the Misfit

between Measured and Theoretical Dispersion Curves ................................................. 166


9.1.3

Misfit between Measured and Theoretical HVSR Curves .............................. 167

9.2

Application of the Monte Carlo Simulation Inversion Method ............................. 168

9.3

Influence of Mean Grain Size (D50) on Estimated Vs Profiles ............................... 172

9.4

Summary ......................................................................................................................... 173

Chapter Ten: Conclusions ........................................................................................175


10.1 Summary of Investigations........................................................................................... 175

- xi -

Table of Contents
10.2 Conclusions .................................................................................................................... 178
10.3 Recommendations for Future Research..................................................................... 181
References....................................................................................................................................... 185
Appendices ..................................................................................................................................... 197
A.

Seismic Waves and Methods........................................................................................ 197

B.

MATLAB Programming .............................................................................................. 206

C.

Deriving Naviers Equation from First Principles.................................................... 219

- xii -

List of Figures

List of Figures
Figure 1.1: Value of engineering construction work completed (Australian Bureau of
Statistics 2012) .............................................................................................................................. 1
Figure 1.2: Compaction methods and assessment. ....................................................................... 3
Figure 1.3: Origin of ambient vibrations. ....................................................................................... 5
Figure 2.1: Phase relationship diagram. ........................................................................................ 14
Figure 2.2: Effect of compaction. ................................................................................................. 16
Figure 2.3: Mohr-Coulomb failure envelope for a saturated soil (Fredlund & Rahardjo
1993). ............................................................................................................................................ 17
Figure 2.4: Relationship between void ratio and internal friction angle (Cha & Cho 2007).17
Figure 2.5: Extended Mohr-Coulomb failure envelope for unsaturated soils (Fredlund &
Rahardjo 1993)............................................................................................................................ 19
Figure 2.6: Correlation between void ratio and permeability (Mesri & Olson 1971). ........... 20
Figure 2.7: Sand cone replacement equipment............................................................................ 21
Figure 2.8: Nuclear density Gauge. ............................................................................................... 22
Figure 2.9: Track mounted drill rig with SPT capabilities (Jeffrey and Katauskas Pty Ltd
2012). ............................................................................................................................................ 24
Figure 2.10: Seismic cone configuration (CPT attachment) (After Brouwer 2008)............... 26
Figure 2.11: Shear wave velocity correlated with void ratio (Heitor et al. 2012).................... 28
Figure 2.12: Shear wave velocity correlated with void ratio (Cha & Cho 2007). ................... 28
Figure 2.13: Effect of matric suction on shear wave velocity (Heitor et al. 2012)................. 31
Figure 2.14: Horizontal to Vertical Spectral Ratio (HVSR) method........................................ 35
Figure 2.15: TrominoTM velocimeter used for the H/V spectral analysis (HVSR). ............... 36
Figure 2.16: Source model (after Lachet & Bard 1994). ............................................................ 38
Figure 3.1: Flowchart of the process involved in estimating the Vs profile of the ground. . 46
Figure 3.2: TrominoTM velocimeter with its carry case............................................................... 47
Figure 3.3: Horizontally layered soil model. ................................................................................ 55
Figure 3.4: Microtremor source model (after Arai and Tokimatsu (2000, 2004)). ................. 55
Figure 3.5: Proposed methodology of the HVSRplus technique. ............................................ 60
Figure 3.6: Example test layout for MSOR technique, also taking advantage of the HVSR
techniques sensor. (a) pseudo-linear array, (b) layout for shot 1 and (c) layout for shot 2.
.......................................................................................................................................................61
Figure 3.7: Shear wave velocity effected by confining pressure (Heitor et al. 2012). ............ 62
- xiii -

List of Figures
Figure 3.8: Surface layer overlying bedrock: One end free and the other end fixed. ............ 65
Figure 3.9: Surface layer overlying bedrock: One end free and the other end fixed
Boundary conditions. ................................................................................................................. 66
Figure 4.1: Sensitivity of the layer thickness (y) parameter on the theoretical HVSR curve.
(a) Layer 1, (b) Layer 2, and (c) Layer 3. ................................................................................. 70
Figure 4.2: Sensitivity of the layer Vs parameter on the theoretical HVSR curve. (a) Layer 1,
(b) Layer 2, and (c) Layer 3. ......................................................................................................71
Figure 4.3: Sensitivity of the layer Vp parameter on the theoretical HVSR curve. (a) Layer 1,
(b) Layer 2, and (c) Layer 3. ......................................................................................................72
Figure 4.4: Sensitivity of the layer density () parameter on the theoretical HVSR curve. (a)
Layer 1, (b) Layer 2, and (c) Layer 3. ....................................................................................... 73
Figure 4.5: Sensitivity of the layer quality factors (Qp = 2Qs) parameter on the theoretical
HVSR curve. (a) Layer 1, (b) Layer 2, and (c) Layer 3. ......................................................... 74
Figure 4.6: Surface layer (10 m thick at 250, 300 and 350 m/s) overlying the half-space
(bedrock at 600 m/s) the higher the resonance peak amplitude (H/V), the greater the
impedance contrast between layers. ........................................................................................ 75
Figure 4.7: Surface layer (10, 15 and 20 m thick at 250 m/s) overlying the half-space
(bedrock at 600 m/s) the higher the resonance peak frequency (on the frequency axis),
the shallower the location of the impedance contrasts......................................................... 76
Figure 4.8: Impedance contrast to HVSR peaks: a two layer system. (a) theoretical HVSR
curve and (b) Vs profile. ............................................................................................................ 77
Figure 4.9: Impedance contrast to HVSR peaks: a three layer system. (a) theoretical HVSR
curve and (b) Vs profile. ............................................................................................................ 77
Figure 4.10: Impedance contrast to HVSR peaks: a four layer system. (a) theoretical HVSR
curve and (b) Vs profile. ............................................................................................................ 78
Figure 4.11: Impedance contrast to HVSR peaks: superimposing the two, three and four
layer systems. (a) theoretical HVSR curves and (b) Vs profiles. .......................................... 79
Figure 4.12: Impedance contrast to HVSR peaks. (a) theoretical HVSR curve and (b) Vs
profile. .......................................................................................................................................... 79
Figure 4.13: Sensitivity analysis: The influence of the first (shallowest) impedance contrast
on the first (highest frequency) peak. (a) theoretical HVSR curves and (b) Vs profiles. . 81
Figure 4.14: Sensitivity analysis: The influence of the second (centre) impedance contrast
on the second (middle) peak.....................................................................................................81

- xiv -

List of Figures
Figure 4.15: Sensitivity analysis: The influence of the third (deepest) impedance contrast on
the third (lowest frequency) peak............................................................................................. 82
Figure 5.1: Trial geotechnical investigation location (Image from Google Maps). ................ 84
Figure 5.2: Test grid at the Penrith (Kingswood) campus of UWS. ........................................ 85
Figure 5.3: An example of measured HVSR curve..................................................................... 86
Figure 5.4: Time-history plot of recording using TrominoTM velocimeter.............................. 86
Figure 5.5: A summary of the measured HVSR curves from the proposed UWS
development site......................................................................................................................... 88
Figure 5.6: Measured and theoretical HVSR curves at stations (a) X5, (b) Z3 and (c) Y5. .. 91
Figure 5.7: Estimated Vs profiles for stations X5 (solid line), Z3 (broken line) and Y5
(dotted line). ................................................................................................................................ 92
Figure 5.8: Verification Boreholes at X1 (left), X3 (centre fitted blind) and X5 (right)
with their respective estimated Vs profiles. ............................................................................ 93
Figure 5.9: Verification Boreholes at X5 (left) and Z5 (right) with estimated Vs profiles at
HVSR stations X5 (left), Y5 (centre) and Z5 (right). ............................................................ 94
Figure 5.10: Verification MASW projected Vs along Line 5 (a) Phase velocity dispersion
curve with active sources at both ends, and (b) MASW estimated Vs profile at line 5
against HVSR estimated Vs profile at station Y5. ................................................................. 95
Figure 5.11: HVSR estimated Vs versus measured SPT-N values. ........................................... 96
Figure 5.12: An example of synthetic multichannel field recording by the MSOR technique.
The vertical axis shows the geophone number and the horizontal axis shows the time. 97
Figure 5.13: Measured and theoretical (dashed line) dispersion curves obtained by the
MSOR technique. ....................................................................................................................... 97
Figure 5.14: Measured (solid) and theoretical (dashed line) HVSR curves. The theoretical
HVSR curve produced using the initial guess Vs profile by the MSOR technique. ......... 98
Figure 5.15: Measured (solid) and theoretical (dashed line) HVSR curves. The theoretical
HVSR curve was produced after minimising the fit between measured and theory. ...... 99
Figure 5.16: Estimated Vs profiles obtained by the MSOR (used for the initial guess for the
HVSR technique) (dashed) and HVSR (solid) techniques. ..................................................99
Figure 5.17: Verification of the Vs profile produced by the HVSRplus technique with
borehole and SPT data. ...........................................................................................................100
Figure 6.1: Test locations at Penrith Lakes quarry....................................................................104
Figure 6.2: Evolution of HVSR curves with compaction (also showing the 2 standard
deviation of the mean (SDOM) for each curve)..................................................................106
- xv -

List of Figures
Figure 6.3: Compaction evolution of measured and theoretical HVSR curves. (a) Very loose
top layer, (b) Slightly loose top layer, and (c) Compacted top layer. ................................107
Figure 6.4: Compaction evolution of normalised Vs profiles. ................................................108
Figure 6.5: 24 hour (a) horizontal and (b) vertical spectra at the MCP area. ........................109
Figure 6.6: Consistency of HVSR curves at MCP area. ...........................................................109
Figure 6.7: Consistency of HVSR curves at WLW area...........................................................110
Figure 6.8: MCP area test locations. ...........................................................................................110
Figure 6.9: HVSR curves summary MCP area. ......................................................................111
Figure 6.10: MCP area station Z3 (a) measured and theoretical HVSR curves, and (b)
normalised HVSR Vs and estimated (before normalisation) HVSR Vs. ......................113
Figure 6.11: MCP area measured and theoretical HVSR curves (a) A1, (b) B1, and (c) B3.
.....................................................................................................................................................114
Figure 6.12: MCP area normalised Vs profiles summary. ........................................................115
Figure 6.13: WLW area test locations. ........................................................................................116
Figure 6.14: HVSR curves summary WLW area. ..................................................................117
Figure 6.15: WLW area measured and theoretical HVSR curves (a) A3, (b) B3, and (c) E2.
.....................................................................................................................................................118
Figure 6.16: Normalised Vs profiles of WLW area...................................................................119
Figure 6.17: Mean Vs with depth with 95% confidence interval. ...........................................120
Figure 6.18: Mean Vs of the compacted material versus standard deviation........................121
Figure 7.1: DCPT-9 area test locations. .....................................................................................126
Figure 7.2: Compaction curve from the DCPT-9 area (Heitor et al. 2012). .........................127
Figure 7.3: Consistency of HVSR curves over 24 hour period at DCPT-9 area. ................128
Figure 7.4: HVSR curves over 24 hour period at DCPT-9 area with peaks of (a)
stratigraphic origin and (b) stratigraphic and artificial origin. ............................................128
Figure 7.5: Hourly HVSR curves over 24 hour period 11 am to 10 am next day in order
from left-to-right and top-to-bottom. ...................................................................................129
Figure 7.6: Time history of recordings at 6 pm (left) and 2 am (right). .................................130
Figure 7.7: Directional HVSR at 6 pm (left) and 2 am (right). ...............................................130
Figure 7.8: Single component spectra at 6 pm (top) and 2 am (bottom). .............................131
Figure 7.9: HVSR curves during time window 8 am 4 pm at DCPT-9 area. ....................132
Figure 7.10: Comparison of the normalised Vs profiles from the hourly HVSR curves over
24 hours in Figure 7.5. .............................................................................................................132
Figure 7.11: HVSR curves superimposed summary DCPT-9 area. ....................................133
- xvi -

List of Figures
Figure 7.12: Calibration DCPT-9 area (a) measured and theoretical HVSR curves at
station A4, and (b) normalised HVSR Vs (location A4) versus normalised CPT qc
(location 9.6). ............................................................................................................................136
Figure 7.13: DCPT-9 area normalised Vs profiles summary (a) full profile and (b) top 3m
profile (roller compacted material). .......................................................................................137
Figure 7.14: DCPT-9 area measured and theoretical HVSR curves (a) A2 and (b) B4.......138
Figure 7.15: Verification DCPT-9 area normalised HVSR Vs versus normalised CPT qc
versus measured DMT M versus normalised MASW Vs (a) HVSR location A2 versus
CPT location 9.8 versus DMT location 9.3 versus MASW Row A, and (b) HVSR
location Z1 versus CPT location 6.4 versus DMT location 6.4. .......................................139
Figure 7.16:Verification DCPT-9 area MASW projected Vs along Row A (a) Phase
velocity dispersion curve with active sources at both ends, and (b) MASW estimated Vs
profile against HVSR estimated Vs profile at location A2 and A4. ..................................141
Figure 7.17: Mean normalised Vs versus the standard deviation of the stations in the
DCPT-9 area. ............................................................................................................................142
Figure 7.18: Influence of composite compaction at DCPT-9 area. .......................................144
Figure 8.1: DCPT-7 area test locations. .....................................................................................148
Figure 8.2: Superimposed summary of the measured HVSR curves at the DCPT-7 area. 149
Figure 8.3: Flowchart calibration, verification and application. ..........................................151
Figure 8.4: Calibration DCPT-7 area CS7.3 (a) measured and theoretical HVSR curves,
(b) Phase velocity dispersion curve, and (c) normalised HVSR Vs and normalised
MSOR Vs versus normalised CPT qc. ................................................................................154
Figure 8.5: Verification DCPT-7 area: Measured and theoretical HVSR curves at stations
(a) CS7.1, (b) CS7.2, and (c) CS7.4. .......................................................................................155
Figure 8.6: Verification DCPT-7 area: Normalised HVSR Vs versus normalised CPT qc
at stations (a) CS7.1, (b) CS7.2, and (c) CS7.4. ....................................................................156
Figure 8.7: Remaining grid points DCPT-7 area measured and theoretical HVSR curves
(a) station A1, and (b) station C2. ..........................................................................................157
Figure 8.8: DCPT-7 area normalised Vs profiles summary. ....................................................158
Figure 8.9: Compaction assessment Vs at 1 m intervals. ......................................................158
Figure 8.10: Influence of dynamic compaction.........................................................................159
Figure 8.11: Contour plot of Vs profiles along Row E (refer to Figure 8.1). .......................160
Figure 8.12: Normalised Vs versus normalised qc relationship for the DCPT-7 area. .........160

- xvii -

List of Figures
Figure 8.13: Frequency versus depth relationships for the DCPT-7 area, also showing the
standard error. ...........................................................................................................................162
Figure 9.1: Monte Carlo inversion at station CS 7.1 (a) measured and theoretical HVSR
curves, (b) measured and theoretical dispersion curves, and (c) normalised CPT, qc
versus normalised average VS profile with its corresponding 95% confidence interval (n
= 30). ..........................................................................................................................................169
Figure 9.2: Monte Carlo inversion: Measured and theoretical HVSR curves at stations (a)
CS7.2, (b) CS7.3, and (c) CS7.4. .............................................................................................170
Figure 9.3: Monte Carlo inversion: Normalised CPT, qc versus normalised average VS
profile with its corresponding 95% confidence interval (n = 30) at stations (a) CS7.2, (b)
CS7.3, and (c) CS7.4.................................................................................................................171
Figure 9.4: Comparison between average normalised qc and normalised Vs obtained from
the Monte Carlo inversion of the four control stations at the DCPT-7 area. .................172
Figure 9.5: Correlation between mean grain size (D50) and soil behaviour index (Ic) obtained
from average normalised, qc and normalised average Vs obtained from the Monte Carlo
inversion of the four control stations at the DCPT-7 area. ...............................................173
Figure 10.1: Penrith Lakes compaction comparison: Mean Vs of the compacted material
versus standard deviation. .......................................................................................................178

- xviii -

List of Tables

List of Tables
Table 1.1: Fraction of compacted ground covered by mechanical testing at Penrith Lakes. . 8
Table 2.1: Vs correlations with SPT and CPT measurements. .................................................29
Table 2.2: Characteristics of Surface Wave Techniques. ........................................................... 32
Table 2.3: Summary of the nature of ambient vibrations (after SESAME 2004a). ............... 34
Table 3.1: Digital velocimeter specifications. .............................................................................. 47
Table 3.2: Thesis recommendations for using the HVSR technique for compaction
assessment. .................................................................................................................................. 51
Table 4.1: Simple theoretical ground profile used for the sensitivity analysis. ....................... 69
Table 4.2: Final Vs profile used for parametric study. ............................................................... 76
Table 5.1: Example SESAME criteria for reliability of results (Grilla by Micromed). .......... 87
Table 5.2: Summary of Vs profiles from both the MSOR and HVSR techniques, and the
borehole information. ..............................................................................................................100
Table 6.1: Dry density profile in MCP area. ..............................................................................112
Table 7.1: Soil properties of the DCPT-9 area (Heitor et al. 2012). ......................................127
Table 7.2: Students t-test results. ................................................................................................143

- xix -

Abbreviations

Abbreviations
Abbreviations and symbols have been defined where they first appear in the text. For
convenience, the most frequently used abbreviations and symbols are also defined below.

Angle of internal friction

Angular frequency

or b

Bulk density

Density of water

Dry density

Effective unit weight

Mean effective stress

Poissons ratio

or G

Shear modulus

Vertical effective stress

qa

Bearing Capacity

qc

Cone end resistance

Vp

Compression wave velocity, P-wave velocity, primary wave velocity

c'

Effective cohesion

Moisture content

fN

Nyquist frequency

Period

f0

Predominant resonance frequency

Qp and Qs

Quality factors of Vp and Vs, respectively

Pa

Reference stress, typically atmospheric pressure

fr

Resonance frequency

Vs

Shear wave velocity, S-wave velocity, secondary wave velocity

Fs

Sleeve friction during CPT

y or H

Soil layer thickness

Gs

Specific gravity of the solid grains

Speed of wave propagation


- xx -

Abbreviations
SPT-N or N

Penetration resistance from the SPT

Cu

Undrained shear strength

Vertical drained constrained modulus

Void ratio

Wave number

CPT

Cone penetration test

CSWS

Continuous surface wave system

DMT

Dilatometer test

DCPT-7

Dynamic Compaction Prototype Trial Area 7

DCPT-9

Dynamic Compaction Prototype Trial Area 9

FFT

Fast Fourier transform

HVSR

Horizontal to vertical spectral ratio

LD

Leaky Dam

MCP

McCarthys Pit

MASW

Multichannel analysis of surface waves

MSOR

Multichannel simulation using one receiver

OMC

Optimum moisture content

ReMi

Refraction microtremor

SASW

Spectral analysis of surface waves

SMDD

Standard maximum dry density

SPT

Standard penetration test

WLW

Wildlife Lake West

- xxi -

Chapter One: Introduction


1.1 Background
Ground compaction is a widely applied form of ground improvement technique in
engineering construction. The value of engineering construction work done for both the
public and private sectors in Australia was worth $86 billion in 2010-2011 (Australian
Bureau of Statistics 2012), contributing 5.7% to the gross domestic product (GDP). Figure
1.1 shows that engineering construction activity is growing steadily, at least for the last
decade. Engineering construction is a growing and significant industry in Australia, and
ground improvement including ground compaction is a very important part of it.

Figure 1.1: Value of engineering construction work completed (Australian Bureau of


Statistics 2012)
-1-

Chapter One: Introduction


The effect of ground compaction is to cause the expulsion of entrapped air in the soil,
rearrange and pack the soil particles more closely together. In the process it reduces the
void ratio and therefore increases the density of the soil. It also leads to improved
geotechnical properties of the ground, namely in terms of increased load bearing capacity,
minimisation of future settlement and erosion as well as reduced soil permeability
(Venkatramaiah 2006). In consequence, it is vitally important and essential to have available
appropriate means to assess the quality and consistency of ground compaction because of
their potential impact on the design, performance, safety and eventual cost of construction.

Methods of Assessing Compaction


In assessing the quality of a compacted ground, engineers resort to a number of techniques
commonly available for and associated with geotechnical site investigation. In general,
these techniques attempt to convey the quality of compaction by measuring certain
indicative quantities that include the following: (1) the dry density and moisture content
(Proctor 1933) (2) various forms of mechanical resistance pertaining to the measuring
devices, such as SPT, CPT, DMT, etc (e.g. Totani et al. 2001; Karray et al. 2011) (3) the
shear wave velocity, Vs, (e.g. Claria Jr. & Rinaldi 2007; Feng et al. 2010; Harutoonian et al.
2012a), of the compacted ground. Techniques suitable to a given site are dependent on a
number of factors such as: site accessibility, required depth of measurement, area of site,
soil information to be retrieved, cost and time required for testing. All techniques have
inherent strengths and limitations in relation to the above factors. Figure 1.2 shows several
ground compaction techniques (e.g. dynamic compaction, roller compaction), as well as
some widely used methods for compaction assessment (e.g. sand cone replacement, nuclear
density gauge, CPT). It can be noted that, the depth of measurement and equipment size of
the methods used for compaction assessment.
The classical techniques for assessing the quality of a compacted site are generally
categorised as (a) invasive (mechanical) or (b) non-invasive (geophysical) methods
(Durgunoglu et al. 2003). The most common mechanical form of compaction assessment
is by measurement of the dry density (d) and moisture content of the compacted ground.
The appeal of this method is due in a large part to the ease of the Proctor (1933) test in
defining the standard maximum dry density (SMDD) and optimum moisture content
(OMC) of a soil for controlling soil compaction. Measurements of dry density/moisture
content are most expedient when the compacted layers are thin and made accessible by a
-2-

Chapter One: Introduction


surface method like the non-invasive nuclear density gauge or the invasive sand
replacement cone. However, there are limitations with these devices in that the
measurements are localised and typically limited to a depth of half a metre at most, and are
unsuitable for deep compacted sites (e.g. dynamically compacted ground).

Dynamic
Compaction

CPT
Rolling
Compaction

Dry density
Tests

Surface
Wave Tests

Figure 1.2: Compaction methods and assessment.


When compaction of thick lifts needs to be evaluated (e.g. in the case of dynamic
compaction), invasive mechanical tests able to penetrate deeper into the ground to conduct
in situ measurements, like the cone penetration test (CPT), the standard penetration test
(SPT), dilatometer test (DMT) and the self-boring pressuremeter are often used (e.g.
Schmertmann et al. 1986; Totani et al. 2001; Arulrajah et al. 2011; Karray et al. 2011). The
CPT, SPT and DMT are well accepted localised techniques in site investigation, but in
practice only a fraction of the site is actually being assessed due to the relatively high cost
and lengthy time invested in applying such techniques. It is thus possible that a local area of
poor compaction may be completely missed by the measurements, when the mechanical
tests are spatially far apart.
In recent years, attempts have been made to measure the shear wave velocity (Vs) of
compacted ground using non-invasive 1D and 2D surface wave techniques like the
-3-

Chapter One: Introduction


continuous surface wave system (CSWS), spectral analysis of surface waves (SASW),
multichannel analysis of surface waves (MASW) and refraction microtremor (ReMi)
methods (e.g. Kim & Park 1999; Kim et al. 2001; Moxhay et al. 2001; Park & Miller 2004;
Pullammanappallil 2006; Feng et al. 2010; Waddell et al. 2010). These methods ultimately
rely on the estimation and then inversion of the phase velocity dispersion curves to derive
the Vs profile, thus providing an indirect conveyance of the compaction quality. Empirical
relationships relating Vs to more accessible soil properties, including the dry density, have
been established and reported in literature (e.g. Blake & Gilbert 1997; Kim & Park 1999;
Claria Jr. & Rinaldi 2007). Surface wave techniques are able to survey the soil profile down
to depths well beyond the interest of geotechnical engineers over large areas rather
expeditiously and economically. Nonetheless, these methods require the setting up of an
array of interconnecting geophones on the compacted surface, which can sometimes prove
unwieldy, with which to measure the phase velocity dispersion curve. Moreover, the
intrinsic difficulty in distinguishing the various modes of the dispersion curves and
differentiating apparent velocities from real velocities in passive arrays requires more
rigorous investigation.
The main objective of this thesis is to characterise compacted ground using a technique
based on the measurement of the passive horizontal to vertical spectral ratio (HVSR) of
ambient vibrations, also known as microtremors. Figure 1.3 illustrates the origins of
ambient vibrations or microtremors using a snapshot of everyday life. In this figure,
several microtremor sources are observed which can contribute to the measured HVSR
data. These sources comprise both low and high frequency components. The low
frequency sources in this figure include: ocean waves and wind, whereas the high frequency
sources result from human activities, traffic, trains and activities in buildings (commercial,
residential and industrial).
A comprehensive survey of literature suggests that the HVSR technique has not yet been
applied to characterise field compaction, so it would appear this is the first time this
technique has been studied for this purpose. Although the HVSR technique is widely
considered as a surface wave method, it is also unique in using a three component (two
horizontal and one vertical) sensor to measure the ambient vibrations at the observation
point on the ground. It uses the horizontal to vertical spectral ratio of the surface
displacements rather than the phase velocity dispersion curves of the surface waves as the
theoretical basis to infer the layered soil profile of the compacted ground. The background
-4-

Chapter One: Introduction


to this method can be traced back to the 1950s when ambient vibration measurements
were pioneered (Kanai & Tanaka 1954; Aki 1957; Kanai 1957), but it was not until the
studies of Nogoshi and Igarashi (1971) and, in particular, Nakamura (1989) that the
possibility of employing this technique to establish the resonance frequencies of the ground
was seriously contemplated. This was then followed by more recent developments of
methods and algorithms to invert the measured HVSR curves to infer the Vs profile and
the associated soil stratigraphy of the site (e.g. Fah et al. 2003; Arai & Tokimatsu 2004;
Lunedei & Albarello 2009). Unlike the other surface wave methods, the passive HVSR
ambient vibration technique does not require the setting up of unwieldy connecting cables,
and thus the logistics and time for undertaking the field tests are at a minimum. This is a
significant advantage, however, there are also several key challenges in applying this
technique to characterise compacted ground.

Activities in
Buildings
Ocean
Waves

Traffic

Human
Activities

Trains

Figure 1.3: Origin of ambient vibrations.

Challenges of the HVSR technique


The primary challenge to this technique, like all other non-invasive surface wave methods,
is to address the issue of non-uniqueness of the inversion solution; that is, dissimilar soil
profiles may be inferred from the same measured HVSR curve (e.g. Castellaro & Mulargia
2009a; Harutoonian et al. 2012a). A second challenge is in making use of the higher
frequencies of the HVSR curve to resolve the ground velocity structure, since hitherto
-5-

Chapter One: Introduction


other researchers have primarily been focussing on the lower frequency section and the
predominant resonance peak of the HVSR curve, for seismological studies. Since the
HVSR technique is a non-invasive surface technique, the effects of anthropic noise (i.e.
unwanted noise caused by artificial activities) may create a third challenge: identifying and
recognising the effects of the artifacts in the analysis. Fourthly, the effects of higher modes
may also cause secondary resonance peaks to appear at higher frequencies while features of
the Rayleigh wave ellipticity may result in the presence of both secondary peaks and
troughs at higher frequencies. These effects need to be examined and a means found to
avoid misinterpretation of the HVSR curve. Fifthly, the sensitivity of the methodology to
sufficiently discriminate different layers of ground compaction needs to be amply
demonstrated on the basis of rigorous field investigation, modelling and analysis.

1.2 Impetus for Research


This research is undertaken because of a special need to develop a low-cost and effective
approach for assessing the quality of extensive compacted fills at Penrith Lakes. The sensor
used to carry the HVSR measurements is a very low cost investment relative to the test rigs
for the invasive mechanical (e.g. SPT, CPT and DMT) tests. In 2012, a sensor unit costs
between $10,000 and $15,000, while the mechanical rig would typically cost in the
hundreds of thousands of dollars. To recover high overheads, an invasive test would cost a
few thousand dollars, whereas a HVSR measurement could be carried out for not more
than a few hundred dollars. Moreover, the setup and testing time for the HVSR (normally
lasting not more than 30 minutes) is significantly faster than the comparative times for an
invasive test (from a couple of hours to longer depending on the depth of test).
Penrith Lakes is a very large sand and gravel quarry occupying some 1,935 ha of land in
Western Sydney. The quarry is reaching the end of its production life and the site is
continuously being remediated to return land affected by quarrying to a satisfactory state.
This means creating parklands, recreational lakes, and land earmarked for potential
infrastructure and housing development. It was originally intended to remediate the quarry
to parkland status only, and in some areas to even less than this, as no housing
development on the land was envisaged in the early stages of quarrying. Subsequently,
however, plans were drawn for future infrastructure and housing development in some
areas of the quarry site. As a result, land earmarked for potential development, including
-6-

Chapter One: Introduction


some areas previously already compacted to parkland status only or even less, have to be
improved to achieve the standards required for development. The quarry site has
undergone, and is still undergoing, at least two forms of roller compaction, as well as
dynamic compaction.
A major challenge has been to find a suitable technique or techniques for Penrith Lakes, to
assess not only differently compacted areas, the methods must also be able to evaluate
different depths and across laterally extensive compacted ground efficiently. Traditional
methods of assessing compaction would commonly include in situ dry density and moisture
content tests (e.g. by means of nuclear gauge, sand cone replacement) for the roller
compacted areas, and invasive penetration tests (e.g. CPT, SPT, DMT) for dynamically
compacted areas. While dry density and moisture content tests are well suited to assess
compaction very close to the surface, typically not more than half a metre deep as
mentioned above, these tests cannot reach the depths required for dynamically compacted
areas without sinking an invasive borehole. Invasive tests are commonly utilised to assess
compaction of greater depths, but such tests are not only expensive, the localised results
will on average cover less than 1% of the actual compacted area (Mooney et al. 2010). Thus
areas of poor compaction not meeting the required standards may be easily missed.
Table 1.1 below shows the intensity of mechanical tests applied to 4 compacted areas of
Penrith Lakes which have been investigated in this thesis: (1) McCarthys Pit (MCP) - roller
compaction, (2) Wildlife Lake West (WLW) roller compaction, (3) Dynamic Compaction
Prototype Trial Area 9 (DCPT-9) combination of dynamic and roller compaction and (4)
Dynamic Compaction Prototype Trial Area 7 (DCPT-7) dynamic compaction. The
statistics in the table reinforces the earlier assertion that the actual compacted ground
covered by each of the mechanical tests occupies only a very small fraction of the entire
compacted ground. Table 1.1 shows that there will be many gaps between the mechanical
tests applied in the compacted grounds. Thus some means to cover the gaps missed by the
mechanical tests would be necessary in order to achieve a more complete and
comprehensive assessment of the compacted ground.

-7-

Chapter One: Introduction


Table 1.1: Fraction of compacted ground covered by mechanical testing at Penrith
Lakes.
Roller Compaction at Penrith Lakes
Tested Surface
Mechanical
Volume
Location
Test Type
Area (m2)
Test Points
Tested
1 test per 646*
MCP
1,200
26
m3
Density
(Nuclear
WLW
1,800
N/A
N/A
Gauge)
DCPT-9**
1 test per 409*
6,000
44
(Top 3 m)
m3
Dynamic Compaction at Penrith Lakes
Location

Tested Surface
Area (m2)

DCPT-7

9,500

DCPT-9**
(Below 3 m)

6,000

Test Type
Borehole, CPT,
DMT

Mechanical
Test Points
4
9

Area Tested
1 test per 2375
m2
1 test per 667
m2

*Since the density tests were carried out at all depths (i.e. bedrock to surface), the Volume Tested was used
instead of Area Tested to determine the frequency of tests. **The DCPT-9 area underwent an increased
amount of testing because it was the test bed for the remaining dynamically compacted areas.

1.3 Objectives of Research


The passive ambient vibration HVSR technique has already been shown as a method able
to infer the ground velocity structure to depths of several hundreds of metres below the
surface. Geophysicists have quite successful applied it in assessing the seismic response (i.e.
resonance frequency and site amplification) of a site (e.g. Nakamura 1989; Lachet & Bard
1994; Castellaro et al. 2005; Bonnefoy-Claudet et al. 2006a; Albarello & Lunedei 2009). In
geotechnical engineering, the ground layers of most interest are typically less than 15
metres below the surface while the shear wave velocity (Vs) of compacted soil also varies
within a relatively smaller range. Both sets of circumstances are quite different to the ones
typically experienced by geophysicists, who in the main must deal with deeper and more
varied ground.
In consequence, the main objectives of this thesis are to:
1. Refine the HVSR technique and develop a methodology to facilitate the derivation
of the shear velocity (Vs) profile of the compacted soil layers,
2. Justify and demonstrate the use of Vs as an alternative geotechnical measurement
for assessing compacted ground,

-8-

Chapter One: Introduction


3. Achieve a significantly better ground characterisation technique in terms of cost
and speed than equivalent classical surface wave methods for the characterisation
of deep and large sites,
4. Demonstrate that the HVSR based approach can be reliably and economically
employed to carry out a site characterisation, particularly of the compacted layers,
when proper calibration and verification against independent test results are carried
out.
The more specific goals of this thesis are to:
1. Optimise the use of invasive mechanical tests by applying the HVSR based
approach to cover the gaps in areas which are not subjected to the mechanical tests,
2. Substantiate the use of the measured HVSR curve to provide preliminary indication
of the compaction quality It is already recognised that the predominant resonance
peak of the HVSR curve is a reflection of the impedance contrast between surface
layers and the bedrock. However, this thesis seeks to validate the conjecture that
the secondary resonance peaks of the HVSR curve at higher frequencies may
reflect impedance contrasts within surface (i.e. compacted) layers,
3. Facilitate and refine the calibration and verification procedures for the HVSR
forward model by incorporating other geophysical techniques,
4. Propose site specific correlations of Vs with available mechanical penetration data
(CPT and SPT) for use in locations without mechanical testing,
5. Demonstrate that the relationship between the predominant resonance frequency
and the bedrock depth can be extended to include the relationships between
secondary resonance peaks at higher frequencies and the depths of strong
impedance contrasts of surface layers.

1.4 Organisation of Thesis


This thesis has investigated the possibility of using a simple high resolution portable device
to perform ground compaction investigations. It has adopted a systematic approach: first
introducing the nature of research and research question, then providing appropriate
literature review and background information before going on to establish the research
methodology, discussion of the results and elaboration of the research outcomes. Above
all, recommendations for future studies have been suggested so that the objectives achieved
in this study could be further developed. The organisation of the thesis is as follows:
-9-

Chapter One: Introduction


Chapter One: Introduction This chapter introduces the research topic and research
question as well as outlining the motivation and the objectives of research.
Chapter Two: Literature Review In this chapter, a literature and background review
has been undertaken on current techniques utilised for evaluation of ground compaction.
In particular, the reason behind employing the dry density at a given moisture content as
the most common method for compaction evaluation is explored. In addition, other
stiffness parameters (e.g. CPT, SPT, DMT) used in compaction evaluation of deep fills are
discussed. The reasoning and ability of utilising Vs for compaction evaluation is presented,
along with introducing the fast and operationally simple HVSR technique as a
complimentary tool for assessing the quality and consistency of roller and dynamic
compaction. Furthermore, a survey of the history and current trends for the theoretical
modelling of the HVSR curve is included, which have the main aim of inferring the Vs
profile of the ground.
Chapter Three: Methodology The approach and methodology is established in this
chapter. The component methods which are often the outcome or an adaption of the work
of previous investigators are discussed. Here a more in-depth review of the HVSR
technique has been undertaken and the important forward modelling (identification)
procedures described to shed light on what is often construed as a black box. The concept
of normalising estimated Vs profiles for confining pressures due to overburden stresses is
discussed. The resonance frequencies (or harmonics) of a two layer system (i.e. surface
layer overlying the bedrock) are derived by mathematically modelling wave propagation
through an elastic medium. An innovative two-step inversion procedure is introduced, the
HVSRplus technique, which involves combining the HVSR and MSOR techniques to
enhance the estimated Vs profiles. Recommendations from the experience and knowledge
gained from this thesis, for the use of the HVSR technique for the evaluation of ground
compaction are presented.
Chapter Four: Parametric Studies and Analysis of HVSR Curves A sensitivity
analysis is performed on the parameters used to produce the theoretical HVSR curve, to
better examine and understand the general landscape of the forward model problem.
Furthermore, a parametric study is initiated to study the influence of impedance contrasts
in the ground on theoretical HVSR curves.
- 10 -

Chapter One: Introduction


Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique
Prior to adopting the proposed methodologies and techniques for compaction assessment,
they were trialled at an area made available at the Penrith (Kingswood) campus of the
University of Western Sydney (UWS). The purpose of this exercise was to test the
proposed methodologies and techniques at a local and more flexible site before applying
them at the Penrith Lakes quarry in Western Sydney. Here the future potential
development site was characterised using HVSR and HVSRplus techniques, which
ultimately allowed for a site specific Vs and SPT-N correlation.
Chapter Six: Investigation of Rolling Compaction using the HVSR Technique
The Penrith Lakes quarry test site in Western Sydney is introduced. Three roller compacted
areas from the Penrith Lakes site have been evaluated for compaction quality and
consistency by HVSR techniques. First, the measured HVSR curves were interpreted to
provide a preliminary indication of the ground compaction. Second, forward modelling of
the HVSR curves allowed for a more informative assessment by estimating the Vs profiles.
The first roller compacted area Leaky Dam was used to establish the HVSR techniques
ability for differentiating ground compaction at different stages of roller compaction. The
remaining two roller compacted areas McCarthys Pit and Wildlife Lake West were then
evaluated to provide a more thorough compaction evaluation. The results from all three
roller compacted areas are presented and discussed.
Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction
using the HVSR Technique A more complex area at the Penrith Lakes quarry which
had undergone both dynamic and roller compaction was investigated using HVSR
techniques and also the active surface wave MASW technique. The array based MASW
technique was used for verification of HVSR technique inferred Vs profiles. The chapter
articulates the main idea of utilising the HVSR technique, that is, to complement the more
costly and expensive penetration techniques for evaluating the compaction of deep fills.
The results from the compaction evaluation of a composite dynamic and roller compacted
site are presented and discussed.
Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus
Technique A purely dynamically compacted area was studied in this chapter. In keeping
with the results and methodologies from previous chapters, the Vs profiles were calibrated
- 11 -

Chapter One: Introduction


and verified with mechanical test data, before applying HVSR techniques to the untested
areas, to fill in the gaps. The sequential inversion of MSOR and HVSR measured data, that
is, the HVSRplus technique, was investigated to identify the potential advantages of this
combination. The purpose of HVSRplus technique is not a joint inversion but instead the
active MSOR technique providing constraining information to the HVSR technique when
left with insufficient mechanical tests (for calibration and verification purposes).
Furthermore, relationships have been proposed relating HVSR data with CPT (cone end
resistance - qc) data, as well as a frequency-depth relationship utilising not only the
predominant resonance peak but also secondary resonance peaks (which are indicative of
the near surface layers). The results from the compaction evaluation of a purely
dynamically compacted site are presented and discussed.
Chapter Nine: Inversion of HVSR and Dispersion Curves using Monte Carlo
Simulation Limitations in the resolution of surface wave techniques can be overcome by
combining active and passive measurements, for an enhanced HVSR and dispersion curve
inversion. Thus, the combined HVSR and Rayleigh wave dispersion inversion by a Monte
Carlo simulation approach was applied to estimate the Vs profiles of the dynamically
compacted ground discussed in Chapter 8. Available CPT data from the DCPT-7 area was
used as reference to determine the appropriate number and thickness of layers, as well as,
to verify the Vs profiles estimated for the control stations in a relative sense. Furthermore,
the influence of the mean grain size, D50, on the Vs values was confirmed after comparing
the average normalised CPT - qc tip resistance, with the average normalised Vs estimated
for all layers.
Chapter Ten: Conclusions In this chapter the main conclusions and outcomes of this
study are presented and discussed. Recommendations for future studies are proposed in
the light of the research conclusions and outcomes obtained in this thesis.

- 12 -

Chapter Two: Literature Review


Ground compaction has been a critical part of civil engineering construction for centuries.
Ancient Rome, for instance, is renowned for building roads on compacted sand and dried
earth, and it is a testament to the quality of the compaction then that some of the roads still
exist even today. At present, ground compaction is commonly applied to remediate fills
and compressible soil areas as part of civil engineering construction for foundations, roads,
embankments, slopes etc. Although the objectives of ground compaction have remained
intact through the centuries, several modern techniques and equipment have since been
introduced for compacting the ground and for the assessment of compaction. The
assessment of ground compaction constitutes an extremely important part of the quality
control of the construction process, and sits at the core of this thesis. This chapter thus
reviews the basis of the current methods for assessing ground compaction, and it is divided
according to the three categories mentioned before in Chapter 1:
(1) the dry density and moisture content (Proctor 1933)
(2) various forms of resistance pertaining to the measuring device such as SPT, CPT,
DMT, etc (e.g. Totani et al. 2001; Karray et al. 2011)
(3) the shear wave velocity, Vs, (e.g. Claria Jr. & Rinaldi 2007; Feng et al. 2010;
Harutoonian et al. 2012a), of the compacted ground.
A review of the passive ambient vibration HVSR technique, the method of choice of this
thesis is then undertaken later in the chapter. Moreover, a survey of the methods used for
- 13 -

Chapter Two: Literature Review


the theoretical modelling of HVSR curves, for the purpose of inferring the Vs profile of the
site, is presented.

2.1 Dry Density and Moisture Content of


Compacted Ground
Measuring the dry density and the moisture content is a widely applied method of
determining the quality of compacted ground. In many respects, ensuring that the
measured dry density and moisture content comply with specified ranges of the maximum
dry density (MDD) and optimum moisture content (OMC) of the soil is widely regarded as
the gold standard for assessing ground compaction. There is already an abundance of
literature discussing the determination, and the specified ranges, of the MDD and OMC of
a soil (e.g. Proctor 1933; Standards Australia 2003a; Venkatramaiah 2006; Yilmaz 2009).
Surprisingly, though, there is much less scientific discourse that outlines from first
principles the relationships between achieving a high dry density at specified moisture
content, and that of achieving less settlement, higher shear strength and lower soil
permeability. After all, the latter are the very attributes ultimately sought after in a
compacted ground. The discussion below is thus partly an attempt to fill in some of the
important and missing details. At a fundamental level, this discourse needs to begin at the
phase relationships, as illustrated in Figure 2.1, that help to define the physical properties of
the soil skeleton.

Voids

Air
Air

Water
Water

Se

Solids

Solids
Solids

Soil
Skeleton

Solids

Dry Soil

Fully
Partially
Saturated
Saturated
Figure 2.1: Phase relationship diagram.
- 14 -

Chapter Two: Literature Review


Here, e and S are the interstitial void ratio and degree of saturation respectively. From the

phase relationship diagram, the dry density (d) and moisture content (w) may be deduced,

giving:

=
=


1+

(2.1a)

(2.1b)

where, Gs is specific gravity of the solid grains and w is the density of water. The dry
density is further related to the bulk density (b) as follows:

1+w

(2.2)

It may be inferred from Equation (2.1a) that the dry density of a soil is inversely related to
its void ratio, meaning that a soil with a low dry density at given moisture content has a
correspondingly high void ratio. A soil with a high void ratio is also synonymous to a loose
material, implying high compressibility or low stiffness (e.g. Panayiotopoulos 1989; Yilmaz
2009). Generally speaking, this results in higher settlement. On the other hand, a soil
having a high dry density at given moisture content has a correspondingly low void ratio,
which is synonymous to a dense material (implying low compressibility, high stiffness)

(ibid.), resulting in a lower settlement. Consequently, for a given soil (where Gs and w may
be assumed as constants), the dry density of the bulk soil provides, by extension, a good
indicator of the soil stiffness and its susceptibility to settlement (e.g. Stephenson 1978;
Yoon & Lee 2010).
As air is expelled during compaction, this allows the addition of more solids into the
ground, and has the effect of decreasing the void ratio and increasing the dry density. This
is illustrated in Figure 2.2. The addition of water (or moisture) in the compaction process
allows the soil particles to slip more easily (acting as a lubricant), and therefore aid the
compaction effort to achieve a smaller void ratio. However, there is a limit beyond which
the addition of more water will not help to reduce the voids any further, for a given
compaction effort. At this point, as Equation (2.1) shows, the dry density will also not
- 15 -

Chapter Two: Literature Review


increase any further and reaches a maximum. This is known as the maximum dry density
(MDD) and its corresponding moisture content, the optimum moisture content (OMC).

Solids
Air

Air

Water

Air

Air

Water

Water

Water

Solids

Solids

Solids

Solids

Before
Compaction

After
Compaction

After More
Compaction

Increasing
Volume of Solids

Figure 2.2: Effect of compaction.


As mentioned above, a second important objective of compaction is to achieve increased
shear strength of the ground. The shear strength of saturated soils is often defined by the
Mohr-Coulomb failure criterion and the effective stress concept as given in Equation (2.3)
(Terzaghi 1936).
= c + tan

(2.3)

where, ff is the shear stress on the failure plane at failure, c is the effective cohesion the
shear strength intercept when the effective normal stress is equal to zero, is the
effective normal stress on the failure plane at failure, and is the effective angle of
friction. The Mohr-Coulomb failure envelope for a saturated soil is shown in Figure 2.3.
Typically referred to as the failure envelope; the line produced by Equation (2.3) represents
possible combinations of shear and effective normal stresses on the failure plane at failure.
The relationship between void ratio and shear strength can be sought by observing the
relationship between void ratio and angle of friction.
Experimental evidence, similar to the results shown in Figure 2.4 by Cha and Cho (2007)
shows that the internal angle of friction generally increases with decreasing void ratio (e.g.
- 16 -

Chapter Two: Literature Review


Oda 1977; Lambe & Whitman 1979; Ueda et al. 2010). Cha and Cho (2007) suggest that
the internal friction angle is governed by interparticle friction, dilation (void ratio and
confining pressure effects), rotation (rolling and rotational frustration), structural
rearrangement, and particle crushing among other factors, particularly for sandy soils.

Figure 2.3: Mohr-Coulomb failure envelope for a saturated soil (Fredlund &
Rahardjo 1993).

Figure 2.4: Relationship between void ratio and internal friction angle (Cha & Cho
2007).
- 17 -

Chapter Two: Literature Review


In practice, a compacted soil remains in an unsaturated state. Unsaturated soils are
commonly defined as having three phases, namely, 1) air, 2) water, and 3) solids. However,
it is also important to recognize the existence of a fourth phase, namely, that of the airwater interface or contractile skin (Fredlund & Rahardjo 1993). This interaction caused by
surface tension, generates capillary stresses, or matric suction. Matric suction or capillary
stresses are predominantly dependant on pore size, void ratio and fabric in a given soil;
they can considerably contribute to the increase of mechanical properties (i.e. shear and
compressive strengths) of the soil (Heitor et al. 2012). The correlation between void ratio
and shear strength (which is defined by the soil cohesion and internal angle of friction) in
an unsaturated state is more complex and is thought to relate to the development of soil
suction resulting from the expulsion of air during compaction. The extended MohrCoulomb failure criterion for unsaturated soils concept in Equation (2.4), is used to define
the shear strength of unsaturated soils (Fredlund & Rahardjo 1993; Khalili & Khabbaz
1998).
= c + tan + ( ) tan

(2.4)

where, c is the effective cohesion intercept of the extended Mohr-Coulomb failure


envelope on the shear stress axis where the net normal stress and the matric suction at
failure are equal to zero, is the net normal stress state on the failure plane at

failure, is the angle of internal friction related to the net normal stress state variable,
( )

is the matric suction on the failure plane at failure, and b is the angle indicating

the rate of increase in shear strength relative to the matric suction. The extended Mohr-

Coulomb failure envelope for an unsaturated soil is shown in Figure 2.5.


Comparing shear strength Equations (2.3) and (2.4), it is observed that the equation for
unsaturated soils is an extension of the equation for saturated soils. The equation for
saturated soils has only one stress state variable, i.e. effective normal stress,

whereas the equation for unsaturated soils has two stress state variables, with one being for
the effect of matric suction, ( ) . As mentioned above, compacted soil remains in an

unsaturated state in practice, and since achieving a high shear strength value is a
fundamental attribute sought after in compacted ground, it is essential to also understand
the influence of matric suction. The review suggests that additional research is needed to
clarify the means for measuring the effects of compaction on the matric suction, and that
this cannot be elucidated by determination of the dry density and moisture content alone.
- 18 -

Chapter Two: Literature Review

Figure 2.5: Extended Mohr-Coulomb failure envelope for unsaturated soils


(Fredlund & Rahardjo 1993).
A third major objective of compaction is to achieve reduction in permeability of the
ground. It may be inferred from Equation (2.1) that the consequence of achieving MDD,
or close to MDD, of a soil during compaction is also to achieve a minimum or close to
minimum void ratio of the soil for a given compactive effort. It is well known that the
coefficient of permeability, for a given soil, is a function of void ratio (e.g. Huang et al.
1998), where minimum permeability is achieved at or slightly above the OMC (generally
when the void ratio is at its lowest) (e.g. Bell 2004; Lay 2009). Experimental evidence,
similar to the results shown in Figure 2.6 by Mesri and Olson (1971) shows that the
permeability of the soil decreases with decreasing void ratio (e.g. Terzaghi et al. 1996;
Churchman et al. 2002). This results in a lower tendency for the compacted soil to absorb
water, and thus reduces the amount of ground seepage.
It is argued that even though the empirical relationships existing between dry density (i.e.
void ratio) and soil stiffness, dry density (i.e. void ratio) and permeability, and dry density
(i.e. void ratio) and internal angle of friction as discussed above are quite clearly evident,
these are not sufficient to reflect the complex changing geotechnical properties accruing
from compaction effort. The complex interactions of water and air at the soil interface
leading to matric suction, which plays a significant role especially in influencing the shear
strength of soil, have not been properly accounted for. It would seem that this cannot be
accounted for in a soil by merely measuring the dry density and moisture content. This
point illustrates the shortcoming, and quite a serious one at that, for even a technique
considered as the gold standard.
- 19 -

Chapter Two: Literature Review

Figure 2.6: Correlation between void ratio and permeability (Mesri & Olson 1971).

2.1.1 Measurement of Dry Density (d) and Moisture


Content
The practice of determining the dry density and moisture content of compacted ground has
remained largely intact. Some commonly used methods for the determination of dry
density and moisture content, both in the laboratory and field are given below:

Proctor Test
In the laboratory, the standard Proctor (1933) and modified Proctor tests (as well as other
variations of the Proctor test) have been designed to determine the maximum dry density
and the optimum moisture content of a soil to be used for controlling field compaction.
The original standard Proctor tests as well as the newer variations, determine the
relationship between the moisture content and dry density of the soil. Generally speaking,
the relationship between these two parameters produces a bell shaped curve, where the
peak of the curve normally indicates the standard maximum dry density (SMDD) and
optimum moisture content (OMC) values for the tested soil.
The standard Proctor test, as defined by Standards Australia in AS 1289.5.1.1 (2003a), relies
on a compactive effort of 596 kJ/m3 using a 2.7 kg free falling steel rammer whereas the
modified Proctor, as defined by Standards Australia in AS 1289.5.2.1 (2003b), relies on a
greater compactive effort of 2703 kJ/m3 using a 4.9 kg free falling steel rammer. The
- 20 -

Chapter Two: Literature Review


modified Proctor test was designed to better mimic the actual greater compactive efforts of
current day compaction methods. Soil is placed in either a 105 or 152 mm diameter
(internal diameter) metal cylindrical moulds with effective heights of 115.5 or 132.5 mm
respectively, and thus producing a nominal mould volume of 1000 and 2400 cm3,
respectively. The choice of mould size depends on the particle size of the tested material.
Compaction of the soil is conducted over a range of moisture contents to determine the
maximum possible mass of dry soil per unit volume using the standard/modified/or other
variation Proctor compactive effort.

Sand Cone Replacement Method


As defined by Standards Australia in AS 1289.5.3.1 (2004a), the sand cone replacement (or
sand replacement) method is commonly used to determine dry density of fine-grained and
medium-grained soil in the field. However, this method relies on slow and strenuous effort
to physically excavate the sampled soil and due to the testing method, is generally only used
for the very near surface (or at the surface). The equipment required for the sand cone
replacement method is shown in Figure 2.7.

Figure 2.7: Sand cone replacement equipment.


The sand cone replacement method involves: (1) excavating a section of the ground in a
small cylindrical shape (a 200 mm diameter hole is a common size, or 150 mm diameter
hole for fine-grained soil), (2) calculating the mass of the portion of soil removed from the
excavated hole, (3) refilling the excavated hole with a sand-cone apparatus using sand of
- 21 -

Chapter Two: Literature Review


known mass and density, (4) determining the excavated hole and extracted soil volume by
knowledge of the amount of sand required to fill the excavated hole, (5) combining the
results from the above steps will allow to estimate the gross soil mass per unit volume (i.e.
bulk density), (6) estimating the dry density of the soil by correcting for the moisture
content of the excavated soil, which can be estimated using laboratory or other methods.

Nuclear Density Gauge Method


As suggested by Standards Australia in AS 1289.5.8.1 (2007), the nuclear density gauge
method can determine the field dry density and moisture content. This is made possible by
measuring both the gross mass per unit volume (i.e. field bulk density) and the mass of
water per unit volume (i.e. field moisture content) of the sampled soil. The main concept of
this method is that it relies on the emission of high energy electro-magnetic gamma
radiation and subsequent detection of the attenuated radiation. The emitted gamma
radiation is absorbed and attenuated by the tested soil at a rate depending on the bulk
density of the sampled soil. The lower the amount of gamma radiation detected by the
gauge, the greater the bulk density of the sampled soil. The addition of neutron sources in
the nuclear gauge allows for the estimation of the sampled soil moisture content, since
neutron sources are attenuated by the presence of hydrogen atoms. Hence, the ability of
the nuclear gauge to estimate the bulk density and moisture content will allow inferring the
dry density of the sampled soil. A typical nuclear density gauge is shown in Figure 2.8.

Figure 2.8: Nuclear density Gauge.


- 22 -

Chapter Two: Literature Review


The nuclear gauge is arguably simpler and faster to employ in the field than most other
density estimating techniques, particularly for shallow layers and thin lifts (for roller
compaction). Since the gauge is placed on the ground surface and a probe holding nuclear
source is lowered into the ground, the common depth range of nuclear gauges is perhaps
between 0.3 0.5 m. However, this method has the capacity to expose operators and other
people close by to nuclear radiation.

2.2 Resistance of Compacted Ground


The second category of methods for assessing the quality of ground compaction is by
measuring the resistance of the compacted ground against a penetrating device. Although
these techniques are regarded as invasive, they are widely accepted in the geotechnical
community for site investigation work. The information obtained from a penetration-based
technique is localised at the point where the test has been undertaken, and such a technique
is usually applied to assess thick fills (e.g. in dynamically compacted ground).
It is ostensibly evident that the resistance encountered by a penetrating device should
reflect the strength of the compacted ground in some sense. This is because the device
must break, that is overcome, the strength of the ground in its immediate vicinity in order
for it to advance any further. Secondly, it is also entirely plausible that the stiffness of a
ground would be related to the resistance of a penetrating device, this being based on the
common observation that a hard (stiff) material is more difficult to penetrate than a soft
(compressible) material. Thirdly, it is intuitively plausible to reason that a dense ground
(implying low void ratio and low permeability) would result in higher penetration resistance
than a loose ground (implying high void ratio and high permeability). On account of the
above philosophical arguments, one should generally expect good correlations between the
resistance encountered by a penetrating device, and the shear strength, stiffness and
permeability of the ground. Indeed, as the review below shows, a number of correlations
have been developed especially between the penetration resistance and shear strength
parameters, and to a lesser extent between the penetration resistance and ground stiffness.
Three of the most common invasive techniques are briefly reviewed below as a vast
amount of literature is already available in public domain.

- 23 -

Chapter Two: Literature Review

Standard Penetration Test (SPT)


The very common geotechnical SPT determines the resistance of soils to the penetration of
a sampler, driven into the ground by a free-falling weight. Figure 2.9 shows typical
equipment required for conducting SPT measurements. As defined by Standards Australia
in AS 1289.6.3.1 (2004b), the SPT relies on a 63.5 kg drop hammer in a 760 mm free-fall to
drive a split spoon sampler into the soil, blow by blow. The split spoon allows for the
recovery of the disturbed soil, for further laboratory testing. The sampler is driven a total
450 mm into the ground, with the number of blows counted for every successive 150 mm
of penetration. The first 150 mm of the samplers penetration is known as the seating
drive. The blow count for the second (150 300 mm) and third (300 450 mm) 150 mm
of penetration are summed, to provide the penetration resistance (N).

Figure 2.9: Track mounted drill rig with SPT capabilities (Jeffrey and Katauskas Pty
Ltd 2012).
The SPT-N value which is measured at large strains is generally correlated to other
geotechnical properties, such as: angle of internal friction (e.g. Peck et al. 1974; Hatanaka &
Uchida 1996), bulk density (e.g. Meyerhof 1957; Mullins 2006) and shear wave velocity (e.g.
Sykora & Koester 1988; Wride et al. 2000; Hasancebi & Ulusay 2007; Hanumantharao &
Ramana 2008; Harutoonian et al. 2012b). The measured SPT-N values at a compacted
- 24 -

Chapter Two: Literature Review


ground allow for an appreciation of the quality of compaction with depth in the relative
sense, but once the correlations are applied, the geotechnical properties of the compacted
layers may be inferred. The use of this technique for assessing ground compaction has been
reported widely in literature (e.g. Wride et al. 2000; Karray et al. 2011).
SPTs are applied at different depths and are generally following behind borehole drilling,
and are terminated when impenetrable rock is encountered. They are often carried out on
truck and track mounted rigs, due to the size and weight of the SPT apparatus. Thus site
access and transportation of the rigs and equipment can limit testing locations. Similar to
borehole, SPTs provide stiffness information at localised spots and would be quite costly
and time consuming to thoroughly cover a large and deep compacted site.

Cone Penetration Test (CPT)


The CPT differs from the above mentioned SPT, where instead of a drop hammer driving
a sampler blow by blow, the CPT involves continuously pushing a cone tip at a constant
rate to reflect soil properties at large strains. This technique has the capability to evaluate
soil compaction to greater depths than just the first metre of the subsurface (e.g. by means
of density tests). To reach great depths, cylindrical steel rods are interconnected to the cone
tip, where it is pushed into the ground until encountering impenetrable rock. Upon pushing
the cone tip into the ground, the resistance against the cone tip, that is, the cone resistance,
qc, (often expressed in unit of MPa) is measured. Also, the frictional resistance acting against
the rods outer sleeve surface, that is, the sleeve friction, Fs, (also expressed in MPa) is
measured.
Similar to the SPT, it is common for CPT systems to be mounted on track and trucks due
to the size and weight of the CPT apparatus. Regular updates see the addition of sensory
equipment on the CPT including devices for: electrical resistivity, seismic wave detection,
radioactive source detection. The addition of this equipment combines both geotechnical
and geophysical tests, and thus allows the gathering of more valuable soil data. Figure 2.10
shows the seismic cone configuration which will allow estimation of the Vs, qc and Fs. The
CPT has similar advantages to the SPT, and also has similar disadvantages. The advantages
include the ability to collect in situ accurate continuous information of the stiffness of the
tested soil to great depths, sometimes in excess of 100 m (in soft ground). However, the
- 25 -

Chapter Two: Literature Review


collected data is localised to the tested spot thus it would be quite costly and time
consuming to cover extensive and deep sites. The requirement of track or truck mounted
rigs to carry the CPT apparatus could hinder portability and its ability to enter poorly
accessible sites. Furthermore, in some cases where hard ground is encountered the CPT
requires pre-drilling the test penetration location to prevent damage to the equipment
(further increasing the cost and time required for testing).
Over the years, a substantial number of empirical relationships have been developed to
ascertain in situ soil properties including soil modulus (E and G0), peak friction angle (p),
constrained modulus (DMT M), etc from CPT qc data (e.g. Robertson 2009 and
references therein). Normalised (see Section 3.4) Vs and normalised qc correlations have
also been previously established by a few investigators (e.g. Robertson et al. 1992a; Fear &
Robertson 1995; Wride et al. 2000; Karray et al. 2011). The use of the CPT method for
assessing ground compaction has been reported widely in literature (e.g. Schmertmann et
al. 1986; Bo et al. 2009; Waddell et al. 2010; Karray et al. 2011).

Oscilloscope
Trigger
Static
Load

Hammer

Shear Wave source


(perpendicular to cone)

Shear Wave

Figure 2.10: Seismic cone configuration (CPT attachment) (After Brouwer 2008).

Dilatometer Test (DMT)


Pioneered and developed by Marchetti (1980), the DMT has since grown to become a
regular in situ soil investigation technique. The DMT is somewhat similar to the CPT,
- 26 -

Chapter Two: Literature Review


where it relies on forced penetration into the ground and in turn measurement of the soil
resistance. The DMT however, relies on the ground penetration to access the subsurface
from where the DMT begins testing. Once the DMTs flat stainless steel blade tip is
penetrated to the desired depth, a small circular disc shaped membrane is pneumatically
inflated. The resisting soil pressures against the inflated membrane are then calculated and
analysed to provide important soil properties, such as: vertical drained constrained modulus
(M), undrained shear strength (cu), material index (ID), etc. In much the same way as the
SPT and CPT described above, the DMT has similar advantages and disadvantages. Even
though the DMT can provide accurate and reliable information, the data is localised to the
tested spot. This limits the number of DMTs (including SPTs and CPTs) conducted at
sites, particularly due to the availability of funds and time. Therefore vital soil data from
DMTs (including SPTs and CPTs) are extrapolated to cover the remaining area of the site.
This could create massive problems, particularly for compaction sites with deep fills (such
as the areas covered in Chapters 7 (DCPT-9) and 8 (DCPT-7) of this thesis), as
problematic or poorly compacted areas could be missed by the DMT (including SPT and
CPT).
Totani et al. (2001) provide numerous formulae to deduce vital geotechnical properties
when employing the DMT, such as: material index (ID), undrained shear strength (cu),
friction angle (), coefficient of consolidation (ch), coefficient of permeability (kh) and
vertical drained constrained modulus (M). The use of this technique for assessing ground
compaction has been reported in literature (e.g. Schmertmann et al. 1986; Bo et al. 2012).

2.3 Shear Wave Velocity (Vs) of Compacted Ground


The shear wave velocity (Vs) was more recently applied to assess soil compaction, initially
based on experimental evidence that the Vs achieved good correlations to the dry density
of the soil (e.g. Kim & Park 1999; Kim et al. 2001; Karray et al. 2011; Heitor et al. 2012).
Heitor et al. (2012) recently presented a correlation (Figure 2.11) between Vs and void ratio
based on laboratory testing of soils from the DCPT-9 area within the Penrith Lakes quarry
(see Figure 6.1). Their correlation in Figure 2.11, shows a linear inverse relationship
between Vs and void ratio, where the lower the void ratio (higher dry density) the higher
the Vs. Similarly, higher void ratio (lower dry density) leads to lower Vs values.

- 27 -

Chapter Two: Literature Review

Figure 2.11: Shear wave velocity correlated with void ratio (Heitor et al. 2012).
Figure 2.12 shows four correlations from four different areas studied by Cha and Cho
(2007) and they also found a linear inverse relationship between void ratio and Vs, that is, a
high void ratio leads to low Vs values, and low void ratio leads to high Vs values. However,
they also showed the influence of confining pressure due to overburden stress on estimated
Vs values measured at the same void ratio. This effect needs to be recognised in order to
produce true Vs estimates, that is, without the effect of confining pressure of the
compacted ground (see, Section 3.4).

Figure 2.12: Shear wave velocity correlated with void ratio (Cha & Cho 2007).
- 28 -

Chapter Two: Literature Review


In other geotechnical investigations mostly not related to soil compaction, Vs has also been
empirically correlated to the invasive SPT and CPT measurements (e.g. Table 2.1). These
findings suggest similar correlations could also be applied to soil compaction but some care
and caution are necessary. This is because the invasive SPT and CPT are large strain
measurements whereas the Vs is usually obtained from small strain dynamic tests.
Table 2.1: Vs correlations with SPT and CPT measurements.
Correlations between Vs and SPT-N
Source
Equation
0.43
Vs = 63N
Sykora and Koester (1988)
Wride et al. (2000)

0.25
(74.8 X 102.1)
Vsn = X (N 1 ) 60

Hasancebi et al. (2007)

Vs = 90N 0.31

Fear and Robertson (1995)

Vsn = 135q cn0.23

Hanumantharao and Ramana (2008) Vs = 82.6 N 0.43


Correlation between normalised Vs (Vsn) and normalised q c (q cn)
Source
Equation
0.23
Vsn = 102q cn
Robertson et al. (1992b)
Wride et al. (2000)
Karray et al. (2011)

0.25
(95.6 Y 110.8) (0.16 D50 0.25)
Vsn = Yqcn

Vsn = 125.5q cn0.25 D500.115

Note: Vs = Shear wave velocity (m/s), Vsn = normalised Vs (m/s), N = Correlated SPT blow count,
(N1)60 = blow count corrected for the effective stress and energy level used in the SPT, qcn = normalised
cone end resistance (qc) (kPa), D50 = mean grain size (mm).

As mentioned above, the small strain measurements of the Vs require discretion on its use
but it should not be a rejection of its appeal. In many cases, the very small strain dynamic
properties of soils are sufficient, since there are often circumstances in seismology and soil
dynamics where the assumption of linearity is an acceptable approximation. In geotechnical
engineering, low strain dynamic properties are commonly used to characterise the dynamic
response of soils at very small strain levels (Lai & Rix 1998). In other words, if an
assessment of the compacted ground is required as a vital part of investigating the grounds
seismic response then it is evident that measurements of the Vs using the small strain
techniques are eminently suitable. Today, various small strain techniques are being applied
to obtain the depth-averaged Vs in the first 30 m of a site known widely as Vs30. This is
then commonly applied for site specific risk analysis to mitigate against soil liquefaction,
landslide and rock fall hazards, building damages resulting from seismic activities (e.g.
Cotton et al. 2006; Gallipoli & Mucciarelli 2009; Castellaro & Mulargia 2009a). However, it
is also increasingly recognized that the Vs profile could potentially reveal valuable
- 29 -

Chapter Two: Literature Review


information on the stiffness and associated geotechnical properties at the near surface (Lai
& Rix 1998; Xia et al. 1999). This may be inferred in part from the theoretical definition of
Vs of a soil, which is given by:

(2.5)

where, G is the shear modulus. As shown in Equation (2.5), the shear wave velocity is a
geotechnical property that is theoretically related to the modulus of a soil. This relationship
at once supports the notion that it is theoretically justifiable in using the Vs to measure the
modulus (or stiffness) of a soil and it may be argued that it serves as a good indicator of
how well the modulus of the soil may be improved with compaction effort.
However, it is common for shear strength (an important geotechnical property) to be
associated with large strain phenomena, whereas as mentioned above, Vs measurements are
inferred from small strain techniques. It may lack the physical justification to produce a
direct relationship between shear strength (large strain) and Vs (at very small strains).
Nevertheless, effective stress (or confining pressures) and void ratio, both of which are
important parameters for the estimation of shear strength, have been shown by several
investigators (as discussed above) to be directly related to the Vs. Thus it could be argued
that shear strength is indirectly but intrinsically related to Vs through the effective stress (or
confining pressure) and void ratio of the ground (e.g. Cha & Cho 2007).
Direct use of Vs in geotechnical applications is seen in the work of Tezcan and his coworkers (Tezcan et al. 2006) who have developed bearing capacity equations for shallow
foundations that are directly related to the Vs as opposed to the shear strength of the soil.
Their work reinforces the notion that developing the Vs as a means to assess ground
compaction has the side benefit of allowing a direct estimation of the soil stiffness and
bearing capacity (qa) as a function of compaction effort. Blake and Gilbert (1997) have
developed a relationship between Vs and the undrained shear strength (Cu) of normally
consolidated clays. Their work allows the estimation of a highly sought after geotechnical
property for foundation design, Cu, via the measurement of Vs. Cha and Cho (2007) have
also developed relationships between Vs and shear strength for sandy soils through
experimental tests by using Vs void ratio shear strength correlation. Other studies have
- 30 -

Chapter Two: Literature Review


been successful in correlating Vs with geotechnical properties, particularly with SPT-N
values (e.g. Table 2.1 and (Wei et al. 1996; Inazaki 2006; Hasancebi & Ulusay 2007 and
references therein)), as well as Vs and CPT values (e.g. Table 2.1 and (Karray et al. 2011
and references therein)).
More recently, Claria Jr. and Rinaldi (2007) and Heitor et al. (2012) have shown in their
experimental studies that, in general, the propagation of Vs in compacted soil is governed
by the soil skeleton (therefore related to its particle arrangement or void ratio), applied
confining pressure and suction stresses. The influence of matric suction should not be
ignored since, as explained above, compacted soil is most commonly unsaturated, and
increased matric suction will result in increased shear strength. Heitor et al. (2012) showed
in Figure 2.13, the relationship between matric suction and the Vs of compacted soil with a
constant void ratio and a moisture content of 16.5 % (on the wet side of the Proctor
compaction curve) before the drying process. There is an obvious relationship in Figure
2.13, where an increased matric suction leads to an increase in Vs. An increase of matric
suction contributes to the strengthening of the soil skeleton by increasing interparticle
contact stresses, and thus increases the Vs. However, as the moisture content increases, it
causes the capillary forces to weaken, concurrently reducing the matric suction stresses
(Claria Jr. & Rinaldi 2007). Hence this increase in moisture content, particularly once
exceeding the OMC, affects the soil structure and reduces the dry density of the soil, which
subsequently reduces the Vs. Moreover, the importance of matric suction on measured Vs
is highlighted by the fact that it plays a major role on the end value of the Vs value.

Figure 2.13: Effect of matric suction on shear wave velocity (Heitor et al. 2012).

- 31 -

Chapter Two: Literature Review

2.3.1 Measurement of Shear Wave Velocity (Vs) by


Geophysical Methods
It is thought that one of the reasons why Vs have not been used more often in the past is
because it has been rather difficult and costly to determine this property using the more
classical geophysical techniques (e.g. crosshole/downhole, P-S logging, etc). A new
generation of geophysical techniques such as CSWS (Menzies & Matthews 1996), SASW
(Nazarian & Stokoe 1984), MASW (Park et al. 1999), ReMi (Louie 2001), HVSR
(Nakamura 1989) have improved the practicality and ease of measuring the Vs of a soil in
the field. Appendix A.1 briefly summarises the background of the Crosshole, SASW,
MASW, CSWS and ReMi techniques, with the seismic CPT briefly summarised in Section
2.2. Table 2.2 shows a summary of the fundamental differences between the most
commonly used non-invasive surface techniques and the HVSR technique.
Table 2.2: Characteristics of Surface Wave Techniques.
Feature

SASW

CSWS

ReMi

MASW

HVSR

Noise Source

Active

Active

Passive

Active

Passive

Sensor
Configuration

1D array

1D array

1D array

1D array

Sensor(s)
Components

Single
(V or H)

Single
(V or H)

Single
(V or H)

Single
(V or H)

Measured Data

Dispersion
curve

Dispersion
curve

Dispersion
curve

Dispersion
curve

Single
station
Three
(NS, EW
and Z)
HVSR
curve

Note: V = vertical, H = horizontal, NS = North-South, EW = East-West and Z = vertical.

A limited number of investigations have attempted to study the Vs profile of the ground
before and after compaction using the new generation of CSWS, SASW, MASW and
ReMi techniques (e.g. Kim & Park 1999; Kim et al. 2001; Moxhay et al. 2001; Park & Miller
2004; Pullammanappallil 2006; Feng et al. 2010). However, none of these studies have
developed the use of the Vs as a compaction identifier in the field, where a rigorous
identification of the Vs within a practical range of compaction effort is required.
Consequently, the development of the Vs as a compaction identifier, if successful, will
enhance the tools and techniques available for compaction control. Previous geophysical
techniques have relied on the estimation of the Vs geotechnical property for compaction
evaluation, whereas one of the techniques proposed in this thesis allows for an initial
- 32 -

Chapter Two: Literature Review


assessment based on only the measured HVSR curve. Furthermore, Table 2.2 shows that
the HVSR technique is theoretically, operationally and physically different to the previously
studied surface wave techniques for compaction evaluation.

2.4 Horizontal to Vertical Spectral Ratio (HVSR)


Technique
The HVSR techniques applied in this thesis are relatively unfamiliar to the geotechnical
community although occupying a prominent place among geophysicists and seismologists.
The HVSR technique relies on the measurement of the horizontal and vertical spectra of
the ambient vibrations (the microtremors and microseisms) which are ubiquitous in the
Earth surface. Microseisms and microtremors, strictly speaking, refer to ambient vibrations
attributed to the natural atmospheric phenomena (e.g. ocean current, wind, strong
meteorological conditions) and cultural sources (e.g. anthropic activities such as traffic,
construction and plant operations) respectively. However, both have been loosely lumped
as microtremor to the extent that the distinction between the two terms microtremors
and microseisms have often become blurred in the literature (Bonnefoy-Claudet et al.
2006a). The components of ambient vibrations linked to natural atmospheric sources
generally occur in the low frequency range (less than 1 Hz) while the cultural noises (traffic,
machinery etc) will generally include components of higher frequencies (greater than 1 Hz).
It is noted that induced tremors from these noises consist of very small vibrations, much
smaller than those experienced during a perceptible earthquake. SESAME (2004a) has
provided a comparison of some of the characteristics of microtremors and microseisms.
These comparisons are shown in Table 2.3, with a few additions.
Furthermore, Figure 1.3 illustrates the above explained origins of ambient vibrations using
a snapshot of everyday life. In this figure, several ambient vibration sources are observed
which can contribute to the measured HVSR curve. The low frequency noise sources in
this figure include: ocean waves, wind, etc, and the higher frequency noise sources include:
human activities, traffic, trains, activities in buildings (commercial, residential and
industrial), etc.

- 33 -

Chapter Two: Literature Review


Table 2.3: Summary of the nature of ambient vibrations (after SESAME 2004a).
Natural
Cultural
Type
Microseism
Microtremor
Frequency
< 1 Hz
> 1Hz
Origin
Ocean, wind
Traffic, industry, human activity
Incident wavefield
Surface waves
Surface and body
Amplitude
Related to oceanic
Day/Night, Week/Week-end
variability
storms
Comparable amplitude slight
Incident wavefield
Rayleigh / Love
indication that Love waves carry slightly
issue
predominantly Rayleigh
more energy
Possibility of higher modes at high
Fundamental /
Mainly Fundamental
Higher mode issue
frequencies (at least for 2-layer case)
Microtremor studies pioneered by Kanai and Tanaka (1954), Aki (1957) and Kanai (1957)
paved the way for development of present day techniques for inferring the dynamic site
characteristics and associated subsurface soil structure at an observation point. In the early
days, the techniques were developed for the primary purpose of identifying the
fundamental resonance frequency of a site at the observation point. Although Nogoshi and
Igarashi (1971) had been the first to apply the horizontal to vertical (H/V) spectra ratio in
their study of the composition of microtremors, the renewed interest in the horizontal to
vertical spectral ratio (HVSR) technique is attributed to Nakamura (1989), who proposed
that the fundamental resonance frequency be defined as the frequency of the predominant
peak (f0) of the ratio of the H/V Fourier amplitude spectra:
HVSR( ) =

FH ( )
FV ( )

(2.6)

where FV() denotes the vertical and FH() the horizontal Fourier amplitude spectrum.
Nakamura (1989, 2008) also postulated that the predominant peak amplitude of the HVSR
curve is caused by amplification of the horizontal tremor due to multiple reflections of the
S-wave (Appendix A.2 provides a brief background of seismic waves), while the vertical
component remains substantially unaffected, and that the peak amplitude is a consistent
estimate of the amplification factor of the site at the observation point. The first of
Nakamuras supposition that the fundamental resonance of a site, as defined by the
predominant peak of the HVSR curve is widely accepted on the basis of both empirical
data and theoretical simulation (see, e.g. Bonnefoy-Claudet et al. 2006b and references
therein). The second supposition of Nakamura as embodied in amplification of the
- 34 -

Chapter Two: Literature Review


horizontal S-wave (a body wave) does not receive the same consensus of agreement. The
opposing view is that the HVSR curve is mainly constituted by fundamental and higher
modes of surface waves (both Rayleigh and Love waves, and the proportion varies with
frequency and location of sensor). The current application of the HVSR technique is
mostly guided by the surface wave supposition (e.g. Arai & Tokimatsu 2004, 2005; Lunedei
& Albarello 2009, 2010), and vary somewhat from earlier works attributing the appearance
of the HVSR curve to the ellipticity of the fundamental mode of the Rayleigh wave (Lachet
& Bard 1994; Kudo 1995; Bard 1998).
It is worth pointing out at this stage, some of the differences between the HVSR technique
and other surface wave array-based techniques that may rely on the measurement of the
passive ambient or active noise mentioned previously: (1) The former would require a
sensor to measure the three components of vibrations (two horizontal and one vertical) at
the observation points whereas the latter may not (measurement of the vertical or one
horizontal component will suffice), (2) the former is a single-station method meaning that
the sensor needs only to be placed at the observation point, but the latter is an array-based
method which requires a series of interconnecting sensing geophones arranged in a
designated configuration. The advantage of the HVSR technique is articulated in point 2
with its superior set up time and effort relative to the other surface wave techniques.
However, the HVSR technique can provide only a relative stratigraphy. To have an
absolute stratigraphy it requires calibration or a constraint (Castellaro & Mulargia 2009a;
Harutoonian et al. 2012a). Figure 2.14 shows the general schematic of the operationally
simple HVSR technique.

TrominoTM

Analysis

Velocimeter
PC/Laptop
Microtremor
recordings in
North-South,
East-West and
Up-Down (Z)
directions
Figure 2.14: Horizontal to Vertical Spectral Ratio (HVSR) method.
NS
EW
UD

- 35 -

Chapter Two: Literature Review


Figure 2.15 shows the simplicity and equipment used to carry out site investigations when
using the HVSR technique. The device pictured is a TrominoTM velocimeter, a portable
high resolution instrument used in this study.

Figure 2.15: TrominoTM velocimeter used for the H/V spectral analysis (HVSR).
The work herein is new in respect to geotechnical site investigation and follows recent
studies on determining the dynamic characteristics of a site based on the inversion of the
measured HVSR curves of ambient vibrations or microtremors (e.g. Fah et al. 2003; Arai &
Tokimatsu 2004, 2005; Parolai et al. 2005; Picozzi & Albarello 2007; Herak 2008; Lunedei
& Albarello 2009, 2010). The foundational basis of this method rests on empirical
experimental evidence and theoretical modelling results suggesting that the shape of the
H/V (Horizontal/Vertical) spectral ratio curve depends mainly on the local geology (or
layering of subsurface) of a site (e.g. Nakamura 1989; Fah et al. 2003; SESAME 2004b;
Lunedei & Albarello 2010; Harutoonian et al. 2012a). The back analysis procedure (also
known as forward modelling) of matching the synthetic (theoretical) HVSR curve
generated from a theoretical model to the measured HVSR curve allows inferring
information such as the Vs and thicknesses of the soil layers of the ground. The next
section will discuss highlights in the development of the HVSR theoretical model.

2.4.1 Survey of Methods Used for the Theoretical


Modelling of HVSR Curves
Theoretical modelling of the HVSR curve was initially studied to interpret and understand
the composition and characteristics of microtremors. Since then many researchers have
- 36 -

Chapter Two: Literature Review


extended the theoretical modelling to allow the estimation of the Vs profile of a site. The
degree of accuracy obtained by theoretical modelling would be based on the assumptions
and the simplications made in the development of the theoretical model (Foti et al. 2011).
As first suggested by Nakamura (1989), the horizontal to vertical spectral ratio (HVSR) of
the surface microtremors reflects the transfer function of the surface layers. Also
suggesting that the HVSR curve possesses information on the resonance characteristics of
the site which is why geophysicists have since used the HVSR technique as a cost-effective
tool for microzonation studies. In recent times, the paper by Nakamura (1989) has
attracted much attention by investigators to model the HVSR curve which allows for the
possibility of estimating the layered soil substructure by means of the so called HVSR
inversion. This process involves, undertaking a Fast Fourier Transform (FFT) on the
measured raw HVSR data to produce the measured HVSR curve and then inverting the
measured HVSR measurements (curve) to produce the Vs profile using techniques of signal
processing, and minimisation of residual error between the theoretical curves and measured
values. The end result of the HVSR inversion is to obtain the Vs profile of the layered soil
substructure. A number of non-linear inversion methods have been proposed in the past to
estimate the Vs profile of a site by using the dispersion curves of Rayleigh waves, however,
these methods may not be directly utilised for the HVSR curve because its frequency
variation is more complicated than of the dispersion curve (Arai & Tokimatsu 2004).
Lachet and Bard (1994) began to model noise (or ambient vibrations) to understand the
possibilities and limitations of the HVSR technique. Figure 2.16 shows the model
envisaged by them to simulate noise by introducing a circular source area containing
uniformly disposed surface sources with random amplitudes and time delays around a
central receiver. They believed that surface sources should be used to simulate urban noise,
hence suggesting that urban noise consists of surface waves in the form of Rayleigh and
Love waves. They realised the source type (e.g. explosion, crack, unidirectional forces) and
source shape (e.g. dirac, step, ricker, triangle) need to be considered when simulating the
nature of urban noise. Surface sources at a depth of 2 m were used to represent noise
measured in urban areas. These sources were either unidirectional forces or dipoles with
step, Ricker and pseudo-Dirac source shapes, where the signal for each source was
computed using the discrete wave number technique. The model contained 240 or 480
sources with a radius ranging from 100 m to 1 km. This theoretical model showed that the
H/V ratio was strongly correlated to the local geology and not with source excitation
- 37 -

Chapter Two: Literature Review


function because the H/V ratio obtained was constant regardless of the source type and
source function. The influence of the source location and receivers in the theoretical model
was also studied by Lachet and Bard (1994). They suggested that source-receiver distance
can affect the amplitude of the H/V ratio peak by adjusting the source depth, structure
characteristic thickness and maximum source-receiver distance. By creating a larger sourcereceiver distance, it is likely that the Rayleigh wave contribution increases, because as
explained above it was believed that the fundamental peak of the Rayleigh wave
polarisation curve generally corresponds to the peak of the H/V ratio. Lachet and Bard
(1994) also studied the effect of Poissons ratio or Vp/Vs ratio of the sedimentary layer.
They realised that manipulating Poissons ratio effects the amplitude of the H/V peak.
Therefore, since the amplitude of the H/V peak can be influenced by Poissons ratio and
the source-receiver distance, it should be interpreted carefully.

Figure 2.16: Source model (after Lachet & Bard 1994).


In the same year, Yamanaka et al. (1994) relied on microseisms (long-period microtremors)
instead of microtremors (see Table 2.3) to estimate the subsurface characteristics using trial
and error to fit the theoretical fundamental mode Rayleigh wave ellipticities with the
measured ellipticities.
Fah et al. (2001, 2003) used the mode summation method and a time domain finite
difference technique to study the H/V ratio of microtremors. The first method requires
well defined sources and source-receiver depths to investigate the H/V ratio, where in the
latter sources are not treated separately and H/V ratios are computed from the time- 38 -

Chapter Two: Literature Review


domain signals. Fah et al. (2001) first applied the finite difference technique to simulate the
propagation of P-SV waves in horizontally layered 2D models. The model involved
creating randomly generated sources. A number of finite difference models were then
combined so that either all the sources are located in the same direction from the receiver,
or they are randomly distributed around the receiver. The mode summation technique was
then applied which is based on the solution of elastic wave propagation in multilayered
media in a matrix formulation. Their theoretical model consists of randomly assigned
source depths with a large number of sources scattered around a receiver. Their model also
relied on a genetic algorithm due to the non-linear nature of the problem to fit the
theoretical fundamental mode Rayleigh wave ellipticity to the measured, in order to
estimate the Vs profile of the ground. They explain that genetic algorithms are generally
robust and can easily adapt to specific problems, also they do not need explicit starting
models. However, they suggest that borehole data should be used to calibrate the
modelling process. When there is no information to constrain the thicknesses of sediments
the bedrock Vs and layer thickness should be able to be estimated to within a reasonable
range in order to continue with the inversion. Fah et al. (2001) suggest that the amplitude
around the resonance frequency (f0) can be affected by the Vs contrast (or impedance
contrast) between the soft layers above the bedrock and the bedrock, where the larger the
Vs contrast the larger the H/V ratio amplitude.
Arai and Tokimatsu (2004) applied the theoretical solutions proposed by Harkrider (1964)
for surface waves in multi-layered elastic media. Harkrider (1964) used the momentum
equation for homogeneous media to develop solutions of wave propagation through elastic
media. He introduced the matrix formulations to derive integral expressions for time
transformed displacement fields produced by simple sources at any depth in a multi-layered
elastic isotropic solid half-space. These integrals are evaluated to obtain surface wave
displacements in the frequency domain. Arai and Tokimatsu (2004) utilised Harkriders
solutions by assuming that the ambient vibrations are dominated by surface waves
(fundamental and higher mode Rayleigh and Love) to simulate the horizontal to vertical
(H/V) ratios of microtremors, and through inverse analysis estimate the Vs profile of a site.
To model the ambient vibrations in the ground to create the theoretical HVSR curve, it is
believed that the noise sources are randomly distributed with independent harmonic
(Fourier-time-transformed vertical and horizontal) point forces with an angular frequency
, LV() and LH() on the ground surface (e.g. Lachet & Bard 1994), at distances greater
- 39 -

Chapter Two: Literature Review


than one wavelength from the observation point (origin) (Arai & Tokimatsu 2004). Surface
and body waves are produced from each source and propagate through the medium,
however, body waves attenuate more rapidly with distance than surface waves (McKenna
et al. 2008). It is therefore the Rayleigh and Love waves that dominate at distances greater
than one wavelength from the source. A layered soil model is supposed in generating the
theoretical curve which is required for the inversion of the HVSR curve to produce the
estimated soil profile. The subsoil structure is assumed to consist of parallel, homogeneous
and isotropic compacted layers bounded below by a semi-infinite elastic medium. As a
first-order approximation it is commonly assumed that the Earth is isotropic because the
equations required for anisotropic media are more complicated. The generalised leastsquares cost function is shown in (2.7) (Arai & Tokimatsu 2004). The inversion process
involved minimising the cost function which is the cumulative sum of the squares of the
normalised errors between the observed H/V spectral ratios and theoretical H/V spectrum
of surface waves via the Marquardt least square algorithm (Marquardt 1963) for a soil layer
model.

( ) ( )
( )

H
H

1 V m V t
F=
min
H
I i =1

V m

(2.7)

where, (H/V)m = Measured H/V spectral ratio of microtremors, (H/V)t = Theoretical


H/V spectrum of surface waves and I = Frequency of microtremor. As the inversion
process dealt with a highly non-linear problem several iterations were needed in order to
estimate the soil parameters. At the start an initial guess is made of the values of the soil
thickness and/or shear wave velocity of the soil profile. At the end of the final iteration,
the errors between the observed and theoretical (synthetic) values were minimised and the
final values of the soil parameters giving the estimated soil properties.
Arai and Tokimatsu (2005) extended the HVSR inversion to a joint inversion (also
attempted by Parolai et al. 2005) with dispersion data. They proposed the joint inversion
using both HVSR and dispersion (obtained from array based techniques) data. The purpose
of the joint inversion is to utilise the advantages of both methods. They explain that
dispersion curve phase velocity is difficult to determine at low frequencies without the use
of large arrays. Therefore, HVSR technique could be used to assist the dispersion curve
method in resolving information at low frequencies. Also, dispersion data can be used to
- 40 -

Chapter Two: Literature Review


help constrain the number of possible solutions to the inversion of HVSR data. The results
produced by them suggest that more reasonable and accurate Vs profile can be estimated
when using joint inversion. They assumed the dispersion data to be the Rayleigh wave
dispersion curve and the H/V spectrum to consist of surface waves in the form of both
Rayleigh and Love waves, also taking into account the effects of the fundamental and
higher modes. The extended generalised non-linear least squares equation for the inversion
process below takes into consideration both surface wave techniques.
F = FR + FHV

w2
= R
IR

C mi C Ri

C mi
i =1
IR

w2
+ HV
I HV

(H V )mi (H V )Si

(H V )mi
i =1

I HV

min

(2.8)

The variables Cm and (H/V)m represent the measured phase velocities of the vertical
motion of microtremors and H/V spectral ratios of microtremors respectively. CRi and
(H/V)Si represent the synthetic (theoretical) phase velocity of Rayleigh waves and the
synthetic (theoretical) H/V spectral ratio of surface waves, respectively. Also, wR and wHV
are weighting factors for the dispersion curve and H/V spectral ratios, respectively.
Bonnefoy-Claudet et al. (2006b) performed a numerical simulation of H/V data by using a
well-know 1D structure which consisted of a sedimentary layer over a half-space. Random
noise sources were distributed at the surface or subsurface, considering only cultural or
urban noise. The cultural sources were modelled as single body forces with randomly
distributed amplitudes (in x, y, z directions) and source time functions. Different sources at
different spatial locations and depths were used to find a good representation of the
measured H/V curve. Some of the source configurations in their study involved using
different source time functions (delta, pseudo-harmonic), source locations and sourcereceiver distances. The theoretical H/V spectral ratios were calculated using the wave
number based code developed by Hisada (1994, 1995) which calculates Greens functions
due to point sources for viscoelastic horizontally stratified media with sources and receivers
at very close depths. Bonnefoy-Claudet et al. (2006b) confirmed the findings by Lachet and
Bard (1994). In that the H/V ratio is weakly dependent on source type and function type:
its frequency is clearly independent, while the amplitude is weakly sensitive.
In 2007, Picozzi and Albarello (2007) proposed a two-step joint inversion of Rayleigh wave
dispersion and HVSR curves using combined genetic and linearised algorithms. The first
- 41 -

Chapter Two: Literature Review


step involved running a genetic algorithm (after Parolai et al. 2005) to constrain the
parameter space where the minimum misfit function is located to provide the linearised
algorithm (after Arai & Tokimatsu 2005) (second step) a starting model. The second step
would then provide an estimate of the optimal solution. Further reasoning behind this
inversion procedure can be found in the paper by Picozzi and Albarello (2007). To model
the physical background, they follow the formulations proposed by Tokimatsu et al. (1992)
and Arai and Tokimatsu (2004), which is based on the assumption that the microtremor is
dominated by surface waves. The structural model of the theoretical H/V ratios uses the
Thomson-Haskell method shown by Herrmann (2002) for vertically heterogeneous 1D
models. Picozzi and Albarello (2007) explain that introducing a wrong bedrock layer (or
half-space) can bias the theoretical model, and therefore the inversion results. Introducing a
wrong half-space can result in a wrong interpretation of higher modes of both Rayleigh and
Love waves of low frequency. However, they suggest that the presence of high impedance
contrast (in this case, contrast between surface layers and bedrock) should minimise this
problem, where with the presence of low impedance contrast might cause problems. This
is because it is believed that in the presence of high impedance contrasts, the surface waves
are naturally constrained in the softer sediment layers, whereas in the presence of weaker
impedance contrasts, higher modes are not constrained to propagate only in the shallow
subsurface (Picozzi & Albarello 2007). Hence, to limit any biases in the modelling, they
introduced a few deeper layers below the sedimentary layers. Similar to Lachet and Bard
(1994), Picozzi and Albarello (2007) also suggest that Poissons ratio can influence the H/V
peak and can further make it potentially possible to estimate this value using the HVSR
technique.
Herak (2008) also developed an inversion procedure. However, the theoretical H/V ratios
are modelled based on the assumption that the HVSR curve is controlled by body waves
and not the common assumption of surface waves.
In recent years, Albarello and Lunedei (2009) and Lunedei and Albarello (2010) have
developed a general model where ambient vibrations are assumed to be the complete
wavefield. This was made possible by generating a continuous distribution of random,
independent point-like sources acting at the surface of a weakly dissipative layered Earth to
model microtremors generated by anthropic activity, with frequencies ranging from 0.5
20 Hz. The model involves a set of homogeneous, isotropic layers overlying a
homogenous, isotropic half-space. They considered the effect of material damping by
- 42 -

Chapter Two: Literature Review


considering a viscoelastic rheology for the medium. After a study comparing the complete
wave field model and approximate interpretation models (surface wave or body waves),
they concluded that the surface waves approximation produced reliable results for
frequencies higher than the fundamental resonance frequency, where the body waves
approximation produced better results around the resonance frequency. Lunedei and
Albarello (2010) suggest that careful interpretation of HVSR peaks for the estimation of
subsoil properties is required when there are sources in close proximity to the receiver, as it
shows that the HVSR values are sensitive to the strength of these close sources. They
explain that the sources in the source-free area can have a great influence on the HVSR
curve. Hence, they proposed to model the full-wavefield with no source-free area.
However, the modelling of the complete wave field for the purpose of producing a
theoretical HVSR curve has been shown to be quite numerically demanding especially
when the source and receiver is at the same depth. It is noted by Lunedei and Albarello
(2010) that the surface waves approximation is much faster to compute when compared to
the full wavefield model. They showed that the HVSR curve produced by a full wavefield is
quite similar to the HVSR curve produced under the surface wave assumption when a
source-free area (few tens of metres) is introduced. Several source-free area and sourcereceiver configurations have been studied in the past (e.g. Fah et al. 2001; Arai &
Tokimatsu 2004; Lunedei & Albarello 2009). They explain that this source-free area can be
used to limit the effect of body waves. Hence, fundamental differences between the surface
model and full wavefield model include: the first relies on the assumption that there is no
effect of body waves, where the latter models both surface and body waves.
In a more recent paper, Snchez-Sesma et al. (2011) developed a theoretical model which
reflects the contributions of both surface (Rayleigh and Love) and body waves in an elastic
inhomogeneous, anisotropic medium. To produce the effect of microtremors in this model
a set of uncorrelated random forces are applied to the surface. The uniqueness of this
model comes with the use of the relationship between diffuse wave fields and Greens
function to determine the H/V ratio.

Summary Theoretical Modelling of HVSR Curves


There have been many interpretations by past investigators to theoretically model the
HVSR of microtremors. It should be evident that using surface waves to represent the
- 43 -

Chapter Two: Literature Review


entire wavefield is not physically possible, since other seismic sources (e.g. near field, body
waves, etc) also exist and can play a major role (Foti et al. 2011). However, recent
theoretical studies by Lunedei and Albarello (2010) suggested that the surface wave
assumption produces quite similar results to the results from full wavefield modelling.
Moreover modelling the full wavefield is computer intensive and time consuming, thus
highlighting the benefit of using surface waves to represent the HVSR of microtremors.

- 44 -

Chapter Three: Methodology


This chapter describes the general approach and methodology proposed in this thesis to
assess compacted ground based on HVSR techniques. While the HVSR technique has
evolved and is still evolving in time during the last few decades, it is still a relatively new
technique in the context of geotechnical engineering applications. This thesis has not
invented the HVSR technique as such but as will be shown in subsequent chapters, it can
lay claim to having made a contribution to its evolution and a methodology in which to
apply this technique for assessing compacted ground. Moreover, this thesis has approached
the assessment of compacted ground in an unconventional way, by attempting to infer the
Vs profile of the compacted ground and thus laying the foundational basis for a practical
approach yielding direct implications for the seismic response of the ground. A flowchart
(Figure 3.1) has been developed to summarise the overall procedure used in this thesis to
obtain both the measured HVSR curve and the theoretical HVSR curve to infer the Vs
profile. The main steps, illustrated in the flowchart, are as follows:
1.

measurement of the microtremor ground motion at the test station,

2. processing of the measured microtremor data to give the measured HVSR curve,
3. computing the theoretical HVSR curve by constrained forward modelling of the
measured HVSR measurements (curve) based on an assumed soil model, and
4. inferring the Vs profile, by fitting theoretical to measured HVSR curves.

- 45 -

Chapter Three: Methodology

Figure 3.1: Flowchart of the process involved in estimating the Vs profile of the ground.
- 46 -

Chapter Three: Methodology


Steps 3 and 4 rely on an iterative process to achieve an acceptable HVSR fitting. This
chapter will provide detailed methodologies of each of the four main steps. Also, given that
the HVSR technique has not been applied to assess and monitor ground compaction,
recommendations will be made for each step based on the experience and knowledge
gained throughout this study.

3.1 Measurement of Ambient Vibrations


The present study uses highly portable three-component (two horizontal components in
North-South (N-S) and East-West (E-W) directions, and one vertical (Z) direction) highresolution electro-dynamic velocimeters (Figure 3.2) (TrominoTM from Micromed), for the
measurement of the ambient vibrations at the observation point or station.

Figure 3.2: TrominoTM velocimeter with its carry case.


The velocimeters are aligned and firmly secured to the ground by short or long spikes
(depending on the stiffness of the ground surface). The technical characteristics of the
sensors are given in Table 3.1.
Table 3.1: Digital velocimeter specifications.
Specification of digital velocimeter used in this study
Dimensions
Weight
Frequency Response
Dynamic Range
Sample Rate
Output Sampling Rate

10 x 14 x 7.7 (height) cm
1.1 kg (batteries included)
0.1 -256 Hz
180 dB
32 kHz per channel
128, 256 or 512 Hz
- 47 -

Chapter Three: Methodology


Measurements were made in compliance with the guidelines of SESAME (2004a) to satisfy
the criteria for reliable and unambiguous HVSR results. It is common practise to analyse
ground vibration data to estimate the frequency spectrum, where the most common, easiest
and also probably the worst way is to chop out some seismic data (widowing), take the
Fourier transform and determine its magnitude (Mavko et al. 2003). Since the ambient
vibrations are captured in the time domain, it is necessary to convert the time signal into
the frequency domain for simpler interpretation (at least for the HVSR technique). Named
after French mathematician and physicist Joseph (Jean Baptiste) Fourier, the Fourier
transform is commonly used to convey a mathematical function of time as a function of
frequency, that is, converting from representation in the time domain to frequency domain
(i.e. frequency spectrum). The Fourier transform of f (x) is shown by:

() = () 2

(3.1)

Moreover sampling a time series at an interval of t effectively conveys information in the


frequency domain til the Nyquist frequency (fN) (For example, if t = 1/512, that is, a
sampling rate of 512 Hz then fN = 256 Hz):
= 1(2)

(3.2)

Therefore, in order to capture and analyse measured data in the frequency domain til 50 Hz
(used in this thesis for near surface compaction evaluation), it is essential for the
velocimeter (or accelerometer) used to capture ambient vibrations, to be able to output at a
sampling rate of at least 100 Hz. Essentially allowing for adequate statistical sampling in the
range 0.1 100 Hz, which is the common interval of engineering interest. As shown above
in Table 3.1, the TrominoTM velocimeter used in this study is able to provide an output
sampling rate of either 128, 256 or 512 Hz. Hence all three output sampling rates are
sufficient to provide information up to 50 Hz on the frequency domain. Consequently, the
output sampling rates used in this thesis were 128 Hz (Chapter 5 for the trial geotechnical
site characterisation) and 512 Hz (Chapter 6 to 9 for compaction evaluation). An output
sampling rate of 512 Hz was used for the compaction evaluation measurements instead of
128 Hz to obtain finer details of the captured data (i.e. more information per second). This
essentially means obtaining a larger file size (in terms of data storage). The selection of
- 48 -

Chapter Three: Methodology


window size depends on the lowest frequency of interest or the anticipated frequency of
the predominant resonance frequency peak, f0. For example, if an f0 value of 4 Hz is
expected (generally the case in the grounds tested in this thesis), the period, T, is 0.25 s
(using the equation T = 1/f). In order to have stable results (due to the statistical nature of
spectral estimates), there should be at least 10 repetitions of the longest anticipated period
(i.e. T = 0.25 s in this example). Therefore, a window size of at least 0.25 s x 10 repetitions
(2.5 s) is required. The window size used for the measurements in this thesis was 20 s. This
allows for 80 repetitions of the longest anticipated period (or lowest expected predominant
resonance frequency, f0), instead of the minimum of 10 repetitions. Moreover, using a 20 s
window size enables to obtain at least 10 repetitions for measurements captured in areas
with an f0 as low as 0.5 Hz. The recording duration of ambient vibrations essentially
depends on the selected window size. It is normal to aim to obtain no less than 20 clean
(refer to Table 3.2) windows. For example, if the chosen window size is 20 s, then 20 s x 20
windows are equal to 400 s, as a minimum recording duration. Consequently, the recording
durations used in this thesis were 10 minutes for the trial geotechnical site characterisation
(Chapter 5) and 16 minutes for compaction evaluation (Chapter 6 to 9). A recording
duration of 10 minutes is already longer than the minimum of 400 s required. However, an
even longer recording duration of 16 minutes was used for the compaction evaluation
measurements due to the location of the tested areas (i.e. they cannot be readily accessed)
and allowable working hours (set by the Penrith Lakes quarry operators).
The H/V curves were calculated by averaging the H/V obtained by dividing the captured
signal into non-overlapping windows (i.e. 20 s as mentioned above), where each window
was fast Fourier transformed (FFT) and smoothed with triangular windows with a width
equal to 10% of the central frequency. The geometric average was used to combine EW
(i.e. HEW) and NS (i.e. HNS) components in the single horizontal (H) spectrum, and then it
was divided by the vertical component (V) to produce the measured HVSR curve as
shown in Equation (3.3).
HVSRm =

H EW H NS

(3.3)

The commercial software Grilla accompanied with the TrominoTM velocimeter was utilised
to process the measured HVSR data. This software was also used to transfer the recorded
microtremor data from the velocimeter to the laptop. Although the commercial software
- 49 -

Chapter Three: Methodology


Grilla may be used to process the measured ambient vibration data, an in-house program
was written in the MATLAB language to separately analyse the measured HVSR data. This
gave flexibility for the data to be processed in an independent and a different way
compared to Grilla, ultimately yielding the HVSR curve. The developed program can be
found in Appendix B.1. Moreover, as mentioned above, the HVSR technique is new for
near surface geotechnical site characterisation and has not been applied before for
compaction assessment. Hence, Table 3.2 has been developed in this thesis to provide
recommendations to users of this technique for compaction assessment. The table includes
recommendations regarding (1) initial acquisition of ambient vibrations, (2) processing and
analysis of acquired ambient vibrations, and (3) forward modelling of the measured
ambient vibrations to infer the Vs profile (the forward modelling will be described in detail
in Section 3.2). These recommendations are based on the experience and knowledge gained
from this study, and the experience of other investigators. It may be used as a guide to
produce clean, stable and meaningful results. Of course, some modifications might be
necessary depending on the site conditions (e.g. local stratigraphy, slopes, material change,
etc).

3.2 Theoretical Modelling of the HVSR Curve


Generally speaking, methodologies applied by other investigators to obtain the Vs profile
by forward modelling are quite similar, with the two fundamental differences being the
assumptions made to produce the theoretical model and the frequency range of interest.
Firstly, the theoretical model relies on several assumptions, particularly regarding the
composition of seismic waves produced at a source. These assumptions include: (1) body,
surface or full wavefield, (2) if a full wavefield is assumed, the ratio between body and
surface waves must be further stipulated, (3) if a surface wavefield is assumed, the ratio
between Rayleigh and Love waves as well as the number of higher modes need to be
stipulated (5) location of the sources required to create the seismic waves, (6) type of
sources used to generate seismic waves (e.g. dirac delta, ricker wave, sine wave, etc), (7)
number of sources, and (8) the configuration of sources. Secondly, the type of application
of the HVSR technique essentially commands the frequency range of interest. Geophysists
typically rely on the lower frequency range, whereas geotechnical engineers would rely on
the higher frequency range, to provide information about the near surface.

- 50 -

Chapter Three: Methodology

Table 3.2: Thesis recommendations for using the HVSR technique for compaction assessment.
Measurement Parameters

Thesis Recommendations HVSR Technique for Compaction Assessment


Maximum There is no maximum on the recording length, however 24 hours should contain more than enough
information.
o Minimum This depends on the desired window size, it is normal to aim to obtain no less than 20 clean windows (e.g. if
window size is 20 s, then 20 s x 20 windows = 400 s).
o Recommended At least 10 minutes (in the presence of strong artificial noise, increase the recording duration).
To obtain information about the very near surface or up to 50 Hz, use a minimum of 100 Hz output sampling rate (i.e. at
least twice the Nyquist frequency). There are no issues with using a higher sampling rate, except for a larger file size.
The spacing of sensors relies on the users preference. If in doubt, a spacing of anywhere between 5 20 m should be
sufficient for near surface testing.
The sensor should be aligned to the North, simply to avoid confusion for subsequent analysis of directional HVSR data.
Directional HVSR analysis may show the direction of which artifact (or artificial noise) is measured. The sensor should be
levelled satisfactorily to avoid issues with velocimeter (e.g. component problems when the sensor is tilted too far to one side).
Every effort should be made to correctly couple the sensor to the ground. This is more important than slightly improving the
orientation and level of the sensor. The sensor should be pushed downwards into the ground and never pulled up to improve
the levelling. Avoid placing the sensor on tall grass, saturated soil, steep slopes and artificial material.
Avoid measurements close to buildings, since human activities inside the building and wind can influence the results. Also,
avoid measuring above underground structures (e.g. car park, pits, pipes, voids, tanks, etc), unless it is the purpose of testing.
o Rain Slight rain should not have an effect on results, but heavy rain/hail might influence the sensor.
o Wind - Strong winds can influence the lower frequencies
o Temperature Check the sensor manufacturers instructions.
Avoid measurements near machinery, roads, industrial areas, etc. Results from these areas need to be carefully studied by
conducting a continuous 24 hour measurement to identify and isolate the influence of artificial noise. See Section 7.2.
o

Recording Duration

Sampling Rate
Sensor Spacing
Sensor Orientation

Sensor Coupling
Recording Near Structures
Weather Conditions
Artificial Noise

- 51 -

Chapter Three: Methodology


Table 3.2: Thesis recommendations for using the HVSR technique for compaction assessment (continued).
Analysis of Measured
Data

Window Size

Smoothing
Averaging of Horizontal
Components
Cleaning
Identifying Artificial Noise

Forward Modelling

Interpretation

Constraints
Normalisation

Thesis Recommendations To obtain the Measured HVSR Curve


The recording length should be dependent on the lowest frequency of interest or the anticipated frequency of the
predominant resonance frequency peak, f0. For example, if an f0 value of 4 Hz is expected, the period, T is 0.25 s (using the
equation T = 1/f). Hence a minimum recording of 0.25 s is necessary. In order to have stable results (due to the statistical
nature of spectral estimates), there should be at least 10 repetitions of the longest anticipated period. Therefore, a window
size of at least 0.25 s x 10 repetitions = 2.5 s is required.
The Konno & Ohmachi (1998) smoothing (using B = 40) procedure is generally used by other investigators. However from
experience, it is preferable to use triangular windows with a width equal to 10% of the central frequency. It is not
recommended to use a much higher smoothing percentage, as it can distort and manipulate the results.
There are several ways to average the NS and EW horizontal components. This study has used the geometric average. A
survey of current literature has shown that the majority of investigators prefer the geometric average. See Section 3.1.
Certain windows may need to be removed from the average measured HVSR curve, due to contamination by artificial noise
(e.g. footsteps very close to the sensor, etc). See Section 5.2.
Artificial noise can be identified by analysing the single component spectra by observing if there is a presence of spikes on all
three component spectra at the same frequency. It might be necessary to produce a measured HVSR curve with zero %
smoothing for easier observation (i.e. clearer observation of spikes on the single component spectra). See Section 7.2.

Thesis Recommendations To infer the Vs profile


This study has shown that the measured HVSR curve can provide preliminary insights into the soil stratigraphy, without
undertaking any forward modelling. Therefore, it is worth analysing the measured HVSR curve for all peaks, troughs, etc
before beginning the forward modelling procedure. See Chapters 5 8.
The theoretical model does not produce a unique solution, hence it is highly recommended to constrain the Vs profile.
Constraints can include: layering thickness from mechanical penetration tests, relative Vs from mechanical or geophysical
tests, bedrock depth, etc. See Chapters 5 9.
This study and previous investigators have shown the importance of the normalisation of Vs for confining pressures due to
overburden stresses for compaction assessment. See Section 3.4.

- 52 -

Chapter Three: Methodology


The forward modelling method applied in this thesis is attributed to Lunedei and Albarello
(2009), and is based on a theoretical model of the surface waves (Rayleigh and Love)
propagation in multi-layered viscoelastic media and on the suite of numerical algorithms
developed by Herrmann (1987). The method of Lunedei and Albarello has its roots in the
work of Arai and Tokimatsu (2004) who have previously formulated a theoretical model of
surface waves in multi-layered elastic media for HVSR modelling. Here, ambient vibrations
(microtremors) in the ground are assumed to have resulted from randomly distributed
independent noise sources comprising vertical and horizontal harmonic point forces
(magnitude LV() and LH() respectively with an angular frequency ), on the ground
surface. Moreover, the point sources are assumed to be located outside a source free zone
of radius, r, from the observation station so that only effects of surface waves are needed to
be considered in this approach. Formerly, this radius was suggested by Arai and Tokimatsu
(2004) to be equal to one wavelength of the surface waves. This is because body waves
attenuate much more rapidly allowing the Rayleigh and Love surface waves to dominate at
distances greater than one wavelength from the source (energy carried by surface waves
decay at a rate of 1/r (where r is the distance from the source) while body waves decay at
1/r2). Later, Lunedei and Albarello (2009) proposed to calculate the radius r by means of
Equation (3.4).
r ( ) =

2 VS

(3.4)

where <Vs> is the average shear wave velocity of shallower soft layers and is the angular
frequency. It is noted that due to the stochastic random nature of the independent noise
sources, the theoretical HVSR of the ambient vibrations at a given , which is taken as the

ratio of average (absolute) components of vibration in the horizontal and vertical


directions, is expressed in terms of variances in Equation (3.5). This is because when
dealing with stochastic random ambient vibrations with a mean of zero, the average power
(viz. the average of the square of the amplitude) of the horizontal and vertical vibrations
finds a natural interpretation as the corresponding vibrations: vH and vV.
HVSR( )S =

v H ( )
vV ( )

(3.5)

- 53 -

Chapter Three: Methodology


Suppose X is the random variable defining the amplitude of the ambient vibrations with a
mean value of E[X], then the variance is given by:

[ ]

2 = E X 2 (E [X ])2

(3.6)

As mentioned above, when the vibrations have mean amplitude of zero, the variance will
reduce to:

[ ]

2 = E X2

(3.7)

The RHS also defines the average power of the vibrations (signal at receiver). Since power
is given as the square of the amplitude, the HVSR is thus given as:
HVSR( )S =

AH
AV

v H ( )
vV ( )

(3.8)

where AH, AV are the horizontal and vertical amplitudes. Furthermore, as suggested by
Lunedei and Albarello (2009), five higher modes were used to model the theoretical HVSR
curve. They recommended using a minimum of five modes to stabilise computations and
to limit the underestimation of power amplitudes. A greater number of higher modes has
been used (up to 15 in this thesis) but was found not to yield any significant difference in
the theoretical results, while only increasing the computational effort. Thus the theoretical
HVSR curves in this thesis have been produced using only five higher modes.
The ambient vibration HVSR technique is based on the analysis of wave propagation
through elastic (Arai & Tokimatsu 2004) or viscoelastic (Lunedei & Albarello 2009) layered
media which requires the solution of the Naviers equation. Naviers equation is derived by
first principals in Appendix C. The assumed soil model is horizontally layered as shown in
Figure 3.3. Here the soil layers are characterised by shear wave velocity (Vs1,....,Vsn), primary
wave velocity (Vp1,....,Vpn), soil layer thickness (y1,.,yn), bulk density (1,...,n) and quality
factors (Qs1,...., Qsn) and (Qp1,...., Qpn). Vs and y, are the parameters that have the greatest
influence on the fitting of the theoretical HVSR curve to the measured HVSR curve (see
Section 4.1). The soil model is numbered such that the uppermost layer at the free surface
is layer 1 and the bedrock half-space is layer n.
- 54 -

Chapter Three: Methodology


Layer 1

Vs1, Vp1, y 1, 1, Qs1, Qp1

Layer 2

Vs2, Vp2, y 2, 2, Qs2, Qp2

Layer 3

Vs3, Vp3, y 3, 3, Qs3, Qp3

Layer 4

Vs4, Vp4, y 4, 4, Qs4, Qp4

Layer n

Vsn, Vpn, y , n, Qsn, Qpn

Figure 3.3: Horizontally layered soil model.


Figure 3.4 shows a measuring station where the theoretical H/V spectrum is to be
determined and the surrounding ambient vibration sources. It is also assumed and shown
in Figure 3.4 that Fourier-time-transformed vertical and horizontal point forces having an
angular frequency , LV() and LH(), are randomly distributed on the ground (e.g. Lachet
& Bard 1994) at distances greater than one wavelength (Arai & Tokimatsu 2004) (or using
Equation (3.4) for Lunedei and Albarellos (2009) model) from the observation point, that
is, the origin. Surface and body waves are produced from each source (Figure 3.4 marked
as x) and propagate through the medium. However, body waves attenuate more rapidly
than surface waves. Therefore the Rayleigh and Love waves dominate at distances greater
than one wavelength from the source, as shown in Figure 3.4 (Rm or Lm).

Point Source
ri
i

r = Rm or Lm

Figure 3.4: Microtremor source model (after Arai and Tokimatsu (2000, 2004)).

- 55 -

Chapter Three: Methodology


Consider an arbitrary ith point source shown in Figure 3.4 (red x). The vertical and
horizontal powers of the mth mode Rayleigh wave from the ith vertical point source at a
frequency () is given by (Harkrider 1964; Arai & Tokimatsu 2004):
2

() = 2 ()2 () 0 ( ) (2 )

(2)

2
2

() = 2 ()2 () 1(2) ( ) (2 )

(3.9)

(3.10)

Analogous Equations (3.11) and (3.12) below express the vertical and horizontal powers of
the mth mode Rayleigh wave from the ith horizontal point source at a frequency ()
(Harkrider 1964; Arai & Tokimatsu 2004),
2
1
2

() = 2 ()2 () 1(2) ( ) (2 )

2
4
1

() = 2 ()2 () 0(2) ( ) (2 )

(3.11)

(3.12)

while Equation (3.13) expresses the horizontal power of the mth mode Love wave from
the ith horizontal point source at a frequency () (Harkrider 1964; Arai & Tokimatsu
2004),
2
1
(2)

() = 2 ()2 () 0 ( ) (2 )

(3.13)

where,
= Angular frequency
A = Medium response factor (Harkrider 1964)
k = Wave number

= H/V ratio of Rayleigh waves on the free surface (Haskell 1953)

r = Distance between the origin and the point source


(2)

= Hankel function of the second kind of the order n

h = Scattering damping ratio of soil (h>0)

The vertical and horizontal powers of all the waves observed at the origin at a frequency
(), are determined by integration of Equations (3.9) to (3.13) for all point sources and
- 56 -

Chapter Three: Methodology


Rayleigh and Love wave modes. Moreover, it is assumed that there is statistical
independence between the loading phases of all the sources. Thus, the integrated vertical
and horizontal powers of all the waves at a frequency () are derived as:

() = () =

()
()
+

2 2
2
1 +

=0

() = () + ()
() =

()
+

()

() =

()

(3.15)

=0

(3.14)

=0

2 2
2 2
1 +

2

2 2

(3.16)

(3.17)

where, is the H/V ratio of the microtremor loading sources, LH/LV is assumed to be
constant in the area considered, M is the highest mode considered, and
= (2)(4). Using Equations (3.14) to (3.17), the H/V ratio of Rayleigh waves

(H/V)R at a frequency () is given by:

()
() =

()

(3.18)

the H/V ratio of surface waves (H/V)S at a frequency () is given by:

()
() + ()
() =
=

()
()

(3.19)

and the Rayleigh to Love wave amplitude ratio for the horizontal motions (R/L) at a
frequency () is further expressed as:

()

() =
()

(3.20)

- 57 -

Chapter Three: Methodology


In order to solve Equations (3.18) and (3.19) to determine (H/V)R and (H/V)S, the
parameter, , is required in addition to the soil model. As suggested by Arai and Tokimatsu
(2004), the unknown value can be determined using Equation (3.20), provided that the
Rayleigh to Love wave ratio (R/L) is known. In their previous study, Arai and Tokimatsu
(2000) specified the use of a (R/L) value of 0.7, since they established that the (R/L) value
was stable between 0.4 to 1.0 (average of 0.7) using microtremor array data.
As suggested by (Harkrider 1964), the Rayleigh wave, AR, and Love wave AL factors
depend on the properties of the medium layers. Thus by adopting Harkriders (1964)
formulation, Arai & Tokimatsu (2004) were able to relate the AR and AL factors to the
ascertain the soil model (Vs profile) of a layered elastic medium. This allowed them to
relate the results of microtremor measurements to the mechanical properties of layered
medium (Lunedei & Albarello 2009).

3.3 HVSRplus Technique


A limitation of non-invasive geophysical techniques is the necessity to provide constraining
information to avoid the misinterpretation of measured data. To address the issue of
ensuring appropriate constraining information for the forward modelling of HVSR curves,
a two-pronged procedure has been proposed in this thesis. Firstly, in this thesis mechanical
CPT data has been used to assist in providing an initial guess of the material layering
information and a relative sense of the stiffness contrasts between layers. As will be
elaborated later in the thesis, the CPT data is also applied to calibrate the ground model
to the HVSR curves during the forward modelling procedure and to verify the forward
modelled HVSR results. Secondly, a modified version of the MASW technique, the
multichannel simulation with one receiver (MSOR) technique (Ryden et al. 2004) has been
utilised to provide the HVSR forward modelling procedure with further constraining
information to feed into the initial guess for the Vs profile. Here it should be noted that the
general concept of the MSOR technique differs from the more common MASW technique.
This is because the MSOR technique relies on a single receiver and a moving active source
along a 1D array, whereas the MASW technique relies on a 1D array of geophones and a
single active source at the end of the array. In this study, it was advantageous to employ the
MSOR technique, instead of the other array based surface wave techniques for two
reasons: (1) the same three component sensor used for the HVSR technique was also used
- 58 -

Chapter Three: Methodology


as one of the receivers to record the active ambient vibrations, and therefore (2) the
logistics, equipment and testing time was reduced. Even though this technique provides a
faster and simpler alternative to other array based techniques, it may still lack the speed and
portability of the HVSR technique but more importantly in this context, it helps to provide
additional constraining information for forward modelling of the HVSR curves.
The combination of the HVSR and MSOR technique, is given the name HVSRplus
technique, for ease of reference in this thesis. This name is chosen since both techniques
utilised the same velocimeter to obtain both passive ambient vibrations and actively
generated noise. In a two-pronged strategy (the other is the utilisation of mechanical CPT
results), the estimated ground velocity structure (Vs and layering thickness) from the
MSOR test provide additional constraining information: (1) the Vs of the shallow layer(s)
and (2) the relative ranges of the ground velocity structure, to the forward modelling
process, this being part of the HVSR forward modelling procedure. Note that in this study,
the information from the MSOR tests complement available CPT data which constitutes
the basis of the applied constraints to the forward modelling. The advantage of the
information from the MSOR test is that it directly yields the Vs profile at the test location,
especially the Vs of the uppermost surface layer which may be deduced from the asymptote
value of the phase velocity dispersion curve. However, the MSOR (like the MASW) test is
thought to estimate the Vs of the upper surface layers more reliably than the ones below
due to the use of relatively low energy active sources which are composed of higher
frequency components (Foti et al. 2011), especially in this study where the active sources
were created using only a small hammer. As a consequence, the estimated velocity structure
from the MSOR test for the lower half of the surface layers has been used with some
caution, in this study. For ease of understanding, a simplified graphical interpretation of the
HVSRplus technique is illustrated by the flowchart in Figure 3.5.
There are three main steps shown in this flowchart:
1. Estimation of the Vs profile by the MSOR technique.
2. Providing the Vs profile in (1) to the HVSR technique as an initial guess (also to
constrain) of the Vs profile.
3. Inferring the final Vs profile by the HVSR technique.

- 59 -

Chapter Three: Methodology

Figure 3.5: Proposed methodology of the HVSRplus technique.

3.3.1 MSOR Technique


The MSOR technique (Figures 3.6a 3.6c) utilises a single receiver and a moving source,
the latter being the equivalent of multiple sources acting at multiple instances in time, along
a 1D array. In the tests undertaken in this thesis, the sensor for the HVSR recordings was
used as a receiver while a vertical geophone acting as the trigger was initially set up at 1 m
from the sensor and it was then moved a further 1 m along the East-West axis after every
recording until a distance of 20 m. At every recording, a hand held hammer was used to
strike against a solid steel plate at a distance of 1 m from the trigger, on the opposite side
of the receiver, with approximately the same energy at each location. This was enough to
generate a controlled active source for every recording. Following the aforementioned
procedures, a pseudo-linear array the equivalent of 20 single component (vertical)
geophones was thus set up at a spacing of 1 m along the East-West axis.
The software Grilla (2012 version 6.2) by Micromed was used to create a synthetic
multichannel field recording by identifying the trigger locations and the times of the shots
of the recordings. The phase velocity spectra were obtained by slant-stacking the signals
and applying an FFT, after the operations of detrending, tapering and zero-padding. Since
surface wave propagation is a multimodal phenomenon, that is, more than one mode can
exist at each or some frequency, the selection of the fundamental mode was performed by
picking the local energy maxima corresponding to the lowest visible velocity (since no
velocity lower than the fundamental mode can exist). It was deliberately decided to work
- 60 -

Chapter Three: Methodology


with the inversion of the fundamental mode only because assigning the mode number to
the local maxima of energy is not straightforward especially in the case of velocity
inversions in the subsoil. In these cases, in fact, the combination of the source and array
geometry is a key element in the number of modes visible and their order (e.g., it can
happen that mode 1 and 3 are visible and mode 2 is not and there is no way to assign the
correct order number to each visible mode).
The theoretical model developed by Haskell (1953), was used in calculating the theoretical
dispersion curve for each ground model (Vs profile). A trial-and-error approach was
applied to minimise the misfit between theoretical and measured dispersion curves in the
inversion to estimate the ground velocity structure at the test location.
HVSR Techniques Sensor
G1

G2

G1

Shot 1

G3

GN

a)

b)

G2

Shot 2

c)

Figure 3.6: Example test layout for MSOR technique, also taking advantage of the
HVSR techniques sensor. (a) pseudo-linear array, (b) layout for shot 1 and (c)
layout for shot 2.

3.4 Normalisation for Confining Pressure


Confining pressures due to effective overburden stresses can significantly influence data
measured at different depths. Measurements retrieved from shallow depths, that is, with
low confining pressure, lead to a reduction in the measured parameter (e.g. strength,
stiffness, etc). Conversely, measurements retrieved from deeper depths, that is, with high
confining pressure, lead to an increase in the measured parameter (Moss et al. 2006). Thus
for accurate retrieval of measured data, which is unbiased by confining pressure, it is
- 61 -

Chapter Three: Methodology


necessary to normalise the raw data to a reference stress. Hence, the role of normalisation
of raw field measurements to a reference effective overburden stress, is to be able to
compare measurements obtained at different depths (Holzer et al. 2005), and to thus
minimise incorrect evaluation of subsurface soil strength and stiffness characteristics. The
reference stress that is generally used by past investigators, and this study, is 1 atm (or
approximately 100 kPa). For example, assuming the effective unit weight is 20 kN/m3, then
by normalisation it is possible to compare data from all depths, where the overburden
stress is equal to 100 kPa, at 5 m below the surface.
Furthermore, past investigators (e.g. Hardin & Drnevich 1972; Cha & Cho 2007; Claria Jr.
& Rinaldi 2007; Heitor et al. 2012) have shown that the Vs is influenced by the void ratio
and effective confining pressure of the medium in which the waves travel. Therefore, it
could be assumed that a medium with a constant void ratio or density at all depths will
have an increase in Vs with increasing depth. The paper by Heitor et al. (2012) showed the
influence of confining pressure on measured Vs values. The tested material from their
paper is from the DCPT-9 area within the Penrith Lakes quarry (described in Chapter 7).
They estimated Vs for compacted material using the laboratory bender elements tests,
inside a triaxial chamber in order to investigate the effect of isotropic confining pressure on
the shear wave propagation. Figure 3.7 shows the results from their study, in which there is
an obvious relationship between Vs and confining pressure, regardless of the moisture
content of the compacted specimens. This figure shows an increase in the compacted Vs
with increasing confining pressure, and thus the necessity for normalisation.

Figure 3.7: Shear wave velocity effected by confining pressure (Heitor et al. 2012).
- 62 -

Chapter Three: Methodology


This effect of confining pressure or effective stress due to overburden pressure on
measured Vs profiles was also investigated and discussed by other investigators (e.g.
Robertson et al. 1992a; Kim & Park 1999; Wride et al. 2000; Kim et al. 2001; Robertson
2009; Karray et al. 2011). It was found that good correlations between normalised Vs and
dry density (d) of compacted ground could be established (e.g. Kim & Park 1999; Kim et
al. 2001; Karray et al. 2011). Since the dry density is the standard for compaction
evaluation, the studies above indicate that it is necessary to normalise raw Vs data to obtain
a more correct indication of the ground stiffness (or a better correlation with dry density).
As previously mentioned, Vs is a direct measure of the small strain shear modulus (see
Equation (2.5)). Low amplitude shear modulus has been represented by Hardin (1978) as:
n
G max = Af (e )OCR K Pa1 n ( m )

(3.21)

where, A = dimensionless coefficient, f(e) = 1/(0.3+0.7e2), e = void ratio, OCR = overconsolidation ratio, K = function of plasticity index, Pa =

reference stress, typically

atmospheric pressure (approximately 100 kPa), n = stiffness parameter and m is a mean


effective stress which is can be represented by:
m =

1
(1 + 2 K 0 ) V
3

(3.22)

where K0 = earth pressure coefficient at rest, v = z is the vertical effective stress, is


the effective unit weight of the surface layer and z is the depth. For a non-plastic soil, K is
zero and since the compacted material tested in this study is not virgin material, it can be
assumed that the effect of the over-consolidation on Gmax is negligible. Therefore, for the
compacted material used in this study, Equation (2.5) can be re-written as:
Vs =

Af (e )Pa1 n ( m )

(3.23)

This equation verifies the statements made above, that Vs which can be represented by

Gmax, is affected by the void ratio/density and confining pressures ( m ). Consequently, the

estimated Vs values in this thesis have been normalised using Equation (3.24) to eliminate
the influence of confining pressures as shown by Robertson et al. (1992a):
- 63 -

Chapter Three: Methodology

Vsn

P
= Vsm a
v

0.25

(3.24)

where, Vsn is the resulting normalised shear wave velocity, Vsm is the derived/measured

shear wave velocity and, and v = z is the vertical effective stress (in kPa), is the

effective unit weight of the surface layer (in kN/m3) and z is the depth (in m). The unit
weight of the soil layers was estimated based on available bore log data.
The mechanical CPT data for this study have also been normalised to account for
confining pressure using Equation (3.25) from Robertson et al. (1992a):
P
q cn = q cm a
v

0.5

(3.25)

where, qcn is the resulting normalised cone end resistance and qcm is the measured cone end
resistance.

3.5 Resonance Frequencies and Higher Modes


The presence of secondary resonance peaks above the predominant resonance frequency,
f0, may point to strong impedance contrasts within the surface layers. However, this
interpretation must be treated with some caution since there is a possibility that a
secondary resonance peak may also reflect a higher mode of f0 (e.g. 3f0, 5f0, 7f0, etc).
Such a possibility arises when the ground is idealised as a surface layer overlying the
bedrock, where wave propagation is essentially a one-dimensional oscillation in the vertical
direction with one end free and the other end fixed. The resonance frequencies of such a
system (Figure 3.8) may be derived by mathematically modelling wave propagation through
an elastic inertial medium. In this instance, the Navier-Stokes equation reduces to the
specialised one-dimensional wave equation shown in Equation (3.26):

- 64 -

Chapter Three: Methodology

z
ux
,

Figure 3.8: Surface layer overlying bedrock: One end free and the other end fixed.
2

2 2
=
2
2

(3.26)

where, (, ) is the amplitude of the displacement in the horizontal x-direction, and c is

the speed of wave propagation through the medium, typically given by the Newton-Laplace
equation shown in Equation (3.27):

2 =

(3.27)

where, is a stiffness coefficient, the shear modulus and is the bulk density. Applying the
Fourier Transform defined as:
+

(, ) =

(, )

(3.28)

gives:
2
2
=

(3.29)

where, is the angular frequency. It is not difficult to show that the general solution of
(3.29) is:

= cos() + sin()

(3.30)

- 65 -

Chapter Three: Methodology


where = is the wave number, and A and B are unknown coefficients. Taking into

consideration the boundary conditions of the medium shown in Figure 3.9, B = 0 by virtue

of the boundary condition = 0 at z = 0.


= 0, = 0,

=0

= 0

Figure 3.9: Surface layer overlying bedrock: One end free and the other end fixed
Boundary conditions.
The fixed end boundary condition, that is, = 0, is satisfied provided:
= cos() = cos() = 0

where, L is the length of the medium. Thus for non-trivial solutions, it requires:

= (2 1)

(3.31)

where, n is an integer. Recalling that:

2
=

where, fr is the resonance frequency, then Equation (3.31) gives:


2

= (2 1)
2

(3.32)

- 66 -

Chapter Three: Methodology


and by rearranging Equation (3.32), the resonance frequencies of a ground with a surface
layer overlying the bedrock, with the above boundary conditions, is given by:

= (2 1)

(3.33)

Equation (3.33) is commonly written as, in seismological literature (see, e.g. Ibs-von Seht &
Wohlenberg 1999):
fr =

nV s
(where n = 1,3,5,....)
4h

(3.34)

where, n is an odd integer, Vs is the shear wave velocity (m/s), and h is the depth of the
bedrock (m). Thus, the first peak (n = 1) is located at the fundamental resonance frequency
(fr = f0); the second peak (n = 3) is located at the first harmonic (fr = 3f0); the third peak (n =
5) is located at the second harmonic (fr = 5f0), and so forth. In view of this, it is prudent not
to rule out the possibility that secondary peaks occurring on the theoretical HVSR curve at
3f0, 5f0, . may be due to the effects of higher modes rather than impedance contrasts in
the surface layers.

- 67 -

Chapter Four: Parametric Studies and


Analysis of HVSR Curves
In this chapter, a sensitivity analysis was performed on the group of input parameters to
produce the theoretical HVSR curve, namely, layer thickness (y), shear wave velocity (Vs),
compression wave velocity (Vp), density () and the quality factors (Qp and Qs), as part of
an effort to better examine and understand the general landscape of the forward modelling
problem. Furthermore, a parametric study was initiated to study the influence of
impedance contrasts in the ground on theoretical HVSR curves. The study in this chapter
provides a deeper understanding and important insights into the relationship between the
layered ground and resulting HVSR curves. This work facilitates the experimental design to
investigate ambient vibration HVSR of the field compaction to be discussed in the
subsequent chapters.

4.1 Sensitivity Analysis of the Theoretical HVSR


Model
In the following sensitivity study, the influence of the ground profile parameters on the
relative magnitude and the general shape of the HVSR curve have been investigated. A
benchmark theoretical HVSR model was established using the simple theoretical ground
- 68 -

Chapter Four: Parametric Studies and Analysis of HVSR Curves


profile shown in Table 4.1. Here the sensitivity analysis was performed by changing the
value of one of the surface layer parameters incrementally, while keeping the remaining
parameters constant. For example, in the first case, the thickness of the first layer (0.4 m in
Table 4.1) was changed from 0.1 to 10 m using 0.1 m increments. At each incremental
change, its corresponding theoretical HVSR model was calculated. In an attempt to
normalise the theoretical HVSR solutions, the calculated theoretical HVSR model were
divided by the benchmark theoretical HVSR model. Once all 100 (i.e. 10m/0.1m)
theoretical HVSR models were calculated and normalised, they were transferred to a 3dimensional shaded surface, similar to a 3-dimensional contour. Figure 4.1a shows the
solution landscape due to manipulating the first layer thickness. The influence and/or
importance of a particular parameter on the theoretical HVSR curve may be observed by
the size or magnitude of the peaks in the solution landscape.
Table 4.1: Simple theoretical ground profile used for the sensitivity analysis.
Thickness (m)
Vs (m/s)
Vp (m/s)
Qs
Qp
(t/m3)
0.4
50
104
1.8
10
20
2
100
208
1.8
20
40
10
200
416
1.8
30
60
(Bedrock)
400
832
1.9
40
80
The above mentioned procedure was also carried out on the remaining surface layer
parameters. Figures 4.1 4.5 show the solution landscape of the theoretical HVSR model
when changing the surface layer thickness (y), shear wave velocity (Vs), compression wave
velocity (Vp), density () and the quality factors (Qp and Qs), respectively. It should be
noted, that in this simulation, a rule-of-thumb of Qp = 2Qs was assumed since this is
generally regarded as a reasonable assumption (e.g. Lunedei & Albarello 2009; Olsen et al.
2009; Dal Moro & Ferigo 2011).
From this sensitivity analysis several conclusions about the solution space can be made: (1)
the solution space is confirmed to be highly multimodal with varying degrees of
complexity, (2) convergence to the optimum theoretical solution is going to be primarily
dictated by the thickness of the layers, this implies that the correct number of layers is a
critical input as well, (2) to a lesser extent, the Vs of the layers also plays an important role
in influencing the final solution of the theoretical model, (3) the Vp, , Qp and Qs are much
less influential on the final theoretical HVSR solution.

- 69 -

Chapter Four: Parametric Studies and Analysis of HVSR Curves

(a)

(b)

(c)

Figure 4.1: Sensitivity of the layer thickness (y ) parameter on the theoretical HVSR
curve. (a) Layer 1, (b) Layer 2, and (c) Layer 3.

- 70 -

Chapter Four: Parametric Studies and Analysis of HVSR Curves

(a)

(b)

(c)

Figure 4.2: Sensitivity of the layer Vs parameter on the theoretical HVSR curve. (a)
Layer 1, (b) Layer 2, and (c) Layer 3.

- 71 -

Chapter Four: Parametric Studies and Analysis of HVSR Curves

(a)

(b)

(c)

Figure 4.3: Sensitivity of the layer Vp parameter on the theoretical HVSR curve. (a)
Layer 1, (b) Layer 2, and (c) Layer 3.

- 72 -

Chapter Four: Parametric Studies and Analysis of HVSR Curves

(a)

(b)

(c)

Figure 4.4: Sensitivity of the layer density () parameter on the theoretical HVSR
curve. (a) Layer 1, (b) Layer 2, and (c) Layer 3.

- 73 -

Chapter Four: Parametric Studies and Analysis of HVSR Curves

(a)

(b)

(c)

Figure 4.5: Sensitivity of the layer quality factors (Qp = 2Qs) parameter on the
theoretical HVSR curve. (a) Layer 1, (b) Layer 2, and (c) Layer 3.

- 74 -

Chapter Four: Parametric Studies and Analysis of HVSR Curves

4.2 Parametric Study of Theoretical HVSR Curves


In his seminal paper, Nakamura (1989) suggested that the horizontal to vertical spectral
ratio (HVSR) of the surface microtremors reflect the transfer function of the surface layers.
Generally speaking, the peaks observed on the measured HVSR curves are explained by the
presence of impedance contrasts of differing soil types (e.g. contrast between bedrock and
softer surface layers). The impedance contrast of an elastic medium is described as the ratio
of stress to particle velocity (Aki & Richards 1980; Mavko et al. 2003). Hence, impedance is
defined here as the product of the density and the shear wave velocity, that is, Vs.
Typically, an observed peak on a HVSR curve suggests that there is a change in shear wave
velocity (Vs) of the material where the higher the peak, the greater will be the
corresponding change in Vs (as illustrated in the simple two layer system in Figure 4.6).
Hence, the measured HVSR curves reflect the soil layering of the site. The soil layering
data, however, must be extracted from the measured HVSR curves by theoretical
modelling of the HVSR curve (refer to Section 3.2). Although six soil parameters are
required for each soil layer in the model (see, e.g. Table 4.1), the soil parameters that will
most significantly influence the characteristics of the HVSR curve are the layer thickness
and Vs (see Section 4.1).

Figure 4.6: Surface layer (10 m thick at 250, 300 and 350 m/s) overlying the halfspace (bedrock at 600 m/s) the higher the resonance peak amplitude (H/V), the
greater the impedance contrast between layers.
A parametric study was initiated to investigate the influence of impedance contrasts on
theoretical HVSR curves. Generally speaking, each resonance peak observed on the HVSR
- 75 -

Chapter Four: Parametric Studies and Analysis of HVSR Curves


curve indicates a detectable impedance contrast between the soil above and the soil below a
given depth, whereby higher resonance frequencies on the frequency axis indicate
impedance contrasts at depths closer to the ground surface (e.g. the simple two layer
system in Figure 4.7). This thesis particularly seeks to refine the technique to better capture
and interpret the information within the higher frequency ranges which are indicative of
the compaction quality in the vicinity of the ground surface. This will be by refining the
current HVSR technique practices to take cognizance of the higher frequency peaks that
are generally not of significant interest to the geophysicists.

Figure 4.7: Surface layer (10, 15 and 20 m thick at 250 m/s) overlying the half-space
(bedrock at 600 m/s) the higher the resonance peak frequency (on the frequency
axis), the shallower the location of the impedance contrasts.

Relating Impedance Contrasts in the Ground to Peaks on the HVSR


Curve
Two, three and four layer systems (i.e. theoretical Vs profiles) are used to convey how
impedance contrasts observed in the ground will lead to peaks on the HVSR curve. In this
theoretical simulation, the Vs and thickness of each layer are increasing with depth by
factors of two and five, respectively (see, Table 4.2).
Table 4.2: Final Vs profile used for parametric study.
Layer Number

Thickness (m)

Depth (m)

Vs (m/s)

1
2
3
4

0.4
2
10
half-space (bedrock)

0.4
2.4
12.4

50
100
200
400

- 76 -

Chapter Four: Parametric Studies and Analysis of HVSR Curves


The two layer system (i.e. one surface layer above the half-space or bedrock) shown by
Figure 4.8b has been used to generate the theoretical HVSR curve in Figure 4.8a. The
impedance contrast observed on the Vs profile at 0.4 m is portrayed on the HVSR curve as
a resonance peak at approximately 28 Hz as illustrated in Figure 4.8a.

(b)

(a)

Figure 4.8: Impedance contrast to HVSR peaks: a two layer system. (a) theoretical
HVSR curve and (b) Vs profile.
A third layer is introduced into the above two layer system. The three layer system (i.e. two
surface layers overlying the half-space or bedrock) shown in Figure 4.9b has been used to
generate the theoretical HVSR curve shown in Figure 4.9a. Here, the impedance contrasts
observed on the Vs profile at 0.4 and 2.4 m are portrayed on the HVSR curve as resonance
peaks at approximately 28 and 11 Hz, respectively. Several observations are made:
1. The same resonance peak observed on the two-layer system HVSR curve is also
seen on the three layer system HVSR curve,
2. The amplitude (or HVSR) of the resonance peak at 28 Hz has slightly decreased
due to a coupled affect, so that changes in one layer will have an effect on the other
layers as well.

(b)

(a)

Figure 4.9: Impedance contrast to HVSR peaks: a three layer system. (a) theoretical
HVSR curve and (b) Vs profile.

- 77 -

Chapter Four: Parametric Studies and Analysis of HVSR Curves


Finally a fourth layer is introduced into the above three layer system. The four layer system
(i.e. three surface layers above the half-space or bedrock) shown in Figure 4.10b has been
used to generate the theoretical HVSR curve in Figure 4.10a. Here, the impedance
contrasts observed on the Vs profile at 0.4, 2.4 and 12.4 m are portrayed on the HVSR
curve as resonance peaks at approximately 28, 11 and 4 Hz, respectively. Again several
observations are made:
1. The same resonance peaks observed on the three layer system (i.e. at 28 and 11 Hz)
HVSR curve are also seen on the four layer system HVSR curve.
2. Similar to the three layer system, the amplitude (or HVSR) of these resonance
peaks have slightly decreased due to the coupled effect between layers.

(b)

(a)

Figure 4.10: Impedance contrast to HVSR peaks: a four layer system. (a) theoretical
HVSR curve and (b) Vs profile.
Figure 4.11 superimposes all three HVSR curves and Vs profiles from this study and this
figure can be essentially simplified as shown in Figure 4.12. These figures show in
particular: (1) peaks observed on the HVSR curve are typically related to the impedance
contrast in the ground, and (2) the location (i.e. frequency) of the peak observed on the
frequency axis of the HVSR curve is dependent on the depth of the impedance contrast
between soil layers (i.e. the higher the peak on the frequency axis, the shallower the
impedance contrast between layers).
Moreover, this theoretical simulation has identified not only a relationship between the
predominant peak (i.e. at 4 Hz) on the HVSR curve with the impedance contrast between
the bedrock and overlying surface layers (i.e. between soil below and above the yellow
dotted interface (at 12.4 m) in the soil profile), but also relationships between secondary
resonance peaks observed at higher frequencies (at 28 and 11 Hz) and impedance contrasts
at shallower depths (at 0.4 and 2.4 m).

- 78 -

Chapter Four: Parametric Studies and Analysis of HVSR Curves

(a)

(b)
Figure 4.11: Impedance contrast to HVSR peaks: superimposing the two, three and
four layer systems. (a) theoretical HVSR curves and (b) Vs profiles.

(a)

(b)
Figure 4.12: Impedance contrast to HVSR peaks. (a) theoretical HVSR curve and
(b) Vs profile.
It must be stressed that the relationship between impedance contrasts and resonance peaks
is more complex than that shown in Figure 4.12. As mentioned above, the first resonance
peak at 28 Hz is due to the impedance contrast at 0.4 m, that is, the contrast between layer
1 and 2 in Table 4.2. The second resonance peak at 11 Hz is due to the impedance contrast
at 2.4 m, that is, the contrast between layers 2 and 3. However, layer 1 also has some
influence on the final shape of the second resonance peak. In particular, the amplitude of
the second resonance peak can increase or decrease depending on the Vs of layer 1. This is
discussed below and is shown in Figure 4.13a. Also, the third (or fundamental) resonance
peak at 4 Hz is due to the impedance contrast at 12.4 m, that is, the contrast between layers
3 and 4. Similarly to the previous case, layers 1 and 2 also influence the final shape of the
third resonance peak. Furthermore, in real case scenarios, when capturing ambient
vibrations to produce the measured (or experimental) HVSR curve, there can be several
unwanted interferences to the final shape of the HVSR curve (e.g. anthropic noise as
discussed in Section 7.2 and unnatural features embedded in the ground such as voids or
concrete).

- 79 -

Chapter Four: Parametric Studies and Analysis of HVSR Curves

Sensitivity Analysis of Vs on Producing an Impedance Contrast


A sensitivity analysis was performed to determine whether there is common threshold for a
theoretical HVSR curve to spawn a resonance peak due to an impedance contrast in the
ground. The Vs profile and theoretical HVSR curve in Figure 4.12 were used in this study.
The study observed the effects on the resonance peaks by incrementally increasing the Vs
of each of the three surface layers in turn until it matches the Vs value of the layer directly
below. For instance, the original Vs of the first layer at 50 m/s was incrementally increased
until it matches the Vs of the second layer (100 m/s). This process was then repeated
between the second and third layers, and third layer and the bedrock to assess the
sensitivity of the HVSR curve to these changes.
The sensitivity of the first resonance peak at 28 Hz with increasing Vs in layer 1 (by 10% of
Vs in layer directly below, namely layer 2) is shown in Figure 4.13. Figure 4.13a also shows
a diminishing first resonance peak on the theoretical HVSR curve until it is difficult to
observe when the difference in Vs between layer 1 and layer 2 is within 30% of layer 2s.
The resonance peak at 28 Hz disappears when the Vs of layer 1 and layer 2 are equal, that is
when there is no impedance contrast between the layers. Moreover, as explained above,
increasing the Vs of layer 1 caused the amplitude of the third resonance peak at 11 Hz to
noticeably decrease due to the coupled affect between layers. In this case, by increasing the
Vs of layer 1, it reduces the contrast between the Vs of surface layers above 2.4 m (i.e.
layers 1 and 2) and the Vs below (i.e. layer 3).
The sensitivity of the second resonance peak at 11 Hz with increasing Vs in layer 2 (by 10%
increment of Vs value in layer 3) is shown in Figure 4.14. The second peak (at
approximately 11 Hz) becomes difficult to observe when the difference in the Vs between
layer 2 and layer 3 is within 30% of the latter. Increasing the Vs of layer 2 to four times
(200 m/s) the Vs of layer 1 (50 m/s) caused the appearance of an exaggerated resonance
peak due to this impedance contrast at 0.4 m. Similar to Figure 4.13a, Figure 4.14 shows a
decrease in the amplitude (HVSR) of the resonance peak at 4 Hz due to the coupled effect
between layers.

- 80 -

Chapter Four: Parametric Studies and Analysis of HVSR Curves

(a)

(b)

Figure 4.13: Sensitivity analysis: The influence of the first (shallowest) impedance
contrast on the first (highest frequency) peak. (a) theoretical HVSR curves and (b)
Vs profiles.

Figure 4.14: Sensitivity analysis: The influence of the second (centre) impedance
contrast on the second (middle) peak.
- 81 -

Chapter Four: Parametric Studies and Analysis of HVSR Curves


The sensitivity of the third resonance peak at 4 Hz with increasing Vs is shown in Figure
4.15. Similar results reported for the resonance peak at 11 Hz above can be seen in this
figure. Interestingly, the sensitivity study shows that a common threshold of 30 to 40 % of
Vs difference between layers was enough to spawn a resonance peak on the theoretical
HVSR curve due to impedance contrast, in all three figures.

Figure 4.15: Sensitivity analysis: The influence of the third (deepest) impedance
contrast on the third (lowest frequency) peak.

- 82 -

Chapter Five: Trial Geotechnical Site


Investigation Using the HVSR
Technique
A preliminary trial was carried out to evaluate the HVSR technique and ideas proposed in
the methodology of this thesis prior to undertaking the main scope of this study, which is
to geotechnically characterise compacted ground. Here a geotechnical characterisation was
made of the ground below a proposed development site, which involves: (1) estimating the
fundamental resonance frequency and amplification of the site, and the parameters defining
the seismic response of an area (2) interpreting the measured HVSR curves for a
preliminary assessment of the soil layering, and (3) fitting the HVSR curves to a theoretical
model to estimate the shear wave velocity (Vs) profile of the ground. Results from the
HVSR technique have been verified against data from classical invasive methods (viz.
borehole data and SPT). Further verification has also been made against the results from
the array based MASW technique. Also, the HVSRplus technique was trialled.

5.1 Test Site


A future development site (broken line in Figure 5.1) of the Penrith (Kingswood) campus
of the University of Western Sydney (UWS) was made available for this study. Penrith
- 83 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique
campus is one of six campuses of the University of Western Sydney spread throughout the
western part of Sydney. At Penrith, there are 3 precincts: Kingswood, Werrington South
and Werrington North. The Kingswood precinct, where the test site is located, is
considered the centre of student and academic activities at Penrith. The campus lies on the
Cumberland plain to the east of the Blue Mountain plateau within the Sydney Basin. The
geology of the campus site is dominated by Bringelly shale of the Wianamatta group. The
near surface soil is made up of fills from more recent construction activities and weathered
parent material.

Figure 5.1: Trial geotechnical investigation location (Image from Google Maps).
The university has undertaken a geotechnical investigation (including bore logging and
SPT) of the site in anticipation of proposed construction of some new buildings. The
borehole data have provided valuable and reliable information on the underlying soil
profile. Different soil layers have been identified as well as other important soil
characteristics located at the proposed new library location. According to the soil report
(Jeffrey and Katauskas Pty Ltd 2008) based on the bore logs, the underlying soil
stratigraphy consisted of:

Fill - was encountered in all boreholes to highly variable depths ranging from 0.4 m
to 2.6 m. The fill consisted of silty clay of low to high plasticity, with igneous gravel
and root fibres, and occasional tile fragments in places. Based on the SPTs
- 84 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique
(standard penetration test) and visual observations, the fill was assessed to be

variably compacted; mostly in the poorly to moderately compacted range.


Natural Soils - were encountered below the fill. The upper soils, mostly silty and
sandy clays, were inferred to be of alluvial origin. Some of the deeper silty clays, on
approaching the bedrock profile, were assessed to be of residual origin since some
contained remnants of the shale structure. In general, most of the clays were of
medium plasticity, but with a few of low and high plasticity. The silty and sandy
clays were generally of very stiff to hard strength, and in places contained ironstone

gravel.
Shale - bedrock was encountered at depths ranging from 3.6m to 8.8 m. It was
concluded that most of the shale bedrock generally classifies as distinctly weathered
and of very low to low strength, with some significant upper zones of extremely
weathered, extremely low strength shale.

The locations of the HVSR stations (denoted by red squares) and the available borehole
locations (denoted by the green circles) are shown in Figure 5.2. The stations are regularly
spaced and arranged in three rows by five lines making a total of 15 test stations. The
layout of the test stations has been planned taking cognizance of the locations of the prior
borehole investigations, and deliberate situating of some stations at the same locations as
the boreholes. Available in situ borehole and SPT data was used for constraining the HVSR
technique and for the verification of the HVSR results. Also, three MASW projections
were conducted at this site along lines 1, 3 and 5 (denoted by blue lines).

Figure 5.2: Test grid at the Penrith (Kingswood) campus of UWS.


Dimensions are in metres (m).
- 85 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique

5.2 Measurement Reliability


Figure 5.3 illustrates an example of the measured HVSR curve. The thick red line
represents the average HVSR and the two black thin lines indicate the 95% confidence
limit of the HVSR. In Figure 5.3 the peak resonance frequency can be observed at
approximately 4.6 Hz with a site amplification factor of approximately 6.4.

Figure 5.3: An example of measured HVSR curve.


The HVSR analysis is accompanied by its time-history, a plot illustrating the HVSR
computed for successive time intervals. Each time interval corresponds to the window
length. A time window of 20 seconds was selected throughout the analysis. The timehistory of the HVSR analysis is useful to identify possible noisy windows which may
disturb the HVSR analysis. The noisy windows can be removed to minimise errors in the
analysis. Figure 5.4 illustrates the HVSR time-history of this recording. This time-history
plot clearly shows a clear band across the resonance frequency throughout the testing, after
removing the noisy windows from the analysis.

Figure 5.4: Time-history plot of recording using TrominoTM velocimeter.


- 86 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique
SESAME (Site EffectS assessment using AMbient Excitations) is a European based
research project with the goal of producing a universal guideline (SESAME 2004a) to
eliminate the misuse of the HVSR technique by introducing a European and World
standard of consistency during testing. The HVSR analysis has been checked against the
SESAME criteria for reliability. This check recognises reliable recordings as well as clear
HVSR peaks of resonance. Table 5.1 tabulates the criteria of reliability produced by the
HVSR analysis program Grilla. The Table shows the HVSR criteria for reliability check of
the recording above. It can be seen that the HVSR recording has the characteristics of a
reliable HVSR curve and a clear predominant HVSR peak. Furthermore, the other
measurements in this thesis showed they have also met SESAMEs criteria for reliability
and clarity.
Table 5.1: Example SESAME criteria for reliability of results (Grilla by Micromed).
Max. H/V at 4.56 0.03 Hz. (in the range 0.0 - 256.0 Hz).
Criteria for a reliable HVSR curve
[All 3 should be fulfilled]

f0 > 10 / Lw
nc(f0) > 200
A(f) < 2 for 0.5f0 < f < 2f0 if f0 > 0.5Hz
A(f) < 3 for 0.5f0 < f < 2f0 if f0 < 0.5Hz

4.56 > 0.67


3216.6 > 200
Exceeded 0 out of 110
times

OK
OK
OK

Criteria for a clear HVSR peak


[At least 5 out of 6 should be fulfilled]
-

Exists f in [f0/4, f0] | AH/V(f ) < A0 / 2


+
+
Exists f in [f0, 4f0] | AH/V(f ) < A0 / 2
A0 > 2
fpeak[AH/V(f) A(f)] = f0 5%
f < (f0)
A(f0) < (f0)
Lw
nw
nc = Lw nw f0
f
f0

f
(f0)

A0
AH/V(f)

f
+
f
A(f)

logH/V(f)
(f0)

3.563 Hz
0.0 Hz
15498.65 > 2
|0.0| < 0.05
0.0 < 0.22813
0.0 < 1.58

OK
OK
OK
OK
OK
OK

window length
number of windows used in the analysis
number of significant cycles
current frequency
H/V peak frequency
standard deviation of H/V peak frequency
threshold value for the stability condition f < (f0)
H/V peak amplitude at frequency f0
H/V curve amplitude at frequency f
frequency between f0/4 and f0 for which AH/V(f ) < A0/2
+
frequency between f0 and 4f0 for which AH/V(f ) < A0/2
standard deviation of AH/V(f), A(f) is the factor by which the mean AH/V(f)
curve should be multiplied or divided
standard deviation of log AH/V(f) curve
threshold value for the stability condition A(f) < (f0)

- 87 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique

5.3 Interpretation of Measured HVSR Curves


The HVSR measurements from all test stations (Figure 5.2) in the area are superimposed in
Figure 5.5. Variations between the HVSR curves are expected and can be caused by several
factors, including changes in layering thickness and material properties. For the entire site,
two common and identifiable significant characteristics were clearly observed in the
superimposed curves, namely:
1. The largest HVSR amplitude (known henceforth as the predominant peak)
occurring at the fundamental resonance frequency (f0) (ranging from 4.5 6 Hz),
2. A HVSR peak possibly associated with a higher mode of f0 occurring at
approximately 3f0, and other discernible secondary resonance peaks at frequencies
above f0 (or in the range 12 50 Hz).

Figure 5.5: A summary of the measured HVSR curves from the proposed UWS
development site.
In (1), the predominant peak indicates an impedance contrast between the bedrock and
overlying surface layers, and the amplitude of the peak correlates to the strength of the
impedance contrast. The frequency at which the predominant peak occurs has been shown
in previous studies to correspond to the fundamental resonance frequency of the site (see,
e.g. Bonnefoy-Claudet et al. 2006b and references therein). Thus a measurement of the
HVSR would provide a relatively simple means to determine the fundamental resonance
frequency and a measure of the site amplification, collectively known as the site effects,
parameters which define the seismic response of the area. The measured HVSR curves in
this area produce a double peak at some stations, which is believed to be caused by a
combination of cultural activities (i.e. traffic, human activities, etc) and the poorly defined
- 88 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique
transition (impedance contrast) from surface layers to bedrock (i.e. transition zone from
weathered soil to the stiff bedrock). In (2), the presence of secondary resonance peaks if
confirmed as not predominantly associated with the higher modes of f0 would indicate
presence of impedance contrasts within the surface layers (e.g. transition from fill to natural
material at the site). Since there is also a real possibility that the secondary peak at 3f0 may
be in fact due to the first higher mode of f0 (see, Section 3.5), whether this secondary peak
reflects the impedance contrast of the soil layers or not cannot be determined simply by
visual inspection of the measured HVSR curve. Consequently, constrained forward
modelling in conjunction with independent mechanical tests is invoked in this study to
resolve this ambiguity.

5.4 Forward Modelling to Infer the Vs Profile


Constrained Forward Modelling of Measured HVSR Curves
This section outlines the procedures for constrained forward modelling to infer the Vs
profile and to resolve ambiguities of the HVSR curves by taking advantage of soil
information already available from in situ bore logs. For the purpose of this study, the four
corner bore logs (X1, Z1, X5, Z5 in Figure 5.2) were used to calibrate the measured
HVSR curves at the same corresponding locations. The two remaining bore logs (X3 and
Z3) were used for verification of the HVSR results. The calibration methodology is
described first. At the start, and beginning from right to left, from high to low frequency
(i.e. from shallow to deep soil), the amplitude and frequency of the predominant and
secondary peaks of the HVSR curves were identified. These features represent possible
impedance contrasts identifying significant material changes picked up by the ambient
vibration recordings, as well as giving an indication of the number and the likely depth of
the soil layering. The thicknesses of the soil layers revealed by the bore log were used to
provide a reliable guide to constrain the layering thickness of the soil model applied in the
forward modelling. The soil type, the SPT data and depth of the water table moreover
provide feasible constraining ranges for the Vs and Vp values of the soil layers. These
constraining data were used to formulate an initial guess of the soil model while taking
cognizance of the identified features from the HVSR curve. A trial-and-error iteration of
the layered soil properties of the soil model was then undertaken with due regard to the
constraints and the identified features of the HVSR curve to seek an improved fit at each
successive iteration, until the best possible fit deemed by visual inspection is achieved. It is
- 89 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique
noted that during the trial-and-error process, some of the secondary peaks could be fitted
(i.e. the theoretical and measured HVSR curves are in agreement at those points in the
curves) with a soil model that is consistent with the bore log, while other secondary peaks
may not. The ones that did not fit were thus ruled out as reflecting the impedance contrasts
within the surface layers.
The calibrated soil models (namely, the ones with the best fit from the four corner
stations X1, Z1, X5, Z5) were subsequently applied as the initial guess to infer the Vs
profiles of the interspersing test stations inside the site area. Here, the procedures for
forward modelling are essentially same as the above, using the constraints adopted from
the nearest corner bore log, as well as the calibrated soil models as a starting guide.
Adjustments were then made to the initial guess with successive iterations and to the
constraints, if this is required, to achieve the best possible fit. It is noted that some
flexibility was exercised regarding the imposition of the constraints for these stations since
they are deemed not to be bound by the results from the bore logs. The HVSR curves for
stations X3 and Z3 (both with bore logs) were fitted blind assuming no prior knowledge
of the bore logs at these locations. This is because the results from X3 and Z3 were
subsequently used for the verification study as discussed below. The results from the
remaining stations were applied to fill the stratigraphy gaps at the locations not covered by
the bore logs. The trial-and-error visual inspection approach has been found to work well
in this study as exemplified by the goodness-of-fit results at stations X5 (calibration
case), Z3 (verification case) and Y5 (no bore log case) shown in Figures 5.6a, b and c
respectively (solid red lines show the measured HVSR curves and broken blue lines show
the theoretical HVSR curves). Similar levels of curve fitting were achieved for the
remainder of the stations but are not shown to avoid tedious repetition.

- 90 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique

a)

b)

c)

Figure 5.6: Measured and theoretical HVSR curves at stations (a) X5, (b) Z3 and (c)
Y5.
The resulting Vs profiles inferred from fitting the HVSR measurements at stations X5, Z3
and Y5 are shown in Figure 5.7. The constrained forward modelling in conjunction with
the bore log data helped to clarify and/or confirm the key features of the HVSR curves:
- 91 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique
1. A large jump in Vs occurs as the soil transitions from the surface layers to the shale
bedrock at 3.55 m (X5), 6 m (Z3) and 4.9 m (Y5), and this is confirmed by the
layering information from the bore log data. Working backwards from left to right
of the HVSR curve, the predominant HVSR peak with the lowest resonance
frequency (which is the fundamental frequency f0 ranging between 4.5 and 6 Hz for
this site) of the measured and the theoretical HVSR curves are fitted. Thus this
peak is confirmed as the one associated with the deepest impedance contrast
defined by the surface layers and bedrock interface.
2. Sizeable jumps occur in Vs at depths of 0.55 m, 1.2 m and 0.8 m, for stations X5,
Z3 and Y5 respectively, showing smaller impedance contrasts at more shallow
surface layers, which are also reflected in the layering data from the bore log. These
jumps are able to be confirmed as the impedance contrast associated with the fitted
resonance peaks between theoretical and measured HVSR curves, at frequencies of
approximately 37 Hz, 21 Hz and 29 Hz to the right of the fundamental resonance
frequencies.

Figure 5.7: Estimated Vs profiles for stations X5 (solid line), Z3 (broken line) and
Y5 (dotted line).

Verification of Results from HVSR Measurements


A profile-to-profile comparison was made between the HVSR estimated Vs profiles and
mechanical borehole data. Figure 5.8 shows the soil layering data extracted from the bore
logs aligned against the estimated Vs profiles at two locations, X1 and X5. The bore log is
shown to 10 m, the depth of refusal and/or termination of the augering in shale bedrock.
- 92 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique
The borehole data and Figure 5.8 clearly identify three distinct material layers, the
uppermost layer comprising fill material, the middle layer consisting of silty clay/sandy clay
overlying the third layer which is a shale bedrock. Inferred Vs profiles from trial-and-error
fitting of the HVSR curves at stations X1 (left) and X5 (right) are shown to be in good
agreement with the relative stratigraphic layering from the mechanical bore log, which
demonstrate that it is possible to calibrate the fitting of the HVSR curves to the bore log
satisfactorily using the above procedures. Figure 5.8 also shows the inferred Vs profile at
station X3 which has been fitted blind assuming no a priori knowledge of the bore log at
X3. The good agreement between the inferred Vs profile and the bore log at station X3 can
be observed in Figure 5.8.

Figure 5.8: Verification Boreholes at X1 (left), X3 (centre fitted blind) and X5


(right) with their respective estimated Vs profiles.
Figure 5.9 shows the relative stratigraphy of the Vs inferred from the HVSR measurements
along Line 5. X5 and Z5 are the calibration stations while Y5 has no bore log data. In
this instance, HVSR results at Y5 confirm that the relative stratigraphy could also be
reasonably predicted by a linear interpolation of the profiles at X5 and Z5 (joined by
broken lines). Although the HVSR justification of the linear interpolation appears trivial, it
adds confidence to the results from the site characterisation study.

- 93 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique

Figure 5.9: Verification Boreholes at X5 (left) and Z5 (right) with estimated Vs


profiles at HVSR stations X5 (left), Y5 (centre) and Z5 (right).
The estimated Vs profile inferred from the HVSR curve at station Y5 has also been verified
against the profile deduced from an independent MASW method. Here, a total of 19 single
component (vertical) geophones were connected at a spacing of 3 m continuously along
Line 5 (Figure 5.2). A 3.6 kg (8 lb.) sledgehammer was used to strike against a solid steel
plate, at a distance of 5 m from each end of the array, to generate the active source. Here
the data from the geophones were captured by a laptop via a multichannel digital seismic
acquisition system (Soilspy Rosina from Micromed). Several shots of the sledgehammer
were attempted to ensure that reliable and clear dispersion curves were obtained. The
actively generated noise was recorded for three seconds at a sampling rate of 512 Hz. The
measured dispersion curve was generated using a modified version of Herrmanns (1987)
program, which is designed to pick the energy maxima of the f-k spectrum. The theoretical
dispersion curve for each ground model was calculated according to a theoretical model
developed by Haskell (1953). A Monte Carlo algorithm was used to provide an automatic
computer guided fit, to minimise the error between theoretical and measured dispersion
curves. In this inversion procedure, Vs and the layer thickness (y) were the two parameters
that most greatly influenced the theoretical model.
Figure 5.10a shows the fitting of the theoretical and experimental dispersion curves from
the MASW test (theoretical dispersion curve is shown by the white broken line) while
Figure 5.10b above shows the superimposed Vs profiles from HVSR curve at station Y5
and the MASW measurements taken along line 5. HVSR station Y5 is chosen because it is
located at the mid-point of the MASW projection at line 5. The profiles from both
methods showed good agreement with respect to the relative stratigraphy and absolute Vs
values.
- 94 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique

a)

b)

Figure 5.10: Verification MASW projected Vs along Line 5 (a) Phase velocity
dispersion curve with active sources at both ends, and (b) MASW estimated Vs
profile at line 5 against HVSR estimated Vs profile at station Y5.

Empirical Relationship Vs and SPT-N


Due to the availability of SPT data, this study seeks to identify a relationship between
HVSR estimated Vs and SPT data for the proposed development site at UWS. The results
from the UWS Penrith (Kingswood) campus are plotted in Figure 5.11 representing Vs in
the range of 110 300 m/s versus uncorrected SPT-N in the range of 4 21. Ordinary
least-squares regression of the data points yielded the power law relationship (R2 = 0.93):
Vs = 47.5 N 0.60

(5.1)

- 95 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique
This relationship allows for an initial estimate of the SPT-N value when only the Vs value is
known, and may be applied to locations in this area which are not covered by the
mechanical SPT.

Figure 5.11: HVSR estimated Vs versus measured SPT-N values.

5.5 Application of the HVSRplus Technique


As mentioned above, a limitation of non-invasive geophysical techniques is the
requirement of providing constraining information to prevent the misapprehension of
measured data. Normally mechanical data (e.g. borelogs, SPT, etc) is used to provide this
vital constraining information. However, in order to reduce the necessity of mechanical
testing for constraining purposes, a novel two-step approach is proposed. This involves
combining the passive HVSR and the array based active MSOR techniques. This proposed
technique relies on the notion of the phase velocity dispersion curve, array based MSOR
technique providing constraining information to the HVSR technique. Moreover, the
MSOR technique relies on the same sensor used by the HVSR technique (consequently
named the HVSRplus technique) to record the active measurements, thus the required
equipment is minimised leading to lower costs and greater portability. The HVSRplus
technique was thus trialled at the UWS development site, before applying it to the
compacted ground at Penrith Lakes (Chapter 8). Following the methodology proposed in
Section 3.3 for the HVSRplus technique, a synthetic multichannel field recording was
obtained after employing the MSOR technique across a known borehole location. The
synthetic multichannel field recording is shown in Figure 5.12 below, and lasts for a
duration of 1 second.

- 96 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique

Figure 5.12: An example of synthetic multichannel field recording by the MSOR


technique. The vertical axis shows the geophone number and the horizontal axis
shows the time.
The above synthetic field recording was post processed to obtain a measured phase
velocity dispersion curve. Afterwards, a theoretical dispersion curve was fitted by trial-anderror to obtain a reasonable fit to the measured dispersion curve. Once successfully fitted,
the theoretical model used to create the theoretical dispersion curve, provided an estimate
of the Vs profile. Figure 5.13 shows both the measured and theoretical dispersion curves.
The measured dispersion curve is shown by the colour (shaded) contour, while the
theoretical dispersion curve is shown by the dashed line.

Figure 5.13: Measured and theoretical (dashed line) dispersion curves obtained by
the MSOR technique.
- 97 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique
Using only the Vs profile obtained by the above MSOR technique as an initial guess for the
HVSR technique, the theoretical HVSR curve shown in Figure 5.14 below was obtained.
The measured HVSR curve is also shown in Figure 5.14.

Figure 5.14: Measured (solid) and theoretical (dashed line) HVSR curves. The
theoretical HVSR curve produced using the initial guess Vs profile by the MSOR
technique.
Interestingly the fit between measured and theoretical HVSR is not too dissimilar. The
features (peaks) observed by the theoretical HVSR curve (using the MSOR techniques
initial guess) are also seen on the measured HVSR curve. Two predominant peaks are
observed on the measured HVSR curve, at approximately 5 Hz and 16 Hz. Two
predominant peaks are also observed on the theoretical HVSR curve, at approximately 5
Hz and 12.5 Hz.
This shows that the MSOR initial Vs profile guess is quite close to the final HVSR Vs
profile. Fine tuning is required by the HVSR forward modelling procedure to fit both the
frequencies and amplitudes of the peaks observed on the measured HVSR curve. Figure
5.15 shows the theoretical HVSR curve obtained after using a trial-and-error approach to
minimise the misfit between the theoretical HVSR curve and the measured HVSR curve.
The measured HVSR curve is also shown in Figure 5.15. Figure 5.15 shows a much better
fit of measured and theoretical HVSR data. Both the frequencies and the amplitudes of the
two peaks observed on the measured HVSR curve are fitted after fine tuning the HVSR Vs
profile.

- 98 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique

Figure 5.15: Measured (solid) and theoretical (dashed line) HVSR curves. The
theoretical HVSR curve was produced after minimising the fit between measured
and theory.
Figure 5.16 shows the inferred Vs profiles from both the MSOR and HVSR techniques.
The Vs profile inferred by the MSOR technique is shown by the dashed blue line. This Vs
profile was also used as the initial guess for the HVSR technique. The Vs profile inferred
by the HVSR technique after fine tuning the initial guess is shown by the solid red line. It is
evident that there is not much difference between the two Vs profiles.

Figure 5.16: Estimated Vs profiles obtained by the MSOR (used for the initial guess
for the HVSR technique) (dashed) and HVSR (solid) techniques.
The geological structure and the properties deduced from the borehole, MSOR Vs profile
and the HVSR Vs profile are summarised in Table 5.2. Both Vs profiles reasonably
represent the material types observed by the borehole. Contrasts between material types
(e.g. fill to natural material, natural material to shale) are also identified by both Vs profiles.
- 99 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique
Table 5.2: Summary of Vs profiles from both the MSOR and HVSR techniques, and
the borehole information.
Borehole
Depth (m)

Material type

0 1.2

Fill
Silts, sands and
clays
Shale

1.2 8.8
*

MSOR
Depth
Vs (m/s)
(m)
1.5
120
2
210
5
295
*
380

HVSR
Depth
Vs (m/s)
(m)
1.5
110
2
240
5.3
280
*
360

* Denotes the bedrock layer.

To verify the HVSRplus produced results, Figure 5.17 shows a profile-to-profile


comparison between the HVSRplus estimated Vs profiles and mechanical borehole and
SPT data. Figure 5.17 shows the soil layering data extracted from the bore log and the
uncorrected SPT-N values aligned against the estimated Vs profile.

Figure 5.17: Verification of the Vs profile produced by the HVSRplus technique with
borehole and SPT data.
The borehole is shown to the depth of termination of the augering in shale bedrock. The
borehole data clearly identifies three distinct material layers, the uppermost layer
comprising fill material, the middle layer consisting of silty clay/sandy clay overlying the
third layer which is a shale bedrock. The inferred HVSRplus Vs profile is shown to be in
good agreement with the relative stratigraphic layering from the mechanical borehole.
Furthermore, the inferred Vs profile is in agreement with the uncorrected SPT-N strength
parameter.
- 100 -

Chapter Five: Trial Geotechnical Site Investigation using the HVSR Technique

5.6 Summary
This chapter described application of the HVSR technique to characterise a potential
development site at the Penrith (Kingswood) campus of the UWS. The following
components of the HVSR technique for geotechnical site characterisation have been
established in this chapter: (1) determining the site fundamental resonance frequency and
the site amplification, the parameters reflecting the seismic response of the area, (2)
interpreting the measured HVSR curves to give a preliminary insight of the structure of the
measured ground, and (3) trial-and-error forward modelling of the theoretical HVSR
curves to estimate the Vs of the ground in conjunction with available borehole data, to
confirm preliminary assessments and to provide detailed results regarding the soil
stratigraphy. The results of the HVSR technique were verified against data from the
remaining boreholes as well as from an independent MASW method. This chapter has
shown that in collaboration with mechanical in situ tests, the HVSR technique can be
efficiently applied to characterise a site and especially to fill in the gaps of the soil
stratigraphy at the locations not covered by the localised and more costly mechanical tests.
Furthermore, a novel two-step forward modelling procedure, the HVSRplus technique was
trialled to determine its potential for geotechnical site characterisation. The simplified
MSOR technique successfully provided an initial guess for the Vs profile, as well as other
constraining information to the HVSR technique. This highlights the potential of the
proposed method and application at the Penrith Lakes quarry site (refer to Chapter 8).

- 101 -

Chapter Six: Investigation of Rolling


Compaction Using the HVSR Technique
In this Chapter, the HVSR technique and the novel methodologies proposed in this thesis
were applied to evaluate rolling compacted ground. This chapter presents a controlled
study of the evolution of compaction, showing that it is possible to identify levels of field
compaction by observing changes in HVSR data. Results from this study suggest that the
fast and operationally simple HVSR technique could be used to assess near surface ground
compaction. Furthermore, the methodology was applied to assess two methods of rolling
compaction at a very large quarry site:
1. Compaction Method 1 In areas yet to be rehabilitated but identified for future
infrastructure development, the filled grounds were compacted by heavy sheeps
foot (cylindrical drum with protruding spikes) roller compactors in strictly
controlled lifts to achieve compliance of either 95% or 98% Proctor standard
maximum dry density (SMDD), with an acceptable moisture content of 3 % of
the standard optimum moisture content (OMC). Fill material was placed in lifts of
either 400 mm or 450 mm (depending on location) and compacted until the
required surface level was reached. Compaction in this area was strictly controlled
with a large number of density tests conducted to ensure satisfactory compaction
was achieved.
- 102 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


2. Compaction Method 2 In areas intended for parklands and recreational uses only,
the ground has been subjected to a method specification compaction
(predetermined layer thickness of approximately 300 mm and minimum of 4 passes
of approved spreading equipment, e.g. CAT 825C of compaction equipment such
as sheeps foot rollers and graders). Here compaction was largely visually monitored
by the naked eye for compliance to a target density of 95% of SMDD. Density tests
were sparingly conducted (relative to Method 1 compaction) to verify the achieved
dry density.
The key features of the measured HVSR curves have been studied and analysed to infer
useful insights about the compaction achieved by the two methods. Furthermore, the
fitting of the measured HVSR curves by trial-and-error forward modelling, forms the basis
for inferring the shear wave velocity (Vs) profile and layer thicknesses of the compacted
ground. It is shown in this chapter that the process of analysing and interpreting the HVSR
curves, as well as the forward modelling of the HVSR curves can reveal useful information
about the quality and consistency of the compacted ground.

6.1 Penrith Lakes Quarry


The test areas are located at Penrith Lakes, a very large sand and gravel quarry in Australia,
situated on the Hawkesbury-Nepean River floodplain west of Sydney at the foot of the
Blue Mountains. Figure 6.1 shows an aerial photo of the Penrith Lakes quarry. It is
expected that quarrying will cease in a few years time when the site re-develops into 1,935
ha of urban and recreational precincts. During quarrying, the top overburden layer was
removed to expose the sand and gravel middle layer which was mined as a resource. The
bedrock is generally located between 13 and 15 m below the ground surface that prevails
before quarrying. The bedrock substrate within the region generally consisted of shale with
some sandstone beds, all part of the Wianamatta Group (e.g. Bringelly shale, Minchinbury
sandstone and Ashfield shale). Also, the geology within the region generally consists of
quaternary alluvium, gravel, sand, silt and clay thus explaining the presence of the quarry at
this location. The removed overburden which is a mixture of clayey, silty and sandy
material is then re-deposited as a fill at other areas of the site to raise the ground to the
required future level. A number of lakes and tailings dams have also been created as a result
of rehabilitation of past quarry activities.
- 103 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


Urban housing and associated infrastructure development were not anticipated during the
early days of quarrying operations, and the landfill sites were generally rehabilitated to
parklands or recreational use standards, and in some areas even less than this. The current
planning includes proposals for housing and infrastructure developments, thus
necessitating more stringent controlled compaction in several areas which were previously
compacted to parklands, recreational or low standards only. As a result of the mix of
future land use proposals, the history of remediation and size of the quarry site, different
methods of compaction have been applied at the site.

WLW

DCPT-9
MCP

LD

DCPT-7

Figure 6.1: Test locations at Penrith Lakes quarry.


This thesis has selected five different compacted areas from the quarry representing
different methods of compaction, with which to investigate the HVSR-based techniques.
As shown in Figure 5.1, these areas are:
1. Leaky Dam (LD) roller compacted to Compaction Method 2
2. McCarthys Pit (MCP) roller compacted to Compaction Method 1
3. Wildlife Lake West (WLW) roller compacted to Compaction Method 2

- 104 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


4. Dynamic Compaction Prototype Trial Area 9 (DCPT-9) combination of both
dynamic and roller (as an alternative to an ironing pass by dynamic compaction)
compaction
5. Dynamic Compaction Prototype Trial Area 7 (DCPT-7) dynamically compacted
without an ironing pass
The LD, MCP and WLW areas will be discussed in this chapter, whereas the DCPT-9 area
will be discussed in Chapter 7, and the DCPT-7 area in Chapters 8 and 9.

6.2 Compaction Evolution Observed via the HVSR


of Microtremors
This section focuses on an in-depth investigation into the changes of the HVSR curves in
response to changing stiffness of the ground. Hence a controlled study was undertaken in
which the HVSR technique was applied to different stages of roller compaction in an
attempt to capture the evolution of compaction, from freshly laid loose fill material to its
fully compacted state.
The test site for the control study is located in the abovementioned Penrith Lakes quarry,
in an area known as Leaky Dam (LD) as shown in Figure 6.1. The rehabilitative work in
this area required compacting the fill material to the standard of Compaction Method 2.

6.2.1 Interpretation of Measured HVSR Curves


The results are presented in Figure 6.2 showing three single HVSR curves at different
stages of roller compaction. A band of 2 standard deviation of the mean (SDOM) of
each curve characterising 95% confidence interval of the mean for each curve is also
shown. The HVSR curve in the initial stage, when the fill layer was still very loose (precompaction), is represented by the dash-dot line; the one for slightly loose (after initial
compaction) is represented by the broken (dash) line and finally the one for fully
compacted (post-compaction) is displayed as the solid line. Differences in the distinct
features of the HVSR curves reflecting the differences of the compaction stages are clearly
evident. The peak H/V ratio for the initial state (i.e. pre-compaction) is about 7
- 105 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


corresponding to a predominant resonance frequency f0 of about 5 Hz indicating a strong
impedance contrast between bedrock substrate and overlying surface soil. Also, a
significant secondary peak amplitude of 4 at about 30 Hz can be observed as a
consequence of the strong impedance contrast between the very loose top layer and the
underlying layers. However, the H/V spectral ratios for both predominant HVSR
frequencies reduce as the compaction effort was increased (due to decreasing impedance
contrast between overlying layers and bedrock). In particular, the secondary predominant
peaks become quite flat for the fully compacted stage, indicating consistency in the
compacted layers. Directly interpreting the features of the measured HVSR curves is at best
useful only in providing a preliminary assessment. Forward modelling of the HVSR curves
to infer the Vs profiles is essential to ascertain a more complete assessment of the
compaction state.

Figure 6.2: Evolution of HVSR curves with compaction (also showing the 2
standard deviation of the mean (SDOM) for each curve).

6.2.2 Forward Modelling to Infer the Vs Profiles


The general methodology to infer the Vs profile involves a trial-and-error process to model
and fit the theoretical to the measured HVSR curves. Since the theoretical HVSR curve
may not provide a unique solution to fit the measured HVSR curve, the total thickness of
the surface layers and the thickness of the top compacted lift were used as constraints.
Furthermore, the misfit between the theoretical and measured curves was evaluated by
visual inspection and the estimated Vs profiles were then normalised to minimise the
effects of overburden stress. The theoretical HVSR curves are shown overlaying the
measured curves in Figures 6.3a c (solid red lines show the measured HVSR curves and
- 106 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


broken blue lines show the theoretical HVSR curves), while the Vs profiles after
normalisation for the three different stages of compaction are presented in Figure 6.4.

a)

b)

c)

Figure 6.3: Compaction evolution of measured and theoretical HVSR curves. (a)
Very loose top layer, (b) Slightly loose top layer, and (c) Compacted top layer.
The increment of the Vs of the top layer from 130 m/s in the very loose stage to 290
m/s in the final stage of compaction can be observed in Figure 6.4. The secondary peaks
- 107 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


observed on the very loose and loose HVSR curves at approximately 30 Hz and 40 Hz
respectively, can be related to the shallowest layer of the Vs profile. At its fully compacted
stage, the Vs of the top layer is comparable to the Vs of the deeper compacted layers, thus
showing the consistency achieved by the roller compaction. Moreover, it is noted that the
thicknesses and Vs values of the remaining deeper layers remain virtually the same, with the
roller compaction having significant effect only in the top surface layer. This is perhaps due
to the rapid dissipation of the compaction energy within the top surface layer such that
very little energy was dispersed to the subsequent layers. The jump in Vs at the depth of
about 13 m is indicative of the presence of bedrock.

Figure 6.4: Compaction evolution of normalised Vs profiles.


This case study has shown the potential of utilising the non-invasive surface wave HVSR
technique for the evaluation of field compaction in terms of both quality and consistency.
The controlled study of the evolution of compaction has identified that improved levels of
field compaction could be observed by analysing HVSR data. Furthermore, the HVSR
technique appeals as an attractive method to cover the gaps not assessed by the more
expensive, localised and time consuming invasive mechanical testing (e.g. CPT, DMT, etc),
particularly for extensive and deep filled sites.

6.3 Stability of HVSR Curves


The ambient vibrations recorded at an observation point are a function of the ambient
vibration sources, the media (the surface layers and bedrock) and the paths travelled by the
- 108 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


generated waves from the source to the observation point. It is generally acknowledged
that the H/V curve remains quite stable and is the basis of the HVSR technique. However,
anthropic noise (from machinery, human activities, etc) has been known to have a slight
affect on the amplitude of the HVSR (SESAME 2004b). To assess the stability or
variability of the ambient vibrations and to identify the presence of artifacts, measurements
were recorded at the test areas over a 24 hour period. Figures 6.5(a) and (b) show the
horizontal and vertical spectra respectively at the MCP area (see Figure 6.1) over the 24
hours.

a)

b)

Figure 6.5: 24 hour (a) horizontal and (b) vertical spectra at the MCP area.
The horizontal and vertical vibrations from the same recordings were then processed for
each hourly interval to give the H/V spectra shown in Figure 6.6. It is now observed that
the HVSR curves at the MCP area were quite stationary over the 24 hour period. These
results indicate a high level of stability achieved with the measured HVSR curves with
respect to time and frequency of measurement at the observation point.

Figure 6.6: Consistency of HVSR curves at MCP area.


- 109 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


Finally this exercise was repeated for the WLW area (see Figure 6.1) and the results shown
in Figure 6.7 also confirm the stability of the HVSR curves in that area.

Figure 6.7: Consistency of HVSR curves at WLW area.

6.4 McCarthys Pit (MCP) Area


Compaction Method 1 was applied to the MCP area covering 1,200 m2. The fill material in
the MCP area extends down to bedrock substrate. The observation stations of the MCP
area where the HVSR measurements were taken are shown schematically in Figure 6.8. The
reason for the irregular grid layout is due to the spread and locations of prior independent
density tests. Thus HVSR measurements were conducted at strategic locations to take
advantage of these density tests.

Figure 6.8: MCP area test locations.


Dimensions given in metres (m).
- 110 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique

6.4.1 Interpretation of Measured HVSR Curves in MCP


Area
The superimposed HVSR curves from this area are presented in Figure 6.9 constituting the
MCP area signature. The common and identifiable key characteristics of the HVSR
curves of the Compaction Method 1 in the area are:
1. The presence of a peak or predominant resonance frequency, f0, at approximately
5.2 Hz.
2. A trough at approximately 2f0 due to Rayleigh wave ellipticity.
3. Aside from the possible first higher mode of f0 occurring at approximately 16 Hz, a
relatively smooth and flat curve with relatively small resonance peaks was observed
in the frequency range 10-50 Hz where the H/V ratio is generally less than one.

Figure 6.9: HVSR curves summary MCP area.


Useful insights of the compacted ground in the MCP area were inferred by directly
interpreting the key features of the HVSR curves, prior to fitting the HVSR curves to
estimate the Vs profile. Thus in (1) above, this is indicative of the impedance contrast
between the bedrock substrate and the overlying softer surface layers. A direct
interpretation of the peak amplitude of the HVSR curves is used only as an approximation
technique in providing a preliminary assessment. The amplitudes of the HVSR curves at f0
were generally quite consistent suggesting that quite consistent compaction has been
achieved from observation station to station in the area. As the depth to the bedrock is
inversely related to f0, the lower the magnitude of f0 the deeper will be the bedrock. The
trough in (2) is a common feature in HVSR curves due to Rayleigh wave ellipticity (Fah et
- 111 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


al. 2001; Tuan et al. 2011). Taking into consideration that this area has been prepared and
compacted with great detail for quality assurance (see Table 6.1), this trough can be
interpreted as having no other significance than a manifestation of Rayleigh wave ellipticity.
Table 6.1: Dry density profile in MCP area.
Density (t/m3)
No. of Density
Depth (m)
Range
Tests
02
24
46
68
8 10
10 12
12 14

1.90 2.01
1.90 2.00
1.90 1.95
1.99 2.00
1.99
1.91 1.95
1.93 2.01

6
3
4
2
1
2
5

Overall

1.90 2.01

23

Table 6.1 shows a total of 23 density tests with results ranging from 1.90 2.01 t/m3,
between depths of 0.3 m and 13.5 m. There are a further 3 density tests which have been
deemed as outliers, with density readings of 1.81, 1.86 and 1.87 t/m3, located at depths of
9.7, 8.4 and 10.7 m respectively, below the surface. These outliers nevertheless still satisfy
their target density of 95% SMDD and 3 % of OMC. The 26 density tests covering the
MCP area grid represent a testing frequency of approximately 1 test per 646 m3 of ground,
compared to the specified minimum of 1 test per 1,000 m3. Finally, it may be inferred that
the compaction with depth at each station was relatively uniform by attribution to the
relatively smooth and flat upper section of the HVSR curve mentioned in (3) above, and
this was verified by the consistency of the results from the dry density tests. In saying that
the higher frequencies of the measured HVSR curve are generally indicative of the ground
stiffness layering in the near surface, it should also be noted that the secondary peak at
approximately 16 Hz or 3f0 could be attributed to another effect, that of the first higher
mode of f0 (see, Section 3.5).

6.4.2 Forward Modelling to Infer the Vs Profile in the MCP


Area
A more complete assessment of the compacted area by fitting the HVSR curves was then
made to estimate the Vs profile of the ground. An initial calibration was undertaken by
- 112 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


trial-and-error forward modelling to fit the theoretical HVSR curve to the measured HVSR
curve at station Z3. Figure 6.10a (solid red line shows the measured HVSR curve and
broken blue line shows the theoretical HVSR curve) shows the fitting of the HVSR curve
using the 4-layered soil model shown in Figure 6.10b.

a)

b)

Figure 6.10: MCP area station Z3 (a) measured and theoretical HVSR curves, and
(b) normalised HVSR Vs and estimated (before normalisation) HVSR Vs.
The HVSR curve at station Z3 is considered as one of the worst cases in terms of the
uniformity of the compaction with depth due to the appearance of a small secondary
resonance peak at 16 Hz. The forward modelling shows that the secondary peak did not
significantly affect the uniformity of the normalised Vs of the compacted ground which
correlates well to the uniformity in terms of the dry density. The 4-layered soil model was
then used as an initial guess for the fitting of the HVSR curves for the remaining stations in
the MCP area (as well as the WLW area discussed below because of the use of the same fill
material and similar compaction methodology). Examples of the fitted HVSR curves at
- 113 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


stations A1, B1 and B3 in the MCP area are presented in Figures 6.11a c (solid red lines
show the measured HVSR curves and broken blue lines show the theoretical HVSR
curves) to illustrate the goodness-of-fit by trial-and-error forward modelling.

a)

b)

c)

Figure 6.11: MCP area measured and theoretical HVSR curves (a) A1, (b) B1, and
(c) B3.
- 114 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


Similar goodness-of-fit was achieved for the remaining stations of the MCP area but is not
presented in this chapter to avoid tedious repetitions. However, the normalised Vs profiles
inferred from all stations for this area (including the aforementioned stations) are shown in
Figure 6.12.

Figure 6.12: MCP area normalised Vs profiles summary.


The relatively uniform profiles in the surface layers have been validated by the large
number of independent density tests conducted in the area reporting relatively uniform dry
density as discussed above. The normalised Vs profiles in the compacted surface layers help
to clarify and/or confirm the interpretations of the following characteristics of Compaction
Method 1 in the MCP area:
1. The presence of a peak or predominant resonance frequency, f0, at approximately
5.2 Hz is associated with the sharp increase in Vs at the depth of the bedrock
between 13 14 m.
2. The trough at 2f0 in the HVSR curves was confirmed as being attributed to the
Rayleigh wave ellipticity and not because of the embedded softer/looser layers
within the surface layers.
3. The uniformity of the profiles agrees with the interpretation that the relatively
smooth and flat HVSR curves observed at higher frequencies (shallow depths) is
indicative of the relatively consistent compaction achieved. The secondary
resonance peak at approximately 16 Hz is also confirmed as the first higher mode
of f0 rather than a consequence of non-uniform compaction (i.e. impedance
contrast) at the very near surface of the ground.

- 115 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique

6.5 Wildlife Lake West (WLW) Area


The WLW area was compacted to Compaction Method 2 specifications, and in some
parts to a total thickness of more than 14 m above the bedrock substrate. The section
investigated spans over an area of 1,800 m2. The observation points of the WLW area are
shown in Figure 6.13.

Figure 6.13: WLW area test locations.


Dimensions given in metres (m).

6.5.1 Interpretation of Measured HVSR Curves in WLW


Area
The superimposed HVSR curves from this area constituting the WLW area signature are
presented in Figure 6.14. Two common key characteristics identified in the HVSR curves
in this area are:
1. The presence of a peak or predominant frequency f0 at approximately 4.3 Hz
2. The section of curve at frequencies above the secondary resonance peak (between 8
to 12 Hz) is relatively smooth for the majority of the curves. The particular
secondary peak (which occurs more prominently in a minority of curves) may
include the first higher mode of f0.
Once again the presence of the predominant frequency is attributed to the large impedance
contrast between the bedrock substrate and the softer surface soil layers overlying bedrock.
- 116 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


The large variation of the HVSR amplitudes at f0 would suggest less consistent overall
compaction achieved from station to station within the area. The relatively flat section at
frequencies above f0 (except for the secondary resonance peak between 8 to 12 Hz) in a
majority of the curves suggests that a relatively uniform depth-wise compaction has been
achieved at each of these stations even though the less rigorous Compaction Method 2 was
applied in this area, unlike the fluctuating curves with prominent secondary peaks in the
same high frequency range that would occur for highly variable surface layers. However,
the same consistency or flatness of the curve is not achieved when compared to the MCP
area.

Figure 6.14: HVSR curves summary WLW area.

6.5.2 Forward Modelling to Infer the Vs Profile in the WLW


Area
The HVSR curves of the WLW area were also fitted to infer the Vs profiles of the area to
give a more complete assessment of the compaction achieved. The fitted HVSR curves at
grid points A3, B3 and E2 shown in Figures 6.15a c (solid red lines show the measured
HVSR curves and broken blue lines show the theoretical HVSR curves) illustrate the
goodness-of-fit. A summary of the normalised Vs profiles for the stations in the area is
shown in Figure 6.16. Greater station-to-station variability of the Vs profiles is observed in
Figure 6.16 than in Figure 6.12 and this is indicative of the less consistent station-to-station
compaction achieved in the WLW area. However, the relatively uniform normalised Vs
profile at most of the stations would also suggest that the compaction was moderately
consistent with depth except for a minority of stations (namely B2, C2 and E2).
- 117 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique

a)

b)

c)

Figure 6.15: WLW area measured and theoretical HVSR curves (a) A3, (b) B3, and
(c) E2.
Forward modelling of the HVSR curves at stations B2, C2 and E2 yielded outlier Vs
profiles which are shown as the dashed lines in Figure 6.16, conveying some differences of
compaction quality with depth. The forward modelling results in general indicate that the
Vs of the compacted layers for the WLW area is noticeably lower than the MCP area.
These results are consistent with the dry density tests confirming that the 95% SMDD was
- 118 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


achieved at the random and sporadic test spots. However, it is not possible to confirm that
the same standard of compaction was achieved elsewhere in the untested area. Finally, the
large jump in the Vs profile at approximately 13 14 m depth could be identified with the
bedrock layer in this area. The forward modelling results corroborate and/or confirm the
preliminary assessments from interpreting the HVSR curves as follows:
1. The predominant resonance at approximately 4.3 Hz is linked to the bedrock at a
depth of 13 14 m.
2. The curves with a comparatively flat section over the high frequencies have resulted
in moderately uniform Vs profiles in the near surface indicating a moderate degree
of consistency in compaction even though Compaction Method 2 was applied in
this area.
3. The curves with a prominent secondary peak between 8 to 12 Hz are shown to
relate to a large impedance contrast within the surface layers indicating some
differences in compaction achieved with depth.

Figure 6.16: Normalised Vs profiles of WLW area.

6.6 Assessment of Compaction Quality


The results from the microtremor measurements have been applied to generate the metrics
to make an informed assessment of compaction quality. In the absence of Vs profiles, the
amplitude of the predominant resonance frequency, HVSR(f0) could give a preliminary
indication of the quality and consistency of compaction of the area in an overall sense.
HVSR(f0) of the MCP area varied between 3.1 and 5.0 (mean = 4.0, standard deviation =
- 119 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


0.6), and between 3.0 and 12.1 (mean = 7.1, standard deviation = 2.9) for the WLW area.
Although the areas are different in size and number of recording stations (20 and 15 in
MCP and WLW areas respectively), the mean and the standard deviation of the amplitude
HVSR(f0) were both significantly smaller in the MCP area than the WLW area. This is
apparent from a visual inspection of the HVSR curves in Figures 6.9 and 6.14. Since the
amplitude, HVSR(f0), is directly related to the impedance contrast between the (compacted)
surface layers and the bedrock substrate, this metric suggests a better quality and
consistency of compaction was achieved in the MCP area than the WLW area. It should be
noted that this exercise is at best only able to give a preliminary assessment of the quality
and consistency of compaction. Careful interpretation is required when using HVSR(f0),
since it is not solely reliant on the local stratigraphy.
A more informed assessment of the quality of compaction was made by analysing the Vs
profiles of the HVSR observation (measurement) stations in the areas. The plots in Figure
6.17 of the mean Vs and 95% confidence interval against depth convey immediately the
overall sense of compaction quality in the two areas.

Figure 6.17: Mean Vs with depth with 95% confidence interval.


Figure 6.18 plots the overall mean of the normalised Vs at the stations from the MCP and
WLW areas against their standard deviation, giving a sense of the compaction achieved
laterally. As the data in Figures 6.17 and 6.18 show, the following metrics of the Vs profiles
have consistently indicated that a better quality and consistency of compaction was
achieved in the MCP area: (1) mean value of Vs with depth in Figure 6.17, a higher value
indicating higher modulus in the compacted layers of the area. A simple comparison of the
- 120 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


mean Vs profiles at 1 m intervals of the two areas shows the MCP area having
approximately 11 34 % higher Vs values than the WLW area, (2) mean value of Vs in
Figure 6.18, a higher value indicating higher average modulus in the compacted surface
layers of the station in the area, and (3) the standard deviation of Vs in Figure 6.18, a lower
value implying more consistent compaction in the compacted surface layers of the station
in the area. The mean and standard deviation of the normalised Vs of each station, namely
(2) and (3) of the above list, are the metrics chosen to characterise the quality of
compaction in a lateral sense in this appraisal.
The points from the two areas are clustered and bounded by the ellipse formed by solid
(MCP) and broken (WLW) lines. It is noted that there are three outliers (particularly at
stations B2, C2 and E2) which did not fall within the WLW (broken line) cluster. As
mentioned above, these three stations produced Vs profiles which are significantly different
to the majority of WLW Vs profiles. Figure 6.18 shows the MCP cluster situated higher up
and to the left of the WLW cluster thus indicating a higher mean Vs (better quality) and
lower standard deviation (higher consistency) in the MCP area than the WLW area.

Figure 6.18: Mean Vs of the compacted material versus standard deviation.


The metrics for both areas could thus serve as a basis for benchmarking future
Compaction Method 1 and Compaction Method 2 areas within the quarry site. In
particular, the HVSR techniques and metrics described herein would provide an efficient
and cost-effective means to assess future Compaction Method 2 areas which lack the same
stringent monitoring and rigorous control of Compaction Method 1.

- 121 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique

6.7 Summary
This chapter has shown the potential of using the non-invasive surface wave HVSR
technique for the evaluation of field compaction in terms of both quality and consistency.
A controlled study of the evolution of compaction has identified that improved levels of
field compaction could be observed by analysing HVSR data. Furthermore, this chapter has
applied the HVSR technique to assess two rolling compacted areas located within the
Penrith Lakes quarry. Two methods of rolling compaction: (1) Compaction Method 1
strictly controlled rolling compaction and (2) Compaction Method 2 visually monitored
rolling compaction by naked eye applied in two areas of the site were studied.
The key features of the measured HVSR curves from the two areas were interpreted to
make a preliminary assessment of the compaction carried out in the compacted areas.
Interpretation of the HVSR curve by itself can provide only a relative indication of the
compaction quality achieved. A more complete assessment was then made by fitting the
HVSR curves to infer the Vs profile conveying the quality of compaction with depth at the
measuring stations. The forward modelling also helps to clarify and/or confirm the key
features of the HVSR curves. However, as mentioned above the HVSR technique by itself
can provide only a relative stratigraphy and the theoretical HVSR forward modelling
without constraining information does not provide a unique solution to fit the measured
HVSR curve. To have an absolute stratigraphy, it requires calibration or constraining to
independent information (e.g. a priori knowledge of the consistency of compaction, dry
density tests, total thickness of the surface layers, etc). The metrics for the MCP area
revealing better and more consistent compaction were verified by independent dry density
tests. The metrics for the WLW area showed moderately consistent compaction at most
stations and lesser consistent compaction in the remaining stations. The variations of the
compaction from station to station were much wider. These results are also consistent with
the less stringent standard of monitoring applied for Compaction Method 2 in the WLW
area, and verified by the randomly sporadic dry density tests for the area. As a result of the
work in this chapter, an informed and objective assessment methodology has been
proposed for compacted areas in terms of quantifiable compaction metrics. Future
Compaction Method 2 areas lacking rigorous dry density tests or other forms of stringent
monitoring, in particular, would benefit from applying the HVSR approaches described in
the present chapter to assess the quality of compaction achieved.
- 122 -

Chapter Six: Investigation of Rolling Compaction using the HVSR Technique


Several areas in this quarry have undergone dynamic compaction; with some areas
consisting of deep fills reaching a depth of approximately 13 m. Compaction of these deep
filled areas was evaluated using CPT and DMT penetration testing. Due to the relatively
expensive, time consuming and localised nature of the tests, only a small proportion of the
dynamically compacted areas were actually evaluated. This may lead to areas with
problematic or poor compaction remaining unnoticed. The current chapter verified the
potential of utilising the HVSR technique to evaluate compaction for both quality and
consistency (namely, a high standard of compaction is uniformly achieved throughout the
compacted area). It is thought that the HVSR technique proposed in this thesis would be
very suitable for application in these deep filled dynamically compacted areas. Ideally the
HVSR techniques can be utilised to complement the invasive mechanical CPT and DMT
tests and thus cover the large gaps not actually being assessed by the invasive penetration
methods.

- 123 -

Chapter Seven: Investigation of


Composite Dynamic and Rolling
Compaction Using the HVSR Technique
The HVSR technique and the methodologies proposed in thesis were applied to evaluate
composite dynamic and roller compaction. This chapter presents an interesting and unique
case study of a composite compacted site where the upper section of dynamically
compacted material achieved in the first stage of compaction was subsequently removed,
reinstated and re-compacted in lifts in the second stage using conventional roller
compaction. Dynamic compaction was initially employed in this area because of the need
to densify deep fill materials. The key features of the measured HVSR curves were
interpreted to give a preliminary insight on the quality of compaction achieved.
Furthermore, a trial-and-error forward modelling procedure fitting the theoretical HVSR
curve to the measured HVSR curve then allowed the shear wave velocity (Vs) profile
conveying the compaction quality of the compacted ground to be inferred. An initial
calibration was carried out to match the inferred Vs profile in a relative sense against the
CPT data at a test location. Verification was further made by comparing the inferred Vs
profiles against independent CPT, DMT and dry density data. The HVSR technique was
then applied to appraise the consistency and quality of compaction at grid points not
covered by the localised mechanical and other independent tests.
- 124 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique

7.1 Dynamic Compaction Prototype Trial Area 9


(DCPT-9)
The composite dynamic and roller compaction test area, otherwise known as DCPT-9, is
shown in Figure 6.1. This area was initially dynamically compacted because the deep fills
did not previously receive adequate compaction for potential future infrastructure
development. There are limited historical records for the standard of compaction carried
out at the DCPT-9 area, as well as the DCPT-7 area (discussed in the next chapter) before
undertaking dynamic compaction. Hence, the compaction standard before the dynamic
compaction in these areas is generally unknown and probably quite variable. Based on
information provided by Penrith Lakes Development Corporation (PLDC), these areas
were initially backfilled from 1975 to 1976, and were further filled and shaped by PLDC
between 1993 and 1997 (Coffey Geotechnics Pty Ltd 2007a, 2007b). Based on historical
data and old aerial photographs, the areas were likely filled by both scraper and truck
dumping, and spreading methodologies. Thus in order to ensure the uniformity and higher
standard of compaction, the filled material in these areas were subjected to dynamic
compaction. Moreover, it is reasonable to assume that the compaction undertaken at
Wildlife Lake West area (i.e. rolling compaction method 2) is of a higher standard than the
dynamically compacted areas (i.e. DCPT-9 and DCPT-7) prior to the dynamic compaction.
The DCPT-9 study area stands over 6,000 m2 and runs to a depth of approximately
between 12 and 13 m, at the level of the bedrock substrate. The groundwater table (GWT)
in this area is approximately 10 m below the surface. The DCPT-9 area has been
dynamically compacted using a 1.8 m wide, 20 tonne (20 x 103 kg) octagonal shaped
pounder attached to a crane, which was dropped from a height of about 23 m above the
ground, in 3 phases. The weight was dropped 16 times at each station spaced 4.5 m apart,
imposing energy of approximately 460 tonne-m at each drop. After each pass the resultant
craters were surveyed, backfilled and re-levelled with the ground surface, using locally
derived material. High impact energy from the falling weight leads to rearrangement of the
soil particles, closer particle packing and reduction in the air voids, that is densification of
the soil. Within the zone of influence, strains are very high, and are high enough to cause
particle rearrangement and densification. Beyond this zone, the wave still propagates, but at
a smaller strain where the behaviour of the soil is more elastic in nature. The degree of
densification or compaction achieved from this form of compaction depends on the height
- 125 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique
of drop, falling weight, number of drops, drop spacing, soil type and so forth (e.g. Feng et
al. 2010).
After reviewing the mechanical CPT and DMT data subsequent to the initial dynamic
compaction, it was decided to remove approximately 3 m of the upper section, and recompact the reinstated material by roller compaction. This option was chosen instead of
utilising an ironing pass (using a lighter pounder and lower drop height) mainly due to the
availability of rollers, scrapers and other heavy machinery required for the roller
compaction at the quarry site. The roller compacted material was compacted in 450 mm
lifts to achieve compliance of either 95 or 98% (depending on location and depth) Proctor
standard maximum dry density (SMDD), with an acceptable moisture content of 3 % of
the standard optimum moisture content (OMC). This addition of roller compacted material
above dynamically compacted material, presents an interesting and unique case study of a
composite compacted site. Figure 7.1 shows the 100 m x 60 m grid of regularly spaced
observation stations of the HVSR recordings covering the DCPT-9 area. BH indicates a
combination of auger drilling, CPT and DMT, and CPT indicates a standalone CPT.

Figure 7.1: DCPT-9 area test locations.


Dimensions are in metres (m).
Samples taken from locations close to the surface of the DCPT-9 area have been analysed
by Heitor et al. (2012), to reveal several soil properties (Table 7.1) and a standard Proctor
- 126 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique
compaction curve (Figure 7.2). Heitor et al. (2012) also used a 50 mm mould but modified
the specific compaction energy required to meet the standard Proctor compaction energy.
This was made possible by compacting in 3 layers using 24 blows per layer, with an
equivalent compaction energy of 529 kN.m/m3.
Table 7.1: Soil properties of the DCPT-9 area
(Heitor et al. 2012).

1Unified

Silty Sand
SP-SM
25.5
10
79%
21%
5.0
0.99
2.70

20.0

Dry unit weight (kN/m3)

Properties
USCS classification1
Liquid limit
Plastic index
Sand (%)
Fines (%)
Cu (coefficient of uniformity)
Cc (coefficient of curvature)
Specific gravity

Standard Mould
50 mm mould

19.5
19.0

Zero air void


line

18.5
18.0
17.5
6

soil classification soil symbol

10

12

14

16

18

20

Moisture content (%)

Figure 7.2: Compaction curve from


the DCPT-9 area (Heitor et al. 2012).

7.2 Stability of HVSR Curves


As mentioned in the previous chapter, it is generally accepted that the HVSR curve remains
quite stable due to the basis of the HVSR technique. Nevertheless, anthropic noise (or
man-made noise) has been known to have a slight affect on the amplitude of the HVSR.
Therefore the stability and variability of the ambient vibrations were also assessed at the
DCPT-9 area, by conducting measurements over a 24 hour period. The 24 hour HVSR
curves observed at the DCPT-9 area also generally show good stability throughout the
frequency range except for the appearance of double peaks at or slightly above the
predominant resonance frequency for curves recorded during certain hours. The
occurrence of the double peaks is suspected to be linked to the presence of artifacts from
traffic on a well utilised road located roughly 600 m away to the east of the area. In Figure
7.3, the additional peak marked as Peak 2 caused by artifacts also appear to reduce the
amplitude of the HVSR curve at f0, marked as Peak 1. However, the rest of the curve
including the remaining peaks and troughs do not appear to have been affected. Moreover,
it is observed that the peaks (including the predominant resonance frequency) and troughs
of all the curves are consistently measured at the same frequencies.
- 127 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique

Peak 1
Peak 2

Figure 7.3: Consistency of HVSR curves over 24 hour period at DCPT-9 area.
In an attempt to isolate the effects of the artifacts and help identify possible causes, the
superimposed measured HVSR curves from the DCPT-9 area (Figure 7.3) have been split
into two separate figures. Figure 7.4a, from recordings between 10 pm and 5 am, represents
the measured HVSR curves with predominantly stratigraphic origin whereas Figure 7.4b
shows the measured HVSR curves with both stratigraphic and artificial origin (artifacts)
derived from recordings between 6 am and 9 pm.

Figure 7.4: HVSR curves over 24 hour period at DCPT-9 area with peaks of (a)
stratigraphic origin and (b) stratigraphic and artificial origin.
In Figure 7.4(b) an artifact can be identified at approximately 5.8 Hz, with the amplitude of
the peak of the artifact (Peak 2) changing in every curve (or with time). To understand the
origin of the artifacts and to identify the reason for the amplitude of the artifact changing
with time, Figure 7.5 shows the hourly measured HVSR curves that are superimposed in
Figure 7.3. The HVSR curves show frequency (Hz) on the x-axis and HVSR on the y-axis.
The curves start from 11 am on the first day until 10 am the following day and are arranged
in chronological order from left-to-right and top-to-bottom. It may be observed that Peak
- 128 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique
2 produced by the artifact slowly increase in amplitude from 3 pm and reaching a
maximum during evening peak hour traffic (5 pm 8pm) before vanishing after 10 pm.
During the intervening period 10 pm to 5 am the artifact peak is no longer present due to
low utilisation of the road, however it re-emerges at 6 am through to 10 am, a period which
once again corresponds to morning traffic. Therefore, by analysing the hourly HVSR
curves in Figure 7.5, the correspondence between the road traffic and origin of the artifact
at the DCPT-9 area is quite evident.

Figure 7.5: Hourly HVSR curves over 24 hour period 11 am to 10 am next day in
order from left-to-right and top-to-bottom.
(Y-axis HVSR amplitude and X-axis Frequency (Hz))
Further identification of anthropic noise is possible by analysing the time history plots of
the recordings, directional HVSR and single component spectra. According to Figure 7.5,
the recording at 6 pm shows the anthropic peak when it is at its highest and the recording
at 2 am shows virtually no sign of the anthropic peak. Therefore, comparisons of the
recordings at 6 pm and 2 am have used to identify the presence of the anthropic peak. The
time history plots show the change in HVSR of the analysed recording, or in other words
- 129 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique
the HVSR curve for each window (refer to Section 3.1). Figure 7.6 shows the time history
plots for the 6 pm and 2 am recordings. The anthropic peak is clearly shown in the 6 pm
recording where there is a high amplitude band across the entire recording at approx. 5.8
Hz. However, the band is not seen in the recording at 2 am but instead the HVSR of the
predominant peak just above 4 Hz is shown to have higher amplitude. This could suggest
that the anthropic peak may possibly have affected the predominant peak.

Figure 7.6: Time history of recordings at 6 pm (left) and 2 am (right).


The directional HVSR plot takes advantage of the HVSR techniques two horizontal
components (North South and East West) to show the direction in which ambient
vibrations are measured. Since the velocimeter was aligned with the North axis, the N-S
component is represented by the 00 and 1800, and the E-W component is represented by
the 900. Figure 7.7 shows the directional HVSR plots for the 6 pm and 2 am recordings.
The 6 pm directional HVSR plot shows a concentration of noise at 900 or the direction of
the nearby road, at both the predominant peak frequency at approximately 4 Hz and
anthropic peak frequency at approximately 5.8 Hz. However, the directional HVSR plot at
2 am shows a more balanced spread of noise across the predominant peak frequency band
in all directions, with virtually no sign of noise from the anthropic peak.

Figure 7.7: Directional HVSR at 6 pm (left) and 2 am (right).


- 130 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique
The single component spectra show the spectra for each of the three components before
the division of the averaged horizontal components by the vertical component, to produce
the HVSR (see Section 3.1) curve. Anthropic noise can be easily identified when all three
components show a spike at the same frequency (SESAME 2004b).

Figure 7.8: Single component spectra at 6 pm (top) and 2 am (bottom).


Figure 7.8 above shows the single component spectra for both the 6 pm and 2 am
recordings. The single component spectra for the 6 pm recording shows attributes of
anthropic noise at both the predominant peak (approx. 4 Hz) and anthropic peak (approx.
5.8 Hz) frequencies. The spikes observed in the 6 pm recording are not seen in the 2 am
recording, instead a much smoother eye-shape is shown, indicating no influences from
anthropic activities.
At the DCPT-9 area only a limited time window, from 8 am to 4 pm (during normal site
operating hours), is available for testing so it is important to verify that measurements
taken during this period are stable. Although artifact noise is present during this period, the
artifact peak has been identified and isolated to avoid misinterpretation. Shown in Figure
7.9 are the HVSR curves from 8 am to 4 pm extracted from the 24-hour data revealing
stable resonance frequencies and amplitudes. The reasonably high level of stability achieved
gives confidence that the HVSR measurements during the 8 am to 4 pm time window in
the DCPT-9 area should provide a meaningful and consistent basis for interpretation of the
compacted ground at observation points within the area.
- 131 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique

Figure 7.9: HVSR curves during time window 8 am 4 pm at DCPT-9 area.


In order to evaluate the influence of the variance of the ambient vibrations over a 24 hour
period on the normalised Vs profiles, a sensitivity analysis was performed by fitting the
theoretical HVSR curves from the forward model to all 24 hourly measured HVSR curves
from Figure 7.5. Figure 7.10 shows the average normalised Vs profile inferred from the
theoretical-to-measured HVSR fittings, and also the profiles one standard deviation
removed from the average.

Figure 7.10: Comparison of the normalised Vs profiles from the hourly HVSR curves
over 24 hours in Figure 7.5.
This figure shows that the variance of the ambient vibrations is mostly affecting the
uppermost surface layers (less than 1 m depth), with the top three layers producing
standard deviations of approximately 22, 7 and 6 m/s, from first (top) to third layer
respectively. The effect on the deeper surface layers (more than 1 m depth) was quite
- 132 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique
marginal. Even though the standard deviations show some variation in the top three layers,
particularly the first (top) layer, it was concluded that the HVSR technique produced
reasonably consistent and stable Vs layering over a 24 hour recording and any variations
were not significant enough to affect the conclusions drawn throughout this study.

7.3 Interpretation of Measured HVSR Curves in


DCPT-9 Area
The HVSR measurements drawn from all observation stations in the area are
superimposed in Figure 7.11 constituting the signature of the local compacted layers.
Taking into consideration expected variations, the following common and identifiable
salient characteristics for the entire grid area were clearly observed in the superimposed
curves, namely:
1. Peak of large amplitude at predominant peak f0 (approx. 4 to 4.5 Hz), indicating a
strong impedance contrast between the bedrock and overlying surface layers.
2. A trough between 6.5 Hz and 10 Hz (where the H/V ratio dips below one) which
occurs at approximately 2f0 commonly regarded as a feature of the Rayleigh wave
ellipticity (Fah et al. 2001; Castellaro & Mulargia 2009b). It is possible, however,
that the Rayleigh wave ellipticity may have obscured a weak embedded layer lying
just above the bedrock as discussed below.
3. A possible first higher mode of f0 at approximately 12 Hz (corresponding to 3f0)
and other discernible secondary resonance peaks at frequencies above the
predominant resonance frequency.

Figure 7.11: HVSR curves superimposed summary DCPT-9 area.


- 133 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique
A few comments regarding the preceding characteristics are worth highlighting. The link
between the predominant peak and the fundamental resonance frequency of the surface
layers was pointed out in previous studies (see, e.g. Bonnefoy-Claudet et al. 2006b and
references therein).
In item (2), although the existence of a trough was anticipated at 2f0 due to Rayleigh wave
ellipticity, the hallmarks of this trough were consistent with a Vs decrease (pointing to an
embedded softer/loose layer) just above the bedrock substrate. The association between
the existence of a velocity decrease within the surface layers and H/V ratio of a trough
dipping below one was previously pointed out by Castellaro and Mulargia (2009b), who
have referred to this phenomenon as a velocity inversion. They showed that velocity
inversions leave an empirical signature as a general decrease of the H/V ratio to amplitude
below one over a wide range of frequencies, due to the decrease of the horizontal
components below the vertical one. The de-amplification of the HVSR amplitude appears
to be due to the decay of the spectra of the horizontal components, with a smaller effect
on the vertical component spectra. However, they also stressed that a persistent H/V ratio
< 1 does not indisputably indicate the occurrence of a velocity inversion, but instead that a
detectable velocity inversion entails a persistent H/V ratio < 1. It should be noted, that the
logarithmic scale used on the x-axis may optically mislead the occurrence of 2f0, when it is
in fact a velocity inversion. Velocity inversions can be encountered naturally (e.g., gravel
overlying silts/clays, cavities) or artificially (e.g., paving, asphalting above silts/clays, etc)
(e.g. Castellaro & Mulargia 2009b). Moreover, the inference that the embedded
softer/looser layer exists just above bedrock is based on the observation that the trough
occurs at a frequency just above f0, and f0 being associated with the bedrock/surface layers
interface. The presence or otherwise of the softer/looser layer can be clarified by both the
constrained forward modelling of the HVSR curve to reveal the Vs profile and available
independent mechanical data.
In item (3), the presence of secondary resonance peaks which were not the higher modes
of f0 indicate strong impedance contrasts within the surface layers and gives an insight of
the consistency of the compaction. As a consequence of the first higher mode of f0 possibly
occurring at 12 Hz, it could not be concluded simply from the reading of the HVSR curves
whether an actual impedance contrast exists in the surface layers at a depth corresponding
to 12 Hz. Hence, constrained forward modelling of the HVSR curve and/or the use of
independent mechanical tests is required to clarify this ambiguity (see, Section 3.5).
- 134 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique

7.4 Forward Modelling to Infer the Vs Profile in


DCPT-9 Area
Although the HVSR curves revealed useful insights of the compacted ground, some
features of the curves remained ambiguous and needed clarification. Thus in a follow up
step, the HVSR curves were fitted to infer the Vs profile which conveys the compaction
quality of the surface layers of the compacted ground.
Before the main fitting of the HVSR curves for the DCPT-9 area, however, an initial
calibration was carried out by matching the HVSR data from station A4 to the CPT test
data from BH 9.6. Starting from right to left, from high to low frequency (i.e. from shallow
to deep soil), the amplitude and frequency of the predominant peaks and troughs of the
HVSR curves were identified. These represent possible changes in the modulus/stiffness
of the soil profile with depth picked up by the ambient vibration recordings. In this
instance, the measured HVSR curve was observed to contain four possible secondary
resonance peaks (at 36 Hz, 25 Hz, 16 Hz, 12 Hz), one prominent trough (8 Hz) and the
predominant resonance peak (at 4.5 Hz) as shown in Figure 7.12a (solid red line shows the
measured HVSR curve and broken blue line shows the theoretical HVSR curve). In a trialand-error iteration, the forward modelling using a 5-layered soil model (including an
intervening soft/uncompacted layer overlying the bedrock and a single layer to represent
the consistent roller compacted material) was able to fit the HVSR curve at 2 of the 4
secondary resonance peaks (25 Hz and 16 Hz), the trough and the predominant resonance
peak, while also matching the inferred profile of the normalised Vs at station A4 and the
profile of the normalised CPT test data at BH 9.6 in a relative stratigraphic sense. The
results of the fitting of the relative stratigraphy are presented in Figure 7.12b.

- 135 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique

a)

b)

Figure 7.12: Calibration DCPT-9 area (a) measured and theoretical HVSR curves
at station A4, and (b) normalised HVSR Vs (location A4) versus normalised CPT
q c (location 9.6).
The 5-layered soil model from the calibration was then used as an initial guess to fit the
remaining HVSR curves. The calibration helped to clarify and/or corroborate the key
features of the HVSR curves:
1. The large amplitude of predominant peak frequency f0 at approximately 4 to 4.5 Hz
indicating the strong impedance contrast between surface layers and the bedrock
substrate, is a HVSR manifestation that could be correlated to the jump in Vs
profile as the soil transitions from the surface layers to the bedrock. This transition
is clearly observed at a depth of 12 13m in the Vs profiles.
2. The trough between 6.5 Hz and 10 Hz where the H/V ratio dips below one in the
HVSR curves has been verified (by the forward model developed for this area) as
truly bearing the hallmarks of an embedded softer/looser soil layer. This is
apparent from observing the significant drop in Vs from a depth of 7 8 m to
- 136 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique
approximately 12 13 m in the surface layers. This confirmation is instructive and
significant for the reason that the presence of Rayleigh wave ellipticity would
normally also cause a trough to appear in the HVSR curves at 2f0, i.e. approximately
8 Hz obscuring the true character of the trough.
3. Smaller secondary peaks located at frequencies above 12 Hz in theses curves,
indicating smaller impedance contrasts between the shallower surface layers, could
be linked to the sizeable jumps in Vs profiles at depths between 0.5 and 4 m.
The summary of all Vs profiles inferred from the fitting of HVSR measurements in the
area, after normalisation, are shown in Figure 7.13a. Figure 7.13b has also been included to
convey the consistency of compaction achieved from the roller compaction (i.e. the
approximately 3 m upper section).

a)

b)

Figure 7.13: DCPT-9 area normalised Vs profiles summary (a) full profile and (b)
top 3m profile (roller compacted material).
- 137 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique
The dashed Vs profiles in Figure 7.13b indicate outlier stations showing consistent roller
compaction to a depth of 2.2 m only below the ground surface (in contrast to
approximately 3 m at other stations).
To illustrate the back-analysis of the HVSR curves in the DCPT-9 area, the results at
selected grid stations A4 as well as A2 and B4 are shown in Figures 7.12a, 7.14a, and 7.14b
respectively (solid red lines show the measured HVSR curves and broken blue lines show
the theoretical HVSR curves). To avoid tedious repetition, the remaining stations of the
DCPT-9 area are not shown but have achieved similar goodness-of-fit levels.

a)

b)

Figure 7.14: DCPT-9 area measured and theoretical HVSR curves (a) A2 and (b) B4.
A review of Figure 7.13a further shows that the dynamic compaction was generally
effective to a depth of approximately 7 8 m below the surface in the area, and this was
anticipated in the design process. This is manifested as the sharp drop in Vs at a depth of 7
8 m to a depth where the bedrock is encountered. The Vs profile thus reveals that

- 138 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique
approximately 4 to 6 m deep layer of fill material directly overlying the bedrock had not
been significantly impacted by the dynamic compaction.
To further verify the accuracy of the fitting, the Vs profiles have been benchmarked against
the mechanical CPT, DMT and dry density results. Both CPT and DMT tests are well
known and widely accepted as geotechnical means to assess deep compaction (e.g.
Schmertmann et al. 1986; Bo et al. 2009; Karray et al. 2011). The CPT measures the cone
end resistance (qc) in the ground, while the DMT yielded the constrained modulus (M).
Figure 7.15 shows the superimposed Vs profiles (at A2 and Z1) of the CPT data (at
locations CPT 9.8 and BH 6.4 respectively) and closest DMT data from the nearby
locations BH 9.3 and BH 6.4 respectively.

a)

b)

Figure 7.15: Verification DCPT-9 area normalised HVSR Vs versus normalised


CPT q c versus measured DMT M versus normalised MASW Vs (a) HVSR
location A2 versus CPT location 9.8 versus DMT location 9.3 versus MASW Row A,
and (b) HVSR location Z1 versus CPT location 6.4 versus DMT location 6.4.
- 139 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique
The results in Figure 7.15 show the correlations obtained between the derived Vs profiles
and the invasive cone resistance from the CPT between the depth of 3 m and 15 m.
Increased Vs values, in particular, between 3 and 6 m, and again at 12 to 13 m were
confirmed by the corresponding increased cone resistances recorded at the same depths.
The Vs profiles were, moreover, consistent in a relative sense with the constrained modulus
from the DMT. The inferred Vs profiles from 0 to 3 m at A2 and Z1 also show a level of
uniformity that was matched by the consistency of the dry density test data (mean = 1.88
t/m3, standard deviation = 0.06 t/m3).
A profile-to-profile comparison was also made between the estimated Vs profiles of the
HVSR and the MASW methods. The MASW technique employed at the DCPT-9 area
involved, utilising a total of 21 single component (vertical) geophones connected at a
spacing of 3 m continuously along Rows A to E (Figure 7.1). A 3.6 kg (8 lb.) sledgehammer
was used to strike against a solid steel plate, at a distance of 5 m from each end of the array,
to generate the active source. Here the data from the geophones were captured by a laptop
via a multichannel digital seismic acquisition system (Soilspy Rosina from Micromed).
Several shots of the sledgehammer were attempted to ensure that reliable and clear
dispersion curves were obtained. The actively generated noise was recorded for three
seconds at a sampling rate of 512 Hz. The measured dispersion curve was generated using
a modified version of Herrmanns (1987) program, which is designed to pick the energy
maxima of the f-k spectrum. The theoretical dispersion curve for each ground model was
calculated according to a theoretical model developed by Haskell (1953). A trial-and-error
procedure was used to minimise the error between theoretical and measured dispersion
curves.
Figure 7.16a shows the fitting of the theoretical (theoretical dispersion curve is shown by
the white hatched line) and experimental dispersion curves from an independent MASW
test, while Figure 7.16b above shows the Vs profiles estimated from HVSR measurements
taken at grid stations A2 and A4 superimposed on the profile estimated from the MASW
measurements taken along grid row A. The profiles from the surface wave methods
showed good agreement in respect of both the relative stratigraphy and absolute Vs values
thus reinforcing the good correlations already achieved between the HVSR method, and
the mechanical CPT, DMT and dry density test data.

- 140 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique

a)

b)

Figure 7.16:Verification DCPT-9 area MASW projected Vs along Row A (a) Phase
velocity dispersion curve with active sources at both ends, and (b) MASW estimated
Vs profile against HVSR estimated Vs profile at location A2 and A4.

7.5 Assessment of Compaction Quality


Following the fitting of HVSR curves to estimate the Vs profiles of the compacted layers, a
more informed assessment of compaction quality in the area was made by establishing the
metrics from Vs data. In the first appraisal, the mean of the normalised Vs at the stations
associated with the CPTs was plotted against its standard deviation in Figure 7.17, and the
points are shown clustered and bounded by the ellipse formed by broken lines. The mean
and standard deviation of the normalised Vs are metrics chosen to characterise the quality
and consistency of compaction in this appraisal. The same metrics from the stations not
associated with the CPTs representing the areas not covered by the mechanical tests are
also plotted in Figure 7.17 and this cluster is shown bounded by the solid ellipse. It is
- 141 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique
observed that the cluster for the stations not covered by the CPTs matches very closely and
overlaps significantly with the cluster for the stations associated with the CPTs thus
indicating that the quality and consistency of the dynamic and roller compaction of both
clusters are virtually the same.

Figure 7.17: Mean normalised Vs versus the standard deviation of the stations in the
DCPT-9 area.
Statistical inference allows an engineer to make conclusions and help in the decision
making process for a population of data, based on the information contained in a sample
of data. For this study, the students t-test (a common statistical significance test applied to
small population samples to compare two means) was chosen to provide a statistical insight
on the hypothesis that: the mean of the DCPT-9 area stations associated with CPTs (1) is
equal to the mean of the DCPT-9 area stations not associated with CPTs (2). An unpaired
two-sample t-test has been applied to the two data sets (with CPT and without CPT)
assuming unequal variances. The t-test was applied in the form (Kottegoda & Rosso 2008):
Null Hypothesis H0: 2 = 1
Alternative Hypothesis H1: 2 1
Level of Significance: = 0.05
For the assumptions made and the t-test described above, the t-statistic value for this study
is obtained by Equation (7.1) and the degrees of freedom by Equation (7.2):

- 142 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique
T'=

v=

(X

X 2 (1 2 )
n1 + S 22 n 2

S12

(S

2
1

n1

[S

n1 + S 22 n 2

(n1 1) + (S 22
2
1

(7.1)

n2

) (n
2

(7.2)

1)

where, X n is the mean of the data sample, n is the mean of the population, S n2 is the
standard deviation of the sample data, and nn is the number of variables in the sample data.
Table 7.2 shows the results from the t-test and accordingly confirms the null hypothesis
(i.e. 2 = 1), given that the t-statistic value of 0.826 (with 164 degree of freedom) is less
than or equal to the t-critical (two-tailed) value of 1.975.
Table 7.2: Students t-test results.
With CPT

Without CPT

Mean

264 (m/s)

259 (m/s)

Variance

1117.6

1565.7

Observations

70

123

Hypothesized Mean Difference

Degrees of Freedom

164

t Statistic

0.826

P(T<=t) two-tail

0.410

t Critical two-tail

1.975

The Vs appraisal and the students t-test suggest that the quality and consistency of
dynamic and roller compaction in the local areas covered by the CPTs and the local areas
of the stations not associated with the CPTs are virtually identical and statistically the same.
It is thus shown that the fast and operationally simple HVSR technique may be used to
cover the gaps that could not been covered by the localised mechanical tests in the DCPT9 area and in this way could prevent poor compaction spots from being missed in the
quality control tests. The approach of combining the more precise but also more expensive
mechanical tests with a cost-effective but less precise HVSR technique would maximise the
strengths while minimising the limitations of field testing needed to assess an extensive and
deep site more comprehensively.
- 143 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique
Due to scheduling, accessibility and other factors beyond the control of this thesis, a preanalysis before the combined dynamic and roller compaction by the HVSR technique was
not possible. Even though a pre-analysis was not completed, the material located between
the influence limit of the dynamic compaction (approximate depth of 8 m) and the
bedrock, which was largely undisturbed by the dynamic compaction, provided a good
estimate of the soil characteristics before the commencement of dynamic compaction.
Hence, a comparison of the mean Vs for this intact material and that of the composite
dynamic and roller compacted material overlying it provided an estimation of the impact of
the roller and particularly the dynamic compaction. Figure 7.18 shows the mean Vs of the
intact material (solid) and the increase in mean Vs after dynamic and roller compaction
(hatched) for each station (location) in the DCPT-9 area. An increase in mean Vs between
77 and 139 m/s or 48 to 94 % of the Vs of the intact material was achieved due to the
impact of the composite dynamic and roller compaction.

Figure 7.18: Influence of composite compaction at DCPT-9 area.

7.6 Summary
This chapter has applied the HVSR technique to assess a composite dynamic and roller
compacted ground located within an area of the Penrith Lakes quarry. Similar to the
previous chapter the HVSR technique was firstly used to give a preliminary assessment of
the ground compaction by interpreting the key features of the measured HVSR curves.
Secondly, forward modelling of the measured HVSR curves made possible to infer the Vs
profiles conveying compaction quality and consistency of the ground. This entails
- 144 -

Chapter Seven: Investigation of Composite Dynamic and Rolling Compaction using the
HVSR Technique
calibrating and verifying the Vs profiles against available mechanical test data and/or before
the forward modelling is applied to measurement stations not covered by mechanical
and/or other reliable independent tests. In the process, the interpretations of the key
features of the measured HVSR curves would also be confirmed and/or clarified. As
shown in this chapter, the Vs metrics could be applied to appraise the quality and
consistency of the composite compaction to complement the appraisal of the mechanical
(e.g. CPT and DMT) and dry density tests. This approach would, in particular, optimise the
use of expensive mechanical tests and cover the gaps in an area that are not subjected to
mechanical tests.

- 145 -

Chapter Eight: Investigation of Dynamic


Compaction Using the HVSRplus
Technique
In this Chapter, the proposed HVSR-based technique and the methodologies were applied
to evaluate a purely dynamically compacted site. In addition to recognising that the
predominant resonance peak of the HVSR curve is a reflection of the impedance contrast
between the surface layers and bedrock, the study recognizes that the secondary resonance
peaks of the curve at higher frequencies may reflect strong impedance contrast within
surface layers. This concept has been applied to develop a methodology of HVSR-based
approach relying on the measurement of the HVSR of microtremors at measuring stations,
and calibration and verification by independent mechanical and MSOR tests. The use of
MSOR tests is introduced for ground compaction in this chapter to facilitate the calibration
of the HVSR forward model, particularly in terms of providing information for the initial
guess of the shear wave velocity, Vs, profile in the HVSR forward modelling. The chapter
demonstrates the effective use of the HVSR-based approach to assess dynamic compaction
in the gaps away from and not covered by the mechanical tests. The mapping between the
depth of bedrock and the predominant resonance frequency is also extended to include the
mapping of the depths of layers with strong impedance contrasts to the secondary
resonance peaks, after the data have been verified by independent mechanical tests. An
- 146 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique


empirical correlation is also presented based on the HVSR inferred Vs and available CPT
(qc) data.

8.1 Dynamic Compaction Prototype Trial Area 7


(DCPT-7)
The purely dynamically compacted test area, otherwise known as DCPT-7, is shown in
Figure 6.1 and is located next to the DCPT-9 area (see Chapter 7). Similar to the DCPT-9
area, the DCPT-7 area has been dynamically compacted because the deep fills did not
previously receive adequate compaction for potential future infrastructure development.
The mixture of clayey, silty and sandy fill stands over an area of 9,500 m2 and runs to an
approximate depth of between 12 and 13 m to the bedrock substrate. The groundwater
table (GWT) in this area is on average 10 m below the surface. The DCPT-7 area has been
dynamically compacted using an octagonal shaped 1.8 m wide, 20 tonne (20 x 103 kg)
pounder lifted up by a crane and dropped from a height of about 20 m above the ground,
in 3 phases. The weight was dropped 32 times at each station spaced 4.2 m apart, imposing
energy of approximately 400 tonne-m at each drop. After each pass the resultant craters
were surveyed, backfilled and re-levelled with the ground surface, using locally derived
material. An ironing pass was not utilised in this area and at present further treatment of
the disturbed upper surface due to dynamic compaction has not been scheduled.
Figure 8.1 shows the 95 m x 100 m DCPT-7 area with regularly spaced grid of the
observations of the HVSR recordings covering the area. Where the red squares indicate
HVSR only measurements and blue circles indicate a combination of independent
mechanical tests (borehole, CPT, DMT) including HVSR and MSOR measurements. There
are a total of four control stations (CS7.1 CS7.4), with each consisting of a combination
of borelogs, CPTs and DMTs. HVSR and MSOR testing were also conducted at the
locations of these control stations, for calibration and verification purposes (this will be
discussed in Section 8.3). The spread of measurements covers the entire area quite
uniformly. The intention here is to utilise the grid of HVSR measurements and the data
from the four control stations to make a comprehensive verification of the quality of the
dynamic compaction in the area without leaving any gaps uncovered. Thus if the proposed
methodology could provide a good estimate of the compaction at the measurement points,

- 147 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique


a significantly better coverage is achieved than utilising just the four control stations of
mechanical tests from the traditional approach.

Figure 8.1: DCPT-7 area test locations.


Dimensions are in metres (m).

8.2 Interpretation of Measured HVSR Curves in


DCPT-7 Area
The measured HVSR curves from the DCPT-7 area may provide an indicative assessment
of the dynamic compaction in this area. Measured HVSR curves drawn from all
observation stations are thus superimposed in Figure 8.2, to identify common features in
the curves symptomatic of the ground velocity structure beneath the area. There are clearly
variations in the measured curves, which are to be expected for a dynamically compacted
area of this size and nature. Nonetheless, similarly to the superimposed measured HVSR
curves from the DCPT-9 area, three common prominent characteristics are discernible in
the superimposed curves.
First, a large HVSR amplitude at predominant frequency f0 (approx. 4 4.5 Hz), indicates a
strong impedance contrast between the bedrock and overlying surface layers. Under
approximately plane-parallel conditions, a high amplitude f0 is symptomatic of higher
impedance contrast between bedrock and surface layers, and vice versa for a low amplitude
f0. Moreover, f0 corresponds to the fundamental resonance frequency of the surface layers,
- 148 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique


this correspondence having been pointed out in previous studies (e.g. Bonnefoy-Claudet et
al. 2006b).

Figure 8.2: Superimposed summary of the measured HVSR curves at the DCPT-7
area.
Second, a trough between approximately 8 Hz and 10 Hz (where the H/V ratio dips below
one) which occurs at approximately 2f0 would commonly be attributed to Rayleigh wave
ellipticity (Fah et al. 2001; Castellaro & Mulargia 2009b). As mentioned above, this feature,
however, may or may not also suggest a weak layer embedded between two stronger layers
referred to as a velocity inversion (Castellaro & Mulargia 2009b) lying above the bedrock,
since the trough is located at a frequency above f0, where f0 is generally associated to the
impedance contrast between bedrock and surface layers. Furthermore, a less obvious
secondary trough appearing at approximately 22 Hz in some of the measured HVSR
curves, may indicate possible presence of a second embedded weak layer located at a depth
above the first embedded weak layer.
Third, occurrence of secondary peaks above f0 may relate to the strong impedance contrasts
within the compacted surface layers and may thus give an insight of the consistency of
compaction. However, this must be interpreted with some caution since there is a
possibility that a secondary resonance peak may also reflect a higher mode of f0, especially
the first higher order mode (corresponding to 3f0) at approximately 12 15 Hz in this case
(see, Section 3.5).
It is evident from the above that the measured HVSR curves may provide an indication,
but not a confirmation, of the layering of the ground velocity structure. It is seen that some
- 149 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique


of the key features may be interpreted in more than one way, and so constrained forward
modelling of the HVSR curve and/or use of independent mechanical tests are needed to
help elucidate the uncertainties.

8.3 Forward Modelling to Infer the Vs Profiles in


DCPT-7 Area
Although a relative and indicative sense of the ground velocity structure (namely, the
impedance contrast between bedrock and surface layers, possible strong impedance
contrasts and embedded weak layers within the surface layers) may be inferred by studying
the salient features of the HVSR curves, constrained forward modelling of the HVSR
curves as shown below will elucidate this structure in a more absolute sense. Here, the
theoretical HVSR curves from the forward model are fitted to the measured HVSR curves
from the DCPT-7 area to infer the Vs profile conveying the compaction quality of the
surface layers. However, forward modelling of the measured HVSR curves is not a
straightforward matter since the fitting of the theoretical solutions to measurements does
not produce a unique solution, as it happens for all geophysical techniques. Therefore, to
ensure correct interpretation of the measured HVSR curve will lead to a sensible inference
of the Vs profile, both calibration and verification with mechanical and/or other
independent tests are required before applying the HVSR technique in the interspersing
areas not covered by mechanical tests. A flowchart of the proposed methodology applied
in this chapter is presented in Figure 8.3 outlining the required procedures.

- 150 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique

Figure 8.3: Flowchart calibration, verification and application.

- 151 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique


There are three general steps:
(1) Calibration of the HVSR curve using constraining information in the form of
independent mechanical tests and the array based MSOR technique. Mechanical test data
(e.g. CPT data in this instance) provide reliable material layering information, and also
identify contrasts between material layers in a relative sense. Furthermore, utilising the
MSOR technique provides the HVSR technique with an initial guess of the Vs profile
(HVSRplus technique), especially in the uppermost layer, and also constraining ranges for
Vs values to feed into the forward model. In this step the key features of the HVSR curves,
namely the resonant peaks and the troughs are ruled in or ruled out as representations of
impedance contrast by comparing the fitted ground model to the mechanical data. Thus if
the theoretical curve fitted to a secondary resonant peak produced a velocity increase
reminiscent of impedance contrast at a certain depth but this is not supported by
independent mechanical tests then this peak is ruled out as of any significance.
(2) Verification of the calibrated model means to confirm that the ground velocity model
of the HVSR curve fitted to the features of significance (identified in the calibration step)
of the measured HVSR curve at a different location to the ones used for calibration, is in
agreement with the independent mechanical data of that location. This step gives added
assurance that the interpretation of the features of the HVSR curve and the constrained
forward modelling are correctly aligned with independent test data.
(3) Once the calibrated model has been verified, constrained forward modelling of the
HVSR technique may then be applied to the HVSR measurements taken in areas not
covered by the costly mechanical tests, to infer the Vs profiles at these locations. The
proposed methodology thus enables the use of HVSR techniques to take advantage of as
well as to complement the available mechanical test data in such a way that allows an
intimate and rigorous assessment of the entire compacted area to be achieved.

Calibration with Independent Mechanical Data


In a trial-and-error iterative process, the forward modelling was calibrated to fit the
theoretical HVSR to the measured HVSR curve at station CS7.3 in such a way so as to
ensure the ground velocity structure is consistent with the CPT data taken at the same
- 152 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique


location. Starting from high to low frequency (i.e. from shallow to deep soil), the amplitude
and frequency of the resonance peaks and troughs of the HVSR curves are identified. As
discussed above, these features represent possible impedance contrast, that is, changes in
the modulus/stiffness of the soil profile with depth identified by the HVSR measurements.
Examination of the measured HVSR curve at CS7.3 (Figure 8.4a solid red line shows the
measured HVSR curve and broken blue line shows the theoretical HVSR curve) reveals
four possible secondary resonance peaks (at 42 Hz, 26 Hz, 14 Hz, 6.7 Hz), two possible
velocity inversion troughs (at 21 Hz, 8.5 Hz) and the fundamental resonance peak (at 4.3
Hz). Using constraining information from the CPT test data and the MSOR technique,
including an initial guess of the Vs profile based largely on the inversion of the Rayleigh
wave dispersion curve of the MSOR technique (Figure 8.4b theoretical dispersion curve
is shown by the white hatched line), a trial-and-error iterative forward modelling procedure
using a 6-layered ground model (including the bedrock half-space) achieved a theoretical to
measured HVSR fit where the Vs profile of the ground model is in good agreement with
the CPT data. The inferred profile of the normalised HVSR Vs, the normalised MSOR Vs
(to derive initial guess for the HVSR forward modelling) and the normalised CPT (qc) test
data from station CS7.3 are superimposed and shown in Figure 8.4c to illustrate this.
Here it is noted that this result was achieved by requiring theoretical to measured HVSR fit
to the predominant resonance peak, the first secondary resonance peak (at 14 Hz), and to a
lesser extent the two secondary resonance peaks (26 Hz and 42 Hz) and the trough at 8.5
Hz. The fourth secondary resonance peak at 6.7 Hz was found to be of no significance,
and its presence is most likely due to artifact noise from anthropic activities (e.g. industrial
and traffic) near the measurement point. The attribution of the peak at 6.7 Hz to artifact
noise is based on the observation that there were spikes at 6.7 Hz on the amplitude spectra
of all three (two horizontal and one vertical) components of the microtremors (SESAME
2004a). In addition, the DCPT-7 area is located adjacent to the DCPT-9 area and the
influence of anthropic activities has already been confirmed in Section 7.2. Nonetheless,
the inferred Vs profile from fitting the peak at 6.7 Hz did not match well with the CPT
data so this was all the more a good reason not to include this peak in the calibration
fitting. The final HVSR Vs profile was reasonably close to the MSOR Vs profile at this
location which suggested that the initial guess was a good one. The 6-layered soil model
from the calibration was then used as an initial guess to fit the remaining of the HVSR
curves. However, in some cases either a lower or higher number of layers were necessary to
- 153 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique


provide a reasonable fit between theoretical and measured HVSR curves. In any case, the
lowest possible number of layers required to provide a reasonable fit between theoretical
and measured curves was used as the final solution.

(a)

(b)

(c)

Figure 8.4: Calibration DCPT-7 area CS7.3 (a) measured and theoretical HVSR
curves, (b) Phase velocity dispersion curve, and (c) normalised HVSR Vs and
normalised MSOR Vs versus normalised CPT q c.
- 154 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique

Verification against Independent Mechanical Data


The inferred Vs profiles from the remaining control stations (CS7.1, CS7.2 and CS7.4) have
been further benchmarked against the mechanical CPT and DMT results to verify the
above calibration. Figure 8.5 shows the theoretical-to-measured fits between the theoretical
and measured HVSR curves (solid red lines show the measured HVSR curves and broken
blue lines show the theoretical HVSR curves) at locations CS7.1, CS7.2 and CS7.4, from
which the Vs profiles have been inferred.

(a)

(b)

(c)

Figure 8.5: Verification DCPT-7 area: Measured and theoretical HVSR curves at
stations (a) CS7.1, (b) CS7.2, and (c) CS7.4.
- 155 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique


Figure 8.6 shows superimposed Vs profiles, CPT data and DMT data from these locations.
These results show good agreement between the derived Vs profiles and the invasive cone
resistance (qc) from the CPT. The Vs profiles are, moreover, consistent in a relative sense
with the constrained modulus from the DMT. The verification shows that the calibration
above is acceptable, and the features in the HVSR curves considered as significant in trialand-error forward modelling may also be applied to the remaining HVSR measurements
not adjoined to the mechanical tests.

(a)

(b)

(c)

Figure 8.6: Verification DCPT-7 area: Normalised HVSR Vs versus normalised


CPT q c at stations (a) CS7.1, (b) CS7.2, and (c) CS7.4.
- 156 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique

Evaluation of Dynamic Compaction Complemented by the HVSRplus


Technique
In the final general step, results from the remaining HVSR stations not adjoined to any
independent mechanical tests in the DCPT-7 area are processed. Examples of the
theoretical to measured HVSR fittings at stations A1 and C2 in the DCPT-7 area are
presented in Figures 8.7a and b to illustrate the goodness-of-fit for the features construed
as significant (solid red lines show the measured HVSR curves and broken blue lines show
the theoretical HVSR curves). A similar level of goodness-of-fit was generally found to
have been achieved for the remaining stations in the area but to avert the risk of
monotonous repetition, these results are not shown.

(a)

(b)

Figure 8.7: Remaining grid points DCPT-7 area measured and theoretical HVSR
curves (a) station A1, and (b) station C2.
The normalised Vs profiles for all stations in the DCPT-7 area are shown in Figure 8.8. A
review of Figure 8.8 shows that the dynamic compaction was generally effective to a depth
- 157 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique


of approximately 4 7.5 m below the surface in the area as manifested in the sharp drop in
Vs at around this depth. In general there is some variability in the compaction quality of the
near surface, but nothing that is unusually concerning. However, the above demonstrates in
principle how the procedures applied in this chapter using the HVSR technique could be
used to minimise the risk of missing out on detecting problematic and poorly compacted
areas.

Figure 8.8: DCPT-7 area normalised Vs profiles summary.

8.4 Evaluation of Compaction Quality


With inferred Vs profiles of the compacted layers available, it is now possible to make a
more informed assessment of compaction quality in the area by establishing relevant
metrics from the Vs data.

Figure 8.9: Compaction assessment Vs at 1 m intervals.


- 158 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique


Thus, Figure 8.9 above shows the mean and standard deviation of the Vs data for every
metre of compacted soil averaged over all control stations (CS7.1 CS7.4) and remaining
HVSR stations in the DCPT-7 area, from the ground surface to the depth of influence of
the dynamic compaction.
A pre-analysis before the dynamic compaction by the HVSR technique of the DCPT-7 area
was not possible, due to scheduling, accessibility and other factors beyond the control of
this thesis. Unfortunately, none of the surrounding areas can provide a reasonable
representation of the state of the ground pre-dynamic compaction. However, as it turns
out, a good estimate of the soil characteristics prior to dynamic compaction can in fact be
provided by the largely undisturbed fill material located below the depth of influence of the
dynamic compaction. In light of this, a comparison of the mean Vs of this intact material
and that of the dynamically compacted material overlying it would help to elucidate the
impact of the dynamic compaction. Figure 8.10 shows the mean Vs of the intact material
(solid) and the increase in mean Vs after dynamic compaction (hatched) for each station
(location) in the DCPT-7 area. An increase in mean Vs between 43 and 97 m/s (average 62
m/s) or 24 to 62 % (average 38 %) of the Vs of the intact material was thus estimated to be
achieved due to the impact of the dynamic compaction.

Figure 8.10: Influence of dynamic compaction.


A contour plot based on ordinary kriging of the Vs profiles along Row E is also presented
in Figure 8.11. The plot, which combines available Vs data from both the control stations
(i.e. CS7.1 and CS7.3) and other HVSR stations (i.e. E1 E5), highlights the possibility of
utilising the metrics to illustrate compacted layering, in this case along Row E.
- 159 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique

Figure 8.11: Contour plot of Vs profiles along Row E (refer to Figure 8.1).

8.5 Empirical Relationships


As mentioned above, the CPT test involves pushing a cone tip into the ground to measure
the cone resistance, qc, which is generally thought to reflect soil properties at large strains.
Over the years, a substantial number of empirical relationships have been developed to
ascertain in situ soil properties including soil modulus (E and G0), peak friction angle (p),
constrained modulus (DMT M), etc from qc data (e.g. Robertson 2009). Normalised Vs
and qc correlations have also been previously established by a few investigators (e.g. Wride
et al. 2000; Karray et al. 2011), although these results must be used with caution since, as
stated above, qc is a large strain parameter while Vs has been estimated from small strain
techniques. Using available data from the HVSR and CPT tests, the normalised Vs is
plotted against normalised qc from the DCPT-7 area as shown in Figure 8.12.

Figure 8.12: Normalised Vs versus normalised q c relationship for the DCPT-7 area.
- 160 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique


In keeping with the work of others (e.g. Robertson et al. 1992a; Fear & Robertson 1995;
Wride et al. 2000; Karray et al. 2011), these results have been regressed to yield a power law
relationship (R2 = 0.82):

Vs = 118q c

0.26

(8.1)

This equation may provide an estimate of qc (kPa) when only the Vs values are known,
within the DCPT-7 area, and more broadly within the Penrith Lakes quarry where the soil
stratigraphy, soil type (comprising a mixture of poorly graded sand, silty sand to clayey sand
material) and use of compaction to improve the ground, are generally similar.
Previous authors have attempted to map the total depth (thickness) of sediments above the
bedrock to the predominant resonance frequency inferred from microtremor
measurements (e.g. Ibs-von Seht & Wohlenberg 1999; Delgado et al. 2000; Parolai et al.
2002). This mapping is underpinned by the well known relationship between the
predominant frequency (f0), the shear wave velocity (Vs) and thickness (H) of the surface
layer overlying bedrock: f0 = Vs/(4H), for a two-layered (surface layer-bedrock) system.
Thus for generally similar material in the surface layer (i.e. constant Vs), this relationship
suggests that the thickness of surface layer, H (i.e. depth to bedrock) is inversely
proportional to the predominant frequency, f0. In this study the mapping has been
extended to apply in respect of compacted surface layers where more than one discernible
impedance contrast has been measured. Here, the results of the invasive mechanical tests
and the inferred Vs profiles in the DCPT-7 area have been utilised to map not only the
total depth of the surface layers overlying bedrock to the predominant frequency, but also
the secondary resonance peaks observed in the HVSR curves to the depths the compacted
layers defined by significant impedance contrasts. Generally speaking, secondary peaks
represent impedance contrasts within the surface layers, however, care is required to avoid
misinterpretations as some of these peaks may in fact be due to the effects of higher modes
(refer to Section 8.2). In the DCPT-7 area, the CPT and DMT data are available to help
verify whether the peaks are manifestations of actual physical contrasts of the compacted
ground or the higher modes of surface waves. Results from the DCPT-7 area after
verification are plotted in Figure 8.13 for frequency of 4.2 27 Hz versus depth of 1.2
13.3 m, with a regressed relationship (R2 = 0.94) given by:

- 161 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique


z = 73 f r

1.17

(8.2)

where z is the depth (m) and fr is resonance frequency.

Figure 8.13: Frequency versus depth relationships for the DCPT-7 area, also
showing the standard error.
Equation (8.2) may thus provide a preliminary estimate of the depth of influence of the
dynamic compaction for the DCPT-7 area as well as the depths where discernible
impedance contrasts in the compacted ground may have occurred. According to Equation
(8.2), HVSR features occurring at frequencies higher than 30 Hz are required to resolve
compacted layers less than 1.0 m deep at the DCPT-7 area. Conversely, an (extrapolated)
estimate of the depth of compaction resolvable by the HVSR technique may be determined
by substituting fr = 0.1 (assumed detectable low end frequency of microtremors) into
Equation (8.2), giving a depth greater than 1 km. This is clearly much greater than the
depth of normal compacted ground or of near surface layers of interest in geotechnical site
investigation.

8.6 Summary
This chapter proposed a more thorough HVSR-based technique for the evaluation of deep
dynamically compacted ground. It recognises that the predominant resonance peak of the
HVSR curve is a reflection of the impedance contrast between the surface layers and
bedrock, but further recognises that the secondary resonance peaks of the curve at higher
frequencies may reflect strong impedance contrasts within the surface layers. This concept
- 162 -

Chapter Eight: Investigation of Dynamic Compaction using the HVSRplus Technique


is then applied to develop a methodology to comprehensively assess the consistency of
deep dynamically compacted fills in combination with independent mechanical and MSOR
tests. In a proposed 3-step approach (calibration, verification and application), the HVSRbased technique may be used to optimise expensive mechanical tests in the sense of
applying it to assess gaps of an area not subjected to mechanical tests after calibration and
verification of forward modelling have been achieved. An empirical correlation has also
been presented based on HVSR inferred Vs and CPT (qc) data. This correlation allows an
indicative estimation of the stiffness of the ground (qc) in areas not covered by the CPT
testing in the DCPT-7 area and more broadly within the Penrith Lakes quarry site. Finally,
this chapter proposes a depth-frequency relationship associating not only the predominant
frequency to the depth of the bedrock, but also the frequencies of the secondary
resonances to the depths of the strong impedance contrasts within the surface layers. This
relationship, developed for the study area utilising HVSR measurements, mechanical and
inferred Vs data from the DCPT-7 area, may be useful in conveying preliminary insights
regarding the depths of the significant stiffness changes within the surface layers by
association to the secondary resonance frequencies of the measured HVSR curves.

- 163 -

Chapter Nine: Inversion of HVSR and


Dispersion Curves Using Monte Carlo
Simulation
Limitations in the resolution of surface wave techniques can be overcome by combining
active and passive measurements, for an enhanced HVSR and dispersion curve inversion.
Thus, the combined HVSR and Rayleigh wave dispersion inversion using Monte Carlo
simulation was applied to estimate the Vs profile of the DCPT-7 dynamically compacted
ground discussed in Chapter 8. The assessment on the standard of the ground compaction
at the site was facilitated and constrained by the calibration of the first and the secondary
predominant peaks and troughs observed by the measured HVSR curve. Available CPT
data from the four control stations (i.e. CS7.1 CS7.4) in the DCPT-7 area (refer to Figure
8.1) were used as references to determine the appropriate number and thickness of layers,
as well as, to verify the Vs profiles estimated for the control stations. The addition of the
MASW technique allows obtaining the Rayleigh wave dispersion curve and subsequent
inversion allowed reducing the number of possible solutions used in the HVSR inversion.
Furthermore, the influence of the mean grain size, D50, on the Vs values was confirmed
after comparing the average normalised CPT - qc tip resistance, with the average normalised
Vs estimated for all layers at the control stations.

- 164 -

Chapter Nine: Inversion of HVSR and Dispersion Curves using Monte Carlo Simulation

9.1 Monte Carlo Inversion Procedure


The Monte Carlo simulation approach was used to estimate the Vs ground profiles by
conducting an enhanced inversion using both HVSR and dispersion curves. This approach
is especially useful when there are a large number of uncertain input variables, such as:
thickness, shear wave velocity, Poissons ratio, and density for several layers for each Vs
profile. The Monte Carlo approach allows for uniform random distribution for all
unknown parameters and has the potential to reduce the number of local minimum
solutions (usually obtained by deterministic approach). This approach involves generating
discrete Vs profiles within lower and upper bounds chosen using a priori information (e.g.
knowledge from independent mechanical test data). The randomly generated Vs profiles
produce theoretical (in this case HVSR and dispersion) curves which are compared for
their fit against the measured curve using a misfit value. The main appeal of Monte Carlo
simulation is that it avoids the assumption of linearity between the unknown variables and
the observed curve (Sambridge & Mosegaard 2002), where the assumption of linearity is
generally not the case for inverse geophysical problems.

9.1.1 Constraining Information for the Inversion


Procedure
The appropriate number of layers and their corresponding fixed layer thicknesses, Hi, were
determined based on CPT data available for each control station at the DCPT-7 area. Also,
the range of shear wave velocity, Vs, values for each layer was determined in a relative
sense using CPT data. Here a realistic constraining range was used for each layer to
eliminate unrealistic solutions and values. For example, an unrealistic solution would
involve having a Vs value of 1000 m/s for the first layer of compacted ground, followed by
a Vs value of 250 m/s for the second layer of compacted ground.
The value for Poissons ratio, , was determined randomly between 0.25 and 0.49, and used
for calculating the value of the primary (P-) wave velocity, Vp (Equation (9.1) below).
Vp = V s

1
1
2

(9.1)

- 165 -

Chapter Nine: Inversion of HVSR and Dispersion Curves using Monte Carlo Simulation
The density, , which is another significant parameter for calculating surface wave
dispersion characteristics (Xia et al. 1999), was calculated according to the empirical
relationship by Gardner et al. (1974):
= 310 Vp0.25

(9.2)

where, is the density in kg/m3, and Vp is the P-wave velocity in m/s. Since, Vp and are
determined using Equations (9.1) and (9.2), the Monte Carlo simulation will therefore only
alter the Vs, H and values for each layer of every possible Vs ground profile.

Selecting Potential Ground Models


To improve the efficiency for selecting potential ground models (i.e. Vs profiles) in the
Monte Carlo simulation process, the value of the theoretical HVSR predominant
frequency, f0S, was roughly calculated using Equation (9.3) for each potential ground model.

f 0S = 1 4

(H
n

i =1

VS i )

(9.3)

where i is the layer number, Hi is the thickness and Vsi is the shear wave velocity of layer i,
and n is the total number of layers overlying the half-space (bedrock). The ground models
with an f0S value within the range (1-)*f0 (1+)*f0 (where, f0 denotes the value of the
measured HVSR frequency in Hz, and is the half bandwidth either side of f0) are filtered
through to the next stage of this inversion method. Additionally, in order to obtain more
reliable Vs profiles, relative constraints for Vs values of all layers were included based on
CPT data and the features of the measured HVSR curve (refer to Sections 8.2 and 8.3).

9.1.2 Obtaining the Measured Rayleigh Wave Dispersion


Curve and the Misfit between Measured and
Theoretical Dispersion Curves
The MASW technique was applied to obtain the measured Rayleigh wave dispersion curves
along the four control stations (i.e. CS7.1 CS7.4) (refer to Figure 8.1) at the DCPT-7 area.
- 166 -

Chapter Nine: Inversion of HVSR and Dispersion Curves using Monte Carlo Simulation
A total of 21 single component (vertical) geophones were connected at a spacing of 1 m
continuously along each control station (Figure 8.1). A 3.6 kg (8 lb.) sledgehammer was
used to strike against a solid steel plate, at a distance of 2 m from each end of the array, to
generate the active source. Here the data from the geophones were captured by a laptop via
a multichannel digital seismic acquisition system (Soilspy Rosina from Micromed). Several
shots of the sledgehammer were attempted to ensure that reliable and clear dispersion
curves were obtained. The actively generated noise was recorded for three seconds at a
sampling rate of 512 Hz. The measured dispersion curve was generated using a modified
version of Herrmanns (1987) program, which is designed to pick the energy maxima of the
f-k spectrum. The theoretical dispersion curve for each Vs profile was then calculated using
a theoretical model developed by Haskell (1953).
After obtaining the theoretical dispersion curve, the inversion of these theoretical curves
against the measured dispersion curves provided possible solutions for the HVSR inversion
(described below). However, in order to proceed to the HVSR inversion step, the misfit
between the measured Rayleigh wave dispersion curve and the theoretical Rayleigh
dispersion curves (multimodal analysis) of possible Vs profiles were calculated using
Equation (9.4).
misfit =

[ (c
m

j =1

oj

ct j ) c o j

(9.4)

where, m is the number of assessed dispersion frequencies, coj and ctj are experimental and
theoretical Rayleigh phase velocities, respectively, for the jth frequency. The theoretical
curves, that is, the Vs profiles with acceptable misfit values were filtered through to be used
by the HVSR inversion in the next step.

9.1.3 Misfit between Measured and Theoretical HVSR


Curves
The inversion of measured HVSR curve was carried out using the theoretical HVSR model
outlined in Section 3.2. The theoretical HVSR curves were calculated using a frequency
step of 0.125 Hz and incorporating up to 5 higher modes. Also, the HVSR inversion relied
on using the filtered Vs profiles from the Rayleigh wave dispersion curve inversion in the
- 167 -

Chapter Nine: Inversion of HVSR and Dispersion Curves using Monte Carlo Simulation
previous step. The misfit between the measured HVSR curve and the theoretical HVSR
curves obtained using the Vs profiles filtered by the dispersion curve inversion, were
calculated using Equation (9.5).
misfit =

[ (hvsr
p

j =1

oj

hvsrt j ) hvsro j

(9.5)

where, p is the number of assessed HVSR curve frequencies, hvsroj and hvsrtj are measured
and theoretical HVSR amplitudes, respectively, for the jth frequency. The first 30 Vs profiles
with misfit values lower than 0.2 obtained using Equation (9.5) were considered as the
possible Vs profiles for the assessed location (i.e. CS 7.1 CS 7.4).

9.2 Application of the Monte Carlo Simulation


Inversion Method
Figure 9.1a shows the comparison between the measured HVSR curve and the theoretical
HVSR curves obtained by 30 possible Vs profiles using the Monte Carlo simulation
inversion method at station CS 7.1, where Figure 9.1b shows the comparison between the
measured dispersion curve and the theoretical dispersion curves at the same location. It can
be seen from Figure 9.1a, that the fit between the predominant peaks and troughs are in
quite good agreement. Figure 9.1c shows the comparison between the normalised CPT
cone end resistance, qc, (shown by the dashed blue line), the normalised average (of the 30
possible Vs profiles) Vs profile (shown by the solid red line) and its corresponding 95 %
confidence interval of the 30 possible solutions (shown by the dotted black lines). This
figure shows good agreement between normalised Vs and normalised qc profiles, with
respect to both the material layering (layer thicknesses) and relative stiffness of each layer.
The estimation of the Vs profiles for the remaining control stations CS 7.2, CS 7.3 and CS
7.4 was carried out following the same Monte Carlo inversion method mentioned above.
Figure 9.2 shows the goodness-of-fit between the theoretical HVSR and measured HVSR
curves for the remaining stations. Figure 9.3 shows the comparisons between the
normalised CPT cone end resistance, qc, (shown by the dashed blue line), the normalised
average (of the 30 possible Vs profiles) Vs profiles (shown by the solid red line) and its
- 168 -

Chapter Nine: Inversion of HVSR and Dispersion Curves using Monte Carlo Simulation
corresponding 95% confidence intervals of the 30 possible solutions (shown by the dotted
black lines) for the remaining stations. Similar to the results for CS 7.1, these figures show
good agreement between the normalised Vs profiles and normalised qc profiles, with
respect to both the material layering (i.e. layer thicknesses) and relative stiffness of each
layer.

(a)

(b)

(c)

Figure 9.1: Monte Carlo inversion at station CS 7.1 (a) measured and theoretical
HVSR curves, (b) measured and theoretical dispersion curves, and (c) normalised
CPT, q c versus normalised average VS profile with its corresponding 95%
confidence interval (n = 30).
- 169 -

Chapter Nine: Inversion of HVSR and Dispersion Curves using Monte Carlo Simulation

(a)

(b)

(c)

Figure 9.2: Monte Carlo inversion: Measured and theoretical HVSR curves at
stations (a) CS7.2, (b) CS7.3, and (c) CS7.4.

- 170 -

Chapter Nine: Inversion of HVSR and Dispersion Curves using Monte Carlo Simulation

(a)

(b)

(c)

Figure 9.3: Monte Carlo inversion: Normalised CPT, q c versus normalised average
VS profile with its corresponding 95% confidence interval (n = 30) at stations (a)
CS7.2, (b) CS7.3, and (c) CS7.4.
- 171 -

Chapter Nine: Inversion of HVSR and Dispersion Curves using Monte Carlo Simulation

9.3 Influence of Mean Grain Size (D50) on


Estimated Vs Profiles
An additional study was undertaken to determine the influence of the mean grain size (D50)
on the normalised average Vs values and the average normalised qc values of each layer of
the four control stations, using the correlation proposed by Karray et al. (2011) for sandy
soils:
0.115
Vs = 125.5 (q c ) 0.25 D50

(9.6)

where Vs is in m/s, qc in MPa, and D50 is the mean grain size in mm. Figure 9.4 shows
these data set pairs (i.e. normalised Vs and normalised qc) in comparison with the
correlation curves for mean grain size D50 = 0.1, 0.48, 1.8, 5.7, and 10 mm. It can be seen
that most layers without coarse gravel are within fine to medium mean grain size range (D50
between 0.1 and 1.8 mm). Also, most layers containing coarse gravel have a higher average
normalised qc than the layers not containing gravel, and some of these layers are within
medium to coarse mean grain size range (D50 between 1.8 and 10 mm), showing high
normalised Vs, which are possibly due to the presence of these large particle size material
(Karray et al. 2011).

Figure 9.4: Comparison between average normalised q c and normalised Vs obtained


from the Monte Carlo inversion of the four control stations at the DCPT-7 area.
- 172 -

Chapter Nine: Inversion of HVSR and Dispersion Curves using Monte Carlo Simulation
The three red points enclosed by red dashed ellipse, represent the second layer of the Vs
profiles with low velocities obtained by the Monte Carlo simulation inversion method for
the control stations CS 7.1. CS 7.2 and CS7.3, and are shown to have large presence of
clayey soils (mainly silty clay, and clayey sandy silt). A further correlation using the above
results can be made against the soil behaviour index, Ic (see, e.g. Robertson 2009), using the
relationship proposed by Karray et al. (2011).
0.12
I c = 1.69 D50

(9.7)

It should be noted that this correlation was also based on sandy soils. Figure 9.5 shows the
soil behaviour index, Ic and the corresponding mean grain size, D50 for all layers of the four
control stations (excluding the three abovementioned points enclosed by red dashed eclipse
which contain a large presence of clays and/or silts). According to this figure, the material
of all the layers at the four control stations, are predominantly composed of sandy material
(2.35> Ic >1.3), with some presence of coarse gravel.

Figure 9.5: Correlation between mean grain size (D50) and soil behaviour index (I c)
obtained from average normalised, q c and normalised average Vs obtained from the
Monte Carlo inversion of the four control stations at the DCPT-7 area.

9.4 Summary
The Monte Carlo simulation inversion scheme applied in this chapter, allowed for an
enhanced HVSR inversion by incorporating the inversion of Rayleigh wave dispersion
- 173 -

Chapter Nine: Inversion of HVSR and Dispersion Curves using Monte Carlo Simulation
curves. The measured Rayleigh wave dispersion curve obtained from the MASW technique
was useful to filter unrealistic Vs profiles, and to improve the efficiency of the HVSR
inversion. The Vs profiles obtained by the Monte Carlo inversion method were compared
against independent mechanical CPT (qc) data. Good levels of agreement were observed
between the Vs and qc profiles, with respect to both the material layering (i.e. layer
thicknesses) and relative stiffness of each layer. This semi-automated inversion method
allowed estimating the Vs profile with more detail (i.e. use of more layers) when compared
to the trial-and-error forward modelling approach applied in Chapter 8. However, the work
(or cost) involved in applying the Monte Carlo simulation inversion method was quite user
and computer intensive. Thus reducing the appeal of using the Monte Carlo approach
instead of the trial-and-error forward modelling approach in Chapter 8. Furthermore, the
influence of mean grain size D50 on the estimated Vs and qc values was examined and
verified.

- 174 -

Chapter Ten: Conclusions


10.1 Summary of Investigations
This thesis has proposed and implemented three HVSR-based techniques, namely the
HVSR technique, the HVSRplus technique and the HVSR-MASW Monte Carlo method
for assessing ground compaction at the Penrith Lakes quarry. The logistics, time and effort
for implementing each method will increase going from the HVSR technique to the
HVSRplus technique and finally to the HVSR-MASW Monte Carlo method, in that order.
The increase is much more significant between the HVSRplus technique and the HVSRMASW Monte Carlo method because of the significantly greater time required for setting
up an array of geophones in the field and the intensive computational effort required for
HVSR inversion, for the latter method. Conversely, the HVSRplus and the HVSR-MASW
Monte Carlo methods will yield valuable additional information for constraining the HVSR
inversion, namely the fitting of the theoretical HVSR curve from a forward model to the
measured HVSR curve. All three HVSR-based techniques have been demonstrated to work
quite well provided sufficient constraining and/or discriminatory information are available
to address the issue of non-uniqueness of the HVSR inversion. The problem of nonuniqueness in inversion is not confined to the HVSR technique alone, it is in fact a
common trait in all methods that required back figuring including the better known surface
wave methods. The question as to which technique or method should be applied to an
- 175 -

Chapter Ten: Conclusions


area, though, must depend on the availability and quality of constraining and/or
discriminatory information from independent tests. Thus, in an area where the availability
of independent and good quality mechanical tests are in relative abundance (e.g. DCPT-9
area), such that there are sufficient information to constrain and discriminate the correct
solution from other possible inversion solutions, then a sound method of choice may be
the HVSR technique, that is the one with a relatively minimal investment in logistics, time
and effort. On the other hand where the availability of independent mechanical tests is not
in relative abundance, but the test data are nevertheless of a good quality (e.g. DCPT-7
area), then a sensible choice is seemingly the HVSRplus technique as the plus part, the
MSOR technique, would be able to furnish additional information on the state of the
compaction. Finally, where the availability of both independent tests and good quality data
are scarce, the HVSR-MASW Monte Carlo approach may be a prudent choice since it will
give a good estimate of the confidence interval of the results.
This thesis has evaluated four unique areas within Penrith Lakes. Since the fill material, the
type and depth of the bedrock substrate, and the depth of fill are generally similar in the
MCP, WLW, DCPT-9 and DCPT-7 areas, a comparison between these areas has revealed
interesting stratigraphic differences mainly due to different compaction methods. Initially
all the HVSR measurements from Penrith Lakes were interpreted to make a preliminary
evaluation, and then forward modelled to furnish a definitive quantitative comparison
between the quality and consistency of the compaction achieved at these areas. The
comparative results by interpretation of the summarised measured HVSR curves in Figures
6.9, 6.14, 7.11 and 8.2, suggest that the MCP area is a better compacted ground than the
WLW, DCPT-9 and DCPT-7 areas for the following reasons:
1. This thesis has shown in Section 6.2 that as a very loose fill layer is transformed into a
compacted fill layer with increasing compaction, the high frequency part of the HVSR
curve lying to the right of f0 flattens out to a H/V ratio of roughly one with
corresponding diminution of the secondary resonance peak(s). On the basis of this
finding, the MCP area is inferred to be generally more evenly or uniformly compacted
depth-wise due to the relative flatness of its HVSR curves to the right of f0 vis--vis the
corresponding curves from the WLW, DCPT-9 and DCPT-7 areas.
2. The summarised HVSR curves from the MCP area (i.e. Figure 6.9) have the lowest
variability indicating a better level of consistency of compaction was achieved laterally.

- 176 -

Chapter Ten: Conclusions


3. The MCP area has the lowest peak amplitude at f0 (3.1 to 5.0) while its f0 value
(approximately 5.2 Hz) is the highest. These compare with the peak amplitude of 3.0 to
12.1 and f0 of approximately 4.3 Hz for the WLW area, 3.2 to 10.8 and approximately 4
4.5 Hz for the DCPT-9 area, and corresponding values of 3.3 to 8.3 and
approximately 4 4.5 Hz for the DCPT-7 area. The first measure, the peak amplitude
of the HVSR curve at f0, is an indication of the strength of the impedance contrast
between the bedrock and surface layers; that is, the higher the amplitude, the stronger
will be the impedance contrast between bedrock and surface layers. This implies that as
the surface layers become increasingly stiff relative to the bedrock, a corresponding
reduction of the peak amplitude at f0 is observed. The second measure, the value of f0,
may be as a first approximation related to compaction on the basis of Equation (10.1):
f =

Vs
4h

(10.1)

where h is the thickness of the surface layers above the bedrock. For the same
thickness h (about 13 m for all areas), it follows that the Vs of the surface layers is
proportional to predominant frequency f0. In consequence, according to Equation
(10.1) the average Vs for the MCP area is the highest indicating the highest level of
compaction among the four areas.
A more definitive quantitative comparison is obtained in Figure 10.1 showing the overall
mean of the normalised Vs at the stations from the four areas at Penrith Lakes. Here, the
normalised Vs values are plotted against their standard deviation to give a sense of the
compaction achieved laterally. The points from the four areas are found to be clustered and
bounded by the ellipse formed by red solid (MCP area), blue dashed (WLW area), orange
dotted (DCPT-9 area) and green dash-dot (DCPT-7 area) lines. The points deemed as
outliers, in particular for the WLW (three points) and DCPT-7 (one point) areas, have not
been included within the clusters. The following conclusions can be drawn based on the
data presented in Figure 10.1:
1. Overall the MCP area achieved higher mean Vs values for the compacted layers,
implying a higher mean modulus in the compacted layers at the measured stations in
the area. Moreover, the MCP areas standard deviation of the mean Vs values of the
compacted layers is a lower value indicating that this area has achieved greater
consistency between the compacted layers.
- 177 -

Chapter Ten: Conclusions


2. The roller compacted area clusters are situated to the left of the dynamically compacted
area clusters, thus indicating a lower standard deviation (higher consistency) by the
roller compaction than the dynamic compaction.
3. The DCPT-9 cluster is located higher than the DCPT-7 cluster, indicating a higher
mean Vs (better quality) was achieved in the DCPT-9 area. This may be attributed to
the additional stage of rolling compaction (in place of an ironing pass) in the DCPT-9
area, while conversely the DCPT-7 area has not undergone an ironing pass.

Figure 10.1: Penrith Lakes compaction comparison: Mean Vs of the compacted


material versus standard deviation.

10.2 Conclusions
A comprehensive methodology has been developed utilising the HVSR-based technique to
assess compacted (particularly, deep) fills in combination with a limited number of
independent tests which may also include the MSOR and MASW tests (see, in particular
Section 8.3 and Figure 8.3). The work in this thesis is differentiated from previous studies
in that it is focussing not just on the lower frequencies of the ambient vibrations, the
frequency range predominantly considered by geophysists for microzonation studies, but
also on the higher frequencies of the HVSR curve. This thesis has conjectured and
demonstrated that the higher frequencies above f0 of the HVSR curve are particularly
important for geotechnical applications. Initially, the conjecture was tested by computer
- 178 -

Chapter Ten: Conclusions


simulation using a numerical model of surface wave propagation in weakly dissipative
layered ground proposed by Lunedei and Albarello (2009, 2010). The simulation showed
that the impedance contrast from stiffness layering would result in secondary resonance
peaks of the HVSR curve at frequencies higher than the predominant resonance frequency.
In the sensitivity study, it was also shown that a variation of 30 to 40 % in the Vs (a proxy
measure of ground stiffness) between a top layer and a bottom layer would give rise to
discernible secondary peak in the HVSR curve. The proof of concept investigation was
then extended from simulation to a field trial in which the purpose was to demonstrate the
feasibility of applying the HVSR technique, with a focus also on the higher frequencies, to
characterise a future development site at UWS Penrith (Kingswood) campus. The
experience gained from the field trial helped germinate the concept of the three-step
procedure (calibrate, verify, apply) that was to eventually form the backbone of the
methodology of the HVSR-based techniques proposed in this thesis. In this work the need
to use independent but limited mechanical soil investigation (bore logs and SPT) data to
calibrate and verify the results from the inversion of the measured HVSR curves (the first 2
steps of the procedure) was amply demonstrated. This is because the fitting of the
theoretical HVSR curve to the measured HVSR curve to infer the soil stratigraphy will not
lead to a unique solution, and hence discriminatory and/or constraining information from
independent tests are necessary to tease out the correct solution. The idea to then apply the
HVSR technique in the gaps of the ground not covered by the independent mechanical
tests (namely, the third step) using calibrated information as constraints was also
germinated during the field trial.
The sensitivity of the HVSR technique to identify significantly different stages of
compaction was demonstrated in the LD area of the Penrith Lakes quarry, where ambient
vibration measurements were carried out for three stages of rolling compaction (see,
Section 6.2). It further identified the possibility of making a preliminary assessment of
compaction quality by studying the key features (resonance peaks and the troughs) of the
measured HVSR curve. There are, however, limitations in the direct interpretations of the
HVSR curves of the compacted areas. The presence of an embedded softer/loose layer (a
velocity decrease) in the compacted areas, which would appear as HVSR < 1, may be
obscured by the HVSR trough due to the ellipticity of the Rayleigh waves occurring at 2f0.
The possibility of the first higher mode of f0 occurring at 3f0 could also complicate the
interpretation of impedance contrast at depths of the compacted grounds associated with
- 179 -

Chapter Ten: Conclusions


3f0. Both of these limitations may be addressed by the forward modelling of the HVSR
curve with known constraints to derive the Vs profile of the compacted layers, and in the
process confirming and/or clarifying the interpretations of the key features of the
measured HVSR curves. The forward modelling to associate the Vs profile in the
heterogeneous media to the peaks and troughs of the HVSR curve is not a simple and
straight-forward process. Here engineering judgement and experience are vital to ensure
correct interpretation of experimental data and theoretical simulations, particularly where
automated procedures are involved. Moreover, it is important to have a deep
understanding on both the experimental and theoretical aspects of the surface wave
technique before utilising them in real applications. The HVSR technique was applied to
more comprehensively assess the consistency of both shallow rolling compacted fills, deep
dynamically compacted fills and composite (rolling and dynamically) compacted fills in the
MCP and WLW areas, DCPT-7 area and DCPT-9 area, respectively based on the threestep procedure (calibrate, verify, apply). The use of MSOR tests was introduced later in the
thesis to facilitate the calibration of the HVSR forward model, particularly in terms of
providing additional information for the initial guess of the Vs profile in the HVSR forward
modelling procedure. This enhanced method is given the name, HVSRplus technique. A
second enhanced method, where the Vs profile was inferred by inversion using both HVSR
and MASW dispersion curves in a Monte Carlo approach, was also applied for the
dynamically compacted DCPT-7 area. This approach models the uncertainty in the input
parameters of the ground models applied in the inversion and so gives a probabilistic rather
than deterministic estimate of the Vs profile.
Evaluation of compaction quality and consistency both laterally and depth wise for the
MCP, WLW, DCPT-9 and DCPT-7 areas based on the parameter Vs was demonstrated. In
particular, future areas within the Penrith Lakes site lacking rigorous independent
mechanical testing or other forms of stringent monitoring would benefit from applying the
HVSR metrics described in the thesis to assess the quality of compaction achieved. Site
specific correlations of Vs with mechanical penetration data (CPT, SPT) were presented to
provide an estimate of the resistance (qc or N) of the ground at locations without
mechanical tests, in the respective study areas. A depth-frequency relationship was
established for the DCPT-7 area in Chapter 8, for use at locations without mechanical
testing to provide a preliminary indication of the depth of stratigraphy features (e.g.
bedrock, depth of influence of dynamic compaction, etc) observed on the measured HVSR
- 180 -

Chapter Ten: Conclusions


curve. The mapping between the depth of bedrock and the predominant resonance
frequency was extended to include the mapping of the depths of layers with strong
impedance contrasts to the secondary resonance peaks, after the data have been verified by
independent mechanical test.

10.3 Recommendations for Future Research


The HVSR technique is relatively new amongst other well established surface wave
techniques, especially in respect to geotechnical applications. The experience gained from
this thesis has highlighted several issues with the HVSR technique which can be refined,
improved and extended to help evolve this technique into a more robust tool for
geotechnical engineers.

Removal of Anthropic Noise from Measured HVSR Curves


Section 7.2 demonstrated how the HVSR curve can be influenced by artificial noise
generated by certain activities during traffic time from a well utilised road, located roughly
600 m away. It involved studying individual hourly HVSR curves over a 24 hour period and
relating them to traffic conditions (i.e. peak hour traffic times). It was obvious that the
double peaks observed with the HVSR curves were largely due to artificial noise sources
(i.e. certain activities during working hours, in this case). In particular, the fluctuating peak
at approximately 5.8 Hz was shown to be caused by anthropic traffic activities. The
individual HVSR curves did show some consistency with respect to the frequencies of the
peaks, however, the same consistency was not seen when observing the amplitudes of
some of the resonance peaks, in particular that of the predominant resonance frequency, f0.
This dilemma has the ability to affect the results of the HVSR technique in two ways: (1)
manipulating the measured HVSR curve, therefore the interpretation of the curve (i.e.
characterisation of the compacted ground), and (2) influencing the forward modelling of
the HVSR curve when estimating the Vs profile, particularly when using automated
inversion procedures.
Two filtering methods have been trialled in this thesis and are shown in Appendix B.2.
They provide a crude approach in minimising the affects of anthropic noise, particularly for
automated inversion procedures which do not and cannot differentiate between resonance
- 181 -

Chapter Ten: Conclusions


peaks representing stratigraphic features or resonance peaks due to anthropic noise. Thus it
is envisaged that the identification and elimination of the unwanted artificial noise would
require further research into signal processing procedures to determine the most
appropriate method to remove and/or minimise the influence of anthropic noise.

Automatic/Semi-Automatic Inversion of HVSR Curves


The trial-and-error forward modelling procedure adopted in this thesis is without a doubt
the most and probably the only onerous step of the HVSR technique. This step requires
the most time of all steps to satisfactorily estimate the soil profile of a site. Thus it is useful,
even necessary, to produce an inversion procedure to automatically or semi-automatically
fit the theoretical HVSR curve to the measured HVSR curve, by developing appropriate
signal processing, optimisation algorithms and software to create a more user friendly
method. The statistical Monte Carlo simulation procedure was utilised in Chapter 9 of this
thesis, however, as mentioned above, the intensive computational and human effort and
time required is too demanding. Several optimisation methodologies can be considered for
the HVSR inversion process. A direct search approach such as the Marquardt least square
algorithm (Marquardt 1963), may be an efficient choice as shown by Arai and Tokimatsu
(2004), given enough prior knowledge of the site under examination and expertise.
Evolutionary Algorithms such as genetic algorithms are a popular choice in other
optimisation applications, as the requirements for prior knowledge are somewhat relaxed as
a good initial guess is no longer required. However, some expertise is still required to
sieve through and find reasonable and meaningful solutions. Perhaps a combination of
direct search, statistical and evolutionary optimisation algorithms to create a hybrid
optimisation algorithm is the solution to efficient and effective automatic HVSR inversion.
A fully automated HVSR inversion procedure would be quite difficult to achieve since it
has been shown in this thesis that constraining and/or discriminatory information is
essential. As mentioned above, some prior knowledge of the tested ground is necessary
because inversion of measured HVSR data does not necessarily provide a unique solution;
that is, the same theoretical HVSR curve may be produced using two different soil profiles.
Thus a semi-automated procedure is perhaps the more sensible solution. Human
involvement in the semi-automated inversion procedure could include: providing
constraining ranges and/or initial guess of the Vs profiles, selection of which features of
the curve to fit, etc.
- 182 -

Chapter Ten: Conclusions

HVSR-based Techniques in Different Applications


Structural Health Monitoring
The application of the ambient vibration HVSR technique for characterisation of an
ongoing ground compaction at Penrith Lakes illustrates one form of structural health
monitoring of an important infrastructure on which other infrastructures are supported.
This technique and/or its extension, however, offers more possibilities for structural health
monitoring in geotechnical or geotechnical/structural context that is beyond the
application illustrated in this thesis because of its cost and time efficiency. In the postcompaction scenario, continued health monitoring of the compacted soil structure is
possible with a minimum of effort and cost. It is not only normal compacted ground that
may benefit from the HVSR approach, a highly apt health monitoring application is for the
HVSR surveillance of unstable slope which runs a high risk of failure. This may constitute
an important part of an early warning system in a part of infrastructure where failure is not
only extremely difficult to predict, but often catastrophic. Further examples are for the
structural health monitoring of embedded shallow foundations and deep pile foundations
where visual monitoring is simply not possible and the use of a non-invasive technique is
clearly desired (e.g. Nakamura et al. 1995). The ambient vibration HVSR technique may
also be conveniently applied for inferring the structural integrity of a building by means of
monitoring its dynamic characteristics, namely the resonance modes and frequencies (e.g.
Mucciarelli et al. 2001). This is particularly the case when foundations and structures have
been subjected to potentially catastrophic loadings such as from earthquake or high impact
energy which may lead to resonance- and mode-changing damage. As many the
possibilities for the HVSR technique may be, the above illustrations also highlight
opportunities to integrate it into a robust structural health monitoring system by offering a
fairly unique mode of monitoring currently not available in the other methods.

Mineral, Gas and Oil Prospecting


An interesting paper by Saenger et al. (2009), identifies the potential of utilising the
operationally simple and fast HVSR technique for hydrocarbon reservoir prospecting. It
should be noted that they did not use the Horizontal/Vertical spectral ratio (HVSR) of
microtremors, but instead the Vertical/Horizontal spectral ratio (VHSR). They discussed
the ability of microtremors correlating, with a high degree, to the location of hydrocarbon
- 183 -

Chapter Ten: Conclusions


reservoirs. Thus conveying the potential of using microtremor measurements to better
indicate and appraise the situation of hydrocarbon reservoirs. Conventional seismic
technologies (e.g. refraction and reflection techniques) for hydrocarbon prospecting rely on
very long/large arrays (sometimes up to several kms) of interconnecting geophones and
very loud active sources (e.g. explosives, accelerated weight drops) to reach the very deep
depths. This is where the advantage of using a non-destructive passive ambient vibration
HVSR technique could serve as an alternative for more cost effective and time efficient
hydrocarbon prospecting. Furthermore, the potential of applying the HVSR technique for
the prospecting of hydrocarbon reservoirs could be extended to the exploration of other
valuable and important minerals (e.g. gold, metals, etc).

- 184 -

References

References
Aki, K. 1957, 'Space and time spectra of stationary stochastic waves, with special reference
to microtremors', Bulletin of the Earthquake Research Institute vol. 35, pp. 415 - 456.
Aki, K. & Richards, P.G. 1980, Quantitative Seimology: Theory and Methods, W.H. Freeman and
Co., San Francisco.
Albarello, D. & Lunedei, E. 2009, 'Alternative interpretations of horizontal to vertical
spectral ratios of ambient vibrations: new insights from theoretical modelling',
Bulletin of Earthquake Engineering, vol. 8, no. 3, pp. 519 - 534.
Arai, H. & Tokimatsu, K. 2000, 'Effects of Rayleigh and Love waves on microtremor H/V
spectra', paper presented to the 12th World Conference on Earthquake Engineering, paper
2232.
Arai, H. & Tokimatsu, K. 2004, 'S-Wave velocity profiling by inversion of microtremor
H/V spectrum', Bulletin of the Seismological Society of America, vol. 94, no. 1, pp. 53-63.
Arai, H. & Tokimatsu, K. 2005, 'S-wave velocity profiling by joint inversion of
microtremor dispersion curve and horizontal-to-vertical (H/V) spectrum', Bulletin of
the Seismological Society of America, vol. 95, no. 5, pp. 1766-1778.
Arulrajah, A., Bo, M.W., Piratheepan, J. & Disfani, M.M. 2011, 'In situ testing of soft soil at
a case study site with the self-boring pressuremeter', Geotechnical Testing Journal, vol.
34, no. 4, paper no. GTJ103310.
Australian Bureau of Statistics 2012, Year book Australia, Report Number 92, ABS
Catalogue No. 1301.0, Australian Bureau of Statistics, Canberra.
Bard, P.Y. 1998, 'Microtremor measurements: a tool for site effect estimation?', Proceeding of
the Second International Symposium on the Effects of Surface Geology on Seismic Motion,
Yokohama, Japan, pp. 1251 - 1279.
Bell, F.G. 2004, Engineering geology and construction, 1st edn, Spon Press (imprint of Taylor
Francis Group), New York.
Blake, W.D. & Gilbert, R.B. 1997, 'Investigation of possible relationship between
undrained shear strength and shear wave velocity for normally consolidated clays',
Offshore Technology Conference, Houston, Texas, pp. 411 - 420.
Bo, M.W., Chang, M.F., Arulrajah, A. & Choa, V. 2012, 'Ground investigations for Changi
East reclamation projects', Geotechnical and Geological Engineering, vol. 30, pp. 45 - 62.

- 185 -

References
Bo, M.W., Na, Y.M., Arulrajah, A. & Chang, M.F. 2009, 'Densification of granular soil by
dynamic compaction', Proceedings of the Institution of Civil Engineers: Ground Improvement,
vol. 162, no. G13, pp. 121 - 132.
Bonnefoy-Claudet, S., Cornou, C., Bard, P.Y., Cotton, F., Moczo, P., Kristek, J. & Fah, D.
2006b, 'H/V ratio: a tool for site effects evaluation. Results from 1-D noise
simulations.', Geophysical Journal International, vol. 167, pp. 827-837.
Bonnefoy-Claudet, S., Cotton, F. & Bard, P.Y. 2006a, 'The nature of noise wavefield and its
applications for site effects studies. A literature review', Earth-Science Reviews, vol. 79,
pp. 205-227.
Brouwer, J.J.M. 2008, In-situ soil testing, Lankelma, <http://www.conepenetration.com/
online-book>.
Castellaro, S. & Mulargia, F. 2009a, 'Vs30 estimates using constrained H/V measurements',
Bulletin of the Seismological Society of America, vol. 99, no. 2A, pp. 761 - 773.
Castellaro, S. & Mulargia, F. 2009b, 'The effect of velocity inversions on H/V', Pure and
Applied Geophysics, vol. 166, no. 4, pp. 567 - 592.
Castellaro, S., Mulargia, F. & Bianconi, L. 2005, 'Passive seismic stratigraphy: A new
efficient, fast and economic technique', T & A, no. 3, pp. 51 - 72.
Cha, M. & Cho, G.-C. 2007, 'Shear strength estimation of sandy soils using shear wave
velocity', Geotechnical Testing Journal, vol. 30, no. 6.
Churchman, G.J., Askary, M., Peter, P., Wright, M., Raven, M.D. & Self, P.G. 2002,
'Geotechnical properties indicating environmental uses for an unusual Australian
bentonite', Applied Clay Science, vol. 20, no. 45, pp. 199-209.
Claria Jr., J.J. & Rinaldi, V.A. 2007, 'Shear wave velocity of a compacted clayey silt',
Geotechnical Testing Journal, vol. 30, no. 5, pp. 399 - 408.
Coffey Geotechnics Pty Ltd 2007a, Geotechnical investigation and monitoring report - dynamic
compaction prototype trial compaction area 9, Report for Penrith Lakes Development
Corporation.
Coffey Geotechnics Pty Ltd 2007b, Geotechnical investigation and monitoring report - dynamic
compaction prototype trial compaction area 7, Report for Penrith Lakes Development
Corporation.
Cotton, F., Scherbaum, F., Bommer, J.J. & Bungum, H. 2006, 'Criteria for selecting and
adjusting ground-motion models for specific target regions: application to Central
Europe and rock sites', Journal of Seismology, vol. 10, pp. 137 - 156.

- 186 -

References
Dal Moro, G. & Ferigo, F. 2011, 'Joint analysis of Rayleigh- and Love-wave dispersion:
Issues, criteria and improvements', Journal of Applied Geophysics, vol. 75, no. 3, pp.
573-589.
Delgado, J., Casado, C.L., Estevez, A., Giner, J., Cuenca, A. & Molina, S. 2000, 'Mapping
soft soils in the sergura river valley (SE Spain): a case study of microtremors as an
exploration tool', Journal of Applied Geophysics, vol. 45, pp. 19-32.
Durgunoglu, H., Varaksin, S., Briet, S. & Karadayilar, T. 2003, 'A case study on soil
improvement with heavy dynamic compaction', Proc. XIII ECSMGE, ISBN 8086769-00-3, vol. 1, pp. 651-655.
Fah, D., Kind, F. & Giardini, D. 2001, 'A theoretical investigation of average H/V ratios',
Geophysical Journal International, vol. 145, pp. 535-549.
Fah, D., Kind, F. & Giardini, D. 2003, 'Inversion of local S-wave velocity structures from
average H/V ratios, and their use for the estimation of site-effects', Journal of
Seismology, vol. 7, pp. 449-467.
Fear, C.E. & Robertson, P.K. 1995, 'Estimating the undrained strength of sand: a
theoretical framework', Canadian Geotechnical Journal, vol. 32, no. 5, pp. 859870.
Feng, S.J., Shui, W.H., Gao, L.Y., He, L.J. & Tan, K. 2010, 'Field studies of the
effectiveness of dynamic compaction in coastal reclamation areas', Bulletin of
Engineering Geology and the Environment, vol. 69, no. 1, pp. 129 - 136.
Foti, S. 2000, 'Multistation methods for geotechnical characterization using surface waves',
PhD Thesis, Politecnico di Torino, Italy.
Foti, S., Parolai, S., Albarello, D. & Picozzi, M. 2011, 'Application of surface-wave methods
for seismic site characterisation', Surveys in Geophysics, vol. 32, no. 6, pp. 777 - 825.
Fredlund, D.G. & Rahardjo, H. 1993, Soil mechanics for unsaturated soils, John Wiley & Sons,
Inc., New York, USA.
Gallipoli, M.R. & Mucciarelli, M. 2009, 'Comparison of site classification from Vs30, Vs10,
and HVSR in Italy', Bulletin of the Seismological Society of America, vol. 99, no. 1, pp. 340
- 351.
Gardner, G.H.F., Gardner, L.W. & Gregory, A.R. 1974, 'Formation velocity and density
The diagnostic basics for stratigraphic traps', Geophysics, vol. 39, pp. 770-780.
Hanumantharao, C. & Ramana, G.V. 2008, 'Dynamic soil properties for microzonation of
Delhi, India', Journal of Earth System Science, vol. 117, no. S2, pp. 719-730.

- 187 -

References
Hardin, B.O. 1978, 'The nature of stress-strain behaviour of soils', Proceedings of the
Earthquake Engineering and Soil Dynamics Conference, vol. 1, ASCE, Pasadena,
California, pp. 33 - 90.
Hardin, B.O. & Drnevich, V.P. 1972, 'Shear modulus and damping in soils', Journal of Soil
Mechanics and Foundations Division, vol. 98, no. 7, pp. 667-692.
Harkrider, D.G. 1964, 'Surface waves in multilayered elastic media. Rayleigh and Love
waves from buried sources in a multilayered elastic half-space', Bulletin of the
Seismological Society of America, vol. 54, no. 2, pp. 627-679.
Harutoonian, P., Leo, C.J., Doanh, T., Castellaro, S., Zou, J.J., Liyanapathirana, D.S.,
Wong, H. & Tokeshi, K. 2012a, 'Microtremor measurements of rolling compacted
ground', Soil Dynamics and Earthquake Engineering, vol. 41, pp. 23-31.
Harutoonian, P., Leo, C.J., Liyanapathirana, D.S. & Wong, H. 2012b, 'Site characterisation
by the HVSR technique', Australian Geomechanics Journal, vol. 47, no. 3, pp. 103 - 112.
Hasancebi, N. & Ulusay, R. 2007, 'Empirical correlations between shear wave velocity and
penetration resistance for ground shaking assessments', Bulletin of Engineering Geology
and the Environment vol. 66, pp. 203-213.
Haskell, N.A. 1953, 'The dispersion of surface waves on multilayered media', Bulletin of the
Seismological Society of America, vol. 43, no. 1, pp. 17 - 34.
Hatanaka, M. & Uchida, A. 1996, 'Empirical correlation between penetration resistance and
internal friction angle of sandy soils', Soils and Foundations, vol. 36, no. 4, pp. 1 - 9.
Heitor, A., Indraratna, B. & Rujikiatkamjorn, C. 2012, 'Characterising compacted soil using
shear wave velocity and matric suction', Australian Geomechanics Journal, vol. 47, no.
2, pp. 79-86.
Herak, M. 2008, 'ModelHVSR - a tool to model horizontal-to-vertical spectral ratio of
ambient noise', Computers and Geosciences, vol. 34, pp. 1514 - 1526.
Herrmann, R.B. 1987, Computer programs in seismology, Vol. IV, St. Louis University, MO,
USA.
Herrmann, R.B. 2002, Computer programs in seismology, 3.2 edn, Saint Louis University.
Heymann, G. 2007, 'Ground stiffness measurement by the continuous surface wave test',
Journal of the South African Institution of Civil Engineering, vol. 49, no. 1, pp. 25 - 31.
Hisada, Y. 1994, 'An efficient method for computing Green's functions for a layered halfspace with sources and receivers at close depths', Bulletin of the Seismological Society of
America, vol. 85, pp. 1456 - 1472.

- 188 -

References
Hisada, Y. 1995, 'An efficient method for computing Green's functions for a layered halfspace with sources and receivers at close depths (Part 2)', Bulletin of the Seismological
Society of America, vol. 85, pp. 1080 - 1093.
Holzer, T.L., Bennett, M.J., Noce, T.E. & Tinsley III, J.C. 2005, 'Shear-wave velocity of
surficial geologic sediments in Northern California: Statistical distributions and
depth dependence', Earthquake Spectra, vol. 21, no. 1, pp. 161 - 177.
Huang, S., Barbour, S.L. & Fredlund, D.G. 1998, 'Development and verification of a
coefficient of permeability function for a deformable unsaturated soil', Canadian
Geotechnical Journal, vol. 35, no. 3, pp. 411-425.
Ibs-von Seht, M. & Wohlenberg, J. 1999, 'Microtremor measurements used to map
thickness of soft sediments', Bulletin of the Seismological Society of America, vol. 89, no.
1, pp. 250-259.
Inazaki, T. 2006, 'Relationship Between S-Wave Velocities and Geotechnical Properties of
Alluvial Sediments', Symposium on the Application of Geophysics to Engineering and
Environmental Problems, vol. 19, no. 1, pp. 1296-1303.
Jeffrey and Katauskas Pty Ltd 2008, Report to the University of Western Sydney on additional
geotechnical investigation for proposed campus library at University of Western Sydney campus off
John Flak Avenue, Kingswood, Penrith (Kingswood), Australia.
Jeffrey and Katauskas Pty Ltd 2012, JK250 Drill Rig, viewed 17June 2012.
Kanai, K. 1957, 'Semi-empirical formula for the seismic characteristics of the ground',
Bulletin of the Earthquake Research Institute, vol. 35, pp. 309 - 325.
Kanai, K. & Tanaka, T. 1954, 'Measurement of the microtremor I', Bulletin of the Earthquake
Research Institute of Tokyo, vol. 32, pp. 199 - 209.
Karray, M., Lefebvre, G., Ethier, Y. & Bigras, A. 2011, 'Influence of particle size on the
correlation between shear wave velocity and cone tip resistance', Canadian
Geotechnical Journal, vol. 48, no. 4, pp. 599 - 615.
Khalili, N. & Khabbaz, M.H. 1998, 'A unique relationship for for the determination of
the shear strength of unsaturated soils', Geotechnique, vol. 48, no. 5, pp. 681 - 687.
Kim, D.S. & Park, H.C. 1999, 'Evaluation of ground densification using spectral analysis of
surface wave (SASW) and resonant column (RC) tests', Canadian Geotechnical Journal,
vol. 36, no. 2, pp. 291 - 299.
Kim, D.S., Shin, M.K. & Park, H.C. 2001, 'Evaluation of density in layer compaction using
SASW method', Soil Dynamics and Earthquake Engineering, vol. 21, no. 1, pp. 39 - 46.

- 189 -

References
Konno, K. & Ohmachi, T. 1998, 'Ground-motion characteristics estimated from spectral
ratio between horizontal and vertical components of microtremor', Bulletin of the
Seismological Society of America, vol. 88, pp. 228 - 241.
Kottegoda, N.T. & Rosso, R. 2008, Applied statistics for civil and environmental engineers - Second
Edition, Blackwell Publishing Ltd, United Kingdom.
Kudo, K. 1995, 'Practical estimates of site response. State-of-the-art report', Proceedings of the
fifth International Conference on Seismic Zonation, vol. 3, Nice, France, pp. 1878 - 1907.
Lachet, C. & Bard, P.Y. 1994, 'Numerical and theoretical investigations on the possibilities
and limitations of Nakamura's technique', Journal of Physics of the Earth, vol. 42, no. 4,
pp. 377-397.
Lai, C.G. & Rix, G.J. 1998, Simultaneous inversion of Rayleigh phase velocity and attenuation for nearsurface site characterisation, Georgia Insitute of Technology.
Lambe, T.W. & Whitman, R.V. 1979, Soil mechanics, SI edn, John Wiley and Sons, New
York.
Lamit Co. 2012, Initiation in primary seismology and the SOS-LIFE Earthquake Alarm working
principle, viewed 09/02/2012 <http://www.lamit.ro/earthquake-early-warningsystem.htm>.
Lay, M.G. 2009, Handbook of road technology, 4th edn, Spon Press (imprint of the Taylor
Francis Group), New York.
Louie, J.N. 2001, 'Faster, better shear-wave velocity to 100 meters depth from refraction
microtremor arrays', Bulletin of the Seismological Society of America, vol. 91, no. 2, pp.
347-364.
Lunedei, E. & Albarello, D. 2009, 'On the seismic noise wavefield in a weakly dissipative
layered earth', Geophysical Journal International, vol. 177, no. 3, pp. 1001 - 1014.
Lunedei, E. & Albarello, D. 2010, 'Theoretical HVSR curves from full wavefield modelling
of ambient vibrations in a weakly dissipative layered Earth', Geophysical Journal
International, vol. 181, no. 2, pp. 1093 - 1108.
Marchetti, S. 1980, 'In Situ Tests by Flat Dilatometer', Journal of the Geotechnical Engineering
Division, ASCE, vol. 106, no. GT3, pp. 299 - 321.
Marquardt, D.W. 1963, 'An algorithm for least squares estimation of nonlinear parameters',
Journal of the Society for Industrial and Applied Mathematics vol. 11, pp. 431 - 441.
Mavko, G., Mukerji, T. & Dvorkin, J. 2003, Rock Physics Handbook - Tools for Seismic Analysis
in Porous Media, Cambridge University Press, <http://www.knovel.com/web/portal
/browse/display?_EXT_KNOVEL_DISPLAY_bookid=2312>.
- 190 -

References
McKenna, J., McKenna, M., Yushanov, S., Crompton, J. & Koppenhoefer, K. 2008,
'Computational modeling of wave propagation in a geophysical domain', paper
presented to the COMSOL Conference 2008, Boston, USA.
Menzies, B.K. & Matthews, M.C. 1996, 'The continuous surface wave system: a modern
technique for site investigation', paper presented to the Special Lecture: Indian
Geotechnical Conference, Madras.
Mesri, G. & Olson, R. 1971, 'Mechanisms controlling the permeability of clays', Clays and
Clay Minerals, vol. 19, pp. 151 - 158.
Meyerhof, G.G. 1957, 'Discussion on Research on determining the density of sands by
spoon penetration testing', 4th International Conference on Soil Mechanics and Foundation
Engineering vol. 3, 110, London.
Mooney, M., Rinehart, R., White, D., Vennapusa, P., Facas, N. & Musimbi, O. 2010,
Intelligent Soil Compaction Systems, in Transportation Research Board (ed.), NCHRP
(National Cooperative Highway Research Program) Report 676, Washington D.C., p. 165.
Moss, R.E.S., Seed, R.B. & Olsen, R.S. 2006, 'Normalizing the CPT for overburden stress',
Journal of Geotechnical and Geoenvironmental Engineering, vol. 132, no. 3, pp. 378-387.
Moxhay, A.L., Tinsley, R.D. & Sutton, J.A. 2001, 'Monitoring of soil stiffness during
ground improvement using seismic surface waves', Ground Engineering, January, pp.
34 - 37.
Mucciarelli, M., Contri, P., Monachesi, G., Calvano, G. & Gallipoli, M.R. 2001, 'An
empirical method to assess the seismic vulnerability of existing buildings using the
HVSR technique', Pure and Applied Geophysics, vol. 158, pp. 2635 - 2647.
Mullins, G. 2006, 'In Situ Soil Testing', in M. Gunaratne (ed.), The Foundation Engineering
Handbook, Taylor and Francis, Boca Raton, Florida, pp. 47 - 86.
Nakamura, Y. 1989, 'A method for dynamic characteristics estimation of subsurface using
microtremor on the ground surface', Quarterly Report of Railway Technical Research
Institute (RTRI), vol. 30, no. 1, pp. 25 - 33.
Nakamura, Y. 2008, 'On the H/V spectrum', paper presented to the The 14th World
Conference on Earthquake Engineering, Beijing, China. Paper ID: 07-0033, 12 - 17
October 2008.
Nakamura, Y., Hidaka, K., Sato, S. & Tachibana, M. 1995, 'Proposition of a method for
pier inspection using microtremor', Quarterly Report of Railway Technical Research
Institute (RTRI), vol. 36, no. 2, pp. 65-70.

- 191 -

References
Nazarian, S. & Stokoe, K.H. 1984, 'In situ shear wave velocities from spectral analysis of
surface waves', 8th World Conference on Earthquake Engineering, vol. 3, San Francisco,
USA, pp. 31 - 38.
Nogoshi, M. & Igarashi, T. 1971, 'On the amplitude characteristics of ambient noise (Part
2)', Journal of Seismology Society Japan, vol. 24, pp. 26-40.
Oda, M. 1977, 'Co-ordination number and its relation to shear strength of granular
material', Soils and Foundations, vol. 17, no. 2, pp. 29 - 42.
Olsen, K.B., Day, S.M., Dalguer, L.A., Mayhew, J., Cui, Y., Zhu, J., Cruz-Atienza, V.M.,
Roten, D., Maechling, P., Jordan, T.H., Okaya, D. & Chourasia, A. 2009,
'ShakeOut-D: Ground motion estimates using an ensemble of large earthquakes on
the southern San Andreas fault with spontaneous rupture propagation', Geophysical
Research Letters, vol. 36, no. 4, p. L04303.
Panayiotopoulos, K.P. 1989, 'Packing of sandsA review', Soil and Tillage Research, vol. 13,
no. 2, pp. 101-121.
Park, C.B. & Miller, R.D. 2004, Report: MASW to map shear-wave velocity of soil, Kansas
Geological Survey, University of Kansas.
Park, C.B., Miller, R.D. & Xia, J. 1997, Multi-channel analysis of surface waves (MASW) "A
summary report of technical aspects, experimental results, and perspective", Kansas Geological
Survey.
Park, C.B., Miller, R.D. & Xia, J. 1999, 'Multichannel analysis of surface waves', Geophysics,
vol. 64, no. 3, pp. 800-808.
Parolai, S., Bormann, P. & Milkereit, C. 2002, 'New relationships between Vs , thickness of
sediments, and resonance frequency calculated by the H/V ratio of seismic noise
for the Cologne area (Germany)', Bulletin of the Seismological Society of America, vol. 92,
no. 6, pp. 2521-2527.
Parolai, S., Picozzi, M., Richwalski, S.M. & Milkereit, C. 2005, 'Joint inversion of phase
velocity dispersion and H/V ratio curves from seismic noise recordings using a
genetic algorithm, considering higher modes', Geophysical Research Letters, vol. 32, no.
L11308.
Peck, R.B., Hanson, W.E. & Thornburn, T.H. 1974, Foundation engineering, 2nd edn, John
Wiley and Sons, Inc., New York.
Picozzi, M. & Albarello, D. 2007, 'Combining genetic and linearized algorithms for a twostep joint inversion of Rayleigh wave dispersion and H/V spectral ratio curves',
Geophysical Journal International, vol. 169, pp. 189 - 200.
- 192 -

References
Proctor, R.R. 1933, 'Fundamental principles of soil compaction', Engineering News-Record,
vol. 111, no. 9, pp. 245 - 248.
Pullammanappallil, S. 2006, 'Geotechnical and geophysical case studies involving the
refraction microtremor (ReMi) method for shear wave profiling', paper presented
to the Highway GeophysicsNDE conference, St. Louis, Missouri.
Robertson, P.K. 2009, 'Interpretation of cone penetration tests a unified approach',
Canadian Geotechnical Journal, vol. 46, no. 11, pp. 1337-1355.
Robertson, P.K., Woeller, D.J. & Finn, W.D.L. 1992a, 'Seismic cone penetration test for
evaluating liquefaction potential under cyclic loading', Canadian Geotechnical Journal,
vol. 29, no. 4, pp. 686-695.
Robertson, P.K., Woeller, D.J., Kokan, M., Hunter, J. & Luternaur, J. 1992b, 'Seismic
techniques to evaluate liquefaction potential', Proceedings of the 45th Canadian
Geotechnical Conference, Toronto, Canada, pp. 5-15-9.
Ryden, N., Park, C.B., Ulriksen, P. & Miller, R.D. 2004, 'Multimodal approach to seismic
pavement testing', Journal of Geotechnical and Geoenvironmental Engineering, vol. 130, no.
6, pp. 636 - 645.
Saenger, E.H., Schmalholz, S.M., Lambert, M.-A., Nguyen, T.T., Torres, A., Metzger, S.,
Habiger, R.M., Mller, T., Rentsch, S. & Mndez-Hernndez, E. 2009, 'A passive
seismic survey over a gas field: Analysis of low-frequency anomalies', Geophysics, vol.
74, no. 2, pp. O29-O40.
Sambridge, M. & Mosegaard, K. 2002, 'Monte Carlo methods in geophysical inverse
problems', Reviews of Geophysics, vol. 40, no. 3, pp. 1 - 29.
Snchez-Sesma, F.J., Rodrguez, M., Iturrarn-Viveros, U., Luzn, F., Campillo, M.,
Margerin, L., Garca-Jerez, A., Suarez, M., Santoyo, M.A. & Rodrguez-Castellanos,
A. 2011, 'A theory for microtremor H/V spectral ratio: application for a layered
medium', Geophysical Journal International, vol. 186, no. 1, pp. 221-225.
Schmertmann, J., Baker, W., Gupta, R. & Kessler, K. 1986, 'CPT/DMT quality control of
ground modification at a power plant', paper presented to the Special conference on
"Use of in situ tests in geotechnical engineering", In Situ '86, Blacksburg, VA, ASCE GSP
No. 6, June 23 - 25.
SESAME 2004a, Guidelines for the implementation of the H/V spectral ratio tecnique on ambient
vibrations. Measurements, processing and interpretation. WP12 European Commission Research General Directorate Project No. EVG1-CT-2000-00026 SESAME, Report
Number D23.12.
- 193 -

References
SESAME 2004b, Site effects assessment using ambient excitations: Nature of noise wavefield. WP08
European Commission - Research General Directorate Project No. EVG1-CT-2000-00026
SESAME, Report Number D13.08.
Shearer, P.M. 2010, 'Introduction to seismology: The wave equation and body waves',
unpublished, Institute of Geophysics and Planetary Physics, Scipps Institution of
Oceanography, University of California, San Diego.
Standards Australia 2003a, Methods of testing soils for engineering purposes, Soil compaction and
density tests - Determination of the dry density/moisture content relation of a soil using standard
compactive effort, AS 1289.5.1.1, Available from: Standards Australia Online.
Standards Australia 2003b, Methods of testing soils for engineering purposes, Soil compaction and
density tests - Determination of the dry density/moisture content relation of a soil using modified
compactive effort, AS 1289.5.2.1, Available from: Standards Australia Online.
Standards Australia 2004a, Methods of testing soils for engineering purposes, Soil compaction and
density tests - Determination of the field density of a soil - Sand replacement method using a sandcone pouring apparatus, AS 1289.5.3.1, Available from: Standards Australia Online.
Standards Australia 2004b, Methods of testing soils for engineering purposes, Soil strength and
consolidation tests - Determination of the penetration resistance of soil - Standard penetration test
(SPT), AS 1289.6.3.1, Available from: Standards Australia Online.
Standards Australia 2007, Methods of testing soils for engineering purposes, Soil compaction and density
tests - Determination of field density and field moisture content of a soil using a nuclear surface
moisture-density gauge - Direct transmission mode, AS 1289.5.8.1, Available from:
Standards Australia Online.
Stephenson, W.R. 1978, 'Ultrasonic testing for determining dynamic soil moduli', in,
Dynamic Geotechnical Testing, ASTM STP 654, American Society for Testing and
Materials, pp. 179 - 195.
Sykora, D.W. & Koester, J.P. 1988, 'Correlations between dynamic shear resistance and
standard penetration resistance in soils.', Earthquake Engineering and Soil Dynamics
IIRecent Advances in Ground Motion Evaluation, Proceedings of the Geotechnical
Engineeringg Division Specialty Conference, ed. J.L. Von Thun, American Society of Civil
Engineers (ASCE), Utah, U.S.A, pp. 389404.
Terzaghi, K. 1936, 'The shear resistance of saturated soils', First international conference on soil
mechanics and foundation engineering, vol. 1, Cambridge, U.S.A, pp. 54 - 56.
Terzaghi, K., Peck, R.B. & Mesri, G. 1996, Soil mechanics in engineering practice, 3rd edn, John
Wiley and Sons, Inc, New York.
- 194 -

References
Tezcan, S.S., Keceli, A. & Ozdemir, Z. 2006, 'Allowable bearing capacity of shallow
foundations based on shear wave velocity', Geotechnical and Geological Engineering, vol.
24, pp. 203 - 218.
Tokimatsu, K., Tamura, S. & Kojima, H. 1992, 'Effects of multiple modes on Rayleigh
wave dispersion characteristics', Journal of Geotechnical and Geoenvironmental Engineering,
vol. 118, pp. 1529 - 1543.
Totani, G., Marchetti, S., Monaco, P. & Calabrese, M. 2001, 'Use of the Flat Dilatometer
Test (DMT) in geotechnical design', paper presented to the In situ 2001: International
Conference on In situ measurement of soil, Bali, Indonesia.
Tuan, T.T., Scherbaum, F. & Malischewsky, P.G. 2011, 'On the relationship of peaks and
troughs of the ellipticity (H/V) of Rayleigh waves and the transmission response of
single layer over half-space models', Geophysical Journal International, vol. 184, no. 2,
pp. 793-800.
Ueda, T., Matsushima, T. & Yamada, Y. 2010, 'Effect of particle size ratio on shear
strength of dense binary mixtures', in, Geomechanics and Geotechnics, CRC Press, pp.
507-511.
Venkatramaiah, C. 2006, 'Compaction of Soil', in, Geotechnical Engineering, New Age
International (P) Ltd, New Delhi, pp. 423 - 445.
Waddell, P.J., Moyle, R.A. & Whiteley, R.J. 2010, 'Geotechnical verification of impact
compaction', WIT Transactions on Ecology and the Environment, vol. 141, pp. PII-73 PII-85.
Wei, B.Z., Pezeshk, S., Chang, T.S., Hall, K.H. & Liu, H.P. 1996, 'An empirical method to
estimate shear wave velocity of soils in the New Madrid seismic zone', Soil Dynamics
and Earthquake Engineering, vol. 15, pp. 399 - 408.
Wride, C.E., Robertson, P.K., Biggar, K.W., Campanella, R.G., Hofmann, B.A., Hughes,
J.M., Kpper, A. & Woeller, D.J. 2000, 'Interpretation of in situ test results from
the CANLEX sites', Canadian Geotechnical Journal, vol. 37, no. 3, pp. 505-529.
Xia, J., Miller, R.D. & Park, C.B. 1999, 'Estimation of near-surface shear-wave velocity by
inversion of Rayleigh waves', Geophysics, vol. 64, no. 3, pp. 691 - 700.
Yamanaka, H., Takemura, M., Ishida, H. & Niwa, M. 1994, 'Characteristics of long-period
microtremors and their applicability in exploration of deep sedimentary layers',
Bulletin of the Seismological Society of America, vol. 84, no. 6, pp. 1831 - 1841.
Yilmaz, Y. 2009, 'A study on the limit void ratio characteristics of medium to fine mixed
graded sands', Engineering Geology, vol. 104, pp. 290-294.
- 195 -

References
Yoon, H.-K. & Lee, J.-S. 2010, 'Field velocity resistivity probe for estimating stiffness and
void ratio', Soil Dynamics and Earthquake Engineering, vol. 30, no. 12, pp. 1540-1549.

- 196 -

Appendices

Appendices
A. Seismic Waves and Methods
A.1 Geophysical Methods used to Measure Vs
Cross-hole Seismic Test
The cross-hole seismic test can provide dynamic soil information to depths of more than
100 m by the resolving of seismic wave velocity profiles. Similar to geophysical techniques,
this technique measures P-wave and/or S-wave velocities of the subsurface with the main
focus on the latter. To facilitate the recording of seismic wave velocities of a large site a
number of boreholes are spaced out over the ground, which are required to conduct a
cross-hole seismic test. Seismic waves are generated in the source borehole and recorded in
a receiver borehole. The receiver borehole has a geophone installed, tightly coupled to the
borehole wall. In order to attain a velocity profile with depth of the subsurface, it is
important to keep the source and receiver installed at the same depths along the boreholes.
Figure A.1 illustrates a typical setup of the cross-hole seismic test, this setup consists of one
source borehole and two receiver boreholes. Since the spacing between the boreholes is
known it is possible to determine the velocity of the S-wave's travelling through a medium.

Seismic
Source
Casing

Receivers

Grout
dx
dx
Figure A.1: Cross-hole seismic test.
Figure A.2 illustrates an example of a seismic wave travelling from the source borehole to
the receiver boreholes.
- 197 -

Appendices

Seismic Source

1st Receiver

2nd Receiver

Time (t)
Figure A.2: Example wave travelling from seismic source to receivers.

Spectral Analysis of Surface Waves (SASW) Technique


The spectral analysis of surface waves (SASW) technique is a more versatile and costeffective tool for obtaining shear wave velocity profiles when compared to traditional
dynamic tests. It has the ability to acquire Vs profiles to deeper depths when compared to
traditional dynamic testing. The SASW technique was introduced in the early 1980s, based
on the wave propagation method, to derive the near surface Vs profile of a site (Nazarian
& Stokoe 1984). This method involved analysing the fast Fourier transform (FFT) of
surface waves generated by a dynamic source. A schematic of the SASW test is shown in
Figure A.3.
Computer

- Noise Source
- Geophone

FFT Analyser

dx

2dx

4dx

Figure A.3: SASW technique (After Park et al. 1997).

Multichannel Analysis of Surface Waves (MASW) Technique


The multichannel analysis of surface waves (MASW) technique (Park et al. 1999) has been
developed in response to the limitations of the SASW technique. MASW implements the
- 198 -

Appendices
idea of using multiple receivers to reduce the duration of the test and to improve the
accuracy of the measurements. A comparison of the SASW in Figure A.3 and the MASW
in Figure A.4 will reveal that the earlier is a more tedious procedure, requiring more
readings to be recorded with the energy source at different locations.

Seismic Source

Receivers
Seismograph

Figure A.4: MASW technique (After Park et al. 1997).

Continuous Surface Waves System (CSWS)


The continuous surface waves system (CSWS) technique (Menzies & Matthews 1996)
shown in Figure A.5, relies on the properties of Rayleigh waves in order to determine site
frequencies. The CSWS technique, like the SASW and MASW techniques is an active array
based technique.

Amplifier

Signal Generator
Geophone output to data
acquisition system
Geophones

Vibrator

1
Noise Source

Vph = f

f1

Rayleigh Wave

Figure A.5: CSWS technique (After Heymann 2007).


- 199 -

Appendices
However, unlike the SASW and MASW it does not rely on an impulsive impact source (i.e.
hammer hit), instead it utilises a variable frequency shaker (or vibrator) to generate the
required energy. This feature gives control to the operator to determine the required depth
of penetration and frequency range (i.e. low frequencies generate long wavelengths and can
penetrate deep depths, while high frequencies generate short wavelengths and can
penetrate shallow depths).

Refraction Microtremor (ReMi)


The refraction microtremor (ReMi) technique (Louie 2001) like the HVSR technique is a
form of ambient vibration technique with a number of similarities to the MASW technique
as previously described. The theoretical basis of ReMi is the same as SASW and MASW. It
is considered as passive since the technique does not require an external active excitation
source, but as used by some researchers controlled surface source is instead provided by
jogging or driving a vehicle alongside the geophones. Figure A.6 shows the general
schematic of the ReMi technique, it illustrates a simple site characterisation technique.

Geophone output to data


PC/Laptop

acquisition system
Digitalised
Geophones

Rayleigh and Love Waves

Figure A.6: ReMi technique.

A.2 Background of Seismic Waves


Since the 1920s, characterisation of the Earths interior using seismic waves has been
studied in seismology. In these earlier days, equipment for measuring seismic noise (or
vibrations) was rare and expensive (Shearer 2010). However, the 1950s and 1960s saw an
increased attention in their use due to the increased possibilities of numerical analysis and
- 200 -

Appendices
improvements in instrumentation for recording seismic events (Foti et al. 2011). The last
two to three decades has seen the concept of using seismic waves in engineering
applications receive much more attention, initially due to the introduction of the SASW
technique (Nazarian & Stokoe 1984) and then by using multiple stations (e.g. Park et al.
1999; Foti 2000). These researchers and many others have studied the possibility of
exploiting information gathered from seismic waves to characterise the medium, in which
they travel. These researchers generally rely on a noise source, travelling through a medium
and measured at a receiver to create a theoretical model, with it representing wave
propagation through the ground. Newtons second law of motion (Equation (A.1)) is
commonly used as the basis for developing these theoretical models to estimate the
material layering properties of the medium.

(A.1)

where, F = applied force (N), m = mass (kg), and a = acceleration (m/s2). Theoretical
models developed to study wave propagation through the ground to characterise the
ground profile involve making important assumptions on the composition of seismic
waves. There are several different types of seismic waves, however, the two main types
considered are body and surface waves. Body and surface waves travel through the inner
layers and along the surface of the Earth, respectively. Figure A.7 shows an example of a
measurement and the arrival times of seismic waves. This figure shows a combination of
both body (P-wave and S-wave) and surface waves. Features of this example will be used to
discuss different types of waves modelled to obtain a theoretical model of wave
propagation.

Figure A.7: Example measurement and arrival of seismic waves (in time domain).
- 201 -

Appendices

Body Waves
The first type of seismic wave is the body wave, which travels through the Earths inner
layers. When compared to surface waves, body waves travel faster and oscillate at a higher
frequency. Hence, these waves arrive before surface waves when excited (from e.g.
earthquakes, active sources, etc) and their energy decays faster than surface waves (e.g.
McKenna et al. 2008). There are two types of body waves which are named with respect to
their arrival time (once excited), they are primary and secondary waves.

P Wave
Primary waves (or P-waves) are the first type of body wave, since they are the first to arrive
once the ground is excited. These waves are considered to be longitudinal or compression
waves because particles travel parallel to the waves direction (direction of wave
propagation). P-waves propagate at a faster velocity than secondary (S-waves) waves and
can travel through both solids and fluids. Figure A.8 shows an example of a P-wave
travelling through a medium. This figure illustrates the push pull action this wave type
follows and also shows that displacement of the medium is in only one direction, that is, the
direction parallel to the wave propagation.

Figure A.8: Example of primary wave (Lamit Co. 2012).


As shown in Equation (A.2), the primary wave velocity is a geotechnical property that is
theoretically related to the elastic P-wave modulus of a soil, and in turn the bulk and shear
moduli. M is the P-wave modulus and K is the bulk modulus (resistance to compressibility).

- 202 -

Appendices
4
+ 3
= =

(A.2)

S Wave
Secondary waves (or S-waves) are the second type of body wave, since they are the second
to arrive once the ground is excited. These waves are considered to be transverse or shear
waves which lead particles to travel perpendicular to the waves direction (direction of wave
propagation). The particle motion of S-waves is typically divided into two components, they
are SH-waves and SV-waves. SH-waves are attributed to the horizontal particle motion in
the direction perpendicular to the plane of wave propagation and SV-waves are attributed
to the particle motion within a vertical plane through the plane of wave propagation. Swaves propagate at a slower velocity than P-waves and can travel through solids but not
fluids. Figure A.9 shows an example of an S-wave travelling through a medium. This figure
illustrates the up down or side-to-side action this wave type follows.

Figure A.9: Example of secondary wave (Lamit Co. 2012).


The ratio of Vp and Vs () can be estimated using Poissons ratio (), as shown in Equation
(A.3).

1
= =
1

(A.3)

- 203 -

Appendices

Surface Waves
The second type of seismic wave is the surface wave, which travels in two dimensions along
the surface of the Earth due to the energy concentration at this area. They are composed of
P- and S-waves in a linear combination. Surface waves are easily seen on seismograms
because they travel slower than body waves and oscillate at a lower frequency. Hence, these
waves arrive after body waves when excited (from e.g. earthquakes, active sources, etc) and
their energy decays slower than body waves. There are two common types of surface waves
which are named after the researchers responsible for their finding, they are Rayleigh and
Love waves. Surface waves are typically responsible for the damage caused by earthquakes,
especially if the quake is located closer to the surface of the Earth. The energy carried by
surface waves decay at a rate of 1/r (where r is the distance from the source) while body
waves decay at 1/r2. Therefore, surface waves dominate the energy measured at a near
surface receiver located at a significant distance from the source (McKenna et al. 2008).

Rayleigh Wave
The Rayleigh wave is named after Lord J.W.S. Rayleigh, who in 1885 mathematically
predicted its existence. Rayleigh waves are composed of both P- and S -waves polarised in
the vertical direction. These waves produce a wave that travels in a rolling motion (elliptical
particle motion), hence the reason why they are also referred to ground roll. As shown in
Figure A.10, this rolling action forces the ground to move up and down and side-to-side in
the direction of wave propagation.

Figure A.10: Example of Rayleigh wave (Lamit Co. 2012).


Rayleigh wave velocity (VR) is generally approximated being 90% of Vs (Equation (A.4)).
- 204 -

Appendices

0.9

(A.4)

Love Wave
The Love wave is named after A.E.H. Love, a British mathematician who in 1911
mathematically predicted its existence. Love waves are composed of S-waves polarised in
the horizontal direction or parallel to the Earths surface, thus carrying energy only in this
direction. Therefore, when generating the theoretical HVSR model, Love waves can only
influence the horizontal components. They are the fastest type of surface wave and as
shown in Figure A.11, they move the ground side-to-side, producing entirely horizontal
motion.

Figure A.11: Example of Love wave (Lamit Co. 2012).

- 205 -

Appendices

B. MATLAB Programming
B.1 Measured HVSR Curve
Although the commercial software Grilla may be used to process the measured ambient
vibration data, an in-house program was written in the MATLAB language to separately
analyse the measured HVSR data. This gave flexibility for the data to be processed in
independent and different ways from Grilla, ultimately yielding the HVSR curve. The
developed program can be found below. The program has 4 parts:

B.1.1 Recording Information and User Interface


The first section is the user interface where the required parameter settings are inputted
into the program. The parameters that need to be set by the user include: acquisition
(recording) length, sampling rate, window size.

B.1.2 Raw Measured Microtremor Data


The second step of the program loads the raw measured data from the TrominoTM
velocimeter to extract the microtremors recorded by each of the three components. The
program then produces time history plots of the recorded microtremors (Figure B.1). This
allows the user to determine whether any anthropic or unusual activities were recorded,
which might influence the outcome of the measured HVSR curve.

Figure B.1: Three component recordings of the recorded microtremors over time.
- 206 -

Appendices

B.1.3 Windows, Fast Fourier Transforms (FFT), Smoothing and Single HVSR
The third step involves the majority of the analysis and calculation work. Here the raw
measured data for each component is cut into user defined window lengths, and a
windowing function is applied to smooth the window terminations (window ends). The
user can specify a number of windowing functions (available in the program) in this
section. A fast Fourier transform (FFT) is invoked for each window of each of the three
components, and then smoothed using a moving average filter. The horizontal
components (North-South and East-West) of each window are combined using a
geometric average (Equation (3.3)). The averaged horizontal component is then divided by
the vertical component to create a single HVSR curve for each window. Different types of
averaging for the horizontal components can be called upon in the program including:
geometric average, root mean square and vector sum.

B.1.4 Single Component Spectra and Averaged HVSR


The fourth step involves averaging the spectra of each component to produce the single
component spectra (Figure B.2). Also, the single window HVSR curves from step 3 are
averaged to produce the average measured HVSR curve (Figure B.3).

Figure B.2: Single component spectra.

- 207 -

Appendices

Figure B.3: Measured HVSR curve.

Measured HVSR Curve Sample Input File

- 208 -

Appendices

Measured HVSR Curve MATLAB Source Code


%% HVmeasured
% 10th February 2011
% University of Western Sydney
% The purpose of this program is to produce the measured HVSR curve
%--------------------------------------------------------------------

close all;
clear all; clc

B.1.1
%% Recording Information and User Interface
Aq_Period = 16; % in minutes
Rec_Length = Aq_Period*60; % minutes in seconds
Samp_Rate = 512; % Sampling Rate
Samples = Rec_Length*Samp_Rate; % minutes in seconds times sampling
rate
Win_Size = 20; % 20 second windows
Num_Win = Samples/(Win_Size*Samp_Rate); % Entire recording divided by
single window times sampling rate

B.1.2
%% Load raw data exported from Grilla
raw_noise = importdata(uigetfile('*.dat'), ' ', 25);

% Create new variables in the base workspace from those fields.


vars = fieldnames(raw_noise);
for i = 1:length(vars)
assignin('base', vars{i} raw_noise.(vars{i}));
end

clear ('vars','textdata','raw_noise','i','colheaders')

NSraw = data(:,1);
EWraw = data(:,2);
Zraw = data(:,3);

clear ('data')

- 209 -

Appendices
% raw signal offset
[NScomp] = NSraw-mean(NSraw);
[EWcomp] = EWraw-mean(EWraw);
[Zcomp] = Zraw-mean(Zraw);

% Plot the offseted raw noise captured by the velocimeter


inctime = 1/(Samp_Rate*60);
interval = 0:inctime:(Aq_Period-inctime);

set(0,'Units','normalized');
get(0,'ScreenSize');
figure('OuterPosition',[850 40 515 340])
figure(1);
subplot(3,1,1), plot(interval,NScomp(:),'b');
xlabel('Time (minutes)');
ylabel('NS');
set(gca,'YTick', []);
grid on;
subplot(3,1,2), plot(interval,EWcomp(:),'r');
xlabel('Time (minutes)');
ylabel('EW');
set(gca,'YTick', []);
grid on;
subplot(3,1,3), plot(interval,Zcomp(:),'g');
xlabel('Time (minutes)');
ylabel('Z');
set(gca,'YTick', []);
grid on;

clear ('inctime','interval')

B.1.3
%% Creating Windows from entire trace file
for Window = 1:Num_Win
for i = 1:3
if (i== 1);
Component = NScomp;
elseif (i== 2);
Component = EWcomp;
elseif (i== 3);

- 210 -

Appendices
Component = Zcomp;
end;

Win_Start = 1+((Window-1)*Win_Size*Samp_Rate);
Win_End = (Win_Start-1)+(Win_Size*Samp_Rate);
Win_Data = Component(Win_Start:1:Win_End);

% Applying Window function


% Win_Func = tukeywin((Win_Size*Samp_Rate),0.10); % Tukey
Window Function with cosine tapering
Win_Func = triang((Win_Size*Samp_Rate)); % Triangular Window
Function
% Win_Func = hann((Win_Size*Samp_Rate)); % Hanning Window
Function
% Win_Func = hamming((Win_Size*Samp_Rate)); % Hamming Window
Function
% Win_Func = rectwin((Win_Size*Samp_Rate)); % Rectangular
Window Function
Win_Data = Win_Data.*Win_Func;

% Fast Fourier Transform


% Paramters for FFT from section above - Recording Information
Comp_FFT = fft(Win_Data,(Win_Size*Samp_Rate));
Comp_FFT = abs(Comp_FFT);

% Smoothing of Spectra and Storing of Data - using a moving


average filter with the number referring to the span for the
moving average
if (i==1);
NSfft(:,Window) = Comp_FFT;
NSfft(:,Window) = smooth(NSfft(:,Window),20);
elseif (i==2);
EWfft(:,Window) = Comp_FFT;
EWfft(:,Window) = smooth(EWfft(:,Window),20);
elseif (i==3);
Zfft(:,Window) = Comp_FFT;
Zfft(:,Window) = smooth(Zfft(:,Window),20);

% Calculate HVSR for single time window - Geometric


Average for Horizontal Components

- 211 -

Appendices
Havg(:,Window) = (NSfft(:,Window).*EWfft(:,Window)).^0.5;
Win_HVSR (:,Window) = Havg(:,Window)./Zfft(:,Window);
% Calculate HVSR for single time window - Root Mean Square
Average for Horizontal Components
% Havg(:,Window) =
(((NSfft(:,Window).^2)+(EWfft(:,Window).^2))
/2).^0.5;
% Win_HVSR (:,Window) = Havg(:,Window)./Zfft(:,Window);
% Calculate HVSR for single time window - Vector Sum for
Horizontal Components
% Havg(:,Window) =
((NSfft(:,Window).^2)+(EWfft(:,Window).^2)) .^0.5;
% Win_HVSR (:,Window) = Havg(:,Window)./Zfft(:,Window);
end;
end;
end;

clear
('Win_Data','Win_End','Win_Start','Window','Component','i','Comp_FFT',
'Havg','Win_Func')

B.1.4
%% Average HVSR over all windows
AVG_HVSR = sum(Win_HVSR,2)/Num_Win;

% Standard Deviation of HVSR over all windows


Std_Dev = std(Win_HVSR');
Std_Dev_Plus = AVG_HVSR'+Std_Dev;
Std_Dev_Minus = AVG_HVSR'-Std_Dev;

% Averages of Single Component Spectra


AVG_NSfft = sum(NSfft,2)/Num_Win;
AVG_EWfft = sum(EWfft,2)/Num_Win;
AVG_Zfft = sum(Zfft,2)/Num_Win;

% Frequency Axis
Freq_Axis = 0:(1/Win_Size):(Samp_Rate/2);

% Plot Single Component Spectra


set(0,'Units','normalized');

- 212 -

Appendices
get(0,'ScreenSize');
figure('OuterPosition',[0 40 850 340])
figure(2);
plot(Freq_Axis(1:((Win_Size*Samp_Rate)/2)),AVG_NSfft(1:((Win_Size*Samp
_Rate)/2)));
loglog(Freq_Axis(1:((Win_Size*Samp_Rate)/2)),AVG_NSfft(1:((Win_Size*Sa
mp_Rate)/2)),'b','LineWidth',1.5); hold all;
plot(Freq_Axis(1:((Win_Size*Samp_Rate)/2)),AVG_EWfft(1:((Win_Size*Samp
_Rate)/2)));
loglog(Freq_Axis(1:((Win_Size*Samp_Rate)/2)),AVG_EWfft(1:((Win_Size*Sa
mp_Rate)/2)),'r','LineWidth',1.5);
plot(Freq_Axis(1:((Win_Size*Samp_Rate)/2)),AVG_Zfft(1:((Win_Size*Samp_
Rate)/2)));
loglog(Freq_Axis(1:((Win_Size*Samp_Rate)/2)),AVG_Zfft(1:((Win_Size*Sam
p_Rate)/2)),'g','LineWidth',1.5);
grid on;
xlabel('Frequency (Hz)');
ylabel('Amplitude');
xlim([1 50]);

% Plot Averaged HVSR Curve


set(0,'Units','normalized');
get(0,'ScreenSize');
figure('OuterPosition',[0 380 850 390])
figure(3);
plot(Freq_Axis(1:((Win_Size*Samp_Rate)/2)),AVG_HVSR(1:((Win_Size*Samp_
Rate)/2)));
semilogx(Freq_Axis(1:((Win_Size*Samp_Rate)/2)),AVG_HVSR(1:((Win_Size*S
amp_Rate)/2)),'r','LineWidth',3.0); hold all;
plot(Freq_Axis(1:((Win_Size*Samp_Rate)/2)),Std_Dev_Plus(1:((Win_Size*S
amp_Rate)/2)));
semilogx(Freq_Axis(1:((Win_Size*Samp_Rate)/2)),Std_Dev_Plus(1:((Win_Si
ze*Samp_Rate)/2)),'k','LineWidth',1.0);
plot(Freq_Axis(1:((Win_Size*Samp_Rate)/2)),Std_Dev_Minus(1:((Win_Size*
Samp_Rate)/2)));
semilogx(Freq_Axis(1:((Win_Size*Samp_Rate)/2)),Std_Dev_Minus(1:((Win_S
ize*Samp_Rate)/2)),'k','LineWidth',1.0);
grid on;
xlabel('Frequency (Hz)');
ylabel('H/V');
xlim([1 50]);

- 213 -

Appendices
HVmaxpluspointone =
max(Std_Dev_Plus((Win_Size+1):((Win_Size*Samp_Rate)/2)))+0.1;
HVmaxpluspointone = ceil(HVmaxpluspointone);
if (HVmaxpluspointone < 8);
HVmaxpluspointone = 8;
end;
ylim([0 HVmaxpluspointone]);

clear('ans','HVmaxpluspointone')

Measured HVSR Curve Output


The MATLAB program outputs the data into its internal memory. The following outputs
are determined:
1. Time history plots of the recorded microtremors Raw noise versus time.
2. Single component spectra Component (North-South, East-West and Vertical)
spectra versus frequency.
3. Measured HVSR curve HVSR amplitude versus frequency. The standard
deviations of the averaged HVSR amplitudes are also shown on the frequency axis.

B.2 Removal of Anthropic Noise from Measured HVSR Curves


As mentioned in Section 10.3, two filtering methods have been trialled in this thesis in an
attempt to remove and/or minimise the influence of anthropic noise, they are the: (1)
band-stop filter, and (2) 3-spike filter.
The band-stop filter works by attenuating frequencies in a specific range to very low levels,
whilst leaving the remaining frequencies untouched. An example from the MATLAB
software of a band-stop filter is shown in Figure B.4. Basically, it is a notch filter but
applied over a wider frequency range. Here, it is necessary to know the frequency range
affected by anthropic noise beforehand, since in this case the filter is applied in the time
domain (i.e. before producing the HVSR curve). An example measured HVSR curve from
station B4 (Figure B.5) of the DCPT-9 area (Chapter 7) is used to demonstrate the
application of the band-stop filter. It was shown in Section 7.2 and Figure B.5 that the
anthropic peak was located at approximately 5.8 Hz, however, it is evident that the
- 214 -

Appendices
anthropic peak affected the measured HVSR curve over a wider frequency range.
Therefore, instead of applying a notch filter at only 5.8 Hz, a band-stop filter (Figure B.4)
was applied between 5.1 (Fc1) and 6.3 Hz (Fc2) (and Fs/2 is the Nyquist frequency = 512/2).

Figure B.4: Band-stop filter diagram by MATLAB.

Figure B.5: Measured HVSR curves before and after applying the band-stop filter.
Figure B.5 superimposes the measured HVSR curve from station B4 of the DCPT-9 area,
as well as the band-stop filtered HVSR curve. This figure shows that the band-stop filter
successfully removed the anthropic peak from the measured HVSR curve, without
affecting the rest of the curve. Figure B.6 shows both the measured and filtered single
component spectra of all three components (two horizontal and one vertical). The ellipse
formed by the dashed black line highlights the result of applying the filter, where it forced
the filtered single component spectra to attenuate to the same amplitude on all three
components, thus eliminating it from the filtered HVSR curve.

- 215 -

Appendices

Figure B.6: Measured single component spectra of all three components before and
after applying the band-stop filter.
Figure B.7 simply shows the difference between the measured HVSR curve and its filtered
equivalent. As suggested by the name of the filter, that is, band-stop, the filtered HVSR
curve basically subtracted the anthropic peak from the rest of the measured HVSR curve.

Figure B.7: Data filtered from the raw measured HVSR curve by the band-stop
filter.
In some cases, resonance peaks on the HVSR curve are produced by vibrations from both
anthropic and stratigraphic origins. Therefore, applying a band-stop filter at a specific
frequency range might eliminate the anthropic noise affected resonance peak, which may
also be partially due to ambient vibrations of stratigraphic origin, that portray the
composition of the ground profile. Thus a band-stop filter may not always be the most

- 216 -

Appendices
appropriate option for removing the influence of anthropic noise from measured HVSR
data.
The second anthropic noise filtering method named 3-spike method (for ease of
reference) involves: (1) examining and locating spikes on all three (two horizontal and one
vertical) single component spectra which are typically caused due to anthropic activities
(refer to Section 7.2 and in particular Figure 7.8), and (2) eliminating the spikes on all three
components by simply bypassing these points on the frequency spectra to produce the
measured HVSR curve. Basically, when spikes are observed on all three single component
spectra at a certain frequency, the 3-spike filter method will delete those points causing the
spikes on the spectra, and re-connect the spectra using the points located immediately
before and after the spike. Figure B.8 superimposes the measured HVSR curve from
station B4 of the DCPT-9 area, as well as the filtered HVSR curve. This figure shows that
the 3-spike filter successfully removed the anthropic peak from the measured HVSR curve.

Figure B.8: Measured HVSR curves before and after applying the 3-spike filter.
Figure B.9 shows both the measured and filtered single component spectra of all three
components (two horizontal and one vertical). The ellipse formed by the dashed black line
highlights the result of applying the filter, where it forced the filtered single component
spectra to delete anthropic noise affected points and re-connect the spectra using linear
interpolation of the points located next to the spikes. Figure B.10 simply shows the
difference between the measured HVSR curve and it filtered equivalent. Here the 3-spike
filter not only removed the anthropic peak at approximately 5.8 Hz, but also other
resonance peaks on the measured HVSR curve fully or partially affected by anthropic
- 217 -

Appendices
noise. The 3-spike filter thus provides a better alternative than the band-stop filtering
method, as it can be applied without the need to specify a frequency attenuation range.

Figure B.9: Measured single component spectra of all three components before and
after applying the 3-spike filter.

Figure B.10: Data filtered from the raw measured HVSR curve by the 3-spike filter.

- 218 -

Appendices

C.

Deriving Naviers Equation from First


Principles

The model implemented for this thesis (Lunedei & Albarello 2009) is based on the work of
Arai and Tokimatsu (2000, 2004), who in turn applied the theoretical solutions proposed
by Harkrider (1964) for surface waves in multi-layered elastic media. Wave propagation in
elastic media is governed by Naviers equation, as are ultimately the abovementioned
models. In consequence, the applicability of models is also constrained by the assumptions
and limitations of Naviers equation. Here, Naviers equation is derived from first
principles as this will be instructive to highlight and inform on the assumptions and
limitations of the theory.

n
(, , )
d

(, )

Figure C.1: Material domain subjected to external body and external surface forces
Consider a material domain subjected to external body and external surface forces
(Figure C.1). The elementary material volume d is subjected to the infinitesimal body
force f:
= (, )

(C.1)

where denotes the bulk mass density of d, x is the position vector and t is the time. The
infinitesimal surface force T acting on the infinitesimal surface area da is given as:
= (, , )

(C.2)

where n is the outward normal to da. Through the tetrahedron lemma, the stress vector T
is defined as follows:
- 219 -

Appendices
(, , ) =

(C.3)

where is the Cauchy stress tensor.


Momentum Balance
In a Galilean frame of reference and in the absence of dynamic moment, the instantaneous
balance of linear momentum is written as:

= (, ) + (, , )

where =

is the velocity field of a particle s located at x and

(C.4)
( )

denotes the particle

or material derivative. Using (C.3), and invoking the Divergence theorem, gives:

(, , ) = =

so that,

= {(, ) + }

(C.5)

Assuming infinitesimal transformation such that:


1
where () is the norm and (, ) is the displacement vector of the particle, (s) leads to:
+ (, ) =

where =

(C.6)

is the particle acceleration.

Constitutive Model
It is assumed that under ambient vibrations, the soil material is only subjected to
infinitesimal transformation and the strain tensor can thus be linearised as:
- 220 -

Appendices
1
= ( + )
2

(C.7)

and the material remains linear elastic so that the following constitutive relationship holds:
= + 2

(C.8)

where 1 is the identity tensor and = () = () is the volumetric strain; and (or
G) are the Lam constants.
Governing Equation
Ignoring body forces, using (C.7) and (C.8), and the identity:
( ) = () ()

(C.6) is re-written as:-

+ ( + ) =

(C.9)

So that,

( + 2) () =

where, =

- 221 -

(C.10)

Potrebbero piacerti anche