Innovations in Geotechnical Design: 2021 Australian Geomechanics Society Victorian Symposium

PROCEEDINGS
2021 AUSTRALIAN GEOMECHANICS SOCIETY

VICTORIAN SYMPOSIUM
Innovations in Geotechnical Design
Thursday, 21 October 2021, 8:30am – 5:00pm
An online-only event
VICTORIA CHAPTER
PREFACE
The Victorian chapter of the Australian Geomechanics Society (AGS) is pleased to announce a one-day
symposium titled “Innovations in Geotechnical Design” which is to be held on 21 October 2021. The event
platform changed to an online-only event due to development of new restriction due to COVID-19 pandemic.
This event will bring together geotechnical and other civil engineering professionals to share and discuss
their knowledge and experiences related to geotechnical design and the latest technology and innovative
methods.
Victoria’s construction industry is still up and running, recovering quite well from the hiccups of the pandemic
and several lock-downs. To catch up with the fast pace of global advancement in technology, incorporating
more and more of the emerging and innovative methods in geotechnical design and construction seems to
be the way forward, facing the challenges in the built environment. This symposium will present overviews
of the state-of-the-art practices innovations in design, including new research, case studies, advanced
technology and simulations related to various geo-structure systems, as well as reliability, safety, and
observational design implemented in Victorian projects.
The symposium will bring together professional engineers, researchers, specialist contractors, regulators,
educators and students to share and discuss their experiences on the above topics. This will be a great
networking opportunity post-pandemic.
SYMPOSIUM ORGANISING TECHNICAL REVIEWERS

COMMITTEE* All technical papers in these proceedings,
Alireza Mohammadinia (Co-Chair) excluding the keynote addresses, were peer
reviewed. The reviewers are acknowledged and
Mike Shackleton (Co-Chair) listed below:
Frazer Saunders
Richard Flynn Alireza Mohammadinia (Editor)
Negin Yousefpour Chris Coulson
Yuqi Tan Joel Gniel
Marvin Bedary Daniel King
Stephen Wall Ken Wright
*a sub-committee of the AGS Victoria committee Ramtin Tajeddin
Sanjive Narendranathan
SPONSORS Mahdi Miri Disfani
PLATINUM Jayantha Kodikara
Rob Day
Chadwicks Geotechnics
GOLD
Global Synthetics Pty Ltd
Insitu Geotech Services (IGS)
All right reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form without the permission of the Australian Geomechanics Society.
© 2021 Australian Geomechanics Society.
2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN i

CONTENTS
Preface i SESSION 3 53
Contents ii OPTIMIZATION OF RISK
About the Keynote Speakers iii AND SAFETY IN DESIGN
Schedule iv Using geotechnical innovation to reduce project risk 55
Sponsors v Chris Lyons (Suburban Rail Loop Authority)
SESSION 1 1 Design methodology and input 63

SERVICEABILITY, RELIABILITY parameters applicable to foundation
AND SAFETY IN DESIGN design for large complex towers
Ying Tay (Golder Associates)
Use of probabilistic methods in 3
geotechnical engineering Optimisation of Temporary Support Design 69
Jinsong Huang (The University of Newcastle) for the Northern Portal Cut & Cover Tunnel
Jawad Zeerak (EIC Activities)
Reliability-based geotechnical design 9
in an Australian context
SESSION 4 75
Andrew Lochaden (Golder Associates)
THE ROLE OF DESIGN IN
Application of soil nail wall to roadway 15 INFRASTRUCTURE PROJECTS
widening using GFRP rebars as per Supporting Innovative Design and Construction 77
Australian design guidelines Scott R Taylor (Major Road Projects Victoria)
Jay Lee (GS E&C Australia - formerly AECOM Melbourne)
BIM to Numerical Modelling Interoperability for 83
Risk Prediction Model for Formation of 21 Geotechnical Design of Underground Metro Station
Underground Cavities and Sinkholes Mengqi Huang (Monash University)
due to Defective Sewer Pipes
Samanthi Indiketiya (Chadwick Geotechnics, Formerly Swinburne On best practices for trackbed design 89
University of Technology) Negin Yousefpour (The University of Melbourne)
SESSION 2 27
INSTRUMENTATION AND
MONITORING IN DESIGN
Ground truth, control and design of driven piles: 29
implementing old ways with a new twist
Julian Seidel (Foundation QA)
The Application of 3D Finite Element Method in 41

the Design of Large Piled Foundation System
- Case Study: Melbourne Cement Facility
Kaveh Ranjbar (FSG Geotechnics and Foundations)
Pile Testing Verification – an Alternative Approach 47

Dion Denes (Golder Associates)
ii 2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN

ABOUT THE KEYNOTE
SPEAKERS
JINSONG HUANG CHRIS LYONS
Professor Principal Geotechnical Engineer
The University of Newcastle Suburban Rail Loop Authority
Jinsong Huang is a professor at the Chris has over 24 years experience
Priority Research Centre for Geotechnical as a geotechnical engineer working
Science and Engineering, the University across a wide range of infrastructure
of Newcastle. His research interests and commercial developments
include risk assessment in geotechnical engineering and projects. He has worked on numerous major projects
computational geomechanics. He has published over 100 around the globe and has been involved in all
journal papers on the risk assessment of slope stability stages of project works ranging from investigations,
and landslides, the modelling of spatial variability, stress detailed design for construction and high-level project
integration techniques for elastoplastic models, the contact coordination. Chris has extensive experience covering
dynamics of granular media, the analysis of hydraulic the design of various types of deep and shallow
fracturing and the predictive maintenance of railway foundations, ground improvement, advanced soil
tracks. He received a Regional Contribution Award from modelling, retaining walls and slope stabilisation works
the International Association of Computer Methods and for both static and seismic loadings. Chris’ recent
Advances in Geomechanics at its international conference areas of focus have been in ground movement damage
in Kyoto in 2014 and the GEOSNet Award from the assessments, dynamic rockfall modelling and integration
Geotechnical Safety Network in 2017. He will serve as the of digital technology into geotechnical engineering
conference chair of the 8th International Symposium on workflows. Chris is currently work for Suburban Rail
Geotechnical Safety and Risk to be held at the University of Loop Authority as the projects Principal Geotechnical
Newcastle in December 2022. Engineer.
JULIAN SEIDEL SCOTT TAYLOR

Managing Director Director Engineering
Foundation QA Major Road Projects Victoria
Dr. Julian Seidel has 42 years professional Scott currently leads the Engineering
experience as a geotechnical engineer Team at Major Road Projects Victoria
specialising in deep foundations, with (MRPV), as the Director, Engineering
particular expertise in pile dynamics His role oversees all engineering
and rock socketed piles. Julian graduated dux of Civil and technical aspects of the projects within the MRPV
Engineering at Monash University in 1979, and 15 years portfolio, with the aim of managing and mitigating
later completed his PhD thesis “The analysis and design of technical risks.
pile shafts in weak rock” at Monash University under Chris
Haberfield and Ian Johnston. Scott has worked in both the government and private
sectors with a focus on Bridge Design and Design
Julian has been employed over this time with GHD, Golder Management. Project Highlights include the West
Associates, VicRoads, Pile Dynamics in the USA, Wagstaff Gate Bridge Strengthening Project, leading the Design
Piling and Monash University. In 1999, he commenced team for the M80 Ring Road Upgrade – Sunshine Ave
his own consultancy Foundation QA, which later morphed to Calder Fwy which included the EJ Whitten Bridge
into FSG. In 2020, Julian left FSG to continue his expert widening, and the Ballarat Rail Upgrade Project.
consulting activities under the Foundation QA brand in Scott is currently a member of the Victorian Division
south east Asia on major infrastructure, legal disputes and Committee for Engineers Australia, and a non-Executive
insurance claims. Director of We Ride Australia. Scott also has a strong
research background, having obtained a PhD in 2004
Google Scholar lists 70 of Julian’s journal and conference in the area of structural dynamics and computational
papers, with over 1000 citations. He continues to pursue his methodologies.
personal research interests and to reinvigorate his publishing
activities which have been put aside for many years.
2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN iii

SCHEDULE
8:30 am – Welcome and opening remarks SESSION 3
SESSION 1 OPTIMIZATION OF RISK
SERVICEABILITY, RELIABILITY AND SAFETY IN DESIGN
AND SAFETY IN DESIGN 1:20 PM KEYNOTE ADDRESS
8:40 AM KEYNOTE ADDRESS Using geotechnical innovation to reduce
project risk
Use of probabilistic methods in geotechnical
Chris Lyons (Suburban Rail Loop Authority)
engineering
Jinsong Huang (The University of Newcastle) 2:00 pm Presentation
9:20 am Presentation Design methodology and input parameters
applicable to foundation design for large
Reliability-based geotechnical design in an
complex towers
Australian context
Ying Tay (Golder Associates)
Andrew Lochaden (Golder Associates)
2:15 pm Presentation
9:35 am Presentation
Optimisation of Temporary Support Design
Application of soil nail wall to roadway widening
for the Northern Portal Cut & Cover Tunnel
using GFRP rebars as per Australian design
Jawad Zeerak (EIC Activities)
guidelines
Jay Lee (GS E&C Australia - formerly AECOM 2:30 pm Sponsor presentation
Melbourne) Global Geosynthetics:
9:50 am Presentation Innovative design approach for working
platforms and hard stands
Risk Prediction Model for Formation of
Amir Shahkolahi
Underground Cavities and Sinkholes due to
Defective Sewer Pipes 2:40 pm Questions
Samanthi Indiketiya (Chadwick Geotechnics, Formerly 2:50 pm Afternoon tea
Swinburne University of Technology)
10:05 am Sponsor presentation SESSION 4
Insitu Geotech Services – A Day in the Life of IGS THE ROLE OF DESIGN IN
INFRASTRUCTURE PROJECTS
10:15 am Questions
10:25 am Morning tea 3:15 PM KEYNOTE ADDRESS
Supporting Innovative Design and
SESSION 2 Construction
INSTRUMENTATION AND Scott R Taylor (Major Road Projects Victoria)
MONITORING IN DESIGN 3:55 pm Presentation
10:50 AM KEYNOTE ADDRESS BIM to Numerical Modelling Interoperability
Ground truth, control and design of driven for Geotechnical Design of Underground
piles: implementing old ways with a new twist Metro Station
Julian Seidel (Foundation QA) Mengqi Huang (Monash University)
11:30 am Presentation 4:10 pm Presentation

The Application of 3D Finite Element Method in On best practices for trackbed design
the Design of Large Piled Foundation System - Negin Yousefpour (The University of Melbourne)
Case Study: Melbourne Cement Facility 4:25 pm Questions
Kaveh Ranjbar (FSG Geotechnics and Foundations)
4:35 pm Closing address
11:45 am Presentation 4:45 pm Finish
Pile Testing Verification – an Alternative Approach
Dion Denes (Golder Associates)
12:00 pm Sponsor presentation
Chadwick Geotechnics:
Innovation – Little things = Big WINS
12:10 pm Questions
12:20 pm Lunch
iv 2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN
Go to table of contents
SPONSORS
PLATINUM SPONSOR
GOLD SPONSORS
TECH SERV
EO I
CE
G
INSITU
S
re d u
cin
ge in
ty
g
ote ta
c h nic al u n cer
2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN v
General Services
Chadwick Geotechnics is a leading supplier of services to the
Geotechnical, Civil and Environmental disciplines across Australia, its
Territories, and throughout the Asia Pacific region.
Key capabilities include:

Field Engineering Drilling
• Construction Engineering • Solid/Percussion/Direct Push
• Onsite & Offsite Logging • Geotechnical and Environmental Systems
• Factual Investigations & Reporting • NDD (Non Destructive Digging)
• Bearing Assessments/Plate Load Testing
Instrumentation
• Earthworks Supervision & Monitoring
• Optimisation
Laboratory • Supply
• NATA Accredited • Installation
• Construction Materials Testing • Monitoring
• Triaxial and Consolidation Testing • Telemetry
• Thermal Resistivity Testing
• Rock Testing
Engineering • Laboratory • Drilling • Instrumentation

Chadwick Geotechnics Pty Ltd
Melbourne www.chadwickgeotechnics.com.au info@chadwickgeotechnics.com.au
Innovative
Geosynthetic
Solutions for
Geotechnical
Problems
Global Synthetics is a 100% Australian-owned company,

proud to offer a complete range of high-quality geosynthethic
products backed by over 200 years of combined staff
experience in the industry.
We have supplied products to some of the largest recent
infrastructure works in Australia. Global Synthetics provides
major benefits to any geotechnical engineering project with
the right products and our technical expertise.
Stay on top of the latest Global Synthetics information and

visit our website at globalsynthetics.com.au
SESSION 1
SERVICEABILITY,
RELIABILITY AND
SAFETY IN DESIGN
Keynote Address
Use of probabilistic methods in geotechnical engineering

Jinsong Huang M.AGS M. ASCE
Discipline of Civil, Surveying and Environmental Engineering, College of Engineering, Science and Environment, The University of
Newcastle, Callaghan NSW 2308, Australia; PH: +61 2 4921 5118; email: Jinsong.huang@newcastle.edu.au
ABSTRACT
Due to the intrinsic inhomogeneous nature of soils and rocks, the minimal site investigations, and the need to extrapolate
available information over a large domain, geotechnical designs have inevitable uncertainties. To be conservative,
geotechnical engineers traditionally use a safety factor to account for uncertainties. A more rigorous way of considering
uncertainties is to use probabilistic methods. To promote the use of probabilistic methods in geotechnical engineering, this
paper tries to address the following commonly encountered questions: 1) Why do we need to use probabilistic methods?
2) How can we use probabilistic methods if we don’t have enough test data? 3) How much field/test data do we need? 4)
How can we use multiple sources of information? 5) How can we use monitoring data to predict future performance?
Keywords: probabilistic methods, geotechnical engineering, limited data, Bayesian methods, prediction
1 INTRODUCTION rocks suggests that geotechnical systems are highly

amenable to a statistical interpretation. This is quite a
In recent years, there has been a remarkable increase in different philosophy from the traditional approach
activity and interest in the use of probabilistic mentioned previously. In the probabilistic approach, we
methodologies applied to more traditional areas of input soil properties characterised in terms of their means
geotechnical engineering. This growth has manifested and variances leading to estimates of the probability of
itself in many forms and spans both academe and failure or reliability of a design.
practice within the geotechnical engineering community,
for example, more dedicated conferences, short courses
for practitioners, and new journals and books. However, 2.2 Probability of failure is more meaningful than
there are still some reluctances. This paper tries to factor of safety
promote the use of probabilistic methods in geotechnical
engineering by addressing some commonly encountered It is common to apply the same factor of safety for
questions. different types of application without regard to the degree
of uncertainty or failure consequence involved. In this
case, the factor of safety can be misleading because a
2 WHY DO WE NEED TO USE PROBABILISTIC higher factor of safety doesn’t necessarily mean the
METHODS? structure is safer.
2.1 Limited site investigation data Considering two examples of drained slope stability, they
both have the same geometry as shown in Figure 1. The
Soils and rocks in their natural state are among the most characteristic values of the soil properties in the two
variable of all engineering materials, and geotechnical examples are shown in Eq. (1) and (2) respectively.
engineers must often “make do” with materials that
present themselves at a particular site. In a perfect world
with no economic constraints, we would drill numerous
boreholes and take multiple samples back to the
laboratory for measurement of standard soil properties
such as permeability, compressibility, and shear strength.
Armed with all this in-formation, we could then perform
our design of a seepage problem, foundation, or slope
and be very conﬁdent of our predictions. In reality we must
usually deal with very limited site investigation data, and
the traditional approach for dealing with this uncertainty in
geotechnical design has been through the use of
characteristic values of the soil properties coupled with a
Figure 1. Slope profile
generous factor of safety.
If we were to plot the multitude of data from the

φ=′ 23°
(1)
hypothetical site investigation as a histogram for one of c′
= 0.048
the properties, we would likely see a broad range of γH
values in the form of a bell-shaped curve. The most likely
values of the property would be somewhere in the middle,
but a signiﬁcant number of samples would display higher
and lower values too. This variability inherent in soils and
2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN 3
2
Use of probabilistic methods in geotechnical engineering J. Huang
φ=′ 32° 3 HOW CAN WE USE PROBABILISTIC METHODS

(2) IF WE DON’T HAVE ENOUGH TEST DATA?
c′
= 0.048
γH We actually have two schools of statistics. The first one is
frequentist approach. Frequentist says that we need to
where c′ and φ ′ are drained cohesion and friction angle have enough data to draw a statistical conclusion, for
respectively, and γ is unit weight of soil. example, a probability distribution. This is the dominant
statistical practice during the 20th century.
According to the stability charts in Michalowski (2002), the The other approach is Bayesian methods. Bayesian says
factor of safety for the two slopes can be found as 1.5 and that before we do any test, we have some prior knowledge
2.0 respectively. We may think that the second slope is about the test we are looking at. And data can be used to
safer than the first one because of a higher factor of update our prior knowledge. Bayesian methods started
safety. However, if we use a probabilistic approach, we getting popular in 21st century, although it was published
may find that the mean and standard deviation of factor of 200 years ago. It is now widely used in computer science,
=
safety for the two slopes are µ FS 1.5,
= σ FS 0.18 (slope medical science, social science and so on.
=
1), and µ FS 2.0,
= σ FS 0.5 (slope 2). Assuming the
The Bayesian approach has special significance to
factor of safety is normally distributed, one can easily
engineering design where available information is
calculate the corresponding probabilities of failure of the
invariably limited and subjective judgement is often
two slopes are 0.27% and 2.27% respectively. It is clear
necessary. In the case of parameter estimation, we may
that the second slope is actually less safe than the first
often have some knowledge (perhaps inferred intuitively
one because of a higher probability of failure. The main
from experience) of the possible values, or range of
reason why factor of safety can be misleading is that it
values, of a parameter; moreover, we may also have
uses only one parameter to quantify safety level. In
some intuitive judgment on the values that are more likely
probabilistic approaches, however, two parameters are
to occur than others.
used to quantify not only the mean safety level but also
the standard deviation.
In Bayesian statistics, site investigation information,
laboratory test data, in situ test data and load tests results
The factor of safety cannot provide meaningful
are used as evidence to update prior knowledge,
information for risk assessment. For the two examples we
predictions and designs. The posterior distribution of the
just considered, both factors of safety are bigger than one
parameters is calculated based on Bayes’ theorem:
which means both slopes are safe. By looking at the factor
of safety along, it is hard for engineers to choose among
L ( D | θ ) P (θ )
the two slopes. P (θ | D ) = (3)
P ( D)
On the contrary, when probabilities are coupled with
consequences of design failure, we can assess the risk where θ denotes the parameters to be updated, D the
associated with the design. For the examples we just data collected, P (θ ) the prior probability distribution of
considered, if the construction cost and consequence of
design failure are estimated as shown in Table 1, the risk the parameters, L ( D | θ ) the likelihood probability
of the two designs can be estimated as Risk= construction distribution and P ( D ) the probability distribution of the
cost + probability of failure × consequence.
evidence.
It can be seen from Table 1 that the second slope with a
higher probability of failure has lower risk than the first To demonstrate how Bayes’ theorem can be used when
slope. However, if the consequence increases from 10M test data are limited, let’s consider load testing of pile
to 100M, the first slope with a lower probability of failure capacity. Before any test is performed, engineers can look
will have a lower risk than the second slope. This is in the literature or similar projects for the initial
consistent with the common sense that for important assessment of the pile capacity ( y ) . For simplicity, let’s
project, we need to lower the probability of failure to assume the pile capacity is normally distributed with mean
reduce risk.
µ and standard deviation σ . The standard deviation σ
Table 1: Risk assessment of slope failure can be estimated based on the variability of the site (e.g.,
Kay (1978)). The mean pile capacity µ is uncertain, and
Slope 1 Slope 2 is initially estimated using the prior probability distribution
expressed by
Construction
cost (million 1 0.5
dollar) f ′ ( µ ) ~ N ( µ0 , σ 0 ) (4)
Probability of where
0.27% 2.27% N () represents the Normal distribution, and
failure
µ0 and σ 0 are the mean and standard deviation of µ
Consequence (mean of mean pile capacity, standard deviation of mean
10 100 10 100
(million dollar) pile capacity). The parameters µ0 and σ 0 were assumed
to be associated with prediction methods (a static or
Risk
1.027 1.27 0.727 2.77 dynamic method).
(million dollar)
4 2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN
3
Suppose n pile capacity tests have been performed and

the results are denoted as the vector y . The posterior
distribution of µ can be obtained by Bayes’ rule:

 P (y | µ ) f ′(µ )
f ′′ ( µ | y ) =
P (y) (5)
(a)
  nσ 2 y + σ 2 µ0  
2
 µ − 0 2  
  nσ 0 + σ 2  
∝ exp  − 
 σ 2σ 2 
 2  20 2 

  nσ 0 + σ  
 
(b)
.
where y is the average value of y Figure 2. True in-situ distribution of undrained shear
strength
A detailed derivation of Eq. (5) can be found in Huang et
al. (2016). Before we do the excavation or test, we actually don’t
know the undrained shear strength distribution in Figure
The posterior mean and standard deviation of µ are 2(b). We can look in the literature for the statistics of the
undrained shear strength of the soil in this particular site.
nσ 02 y + σ 2 µ0
µ1 = (6) And based on the statistics, we can use random field
nσ 02 + σ 2 theory to guess the possible spatial distributions of the
undrained shear strengths and perform a set of Monte
and Carlo simulations as shown in Figure 3. We generate a
set of realisations of the slope, some of them will fail (e.g.,
Figure 3 a), some will not (e.g., Figure 3 b). If you divide
σ 02σ 2
σ1 = (7) the total number of failures by the total number of
nσ 02 + σ 2 simulations, we can estimate the probability of failure,
which is 17% in this case.
Let us assume the prior mean and standard deviation of
µ are µ0 =1.3 and σ 0 = 0.5 and suppose five load tests For the same slope, if we perform a cone penetration test
have been conducted. The average capacity of tested as indicated by the red box in Figure 4, we then know the
piles is y = 0.8 . According to Eqs. (6) and (7), the undrained shear strength of this column. Now we have
statistics of the undrained shear strengths plus some test
posterior mean and standard deviation of µ are data. Based on the statistics and data, we can perform a
set of conditional Monte Carlo simulations (e.g., Yang et
nσ 02 y + σ 2 µ0 al. 2019). Conditional means that the values of the known
µ1 = undrained shear strengths will not change from simulation
nσ 02 + σ 2 to simulation, but fixed. From a set of conditional Monte
5 × 0.52 × 0.8 + 0.22 ×1.3 Carlo simulations, we can estimate the probability of
= failure, which is 6% in this case.
5 × 0.52 + 0.22
= 0.82 If we perform a second CPT, we will know two columns of
undrained shear strengths as shown in Figure 5. We can
and perform another set of conditional Monte Carlo
simulations, and estimate the probability of failure, which
σ 0σ
σ1 = is 0.7%. If we perform a third CPT, the estimated
nσ 02 + σ 2 probability of failure is essentially close to zero as shown
in Figure 6. If we keep doing test, eventually we will know
0.5 × 0.2 the undrained shear strength distribution everywhere.
=
5 × 0.52 + 0.22 And it turns out that the slope is actually safe. we can see
based on three CPTs, we can reach a very reliable
= 0.09
prediction of the slope stability. Please note that this
The uncertainty of the mean capacity (µ ) has been conclusion only holds for this particular slope. For a
reduced significantly from σ 0 = 0.5 to σ 1 = 0.09 by the different slope, the slope may fail, and we may need
different number of CPTs to find it out. But the same
five load tests. It is interesting to note that even when we
methodology can be used to study how much tests do we
have only one test, we can still use Eqs. (6) and (7).
need to draw a reliable prediction.
4 HOW MUCH FIELD/TEST DATA DO WE NEED?
Suppose we want to do an excavation on a site as shown

in Figure 2(a) and the final slope profile is shown in Figure
2(b). The grayscale in Figure 2(b) represents different
undrained strengths.
4
.
. Conditional Monte Carlo simulations
(a) .
.
. Monte Carlo simulations
.
Figure 5. Conditional Monte Carlo simulations based on

the statistics and two CPT tests
(b)
Figure 3. Unconditional Monte Carlo simulations based

on the statistics in the literature
Figure 6. Probability of failure verse number of CPT tests
. Generally speaking the approaches that we use to assess

. Conditional Monte Carlo simulations the safety of structures can be summarised in Table 2. It
. can be seen from Table 2 that probabilistic approaches
can use test data directly, thus can reduce the uncertainty
in our prediction. Recently Machine learning approaches
become popular in geotechnical engineering. It is noted
however, that Machine learning approaches use test data
only, while probabilistic approaches can be combined
with mechanical models (e.g., finite element model).
Figure 4. Conditional Monte Carlo simulations based on

the statistics and one CPT test
5
Table 2: Risk assessment of slope failure the soil profile based on conditional random field and the
CPT test only. It can be seen from Figure 8 that at the
Approach Input data place where we have CPT test, the mean cone tip
resistance is the same as we tested by CPT and the
standard deviation of cone tip resistance is low. However,
Deterministic Characteristic values at the places where we don’t have any test, the mean
cone tip resistance is similar to and informed only by the
CTP test results, and the standard deviation of cone tip
Mean, standard resistance is large. This means that we have less
Unconditional
deviation, spatial confidence on the profile of cone tip resistance based on
probabilistic analysis
correlation structure one CPT only.
CPT-6A
Mean, standard MASW-1
0 200
Conditional probabilistic deviation, spatial
Depth (m)
5
analysis correlation structure, test
VS (m/s)
150
data 10
100
15 qt (MPa)
0 2.5 5
0 10 20 30 40 50 60 70
Machine learning Test data Distance (m)
Figure 7. Geotechnical CPT test and geophysical multi

channel surface wave test
5 HOW CAN WE USE MULTIPLE SOURCES OF
(a)
INFORMATION? 0
5
Depth (m)
5 4
qt (MPa)
Site characterisation is a fundamental step of collecting 10
3
geotechnical information for proper design, construction 2

15 1
and long term performance of all types of civil and 0
geotechnical structures. Current site investigation 0 10 20 30 40 50 60 70
Distance (m)
practice in geotechnical engineering often involves two (b)
steps: 1) a geophysical site investigation or a desk-top 0 0.12
study of existing geophysical data; and 2) a dedicated in 0.1

Depth (m)
qt (MPa)
0.08
situ geotechnical investigation including several 10
0.06
boreholes and laboratory testing at the intended location 15 0.04
of the future structure. Each type of investigation explores 0 10 20 30 40 50 60 70
a specific volume of the subsoil and has different degrees Distance (m)
of uncertainty. Geophysics provides a wide variety of tools
which can help to identify the subsoil stratigraphy. Data Figure 8. Soil profile based on one CPT test, a) mean
are obtained on a two- (or three-) dimensional section of cone tip resistance, b) standard deviation of cone tip
the ground, often with low resolution. On the other hand, resistance.
in situ tests are performed along a one-dimensional line
with depth and laboratory tests performed at point If we use the CPT and MASW test results together, we
locations within the ground. These tests provide direct can have a higher confidence on the cone tip resistance
measurements of physical and mechanical properties, but profile. Figure 9 shows the soil profile based on Bayesian
cover only a small volume of the subsoil. In current updating and both CPT and MASW test results. It can be
engineering practice, the integration of geophysical data seen from Figure 9 that at the place where we have CPT
with geotechnical data is done manually, often by visual test, the mean cone tip resistance is still the same as we
inspection based largely on engineering judgement and tested by CPT and the standard deviation of cone tip
experience, which does not only introduce additional resistance is low. However, at the places where we don’t
human error, but also causes loss of information. For have CPT test, the mean cone tip resistance is informed
example, the geophysical data is used to find the by the MASW results, and the standard deviation of cone
desirable locations of geotechnical tests, but often tip resistance is reduced.
ignored in deriving geotechnical properties.
A more scientific integration of geophysical data with

geotechnical data can be done based on Bayes’ theorem.
Figure 7 shows a one dimensional cone tip resistance
profile obtained by a CPT test. On top of that, it also
shows a two dimensional shear wave velocity profile
obtained by multi channel surface wave (MASW) test. If
we can use CPT and MASW test results together, it will
help to reduce the uncertainties in our model.
Huang et al. (2018) developed a Bayesian updating

approach, which can combine multiple sources of
information automatically. If we use the CPT test result
only, we can use random field theory to estimate the
mean, standard deviation and spatial correlation length Figure 9. Soil profile based on CPT and MASW test
using the CPT results and then use conditional random results, a) mean cone tip resistance, b) standard
fields to guess the possible soil profiles. Figure 8 shows deviation of cone tip resistance.
6
6 HOW DO WE USE MONITORING DATA TO consolidation periods for the test embankment built at
PREDICT FUTURE PERFORMANCE? Ballina, New South Wales, Australia. The results show
that surface settlement can be well predicted using 116
In design stage, we predict the safety level of the days of observed settlements, while the pore pressure
structures based in lab and field test results. After can be predicted using 292 days of pore pressure
construction, the behaviour of the structure most of the measurements. The predictions are shown to converge to
time is different from our prediction. For important projects the field measurements, regardless of some assumptions
such as large dams and bridges, we usually install some about the measurement errors. It is also demonstrated
sensors or devices to monitor the behaviour of the that incorporating more monitoring data into the Bayesian
structure. In this case, we need to use the monitoring data updating process enables more accurate predictions.
to reassess the safety level of the structure. This is called
back analysis. 7 CONCLUSIONS
Another example where back analysis becomes Deterministic factor of safety approach has been used in
important is embankment built on soft soils. Due to geotechnical engineering for many decades. Although
variations in a soft soil profile and its properties, the probabilistic approaches can provide more information for
actually settlement behaviour is usually different from the risk assessment, there are still quite some resistances to
performance anticipated in design. If the decision on the the use of probabilistic approaches. The most common
settlement behaviour (i.e., quicker or slower than excuse for not using probabilistic approach is lack of data.
expected) can be made sooner, then a smaller amount of This paper has shown that Bayesian approaches can be
surcharge is required and road construction can start used even when we have only one test result. Bayesian
earlier, which leads to significant financial benefits. approaches are also useful for combining different types
of test results and back analysis based on monitoring
Table 3 lists some of the most commonly used back data. It is anticipated that probabilistic approaches will
analysis methods. The simplest back analysis is manual become more and more popular in geotechnical
calibration where engineers try to minimise the different engineering to supplement, not replace the factor of
between observation D and model output g (θ ) . Least safety approach.
square method is probably the most widely used method
8 ACKNOWLEDGEMENTS
for back analysis because it always provides a solution no
matter how complicated the problem is. Least square is
The research supports by the Australian Government
actually a special case of the Maximum Likelihood
through the Australian Research Council's Discovery
method where measurement error is normally distributed
Projects funding scheme (project DP190101592) and
with zero mean. In contrast to the Maximum Likelihood
Linkage funding scheme (project LP200100367) are
method, Maximum A Posterior method not only consider
acknowledged.
the likelihood, but also the prior distribution of the
parameters. Bayesian back analysis is the most generic
REFERENCES
type of back analysis. Instead of solving for the set of
parameters that maximum posterior probability density as Huang, et al. (2016). "Updating reliability of single piles and pile
in the Maximum A Posterior method, Bayesian updating groups by load tests." Computers and Geotechnics 73:
samples the whole posterior probability density function. 221-230.
The most commonly used method for this purpose is the Huang, et al. (2018). "Probabilistic characterization of two-
Markov chain Monte Carlo method. dimensional soil profile by integrating cone penetration test
(CPT) with multi-channel analysis of surface wave
Table 3: Commonly used methods for back analysis (MASW) data." Canadian Geotechnical Journal 55(8):
1168-1181.
Method Formula Kay (1978). "Safety Factor Evaluation for Single Piles in Sand."
Journal of the Geotechnical Engineering Division-Asce
Manual calibration Min D − g (θ ) 104(1): 148-149.
Kelly, et al. (2015). "Bayesian updating for one-dimensional
consolidation measurements." Canadian Geotechnical
Min ( D − g (θ ) )
2 Journal 52(9): 1318-1330.
Least square
Michalowski (2002). "Stability Charts for Uniform Slopes."
Journal of Geotechnical and Geoenvironmental
 D − g (θ ) − µe  Engineering 128(4): 351-355.
Maximum Likelihood Max L  
 σe 
Yang, et al. (2019). "Importance of soil property sampling
location in slope stability assessment." Canadian
 D − g (θ ) − µe  Geotechnical Journal 56(3): 335-346.
Maximum A Posterior Max L   P (θ )
 σe 
Zheng, et al. (2018). "Embankment prediction using testing data
and monitored behaviour: A Bayesian updating approach."
 D − g (θ ) − µe  Computers and Geotechnics 93: 150-162.
Bayesian updating P (θ D ) ∝ L   P (θ )
 σe 
Bayesian updating can use prior knowledge (e.g., lab and

field test results) and monitoring data in a rigorous
framework. It can also consider measurement errors and
provide confidence interval on the predictions. A proof-of-
concept application of Bayesian updating for
embankment settlement prediction can be found in Kelly
et al. (2015). Zheng et al. (2018) used Bayesian updating
with laboratory data, field test data, and monitoring data
to yield accurate predictions during the construction and
Reliability-based geotechnical design in an Australian context
A. L. E. Lochaden1 and S. Buttling2
1
Principal Geotechnical Engineer, Golder Associates Pty Ltd, Melbourne, PO Box 6079, Hawthorn East, VIC 3121; PH (+61) 3 8862 3500;
FAX (+61) 3 8862 3501; email: alochaden@golder.com.au
2
Principal, National Geotechnical Consultants, PO Box 485, The Gap, QLD 4061; PH (+61) 7 3300 0088; email:
stephen.buttling@ngconsult.com.au
ABSTRACT
The design philosophy adopted by the international geo-engineering community over the last few decades has largely
evolved from working stress design (WSD), in which a single or lumped factor of safety (FoS) is adopted, to a load and
resistance factor design (LRFD) approach. In LRFD, partial factors are applied to actions (i.e. loads), soil parameters and
/ or resistances. These partial factors vary in magnitude depending on the relative uncertainty of each parameter to which
they are applied. The approach adopted in Australia has differed, with both WSD and LRFD approaches being widely used.
In the authors’ experience, there is limited awareness in the Australian geo-engineering community of the relationship
between the concept of reliability and the partial factors adopted in LRFD, and therefore of the potential benefits of
undertaking reliability-based design (RBD). The outcome of this is that RBD, in which the uncertainty of the variables which
may affect the design is individually assessed, is rarely undertaken. This paper discusses the concept of RBD and its place
within the framework of Australian Standards and presents practical means of adopting RBD with accompanying examples
from the literature. The intention of the paper is to encourage practitioners to consider uncertainty in geotechnical design
more rigorously, whilst acknowledging the importance of maintaining engineering judgement in design.
Keywords: geotechnical design, working stress, limit state, reliability.
1 INTRODUCTION 2 THE STATE OF GEOTECHNICAL DESIGN IN

AUSTRALIA
At its simplest, the purpose of any geotechnical design is
to ensure that the risk of failure is sufficiently low to be National Standards exist to ensure that the balance
deemed acceptable. Failure may be the catastrophic between cost and safe / serviceable structures is
collapse of a retaining wall for example, but could also appropriately achieved, and that a similar level of
be an unacceptable degree of cracking in a building due reliability is achieved in design undertaken by different
to the excessive movement of the footing system. practitioners. The state of geotechnical design in
Australia, within the framework of national design
Historically, the geo-engineering professional has standards, is somewhat unusual for a number of related
considered this risk by applying a lumped FoS to a reasons.
calculated ultimate resistance, i.e. WSD. This
methodology has largely been replaced internationally Firstly, whereas the UK, previously through BS 8004
by LRFD, in which various factors are applied to actions (BSI, 1986) which encompassed all types of foundations
(including loads), soil parameters and / or resistances, and now through Eurocode 7 (BSI, 2004), and the United
the magnitude of which are typically defined in national States through AASHTO (2017), have complete and
design standards and are a function of the expected overarching design standards, there are a limited
relative uncertainty of the variable under consideration. number of Australian Standards for the design of various
The introduction of LRFD brought with it a lexicon which geotechnical elements, for example AS2159 (Standards
was new (and in some cases, potentially unnecessary), Australia, 2009) for piled foundations, AS2870
and which remains frustrating to many geo-engineering (Standards Australia, 2011) for residential slabs and
professionals. footings, and AS4678 (Standards Australia, 2002) for
(predominantly gravity) retaining walls.
If LRFD is considered to be an evolution of WSD, RBD
may be considered to be an evolution of LRFD. RBD Secondly, relevant Australian Standards have not been
involves the rigorous assessment of the uncertainty of published for commonly encountered design scenarios
the variables on which a design is dependent. RBD has such as slopes and shallow footings including raft
been discussed in the literature for over 40 years, but its foundations for high rise buildings (other than residential
adoption as a design methodology has been slow. slabs and footings, as described above), although
However, it is now included in some international design AS5100.3 (Standards Australia, 2017a) is sometimes
standards, including the Canadian Highway Bridge adopted for shallow footing design for infrastructure
Design Code (Canadian Standards Association, 2014). projects.
This paper sets out the current state of geotechnical Thirdly, whilst most international design standards have
design in Australia, describes the relationship between moved from WSD to LRFD, Australia has effectively
LRFD and RBD, and presents ways in which RBD can maintained both approaches. For example, pile
be adopted in practice. foundations are typically designed using a limit state
approach to AS2159, whereas slopes are typically
designed using a lumped FoS approach.
8
Reliability-based geotechnical design in an Australian context A. L. E. Lochaden and S. Buttling
To further complicate matters, the analysis tools One of the benefits of LRFD over WSD is that the
commonly adopted in geotechnical design have changed engineering professional is required to apply partial
over the last number of years such that parts of the factors to individual components of load and resistance,
national design standards have lost some of their thereby allowing the higher risk of variation in (say) live
relevance. For example, embedded retaining walls are loads compared to (say) the self-weight of concrete to be
now commonly designed using tools such as WALLAP considered in the design. This is not possible in WSD, as
and PLAXIS 2D. However, both AS4678 (Standards a single lumped FoS is applied.
Australia, 2002) and AS5100.3 (Standards Australia,
2017a) provide insufficient guidance on how such soil- LRFD has been subject to criticism by Australian
structure interaction analyses should be undertaken. practitioners, for example Day (2001), Pells (2011), Day
Although beyond the scope of this paper, the reader is et al. (2007), and Wong et al. (2007). It is acknowledged
directed to Haberfield (2017), for example. that many of the arguments made in these references
are valid, but note that a discussion of these arguments
3 THE RELATIONSHIP BETWEEN RBD AND lies outside the scope of this paper. It is noted that
LRFD despite the criticisms set out in these references, Wong
et al. (2007) state that “…despite these difficulties, there
3.1 Introduction is definitely a place for limit state design in geotechnical
engineering” and “the use of limit state design…has
As stated in Section 2, design must balance cost with the benefits as long as the underlying design principles and
safety / serviceability of a structure. Regardless of the soil-structure interaction effects are properly understood
various design methodologies which may be adopted, or and communicated”.
whether the design is a structural or a geotechnical one,
an appropriate design must establish that the design 3.3 Reliability based design
resistance is not less than the design actions (applied
loads). Reliability theory was introduced initially as a design
concept for structural engineering, and subsequently for
The degree of safety of a structure has traditionally been geotechnical engineering, primarily to counter this lack of
assessed in a deterministic manner by engineering robust consideration of the potential variability. The use
professionals in terms of a calculated FoS. The of reliability theory in the field of geotechnical
engineering professional undertakes this assessment engineering is not new – it was the subject of the 17th
using either what he / she considers to be appropriate Terzaghi Lecture in 1981 (Whitman, 1984).
values for applied loads and resistances to take account
of their potential variability, or what the relevant design The requirement for this more robust consideration is
standard mandates. This is illustrated in Figure 1, in discussed below with reference to Figure 2. This has
which the margin of safety is defined as the difference previously been presented in the literature (including
between the resistance and the loads. However, this Becker, 1996a). Loads and resistances are variables
methodology neglects to robustly consider the potential and are not defined by a single value as considered in
variability of both the applied loads and the resistances. the deterministic approach described in Section 3.1.
Sources of geotechnical uncertainty include spatial
variation, measurement error, transformation error (for
Probability of occurrence
1.0
example, the calculation of undrained shear strength of
a cohesive material from Cone Penetration Test data),
and model calculation uncertainty (i.e. the degree to
which the analytical model represents the true behaviour
Margin of safety = R - S
of the system).
By plotting both the load and resistance frequency

distributions on a single plot, the relatively low probability
0.0 of a combination of high loads and low resistance which
Loads (S) Resistance (R)
would result in failure is represented by the intersection
Magnitude of resistance and loads
of the load and resistance frequency distributions.
Figure 1. Graphical illustration of margin of safety for In Figure 2:
WSD (after Becker, 1996)
• Cases (a) and (b) are two curves representing the
3.2 Load and resistance factor design frequency distributions of resistance which have the
same design value, but differing standard
LRFD requires the engineering professional to deviations, noting that the standard deviation for
demonstrate that serviceability and ultimate limit states Case (a) is less than that for Case (b).
(SLS and ULS, respectively) are not exceeded. In simple • Case (a) represents a condition whereby the
terms, the purpose of the SLS check is to assess if resistance is well defined (i.e. the frequency
movement beyond an allowable limit occurs, whilst the distribution curve is relatively narrow). This could be
purpose of the ULS is to ensure that collapse will not the case where a significant amount of geotechnical
occur. investigation has been undertaken at a site and
where there is relatively small variability in the
As set out in Becker (1996b), the partial factors which geotechnical data. The area of intersection (in
are applied in LRFD may be based on calibration using green) of the load and resistance curves is relatively
one or a combination of: small and therefore the probability of failure is
• Previous experience. relatively low.
• WSD approach. • Case (b) represents a condition whereby the
• Reliability theory (as discussed in Section 3.2). resistance is not well defined (i.e. the frequency
9
distribution curve is relatively broad). This could be • The ellipses are centred about characteristic values
the case where little geotechnical investigation has of φ’ and δ.
been undertaken at a site and where there is • The “critical combination” is the most probable
relatively high variability in the geotechnical data. combination of φ’ and δ which would result in failure,
The area of intersection of the load and resistance and is the shortest distance on the plot from the
curves (the combined green and red areas) is characteristic values of φ’ and δ to the failure
relatively large and therefore so is the probability of surface.
failure. • β is defined as the quotient of the shortest distance
• The higher probability of failure of Case (b) from the centre of the ellipse to the failure surface
compared to Case (a) is not captured by the (R in Figure 3) and one standard deviation (r in
traditional FoS approach, as the mean margin of Figure 3).
safety (and therefore FoS) for both Cases (a) and
(b) is the same. 45 One standard-deviation dispersion ellipse
FoS = 1.75
Distribution of load
Effective friction angle of retained material, φ' (°)

40 = 1.50
(a) Distribution of resistance (high confidence)
(b) Distribution of resistance (low confidence)
= 1.25 β-ellipse
Probability of occurrence
Mean margin of safety = R - S

35
= 1.00 r
30
(i.e. limit
state surface) R
25 Safety
20
Loads (S) Resistance (R) 15 Failure

Magnitude of resistance and loads
10
Figure 2. Variation of loads and resistance (after
Becker, 1996a and Green, 1989) 5 Characteristic values
Critical combination
The intention of RBD is to provide adequate confidence 0
that the probability of failure (Pf) is sufficiently low. RBD 0 5 10 15 20 25 30
allows a more consistent assessment of reliability to be Soil-wall interface friction angle, δ (°)
calculated compared to either WSD or LRFD. Although
the theory of reliability analysis lie beyond the scope of Figure 3. Definition of β (after Low, 2005)
this paper, reliability analysis requires assumptions to be
made in relation to the frequency distributions of the More than two variables can be considered in reliability
loads and resistances (for example the normal analyses, the first step being the replacement of the
distributions shown in Figure 2), and therefore the Pf ellipses by ellipsoids in three dimensions. However, to
calculated should not be considered to be the “true” allow graphical representation in two-dimensional space,
probability of failure, but a reasonable estimate. the example above has been limited to two variables
Furthermore, as with any design, reliability analysis only.
requires the engineering professional to use appropriate
means of calculation, interpretation of geotechnical data, If for some design scenario the calculated β value was
etc.. too low, the engineering professional could increase the
strength and size of structural members while leaving the
As an alternative to Pf, the likelihood of failure may be uncertainty changed. This would result in an increase in
expressed in terms of the reliability index (β), which is a the distance from the characteristic points to the failure
description of the safety of the structure normalised by surface (R) and leave r unchanged, thereby increasing
its uncertainty. A higher β value indicates a more robust the calculated β value. An alternative solution would be
design. The target β for a specific structure should be to leave the strength and size of structural members
based on a number of considerations, but primarily the unchanged and decrease the uncertainty, for example by
consequences of failure (for example, risk to life and undertaking more and / or higher quality geotechnical
economic consequences). The target β for Australian investigations. This would cause a reduction in the radius
structures for various considerations is set out in AS5104 of the one standard-deviation dispersion ellipse (r), also
(Standards Australia, 2017b), and which includes an resulting in an increase in the calculated β value.
appendix focussed on geotechnical engineering.
The partial factors presented in national design
The concept of β is explained in Low (2005) with standards and which are required to be adopted in LRDF
reference to Figure 3 and the First Order Reliability may have been (but are not always) calculated based on
Method (FORM, also known as the Hasofer-Lind reliability analysis, as discussed in Section 3.2. The
method, Hasofer and Lind, 1974), demonstrated using benefit of the presentation of partial factors within
the example of the rotational failure of a gravity wall, as national design standards is that the engineering
follows: professional is not required to undertake reliability
• The ellipses in Figure 3 represent combinations of analyses for a design, and can instead use the provided
the effective angle of friction (φ’) of the retained partial factors to achieve the required level of reliability.
material and of the soil-wall interface friction angle However, in deriving partial factors to be applied for all
(δ) which have the same probability of occurring potential scenarios to achieve adequate reliability,
concurrently. reasonable ranges of uncertainty must be considered.
This means that in some design situations, the partial
factors presented in national design standards will result
10
in a β for a structure which is greater than that required, quotient of the standard deviation and mean of the
i.e. the resulting design will be overly conservative. A sample, and is a description of its relative dispersion.
site-specific RBD therefore presents the engineering
professional with the potential opportunity to adopt lower Baecher and Christian (2003) present a summary of
partial factors, if: various numerical methodologies which can be adopted
• The engineering professional has the expertise to for the practical adoption of RBD, some of which are
undertake such an analysis, summarised below.
• Sufficient geotechnical investigation data are • First Order Second Moment (FOSM) method.
available to undertake a valid statistical assessment • Point Estimate method.
of the data, and • FORM.
• The relevant national design standard allows it. • Monte Carlo simulation.
4 UNDERTAKING RBD IN PRACTICE Many geo-engineering professionals will reasonably

endeavour to adopt reliability based methodologies
4.1 Introduction within the software with which they are familiar and which
is commonly adopted in the industry. The SLOPE/W
The purpose of Section 4 is to demonstrate the way in software, for example, allows the user to undertake
which RBD may be adopted in practice. Practical Monte Carlo simulations. Whilst the application
examples of RBD for various commonly encountered PROBANA has been used in the past to undertake
geotechnical design scenarios are summarised. FORM and Monte Carlo analysis coupled with PLAXIS
2D, the authors understand that this application is neither
Whilst there are many examples in the literature of commercially available nor supported by more recent
reliability assessments which have been undertaken versions of PLAXIS 2D (Bentley Systems, personal
using both non-commercially available finite element communication, 10 August 2021).
software (for example, Fenton and Griffiths, 2003), and
bespoke software coupled with commercially available 4.3 Design examples – shallow footing
software (for example, Schweckendiek et al., 2007), it is
the authors’ opinion that these are unlikely to be practical Cherubini (1990) presents a closed-form solution for the
for the majority of consulting geo-engineering probabilistic evaluation of the bearing capacity of a
professionals considering the current level of maturity of shallow foundation for a cohesionless soil with φ’ ≤ 35°.
RBD in both Victoria and Australia more widely. The φ’ only is considered as an independent variable. The
focus has therefore been on more routine examples. frequency distribution of the ultimate bearing capacity,
calculated using numerical integration, is presented for
4.2 Design approach and methodologies various standard deviations of φ’.
Fenton et al. (2016) present a general methodology for Low and Tang (1997) present a more sophisticated
the adoption of RBD which is common to many reliability analysis of a shallow footing. As the
geotechnical design scenarios. This methodology methodology is similar to that set out in Low (2005) and
includes the following (albeit high-level) steps: discussed in Section 4.5, this is not discussed further
• Identify the limit state under consideration, for herein.
example bearing capacity of a shallow footing.
• Describe the relevant parameters (for example 4.4 Design example – pile group
effective cohesion, c’ and φ’) statistically by using
appropriate frequency distribution curves. Randolph and Buttling (2022) demonstrate how the
• Use a numerical model to assess for many Monte Carlo method, scripted using Python, can be
scenarios if the limit state is exceeded. applied to undertake a probabilistic analysis of a pile
• Assess the probability of failure based on the group analysed using the spreadsheet program PIGLET.
number of cases which result in the limit state being In the example presented by Randolph and Buttling
exceeded. (2022), a number of inputs to the pile group analysis are
treated as random variables, including the shear
Of the above steps, a typical geo-engineering modulus (both axial and lateral) of the soil in which the
professional will likely be least familiar with the piles are constructed, the axial capacity of the piles, and
assessment of frequency distribution curves and with the the normalised lateral pile head displacement at which
identification of a suitable numerical model to assess the the secant stiffness is 50% of the initial tangent value. As
limit states for the purposes of a reliability analysis. the analysis undertaken was a test case (and not a
design to be constructed), the range of the random
An assessment of an appropriate frequency distribution variables used as inputs to the probabilistic analysis was
requires a working knowledge of statistics. Ideally, the expressed through the use of published COV data,
geo-engineering professional would have a thorough rather than a statistical assessment of site-specific
understanding of the site conditions from an appropriate geotechnical investigation data.
site investigation, with many data points to allow an
appropriate frequency distribution to be identified (e.g. The Python subroutine was used to carry out 100,000
normal, log-normal, etc.). However, this is not always analyses for 38 load cases, which took approximately
practical and so the geo-engineering professional may 150 minutes on a standard computer. It was found that
use information presented in the literature in relation to the pile group lateral displacement limit of 50 mm
the mean and coefficient of variation of various soil (considered as a ULS case) was exceeded by 0.33% of
properties (refer Uzielli et al., 2006 for example). It is the simulations, giving a β value of 2.72. This compares
noted that the coefficient of variation (COV) is the with a value of 3.1 recommended in AS5104 (Standards
Australia, 2017b) for Class 2 structures with a high cost
of safety measures. A typical bridge such as a highway
11
bridge would probably be defined as a Class 3 structure

in which case β would be required to be 3.3. As noted
above, for the test case of the subroutine the site specific
soil data was not used, but it was considered that β could
relatively easily be raised to 3.3 if required, such as by a
better site investigation, for example using a dilatometer
or pressuremeter to measure the in situ modulus and its
variability.
4.5 Design examples – retaining walls
Low (2005) presents reliability analyses of two types of

retaining walls which were undertaken using FORM
within Microsoft Excel and its built-in optimisation
program application Solver. The reliability analyses were
undertaken on both a gravity retaining wall (Figure 5) to
assess the ULS failure modes of rotation and sliding, and
Figure 5. Reliability analysis of gravity retaining wall
an anchored retaining wall (Figure 6) to assess the ULS
(from Low, 2005)
failure mode of rotation. Low (2005) provides detailed
information on how the methodology was undertaken
Low (2005) concludes that the calculated β values using
and includes a link from which the relevant Microsoft
FORM are in good agreement with those calculated from
Excel files may be downloaded.
an alternative Monte Carlo simulation, but with
significantly less computational effort.
For the case of the gravity retaining wall, limiting
equations for rotation and sliding (referred to as
5 DISCUSSION AND SUMMARY
performance functions) are derived from basic physics
and geo-mechanics. Three variables are considered: φ’,
If RBD is to be adopted more widely in the geotechnical
δ and base adhesion. The degree of correlation between
industry in Australia, a number of issues need to be
φ’ and δ is also considered, and the degree to which the
addressed. These include:
reliability is sensitive to each variable is assessed. β is
calculated using the Solver application described above. • Increased focus on appropriate statistics in
undergraduate geo-engineering courses.
For the case of the anchored retaining wall, nine • Upskilling of the knowledge of relevant statistics of
variables were considered in the assessment of the currently practising geo-engineering professionals.
required wall embedment to achieve a target β. These • Further research into the benefits that RBD has over
variables included soil unit weight, surcharge pressure, WSD and LRFD, and clear communication of these
depth to groundwater and the yield force of the anchor. benefits to clients such that the additional high-
In a similar manner to the gravity retaining wall case, the quality site investigations required for RBD are
equations for force and moment equilibrium are used as approved.
performance functions, and a reliability assessment
undertaken.
Figure 6. Reliability analysis of anchored retaining wall (from Low, 2005)
12
• The improvements in recent years in the way in Fenton, G.A. and Griffiths, D.V. (2003). Bearing-capacity
which electronic site investigation data are gathered prediction of spatially random c - φ soils. Canadian
and manipulated has allowed significant efficiencies Geotechnical Journal, 40, pp. 54-65.
to be gained in the management of geotechnical Fenton, G.A., Naghini, F., Dundas, D., Bathurst, R.J. and
data. The time and budget gained through these Griffiths, D.V. (2016). “Reliability-based
improved efficiencies needs to be re-invested in geotechnical design in 2014 Canadian Highway
more sophisticated interpretation methodologies to Bridge Design Code”. Canadian Geotechnical
allow RBD to be undertaken. Journal, 53, pp. 236-251.
• Greater engagement with RBD by the Australian Green, R. (1989). Limit states design: some thoughts.
geo-engineering community. Proceedings of the Symposium on Limit States
• Increased co-operation between industry and Design in Foundation Engineering. Canadian
academia in relation to how RBD may be adopted in Geotechnical Society – Southern Ontario Section,
practice, e.g. the Recent Trends in Geotechnical Toronto, May 26-27, pp. 91-116.
and Geo-Environmental Engineering and Education Haberfield, C.M. (2017). Practical applications of soil
(RTGEE) workshops. structure interaction analysis. Gregory
• Development of an overarching Australian Standard Tschebotarioff Lecture 2017, ISSMGE. 19th
for geotechnical design, similar to that which exists International Conference on Soil Mechanics and
in Europe and the USA, which addresses the Geotechnical Engineering, Seoul, pp. 81-100.
concept of RBD. Hasofer, A.M. and Lind, N. (1974). Exact and invariant
second-moment code format. J. Eng. Mech.,
Subject to the above, the authors consider that RBD has 100(1), pp. 111-121.
the potential to be a highly useful design tool for the geo- Low, B.K. (2005). Reliability-based design applied to
engineering professional. Notwithstanding this, the retaining walls. Geotechnique, vol. 55 no. 1 pp. 63-
authors consider that, as described by Becker (1996a): 75.
• Whilst probability and statistics are useful tools if Low, B.K. and Tang, W.H. (1997). Automated reliability
properly applied, they must not become a substitute based design of footing foundations. Proceedings
for trying to understand the behaviour of of the 7th international conference on structural
geotechnical materials. safety and reliability. ICOSSAR ’97, Kyoto, Japan,
• Geotechnical design must not become overly vol. 3 pp. 1837-1840.
distracted by levels of safety and their quantification, Pells, P.J.N. (2011). “Against limit state design in rock”.
but instead focus on the understanding of the basic Tunnels & Tunnelling International, February
failure mechanisms and fundamental material edition, pp. 34-38.
behaviour. Randolph, M.F. and Buttling, S. (2022). “An investigation
of probabilistic analysis in relation to the design of
REFERENCES pile groups”. Proceedings of the 20th International
Conference on Soil Mechanics and Geotechnical
AASHTO (2017). “AASHTO LRFD bridge design Engineering, Sydney, Australia (awaiting
specifications”. Washington DC, USA. publication).
Baecher, G.B. Christian, J.T. (2003). “Reliability and Schweckendiek, T. Courage, W.M.G., and van Gelder,
Statistics in Geotechnical Engineering”. P.H.A.J.M. “Reliability of sheet pile walls and the
Becker, D.E. (1996a). “18th Canadian Geotechnical influence of corrosion – structural reliability analysis
Colloquium: limit state design for foundations. Part with finite elements”. Proceedings of the European
I. An overview of the foundation design process”. Safety and Reliability Conference (ESREL 2007),
Canadian Geotechnical Journal, 33, pp. 956-983. Stavanger, Norway, 25-27 June, pp. 1791-1799.
Becker, D.E. (1996b). “18th Canadian Geotechnical Standards Australia (2002). “Earth-retaining structures,
Colloquium: limit state design for foundations. Part AS 4678:2002”. Sydney, NSW.
II. Development for the National Building Code of Standards Australia (2009). “Piling – design and
Canada”. Canadian Geotechnical Journal, 33, pp. installation, AS 2159”. Sydney, NSW.
984-1007. Standards Australia (2011). “Residential slabs and
British Standards Institution (BSI) (1986). “Code of footings, AS 2870”. Sydney, NSW.
Practice for Foundations”. Standards Australia (2017a). “Bridge design, part 3:
British Standards Institution (BSI) (2004). “Eurocode 7: foundation and soil-supporting structures, AS
Geotechnical design”. 5100.3”. Sydney, NSW.
Canadian Standards Association (2014). “Canadian Standards Australia (2017b). “General principles on
Highway Bridge Design Code”. CAN/CSA-S6-06, reliability for structures, AS 5104”. Sydney, NSW.
Mississauga, Ontario. Uzielli, M., Lacasse, S., Nadim, F. and Phoon, K.K.
Cherubini, C. (1990). “A closed-form probabilistic (2006). “Soil variability analysis for geotechnical
solution for evaluating the bearing capacity of practice”. Characteristic and engineering properties
shallow foundations”. Canadian Geotechnical of natural soils, vol. 3, pp. 1653-1752.
Journal, vol. 27, pp. 526-529. Whitman, R.V. (1984). “Evaluating calculated risk in
Day, R.A. (2001). “Factored material properties and limit geotechnical engineering”. J. Geotech. Engrg.,
state loads – unlikely extreme or impossible 110, pp. 143-188.
pretence”. Geotechnical Engineering, vol. 149, no. Wong, P.K., Day, R.A. and Poulos, H.G. (2007) “Fifteen
4, pp. 209-210. years of geotechnical limit state design in Australia.
Day, R.A., Wong, P.K. and Poulos, H.G. (2007) “Fifteen Part II – foundations”. Proceedings of 10th
years of geotechnical limit state design in Australia. Australia New Zealand Conference on
Part I – soil retaining structures”. Proceedings of Geomechanics, Brisbane.
10th Australia New Zealand Conference on
Geomechanics, Brisbane, pp. 596-601.
Application of soil nail wall to roadway widening using GFRP rebars as per
Australian design guidelines
J. Lee1, PhD, PE, CPEng, FIEAust, E. Lee2, ME, MIEAust, M. Leong3, ME, M. Wei4, PhD, CPEng, FIEAust
1
Executive Director, GS E&C Australia, (formerly AECOM, Melbourne), Level 3, 100 Walker Street, North Sydney, NSW 2060; PH (02)
9135-2961; email: jeawoo.lee@gsenc.com
2
Senior Geotechnical Engineer, AECOM, Level 10, 727 Collins Street, Docklands, VIC 3008; PH (03) 9653-8705; email:
erin.lee@aecom.com
3
Design Manager, CPB Contractors, Level 6, 567 Collins Street, Melbourne, VIC 3008; PH (03) 9228-7000; email:
melvyn.Leong@cpbcon.com.au
4
Technical Principal, EIC Activities, Level 6, 567 Collins Street, Melbourne, VIC 3008; PH (03) 9228-7700; email:
michael.wei@eicactiv.com
ABSTRACT
This paper presents the application of soil nail wall technology to roadway widening. An arterial road upgrade project in
Melbourne consists of widening the freeway in the northern part of Melbourne, including massive cuts into a hillside on the
southern side of the existing freeway. Cuts up to 12 meters necessitated the use of retaining walls at 1H:10V batter to stay
within the right of way. A soil nail retaining wall was adopted to facilitate reduced excavation, less impact on the existing
slope, and improved construction speed with a top-down process. The soil nail retaining wall is 520 m long with a maximum
height of approximately 12 meters, including the undercut for pavement and drainage. The ground comprises clay fill
overlying residual clay and subsequent weathering profiles of basalt from Newer Volcanics group, partly overlying Brighton
Group sediments. Eleven boreholes were drilled sufficiently below the bottom of the wall. Laboratory tests were conducted
to estimate the soil and rock strength, including triaxial compression tests with pore pressure measurement to determine
effective strength parameters for Brighton group clayey soil. The design employed Glass Fibre Reinforced Polymer (GFRP)
bars to enhance the work efficiency by removing encapsulation of steel bars, the durability of which was reviewed for the
100 year design life with the proven data provided by the manufacturer. The soil nail retaining wall was designed as per
AS5100.3 and AS4678 guidelines selectively depending on the importance of the wall with reference to VicRoads
Specification Section 683 and FHWA-NHI-14-007.
Keywords: retaining wall, GFRP permanent nail, roadway widening, Australian design guidelines for soil nail walls
1 INTRODUCTION enhance the work efficiency by removing encapsulation

of steel bars. The durability of GFRP rebars was
An arterial road upgrade project in Melbourne consists of reviewed for the 100-year design life with the proven
widening the freeway in the northern part of Melbourne, data provided by the manufacturer. The soil nail retaining
including massive cuts into a hillside on the southern wall was designed as per AS5100.3 and AS4678
side of the existing freeway. Cuts up to 12 meters guidelines selectively depending on the importance of
necessitated the use of retaining walls to stay within the the wall with reference to VicRoads Specification Section
right of way. The reference design provided a reinforced 683 and FHWA-NHI-14-007. This paper addresses the
concrete wall on piles to form the retaining wall, which implication of Australian design guidelines for soil nailed
provokes construction risks due to significant excavation walls using GFRP bars, soil nail bond strength
to be committed adjacent to the live road. Comparing determination, the impact of excavation on adjacent
various retaining wall options, the detailed design existing structures, and the integration of road barriers
adopted a soil nail retaining wall in light of several into the wall.
advantages over other earth retention systems, including
reduced excavation, less impact on the existing slope, 2 SITE CONDITIONS
and improved construction speed with a top-down
process. The soil nail retaining wall is 520 m long with a 2.1 Ground conditions
maximum height of approximately 12 meters, including
the undercut for pavement and drainage. The ground The ground conditions have been inferred from eleven
comprises clay fill overlying residual clay and boreholes drilled along/close to the wall alignment up to
subsequent weathering profiles of basalt from Newer a depth equal to the wall height below the base of the
Volcanics group, partly overlying Brighton Group wall. Six test pits were dug up to 3 m deep, which informs
sediments. the design of unsupported exaction. A review of
geological maps of the area and the available
Eleven boreholes were drilled sufficiently below the geotechnical information indicates that the soils and rock
bottom of the wall along the wall alignment. Test pits of the Quaternary age Newer Volcanics are underlain by
were dug to inform the depth of unsupported cut. Tertiary age sediments.
Laboratory tests were conducted to estimate the soil and
rock strength, including triaxial compression tests with Boreholes drilled on-site demonstrate ground conditions
pore pressure measurement to determine effective comprising fill varying in depth between 1.5 m and 4.0 m,
strength parameters for Brighton group clayey soil. The overlying natural residual soil and basalt of Newer
design employed Glass Fibre Reinforced Polymer Volcanics Group. The borehole logs long plot is
(GFRP) rebars as opposed to steel bars, which had been displayed along the soil nail wall alignment in Figure 1.
used for most soil nail walls in roadway projects, to As shown in Figure 1, the boreholes near the
Application of soil nail wall to roadway widening using GFRP rebars as per Australian design guidelines J. Lee, E. Lee, M. Leong and M. Wei
western/eastern end of the soil nail wall indicate that the

Newer Volcanics are underlain by cemented sandy clay
of Tertiary age sediments, which forms part of the ground
behind the wall. The residual clays encountered within
these boreholes are of very stiff to hard consistency. The
weathering of the basalt varies from highly to moderately
weathered with low to high strength. Based on the
available geotechnical information, the basalt is
identified to be highly jointed and fractured. Existing
groundwater data available on-site suggests the
groundwater levels are below the base of the soil nail
wall.
2.2 Laboratory soil and rock testing and rock

mass quality
Considering limited information on relationships between

effective soil strength parameters and in-situ sounding
tests, such as SPT “N” on hard sandy clay of Tertiary age Figure 2. P-q’ plot from CU triaxial tests with pore
sediments, consolidated undrained triaxial tests with pressure measurement
pore pressure measurement were conducted on the
undisturbed samples taken from hard sandy clay
encountered in boreholes near the western and eastern
ends (BH01, 02 and 09). The results of the triaxial tests
are displayed in the form of p-q’ plot, as shown in Figure
2.
The effective cohesion tested greater than 40 kPa,

indicating the clay is highly over-consolidated throughout
the boreholes, with plasticity tests on the samples
indicating low plasticity. Observation on rock cores and
laboratory test results on basalt core samples indicate
the rock mass quality of the basalt pertaining to Class IV
Poor rock in Rock Mass Rating by Bieniawski (1989) and
Blocky/Disturbed/Seamy to Disintegrated on GSI chart
by Hoek (1994) as shown in Figure 3.
Uniaxial compression strength (UCS) of basalt tested

between 8 and 35 MPa. The UCS and Point Load
Strength (Is50) relationship appeared to be UCS = 8 x
Is50 sitewide.
Figure 3. Rock mass quality on GSI chart by Hoek
Figure 1. Borehole long plot along the wall
2.3 Restrictions Imposed by Adjacent Structures
The wall alignment neighbours existing structures, which

imposes constraints, including:
• proximity to the existing transmission tower that is
approximately 8 m south of the crest of the proposed
soil nail wall for which the asset owner required the
design to ensure minimum clearance of 0.5 m between
the closest soil nail and the tower footing and
maximum vertical tolerance of 5 mm for the
transmission tower;
• existing soil nail wall, the face of which needs to be
dismantled to tie in with; and
• proximity to the existing bridge at the eastern end of
the wall.
Design and construction measures have been made to
manage the tolerance of the wall deflection in proximity to
the existing transmission tower by complying with the
design requirement as per AS5100 and numerically
predicting the wall deflection, which has been ensured
with the monitoring during the construction. More
stringent design requirements out of AS5100 and AS4678
have applied for the design where the wall extends up to
30 m from the existing bridge abutment or existing assets Figure 4. Typical cross-section of soil nail wall featuring
of high importance with reference to VicRoads Road barrier and shotcrete upstand
Structure Inspection Manual (2018). The implication of
design codes is further discussed in Section 3.2. 3 DESIGN CONSIDERATIONS
In addition, the traffic barriers adjoining the bottom of the 3.1 Design soil/rock parameters and bond
soil nail wall shall be designed to act as independent strength
systems or to prove no detriment to the soil nail wall in the
event of any collision with the traffic barrier where the Geotechnical material parameters adopted for design
barriers and the wall are integrated. To this end, the have been derived based on the available geotechnical
impact load is considered to be taken by the traffic barrier information, including both in-situ and laboratory tests,
designed as an independent system to the soil nail wall, and practically accepted correlations in literature (e.g.
hence no damage to the integrity of the wall or load Burt 2005), as summarised in Table 1. The highly
transfer to the wall. This is further discussed in Section weathered (Hw) to moderately weathered (Mw) basalt
3.3.1. strength parameters are determined based on the
empirical relations between the RMR class and Mohr-
2.4 Other Wall Features Coulomb failure criterion strength parameters and Mohr-
Coulomb strength parameters equivalent to Hoek and
The retaining wall requires a balustrade or other barrier to Brown strength for rock mass.
alleviate the risk of a person falling from the top of it as The soil nail bond strength was determined from the
depicted in Figure 4. The extension at the top of the soil lesser of empirical bond stress presented in literature
nail wall, which forms a safety barrier upstand, is such as CIRIA report C637 (CIRIA 2005) and effective
assumed to be constructed at the same time as the upper- stress based estimation employing the effective strength
most nails. This barrier upstand is subjected to wind design parameters in Table 1.
loading and handrail loading from AS5100 Cl 12.5 (a-c).
Table 1: Geotechnical design parameters

Effective Undrained Soil Nail
Unit Effective Young’s
Friction shear Ultimate Bond stress
Material Weight, Cohesion Modulus,
Angle, strength, Bond Stress, based on
kN/m3 , kPa kPa
degrees kPa kPa
Average normal
Existing Fill 19 5 28 75 20 50 effective stress
Newer Volcanics
Average normal
Residual Clay 19 10 28 100 30 60 effective stress
(Very stiff to hard)
Empirical
Basalt – Hw to
22 80 32 - 300 300 adhesive
Mw strength
Inferred Brighton
Group – Average normal
19 15 35 200 60 80 effective stress
cemented sandy
clay (hard)
16
The adhesion between the nail grout and Hw basalt for method as per AS 5100.3-2017, achieving a minimum
the rotary-drilled hole was determined with reference to global FoS of 1.8 in long-term
empirical literature charts such as those in FHWA Condition. The important structures sections include:
(FHWA 2003). The nail verification pull-out tests
conducted during the installation of nails have verified • located within 30 m of the existing road bridge
the design ultimate bond stress to be adequate. • located within 30 m of the existing transmission tower.
Seismic stability has been analysed based on the
3.2 Soil Nail Wall Design as per AS4678 and working stress method, achieving a minimum global FoS
AS5100 for Soil Nail Retaining Walls of 1.2. The stability analysis has taken into account
possible construction staging with 0.5 m excavation
The soil nail wall design has been completed below each row of nails prior to installing the subsequent
predominantly relying on the stability of soil nail walls, row of nails and 1.5 m over-excavation below the
which has been analysed using the 2D limit equilibrium pavement level at the end of the excavation included.
software Slide2 by Rocscience and the 2D finite element The short term analysis accounting for ‘during-
software PLAXIS 2020 with the strength reduction construction cases’, including those subject to the build-
method employed. up of temporary water pressure and seismic loading, has
been performed employing undrained soil design
Static stability design has primarily adopted the limit parameters, which have been switched to effective
state method with material strength partial factors and stress design parameters for the long term stability
load factors adopted in accordance with AS 4678-2002 analysis.
as summarised in Table 2 and Table 3, achieving a
minimum global “factor of safety” (“FoS”) of 1.0 in both As shown in Figure 5, the long term analysis has been
short-term (i.e. during construction) and long-term static carried out excluding the bottom row of soil/rock nails
conditions. behind the road protection barrier that is considered
sacrificial for collision so that the wall stability is
Table 2: Partial Factors as per AS4678 independent of the bottom row soil/rock nails. As a
Partial Factor Items Existing Fill In-situ Material result, the minimum global “FoS” has been calculated to
Effective friction 0.9 0.85 be between 1.1 and 1.4 for the sections designed as per
angle, ɸɸ AS4678 and 1.9 for those as per AS5100 in long term
Effective cohesion, 0.75 0.7 conditions.
ɸc
Undrained shear 0.5 0.5 On the other hand, the strength reduction method in the
strength, ɸuc finite element models has generated the minimum global
FoS slightly higher than that from the conventional limit
Table 3: Load Factors as per AS4678 equilibrium method. The working stress method analysis
Partial Factor Items Factor adopted against all the sections designed as per AS4678 or
Dead load behind wall 1.25 AS5100 has indicated the minimum global FoS greater
Dead load ahead of wall 0.8 than 1.5 specified as the design criteria in FHWA-NHI-
Traffic load behind wall 1.5 14-007 and CIRIA report C637.
The following design reduction factors have been The nails have been laid out at spacings of 1.5 m in soils
considered for the global stability analysis in as per and 1.8 m in Hw to Mw basalt with lengths up to 0.9 H at
AS4678-2002 or other standards as stated below: maximum. The hole diameter has been set to 125 mm
• Tendon structural capacity reduction factor = 0.4 as in soils and 105 mm in Hw to Mw basalt.
per VicRoads Section 683 (06.b.vii);
• Pull-out (bond) resistance reduction factor for rock
nails = ɸk x ɸb = 0.56;
where:
o Importance category reduction factor, ɸk = 0.8
(AS4678-2002 Table B1);
o Minimum material reduction factor ɸb = 0.7 (AS4678-
2002 Table B2) for the bond between rock and grout;
o Pull-out resistance reduction factor for soil nail = ɸn x
ɸb =0.63;
where:
o Structure classification design factor, ɸn = 0.9 for
structure classification C (AS4678-2002 Table 5.2); Figure 5. Soil Nail Wall Stability Analysis with the
and bottom nail excluded
o Minimum material reduction factor, ɸb = 0.7 (AS4678-
2002 Table B2) for the bond between soil and grout; 3.3 Deformation Analysis Considering
Interaction with Adjacent Structures
The bond capacity within the first 1 m length of nail has
been ignored in the design as per VicRoads Section 683. The Finite Element analysis undertaken has also
assessed the deformation of the soil nail walls, including
For the design sections in proximity to important the influence of collision loading and its impact on the
structures, the design has adopted the working stress existing structures in proximity. The maximum lateral
wall deflection has been estimated to be less than 0.3 %
of the wall height, which results in insignificant settlement
17
behind the wall in terms of serviceability. Particular

attention has been paid to whether the soil nailing
induced ground deformation adversely impacts the
structure adjacent to the wall.
3.3.1 Collision loading impact assessment
As per the VicRoads regulation on external loads acting

on a retaining wall, the road protection barriers in front of
the soil nail wall are to be designed to be structurally
independent of the soil nail wall from the collision loads.
The geotechnical information indicated that the materials Figure 6. Soil Nail Wall Stability Analysis with the
behind road barriers consist of Hw to Mw weathered bottom nail excluded
basalt or highly over-consolidated sandy clay. No
structural component was required as part of the 3.3.3 Interference with Existing Nails
retaining wall to support the ground after excavation. The
facing was assumed to be redundant (i.e. no The presence of existing soil nails was expected to
replacement required) behind the barrier in the event of introduce interface constraints with the soil nails
a vehicle collision. A joint placed in the facing behind the proposed at the section tying in to the existing soil nail
top of the barrier split the facing into two pieces wall. PLAXIS 2D analysis was carried out to assess the
structurally. A 25 mm thick compressible layer was also impact of removing the existing soil nail walls based on
included to prevent the load transfer from the barrier to the information shown on the historical design drawings.
the shotcrete. The road protection barriers have been The existing soil nails exposed within the excavation for
placed without an extended foundation, piled or slab each lift were assumed to be cut prior to the installation
footing where the assumed design ground condition is of the new soil nails, with the existing soil nail behind the
present. new wall remaining in place without ground disturbance
and a new row of soil nails installed.
The impact of collision loading has been assessed using
PLAXIS 2D for the design sections. The road protection The critical section comprising the tallest heights of the
barrier was modelled with concrete properties embedded retaining walls (existing retaining wall height of 6.6 m and
100 mm into the finished road level. The result has new retaining wall height of 8.8 m) was modelled, as
shown no sign of local instability and a negligible shown in Figure 7. The result of the impact analysis using
increase in either ground deformation or working load in PLAXIS 2D has indicated a low likelihood of disturbance
nails. to the stability of the soil nail walls by the removal of the
existing nails in a sequential manner.
3.3.2 Asset impact assessment
A 220 kV transmission tower has been identified where

the wall height becomes the highest. With the proposed
wall slope of 1H in 10V, the closest tower leg is
approximately 8 m behind the crest of the soil nail wall.
The available information on the tower has informed that
each foot of the transmission tower is founded on a
circular trapezoid pad footing with the upper surface
diameter of 1.1 m and the base diameter of 1.4 m at 2.6
m below ground surface at the time of construction. Due
to proximity to the soil nail wall, the transmission tower Figure 7. Soil Nail Wall Stability Analysis with the
might be subject to ground movement induced by the bottom nail excluded
excavation during soil nail wall construction. The asset
owner of the existing transmission tower requested the 3.4 Facing Design
conditions described in Section 2.3 to be met. The
PLAXIS 2D model developed for asset impact The wall facing comprises double shotcrete layers -
assessment is presented in Figure 6. The material structural facing overlain by aesthetic facing. The
underneath the tower footings is assumed to be residual structural facing has been designed in accordance with
soil clay with very stiff to hard consistency. Two FHWA-NHI-14-007. The working stress-based PLAXIS
transmission tower footings with 6.7 m apart centre to analysis outcomes have informed the structural capacity
centre are included in the analysis. assessment with the employment of a lumped load factor
of 1.5 as per AS5100.3.
Due to the uncertainty of the transmission tower loads on
the footing, an allowable bearing pressure of 440 kPa on 4 USE OF GFRP BARS
residual clay has been taken as the footing foundation
load. The upper row of soil nails is 6 m long at an 4.1 Longevity of GFRP Bars
inclination of 15°, maintaining sufficiently more than 500
mm from the tower footing, which has been accepted by The lightweight properties of GFRP rebars make
the asset owner. handling and installation easy and safe compared with
conventional steel rebars. Their inherent non-corrosive
The PLAXIS analysis result has shown the maximum properties have additional benefits from a durability
differential movement under the footings less than 3 mm perspective, as no additional corrosion protection is
across the neighbouring footings, which is within the required (CIRIA 2005). Although the application of GFRP
tolerance of 5 mm. rebars to permanent soil nail walls is yet to be long
18
enough to attest to the longevity over a 100-year design 5 CONCLUSIONS

life, indicative testing conducted by others has
suggested that vinyl ester resin GRP should have a long A soil nail wall up to 12 m high has been successfully
life in a cementitious environment (Johnson 1998). designed and constructed adjacent to one of the busiest
Mechanical testing to inform the 100-year design life is arterial roads in the Greater Melbourne area as part of a
usually based on creep rupture testing. This testing is freeway widening project. The application of soil nailing
used to estimate indicative sustainable loads over a utilising GFRP rebars to the retaining wall has provided
period of 106 hours (>114 years) assuming the linear significant benefit to the project from either the design or
degradation of the ultimate tensile strength (UTS) of the construction perspective, including:
GRP due to creep. This testing is undertaken in the
absence of embedment material around the soil nails. • Meeting the design requirements with innovative
The indicative performance is therefore based on an material replacing conventional steel rebars
allowable applied load on the GRP bars of less than the • Relieving burden on the design concerning corrosion
creep rupture strength estimates at 100 years and the of conventional steel rebars
indicative performance of the Vinyl Ester resin GRP in • Minimising disruption to the adjacent live arterial road
the cement mortar bedding. by reducing temporary earthworks, which is facilitated
by top down staging of soil nailing
The vinyl ester resin GRP Bluey BluGeo GRP60-20 mm • Speeding up the nailing installation progress, leading
bar was identified in the testing performed by Mohamed to construction time saving
and Benmokrane (Mohamed and Benmokrane 2018) to • Enhancing work efficiency by reducing the burden to
have an indicative 106 hour creep rupture strength of handle heavy steel products and encapsulating steel
58% of the original UTS. Theoretically, where the load is rebars
applied less than the 106 hour creep rupture strength
Lastly, although the case study presented demonstrates
and the mortar and soil nail bar bond is maintained over
the applicability of GFRP soil nails for use in retaining
the design life, the properties of the GRP are not
walls with significant benefit design and construction-
considered unduly affected by the environment, which
wise, the use of GFRP nails will be better convinced with
the soil nail should then have adequate durability for the
further research on the long term performance in the
nominated 100-year design life. In addition, according to
future. Monitoring the wall profile at regular intervals over
the manufacture information, the product should comply
a long period may help understand the long term
with VicRoads Section 683.
behaviour of the wall reinforced with GFRP rebars
4.2 Implication of GFRP bars on Soil Nail wall
design
REFERENCES
For the sake of comparison between GFRP and Australian Standard (2002), AS4678-2002.
conventional steel nails as to their implication to the Australian Standard (2017), AS5100-2017.
design of soil nail wall, PLAXIS2D and SLIDE models Bentley Systems (2020), PLAXIS User Manual.
have been analysed over one of the design sections with Bieniawski, Z. T. (1989). Engineering Rock Mass
respect to the minimum global FoS, maximum horizontal Classifications. Wiley-Interscience. p. 272.
wall displacement, and maximum axial nail force CIRIA (2005) Soil Nailing - best practice guidance, CIRIA
adopting the parameters tabulated in Table 4. Report C637, CIRIA.
Hoek, E. (1994) Strength of Rock and Rock Masses, ISRM
News Journal, Vo 2(2), pp. 4-16.
Johnson, P. E. & Carf, G. B. (1998), The use of soil nails for the
Table 4 Properties of GRP and Steel construction and repair of retaining walls, Transport
BluGeo DSI GEWI® Research Laboratory 373.
Items Look, B. G. (2007) Handbook of GeotechnicalInvestigation and
GRP60 Threadbar
Design Tables, Taylor & Francis.
Diameter (mm) 25 20 Mohamed, K. And Benmokrane, P. (2018) Creep Rupture of
BluGeo GRP60-20mm GFRP Bars of Size 6 (Designated
Young’s Modulus 60 200 diameter of 20 mm), NSERC Research Chair in
(GPa) Innovative FRP Composite Materials for Infrastructure,
University of Sherbrooke.
Ultimate Tensile 350 157 VicRoads (2018), Road Structures Inspection Manual.
Strength (kN) VicRoads (2019), VicRoads Standard Specification 683 – soil
Unit Weight (kg/m) 0.9 2.47 nail walls.
Structural Reduction 0.40 (1) 0.648 (2)

Factor
1
As per VicRoads Standard Specification 683
2
As per AS 4678-2002 Earth-retaining structures
The use of GFRP nails in lieu of conventional steel nails

results in the following:
• decreases in axial nail force up to 30% due to lower
elastic modulus of GRP nails than that of steel nails.
• marginal increases of soil nail wall horizontal
displacement up to 7% due to the lower elastic
modulus of GRP nails than that of steel nails.
• no change in the factor of safety against wall
instability.
Risk Prediction Model for Formation of Underground Cavities and Sinkholes due to
Defective Sewer Pipes
S.R. Indiketiya1, P. Jegatheesan2 and R. Kuwano3
1
Former PhD Student, Swinburne University of Technology, Hawthorn, Melbourne, Vic 3122, Australia; email: samanthi.uom@gmail.com
2
Former Senior Lecturer, Swinburne University of Technology, Hawthorn, Melbourne, Vic 3122, Australia; email:
piratheepan.jegatheesan@yahoo.com
3
The University of Tokyo, Meguro-ku, Tokyo 153-8505, Japan; email: kuwano@iis.u-tokyo.ac.jp
ABSTRACT
Sinkhole formation due to internal erosion around defective sewer pipes is identified as a serious threat in urban
infrastructure system. Post-repair and rehabilitation after pipe failure are not effective as emergency pipe repairs are very
costly and pipe failure leads to various public and environmental consequences. Only a few studies have been conducted
on the prediction of the risk of ground erosion around pipe defects. Therefore, the main objective of this article is to propose
a model which can predict the risk of formation of sinkholes around sewer pipelines based on the weighted factors method
when a pipe defect is certain. The proposed methodology relies on different factors which contribute to void development
and severity of the consequences. The Risk of Erosion (ROE) combines the effect of Likelihood of Erosion void formation
(LOE) and Consequence of Erosion cavity formation and ground failure (COE). The LOE rating is related to many
parameters, including soil properties, hydraulic conditions, and pipe defect characteristics, while the COE rating is related
to the environmental, economic, and social consequences of pipe failure. Therefore, this model, which can predict the risk
of developing a sinkhole close to an existing pipe defect, will enable sewer assets maintenance teams to evaluate each
pipe and prioritize the maintenance and rehabilitation work based on the risk to each pipe.
Keywords: Sinkholes, Risk Prediction Model, Sewer Rehabilitation, Internal Erosion
1 INTRODUCTION rehabilitation projects and for sewer and storm water pipe
designers to design high stability backfill design
Wastewater and stormwater collection systems are considering the erosion resistance of the backfill.
critical components of urban infrastructure systems in any
country. Failure of deteriorated buried sewer pipes can 2 METHODOLOGY
occur in two ways. One process is when the soil backfill
in the vicinity of a pipe defect is gradually eroded, the pipe 2.1 Risk assessment method
will bend and collapse due to loss of confinement from the
ground (Balkaya et al. 2012). The second process occurs Most of the risk assessments conducted for sewer failures
when the cracks are formed at either sides or the crown as outlined in available literature have considered some
of the pipe. Voids can develop above the pipe while the form of likelihood or probability of failure, consequence of
pipe is still well supported from the bottom. In this case, failure and risk of failure. Emilio (2015) adopted a similar
erosion may propagate towards the ground surface method to predict the risk of failure of sewer pipes.
causing a sinkhole which eventually breaks most of the Therefore, modified version of the same approach is used
other buried service lines (Sato and Kuwano 2015). in this study to predict the risk of erosion void formation
(ROE) which combines the likelihood of erosion void
Many studies have been conducted on sewer formation (LOE) and consequence of erosion void
deterioration and risk prediction models for pipe failures formation (COE) in sewer pipes. The risk assessment
(Yan and Vairavamoorthy 2003; Baah et al. 2015; Emilio involved identifying influencing parameters for LOE and
2015). To the author’s best knowledge, only one risk COE then quantifying the LOE and COE by assigning
prediction model is available in the literature (Kaddoura critical scores based on the literature and by allocating
and Zayed 2017) which proposed a model that can predict factor weights based on experts’ opinion. After both
the risk of void erosion present outside sewer pipelines. ratings were determined, they were multiplied together to
This is based on weighted factors method using Fuzzy find the risk score for ROE as shown in Equation 1.
theory. However, this model has several limitations. It
does not account few a few other critical factors which ROE = LOE * COE (1)
control the process of cavity formation such as the relative
density or compaction of the ground and the influence of Following an extensive literature review, seven most
rainfall. Moreover, the erosion susceptibility of clay has influencing parameters which control the erosion initiation
been considered as higher than fine sand and silt which and progression through pipe defects were identified for
is contradicting with available literature (Rogers 1986, the LOE function as shown in Equation 2. The contribution
Indiketiya 2019). The model only predicts the probability from each factor to LOE, λi was estimated based on the
of erosion void presence outside sewer pipes and the risk expert's belief where, each 0 ≤ λi ≤ 1 and Σλi = 1.
due to consequences of erosion void formation is not
accounted for the overall risk. LOE = λ1.Soil type score +λ2.Relative density of
backfill score + λ3.Depth of sewer pipe score + (2)
Therefore, it will be useful to develop an efficient model λ4.Location of GWT score + λ5.Pipe defect size
score + λ6.Frequency and magnitude of sewer
which can predict the overall risk of developing erosion
exfiltration score + λ7.Frequency and magnitude
voids in defective sewers which can be useful to sewer of rainfall score
asserts management teams for allocating priorities in
Risk Prediction Model for Formation of Underground Cavities and Sinkholes due to Defective Sewer Pipes S. R. Indiketiya, P. Jegatheesan and R. Kuwano
The consequences of the erosion void formation and 2.2.3 The depth of the sewer
sinkhole formation are very similar to the consequences
of the pipe failures. Therefore, four parameters which Pipes which are closer to ground surface are often
have been commonly adopted in previous risk prediction disturbed and damaged due to heavy traffic, construction
models for pipe deterioration (Baah et al. 2015; Emilio works and maintenance of other underground services
2015) were selected for COE of this model. Selected four (Davies et al. 2001). In addition, O’Reilly et al. (1989)
factors to represent environmental, economic, and social found out that the defect rate decreases with the
impacts of the pipe location are depicted in Equation 3. increasing of the pipeline depth. The chance of erosion
Where, µi was estimated based on the expert's belief initiation and propagation also could increase at a lower
while satisfying 0≤µi ≤1 and Σµi = 1. depth due to lower confining pressure and faster rain
infiltration through the opening. Kaddoura and Zayed
COE = µ1.Environmental-pipe diameter score + (2017), also proposed a depth categorisation for sewer
µ2.Economic-commercial zone score) + (3) pipes in his void erosion prediction model. Based on the
µ3.Social-critical infrastructure score) + µ4.Social- above-mentioned discussion and studies, pipes are
traffic volume score)
scored based on it’s depth in LOE function as displayed
in the Table 3.
2.2 Assigning critically scores for LOE
Table 3. Scoring for depth of the pipeline
The condition of each pipe is evaluated for the LOE based The depth of the sewer (m) Attribute Score
on seven factors and each parameter was assessed and >4 Excellent 0
scored between 0 and 4 as given in Table 1 to Table 7. 2-4 Good 1
Therefore, the overall risk of LOE of pipe will also be 1.2-2 Fair 2
scored out of 4 as 0 ≤ λi ≤ 1 and Σλi = 1. as depicted in 0.6-1.2 Poor 3
Equation 2. The scoring criteria for each of the seven 0-0.6 Worst 4
factors are summarized through section 2.2.1 to 2.2.7.
2.2.4 Location of the GWT
2.2.1 Soil Characteristics If the GWL exists above the sewer pipeline or above any
structural defect, there is a higher possibility of infiltration
The soil type, its particle-size distribution and plasticity are and soil migrating to the sewer pipeline (Davies et al.
identified as key parameters which controls the initiation 2001). Indiketiya (2019) also discussed propagation of
of erosion through pipe cracks (Rogers 1986; Sato and erosion voids considering the geometry of the void and
Kuwano 2015; Indiketiya 2019). Considering all those the relative location of the GWT. Similar studies have
data, soil types were arranged into a hierarchy based on categorised the influence of GWT into two as pipe located
the erosion resistance through underground openings above and below the GWT (Rogers 1986; Davies et al.
and tabulated as shown in Table 1. 2001). However, the most important fact is the location of
the GWT relative to the pipe defect. As crack or the
Table 1. Scoring for soil type opening in a pipe can be at the bottom or the crown,
Soil type choosing the depth of pipe can be misleading for large
Attribute Score
(USCS specification) diameter pipes. Therefore, in this model, sewer depth was
GW, GC, CH Excellent 0 categorised and scored into three sections as shown in
CL-CH, SC, GM Good 1 Table 4.
SW, CL, GP Fair 2
SM, CL-ML, MH Poor 3 Table 4. Criticality scoring for the GWT
SM, ML, SP, Dispersive Clay Worst 4 Groundwater table Attribute Score
Well below the pipeline Excellent 0
Closer to the pipeline Poor 3
2.2.2 Relative density of the backfill Above the pipeline Worst 4
The importance of having a higher relative density to

minimise erosion potential in non-cohesive material has 2.2.5 Effect of pipe defect size
also been discussed by Rogers (1986) and Renuka
(2012). Most of the sewer pipe construction specifications Pipe defect size is one of the most governing factors
have recommended minimum compaction of 70% relative which controls the rate of soil loss. Rogers (1986) first
density for pipe embedment (United States Department of proposed a relationship between ground loss and the ratio
the Interior 1996; WSA 02-2002-2.3-MRWA Edition of B/D85 for a variety of cohesionless soils under the one-
2002). Therefore, considering the experimental results way water flow. Where D85 is the size of sieve through
from above researchers, different scores were assigned which 85% by weight of a soil sample will pass and “B” is
to each pipe backfill density as displayed in Table 2. the crack width. Critical crack width for continuous
migration of soil in monotonic water flow was expressed
Table 2. Scoring for density of backfill as 2.5D85 to 4.5D85. Indiketiya (2019) observed that the
Relative density rate of soil loss is significantly smaller if the “B” is less than
Attribute Score
(Non-Cohesive Soil) Dmax or 2D85 of the backfill material and soil loss was
>90% Excellent 0 critical when “B” was greater than 2Dmax. These findings
70%-90% Good 1 are consistent with the studies available on granular flow
60-70% Fair 2 through orifices. Based on these findings, pipe defects
35-60% Poor 3 were categorised into different groups for critical scoring
<35% Worst 4 as given in Table 5.
Table 5. Criticality scoring for the pipe defect The parameters include pipe diameter, distance from
B/Dmax B/D85 Attribute Score commercial zones, distance from critical infrastructure,
<0.75 <1 Excellent 0 and vehicular traffic volume as parameters for quantifying
0.75-1 1-2 Fair 2 the environmental, economic, and social consequences.
1-2 2-3 Poor 3 The scoring criteria for each of the four factors above are
>2 >3 Worst 4 summarized below in Section 2.3.1 to 2.3.4.
Similar to LOE scoring, COE scoring is assigned under

2.2.6 The magnitude of sewer exfiltration six attributes as excellent, very good, good, fair, poor and
worst. The scores are assigned as 0, 1, 2, 3, 4, 5
Rogers (1986) performed a series of physical model tests respectively. Therefore, the overall risk of COE of erosion
to examine the soil loss from sewer exfiltration from pipe void formation closer to a pipe is scored out of 5 as 0 ≤µi
and groundwater infiltration into the pipe. It was revealed ≤1 and Σµi = 1.
that in fine-granular soil, the frequency of sewer
exfiltration or leakages with a smaller surcharge was more 2.3.1 The size of the pipe diameter
critical than having a single leakage with a larger
surcharge. Therefore, even minor daily leakages through Pipe diameter is a crucial factor in COE function as it
the pipe cracks can cause significant soil loss if the flow represents the volume of sewer flow, the amount of public
occurs cyclically through large defects. The hierarchy coverage and the potential environmental consequences
proposed in WRC (2001) for the magnitude and frequency in a sinkhole event or rehabilitation work. For large
of sewer leakages is used in this model by assigning diameter pipes, pipes cover a larger domestic and
criticality scores as shown in Table 6. industrial buildings and if a pipe breaks the public
inconvenience is higher. When the diameter is increased,
Table 6. Scoring for sewer exfiltration the depth of the pipe is also higher, and the
The magnitude of exfiltration Attribute Score consequences of sinkhole formation will be greater as the
Never Excellent 0 size of the sinkhole is increasing with the depth (Guo et
Occasional, low magnitude Good 1
al. 2013). Therefore, the total cost, effort and time
Occasional, high magnitude Fair 2
required for a repair are higher than in a smaller size pipe.
Frequent, lower magnitude Poor 3
Frequent, high magnitude Worst 4 Table 8 shows the scoring breakdown for the pipe
diameter. This categorisation is referring Emilio (2015)
2.2.7 The magnitude of rainfall and actual sewer pipe sizes available in real practice in
Australia.
Kuwano et al. (2006) reported that majority of sinkholes in
Japan are formed in the rainy season. Rise in GWT, Table 8. Scoring for the size of the pipe diameter
Pipe Diameter (mm) Attribute Score
frequent sewer overflows, and groundwater infiltration
<150 Excellent 0
through existing cavities due to rain can accelerate the 180 - 225 Very Good 1
erosion process and sinkhole development. Referring to 250 - 375 Good 2
the classification introduced by WRC (2001) for sewer 400 - 525 Fair 3
exfiltration which was presented in Table 6, a very similar 550-750 Poor 4
approach was used in this model to classify the frequency >750 Worst 5
and magnitude of rainfall. It is difficult to quantify the
magnitude and the frequency of rainfall for each group as
it is sensitive to location and the country. Therefore, the 2.3.2 The distance from commercial zones
proposed qualitative classification in Table 7 can be
adapted to any location based on a relative judgment. In Australia, different states use different land zone
classifications and Victoria's major zonings are
Table 7. Scoring for the magnitude of rainfall residential, commercial, industrial and rural, in addition to
The magnitude of rainfall Attribute Score sub-zones as specified by the State Government of
No rainfall Excellent 0 Victoria (2014). In commercial zones, there are
Occasional, light shower Good 1 commercial activities which are essential for functioning
Frequent, light shower Fair 2 of a city such as retail shops, offices and childcare
Occasional, heavy rainfall Poor 3 centres. Therefore, the distance from the sinkhole to
Frequent, heavy rainfall Worst 4 these commercial zones is important, the greater the
distance, the lower the risk of consequences. Scores
defined by Emilio (2015) are selected for this model as
2.3 Assigning critically scores for COE shown in Table 9. Based on the scoring, pipes which are
1500m away from commercial activities have less impact
The COE is a representation of values of financial or of while those located closer than 300m have severe impact.
life-loss, and it measures the effect of pipe failure
regarding environmental, economic, and social Table 9. Scoring for the distance from commercial zones
consequences. There are various studies which predict Distance from
Attribute Score
the risk of consequences of pipe failures. Erosion void commercial zones (m)
formation and ground subsidence have very similar >1500 Excellent 0
consequences regarding environmental, social and 1200 - 1500 Very Good 1
900 - 1200 Good 2
economic impacts. Therefore, the four parameters used
900 - 600 Fair 3
by Emilio (2015) for sewer pipe failures are used in this
300-600 Poor 4
model. <300 Worst 5
2.3.3 The distance from critical infrastructure As this theme is still quite new and thorough knowledge
about the pipe deterioration and erosion process is
Critical infrastructure enables the provision of essential required to fill up the questionnaire, 15 researchers were
services such as food, water, health, energy, selected from various countries referring to their
communications, transportation, emergency services and publications which indicates their expertise on this study
banking. Sinkholes, pipe breaks and the subsequent area. All the information required for survey was provided
repairs may influence a city’s ability to respond to and experts were requested to assign λi and µi values
emergencies. Emilio (2015) also counted schools as corresponding to parameters from their expertise so that
critical infrastructure in his model, as schools are large ∑(λi) = 1 for LOE and ∑(µi) = 1 for COE separately.
public gathering points that may be negatively affected.
The same scoring breakdown used by Emilio (2015) is 2.5 Data analysis and determination of weighted
applied in this COE model as shown in Table 10 since influence factors
pipe failures have very similar consequences to sinkhole
events. The proposed λi and µi values by experts were scattered
in a wide range. Therefore, the weighted average (Mean)
Table 10. Scoring for the distance from critical and the mode of each data set for λi and µi are calculated
infrastructure and presented in Table 12 and Table 13 respectively. The
Distance from critical Attribute Score mean and the mode of all seven λi values are close and
infrastructure (m) consistent. Therefore, final λ values were designed clearly
>2500 Excellent 0 as closely following the mean and the mode values while
2499 - 2000 Very Good 1 satisfying the condition of ∑(λi) =1.
1999 - 1500 Good 2
1499 - 1000 Fair 3
However, all four µi, parameters are having higher
999-500 Poor 4
standard deviations while mean and the mode values are
<500 Worst 5
also slightly different. Therefore, considering the mean,
mode and the distribution of the normal frequency,
2.3.4 The traffic volume/road type of the
judgment on final value was decided as presented in
location
Table 13. Emilio (2015) had the same parameters in the
COE model and the judgment was µ1=0.3, µ2=0.3, µ3=0.2
The type of the road or the volume of traffic where the
and µ4=0.2. These have been decided manually based on
sinkhole appears are crucial which affect the
judgment where more emphasis was applied to social
consequences of the event. However, as the traffic
concerns because they directly affect the public.
volume highly depends on the location, the city and the
country, it is difficult to normalise the volume of traffic as
Therefore, the final equations for LOE and COE with the
an indicator for general practice. Therefore, road type
proposed influencing factors can be written as shown
would be beneficial as it is a relative gauge which can be
below in Equations 4 and 5.
easily replaced for different countries. There is a variety
of classifications for road types, and no single variable is
Table 12 The Mean, mode and proposed λi values
available to completely describe a class of roads. The
λ
road classification implemented throughout Victoria Parameter
Mean Mode Proposed
according to (Austroads 2005) are used here. According λ1 0.19 0.2 0.2
to this hierarchy, “M” routes deserve a higher critical score λ2 0.11 0.1 0.1
as it carries the highest traffic volume and has greatest λ3 0.09 0.1 0.1
connectivity within the road network. The scoring system λ4 0.18 0.2 0.2
proposed in this study is given in Table 11. Footpaths are λ5 0.21 0.2 0.2
also added with the lowest critical score as some sewer λ6 0.12 0.1 0.1
pipes are buried in footpaths which are not included in the λ7 0.10 0.1 0.1
national road numbering system. ∑( λ i) 1.0
Table 11. Criticality scoring for the road type Table 13 The Mean, mode and proposed µi values
Distance from critical µ
Attribute Score Parameter
infrastructure (m) Mean Mode Proposed
Footpaths Very Good 1 µ1 0.26 0.3 0.3
C Good 2 µ2 0.25 0.2 0.2
B Fair 3 µ3 0.25 0.3 0.25
A Poor 4 µ4 0.25 0.2 0.25
M Worst 5 ∑( µ i) 1.0
2.4 Data acquisition for weighted influence factors LOE = 0.2*(Soil type score) +0.1*(Relative density
of backfill score) + 0.1*(Depth of sewer score) +
Two of the recommended methods for estimating 0.2*(Location of GWT score) + 0.2*(Effect of pipe
(4)
defect size score) + 0.1*(Frequency and magnitude
influence factors (λ and µ values) in Equation 2 and 3 is
of sewer exfiltration score) + 0.1* (Frequency and
to get the expert's judgment (Joseph et al. 2010) or magnitude of rainfall score)
iterative computation through field data (Emilio 2015;
Kaddoura and Zayed 2017). As collecting actual field data COE = 0.3 * (Environmental-pipe diameter
related to defective sewer pipes and sinkhole events are score) + 0.2 * (Economic-commercial zone (5)
extremely difficult, multiple experts are selected by score) + 0.25 * (Social-critical infrastructure
considering the experience and confidence on the topic. score) + 0.25 * (Social-traffic volume score)
Table14. Risk level categorisation for ROE based on a risk matrix
Since the LOE is scored out of 4 and COE is scored out and the location of the GWT have the highest relative
of 5, this model considers pipes with LOE of 3 to 4 as the weights.
higher vulnerability for ground subsidence with pipe Four parameters which evaluates the consequences
failure and in need of urgent attention. Similarly, COE is of erosion void formation were considered as the pipe
scored out of 5 and pipes with COE score of 4 to 5 are diameter, the distance from commercial zone, the
considered as critical pipes which causes many social, distance from critical infrastructure and the road type
economic and environmental consequences. Various
of the location
studies have graphically illustrated the risk matrix system
(Baah et al. 2015) which combines the probability and • Criticality scores for LOE function was scored out of 4
consequences for an event. Referring those studies, risk and COE function was scored out of 5. Therefore, the
level for a pipe to develop an erosion cavity was overall risk of erosion, ROE is predicted rating out of
categorised through the risk matrix based on the score as 20. Then the level of risk for a defective pipe to
low, moderate, high and very high as illustrated with a develop an erosion cavity is categorised according to
colour code in Table 14. a risk matrix based on the ROE score: low (0 to 4),
moderate (5 to 9), high (10 to 14) and very high (15 to
3 VALIDATION 20).
• Unfortunately, due to lack of published data, and the
The model can be validated one of these methods: (1) legal requirements of different institutes and countries,
comparison of the predicted risk with that of previous
it proved impossible to access real data to validate the
models, (2) validation based on real field data or (3)
model. However, this model development contributes
implementation of the model in a case study. Detailed
reports of investigations of sinkhole accidents are significantly to existing knowledge. Therefore, it can
extremely rare, as the relevant authority’s primary goal is be validated and improved with collaboration from
to restore the infrastructure as quickly as possible to industry partners in a future study.
minimise the public, economic and social inconvenience. • Ultimately, this allows defective pipes to be ranked
General information about the events is usually published based on the risk of erosion void formation and
in media as news reports. All the attempts to access some consequences which assist to organise a
field data to validate the model from local and maintenance and rehabilitation schedule by allocating
international authorities were unsuccessful as the present the priority.
research is a university-based study without any industrial
collaborations.
5 ACKNOWLEDGEMENTS
The ROE model presented in this chapter was developed
following a thorough study of previous research related to Authors would like to acknowledge:
erosion through pipe defects, risk prediction models for • Swinburne University of Technology in supporting this
pipe failures and the experimental investigations available research by offering Swinburne University
in literature. The output of COE is closely aligned with Postgraduate Research Award.
Emilio 2015). Nevertheless, the LOE function needs to be • All the research experts who contributed for the
validated before implementing it in industrial applications. survey to determine the weightage factors for the
model.
4 CONCLUSIONS • Chadwick Geotechnics in providing the financial
support for publishing the article.
Based on the study, the following conclusions were
driven. REFERENCES
• Seven key parameters which affect the likelihood of
erosion void formation are identified. Of those Austroads. 2005. Towards a Nationally Consistent
parameters, the soil type, the size of the pipe defect Approach to Route Numbering. Available from
24
https://www.onlinepublications.austroads.com.au/it Technologies for Urban safety of mega cities in Asia

ems/AP-R224-03 2018]. (USMCA), Phuket, Thailand.
Baah, K., Dubey, B., Harvey, R., and McBean, E. 2015. A Moss, R.E.S. 2013. Applied civil engineering risk analysis.
risk-based approach to sanitary sewer pipe asset National Association of Sewer Service Companies,
management. Science of the Total Environment PACP Assessment Certification Program.,
505: 1011-1017. doi: Charleston: Shedwick Press.
https://doi.org/10.1016/j.scitotenv.2014.10.040. O’Reilly, M.P., Rosbrook, R.B., Cox, G.C., and
Balkaya, M., Moore, I.D., and Sağlamer, A. 2012. Study McCloskey, A. 1989. Analysis of defects in 180 km
of non-uniform bedding due to voids under jointed of sewer pipes in Southern water authority. TRRL
PVC water distribution pipes. Geotextiles and Research Report, Transport and Road Research
Geomembranes 34: 39-50. doi: Laboratory, Berkshire, UK.
http://dx.doi.org/10.1016/j.geotexmem.2012.01.003 Renuka, I.H.S. 2012. Evaluation of ground loosening
. behavior and mechanical properties of loosened
Davies, J., Clarke, B., Whiter, J., and Cunningham, R. sand associated with underground cavities. M.Sc.
2001. Factors influencing the structural deterioration thesis, Department of Civil Engineering, University
and collapse of rigid sewer pipes. Urban Water 3(1): of Tokyo
73-89. Rogers, C.J. 1986. Sewer deterioration studies the
Emilio, C.R. 2015. Criticality and risk assessment for pipe background to the structural assessment procedure
rehabilitation in the city of Santa Barbara sewer in the sewerage rehabilitation manual. Water
system. M.Sc. Thesis, Department of Civil and Research Centre, Swindon SN5 8YF, Great Britain.
Environmental Engineering, California Polytechnic Sato, M., and Kuwano, R. 2015. Influence of location of
State University. subsurface structures on development of
Guo, S., Shao, Y., Zhang, T.Q., Zhu, D.Z., and Zhang, underground cavities induced by internal erosion.
Y.P. 2013. Physical modeling on sand erosion Soils and Foundations 55(4):829-840.doi:
around defective sewer pipes under the influence of http://dx.doi.org/10.1016/j.sandf.2015.06.014.
groundwater. Journal of Hydraulic Engineering State Government of Victoria. 2014. Reformed zones for
139(12): 1247-1257. doi: Doi Victoria. Available from
10.1061/(ASCE)Hy.1943-7900.0000785. https://www.planning.vic.gov.au/policy-and-
Indiketiya, S. 2019. Erosion potential of pipe embedment strategy/planning-reform/reformed-zones-for-
materials through defective sewer pipes. PhD. victoria [accessed 29-07- 2021].
thesis, Department of Civil Engineering, Swinburne United States Department of the Interior. 1996. Pipe
University of technology, Australia, Bedding and Backfilling. In Geotechnical Training
Joseph, S.A., Adams, B.J., and McCabe, B. 2010. Manual No.7. Bureau of Reclamation, Technical
Methodology for Bayesian belief network Service Center, Geotechnical Services, Denver,
development to facilitate compliance with water Colorado.
quality regulations. Journal of Infrastructure Water Research Centre (WRC). 2001. Sewerage
Systems 16(1): 58-65. Rehabilitation Manual. Water Research Centre.
Kaddoura, K., and Zayed, T. 2017. Erosion Void Marlow, UK.
Prediction Model for Sewer Pipelines. In WSA 02-2002-2.3-MRWA Edition. 2002. Sewerage code
International Congress on Underground of Australia, Melbourne Retail Water Agencies
Infrastructure, Water Management and Trenchless edition, Version 1.0 Melbourne, Australia.
Technology, Istanbul. pp. 114-126.
Kuwano, R., Hiorii, T., Kohashi, H., and Yamauchi, K. Yan, J.M., and Vairavamoorthy, K. 2003. Fuzzy Approach
2006. Defects of sewer pipes causing cave-ins' in for Pipe Condition Assessment. In New Pipeline
the road. In 5th International Symposium on New Technologies, Security, and Safety. pp. 466-476
SESSION 2
INSTRUMENTATION
AND MONITORING
IN DESIGN
Keynote Address
Ground truth, control and design of driven piles:

implementing old ways with a new twist
J. P. Seidel1, CP Eng and D. Denes2, CP Eng
Foundation QA Pty Ltd, Managing Director, POBox 4280, Croydon Hills, Victoria, Australia, 3136; email: julian@foundationqa.com
1
Golder Associates Pty Ltd, Foundation/Pile Testing Department, Building 7, Botanicca Corporate Park, 570 – 588 Swan Street,
2
Richmond, Victoria 3121, Australia; email: Ddenes@golder.com.au
ABSTRACT
Piling design and verification is a fraught and risky business. The spread of pile capacity estimates submitted to conference
predictions exercises is often staggering and sobering. This underlines why design of driven piles does not stop at the
design engineer’s desk but continues through construction, and relies on the valuable information provided by the
installation process. Each installation blow is a test - a test of the ground response to hammer input delivered into the pile.
Traditionally, pile capacity has been interpreted from this input-response relationship through various and many pile driving
formulae. Five decades ago, measurement systems were first used to measure and interpret the stress waves in piles
generated from the hammer inputs and reflected from the ground response to infer capacity in a more sophisticated and
reliable way using wave mechanics principles. Today, PDA testing and wave matching are routinely accepted practice.
However, each PDA test has direct relevance only to the individual pile which is tested. This paper will argue that our
fundamental task as designers and supervisors is to establish ground truth, by synthesizing the results of PDA tests into a
locally-evidenced and locally-targeted dynamic formula. Therefore, only dynamic formulae, properly modified and
correlated, must be the vehicle for delivering local ground truth and ultimately being the basis for sign-off. On a foundation-
wide basis, the role of PDA tests is critical but subservient, and principally to provide the evidence on which a correlated
dynamic formula is developed. Consequent implications for the foundation sign-off process, and for a proposed new
approach to establishing capacity reduction factors for driven piles will also be discussed.
Keywords: pile driving formulas, pile acceptance, PDA testing, capacity reduction factors, wave equation analysis
1 INTRODUCTION Lawton et al., 1986; Allin et al., 2015) amongst many

others. A particularly good summary can be found in
Piling design is a fraught and risky business. The spread Groom, (2019).
of pile capacity estimates submitted to conference
predictions exercises is often staggering and sobering. The underlying assumption of any pile driving formula is
Fellenius (2013) summarizes the results of such a that it should reflect a universal ground truth and by
prediction exercise for a continuous flight auger (CFA) pile application of this formula, pile capacity can be inferred
installed in clay till with sand and gravel lenses. The by measurement or assumption of some key parameters.
predicted load-movement responses from 41 invited The many studies which assessed the comparative merits
foundation engineers are shown in Figure 1 with the of different formulae by comparison with static load tests
actual load test results. The test had to be terminated all indicated large scatter and poor reliability which
prematurely for safety reasons and two possible resulted in recommended factor of safety of up to 6 (for
extrapolations are shown in the solid lines. 35 predictions the Engineering News Record formula).
included a full load-settlement curve; 6 provided only a
capacity estimate at 50mm movement indicated by the However, it should be considered that at the time of many
black circles. The average capacity prediction was of these comparison studies, the effect of pile setup could
1920kN, with the range of predictions between 700kN and not be assessed, and the true energy delivered by pile
in excess of 5000kN. For the purpose of this paper, the driving hammers could not be measured. These are
issue is not how well the predictions compare with the critical factors which the authors of early studies did not
measured (and extrapolated) test response, but the have at their disposal, and would undoubtedly have
extreme range of predictions of both pile capacity (7-fold) influenced their finding
and pile stiffness (8-fold).
This paper is concerned with driven piles, and in contrast

to CFA piles, there is another opportunity to evaluate
capacity for individual piles based on monitoring of key
installation characteristics. The potential uncertainty of
geotechnical design underlines why design of driven piles
does not stop at the design engineer’s desk but continues
through construction, and relies on the valuable
information provided by the installation process.
For pile driving, each installation blow is a test. A test of

the ground response to hammer input delivered into the
pile. Traditionally, pile capacity has been interpreted from Figure 1. Load-movement predictions from Fellenius
this input-response relationship through various and (2013)
many pile driving formulae (e.g. Hiley, 1930; Chellis,
1961; Olson and Flaate, 1967; Fragaszy et al., 1989;
27
Ground truth, control and design of driven piles: implementing old ways with a new twist J. P. Seidel and D. Denes
Fifty years ago, electronic systems were first used to subsequent argumentation regarding the role of dynamic
measure and interpret the stress waves in piles generated pile formulae.
from the hammer inputs and reflected from the ground
response to infer capacity in a more sophisticated and In the analysis phase of our pile design case, activities
reliable way using wave mechanics principles (Rausche include:
et al., 1972; Goble et al., 1975). Today, PDA testing and • Desktop studies
wave matching are routinely accepted practice • Site investigation
(Hannigan, et al., 2016). But prior to 1985, pile driving • Insitu testing
acceptance was based only on static load testing and pile • Laboratory testing
driving formulae.
The synthesis phase activities include:
Our fundamental task as designers and supervisors is to • Development of simplified site stratigraphy
ensure that capacity and integrity of each pile installed • Assigning characteristic properties to the layers
meets the demands of the structure that it supports. To • Selecting and applying a pile design method
do this we need to establish a ‘ground truth’, albeit a
• Considering the structural loads to establish a
locally-evidenced and locally-targeted ground truth
design pile size and length
capable of application to all piles in a foundation system.
Dynamic formulae, properly modified and correlated,
Let’s suppose that the pile is a floating pile in a deep clay
must be the vehicle for delivering local ground truth and
layer, and that the design method adopted is the
ultimately being the principal basis for sign-off.
𝛼𝛼 method, in which the pile adhesion, 𝜏𝜏 at any depth is
computed based on the local undrained shear strength,𝑐𝑐�
It is important to remember that each PDA test has direct
and an adhesion factor, 𝛼𝛼
relevance only to the individual pile which is tested. On a
foundation-wide basis, the role of PDA tests is critical but 𝜏𝜏 � 𝛼𝛼𝛼𝛼� (1)
subservient, and principally to provide the evidence of
ground truth on which a correlated dynamic formula is Figure 2, reproduced after Coduto (1994), is typical of
founded. design charts which relate 𝛼𝛼 to 𝑐𝑐� . The Coduto curve is
similar to the design chart which was included in the
Consequent implications for the foundation sign-off Appendix to AS2159 (1978).
process, and for a new approach to establishing capacity
reduction factors for driven piles will also be discussed. Such design curves are often used by designers in the
belief that they are based on some universal truth,
2 ANALYSIS AND SYNTHESIS unaware of the significant scatter which lies behind this
design relationship.
The engineering design process can be generalized to
comprise two phases Figure 3, includes the 124 test results from which the
Coduto relationship was developed. Based on the
1. An analysis phase which includes collecting reference tests alone, it can be seen that if a designer
evidence, organization and processing of data uses this design curve, they could easily overestimate or
2. A synthesis phase in which the data collected and underestimate adhesion by a factor of 2.
analysed is transformed into a connected design
Elsewhere analysis is defined variously as “The process

of studying or examining something in an organized way
to learn more about it, or a particular study of something”
(Cambridge, 2021) or “The process of separating
something into its constituent elements” (Oxford, 2021).
Similarly, synthesis is defined variously as “The act of

combining different ideas or things to make a whole that
is new and different from the items considered separately”
(Cambridge, 2021) or “The combination of components or
elements to form a connected whole” (Oxford, 2021).
2.1 Geotechnical design Figure 2. Adhesion factor α after Coduto (1994)

The geotechnical design process is no different, and The scatter has an equivalent coefficient of variation
proceeds through the same analysis and synthesis (COV) of 29% about the best-fit curve, which is within the
phases. We explore, observe, sample and test (analysis) bounds of normal expectation for soil investigations.
and then draw all this together as best we can in the Kulhawy and Phoon (2002) suggest typical COVs of 10 to
design process, trying to transform the diverse 30% for laboratory UU triaxial tests; 10-40% for field vane
information into simple models which are amenable to shear tests and 25 to 50% for SPT N tests, amongst
solution. others.
In this section, we will consider the design for a project to The development of this design curve is the embodiment
be supported on driven piles, and the process of of the synthesis process, in which the engineer
designing a single pile to support a column load. necessarily has to reduce the diverse test results into a
design algorithm, despite the up to four-fold range of the
The purpose of this section is only to provide an analogy data on which the algorithm is based.
to a process that is well understood as a basis for
28
type, then that will further increase the reliability of

prediction.
Figure 3. Data sets for adhesion factor α after Coduto

(1994)
Coduto is only one of many researchers who have

attempted to determine the relationship between 𝛼𝛼 and Figure 5. Prediction of adhesion taking into account
𝑐𝑐� . Figure 4 compares the Coduto recommendations to OCR and plasticity index (IP)
other published relationships (Cherubini and Vessia,
2007). It would be impractical for all but the most significant
projects to undertake a program of static load tests to
develop site-specific relationships. On the other hand,
dynamic load tests conducted adjacent to bore locations
provide a practical opportunity to develop the basis for
evaluation of site-specific or region-specific 𝛼𝛼 values. Our
design-and-construct piling subcontractors have been
doing this for decades in order to ensure that they can
provide efficient and cost-effective designs. Of course,
such an approach must be based on thoughtful, and
geotechnically-informed Wave Equation analysis.
Figure 6 is a simulation of the sort of data set which might

be generated for a set of tests within a particular
geological formation and for a specific pile type. The
coefficient of variation of this data set is now less than 5%.
Figure 4. Comparison of published adhesion factor α
The grey zone represents the 95% confidence limit
relationships
bounds. The lower bound aligns with the definition of
characteristic value adopted in Eurocode EC7 (Bond et
It is clear that other researchers, investigating the same al., 2013).
relationship but in other geologies and with other pile
types have determined and propose distinctly different
empirical relationships. This demonstrates that empirical
relationships are not universal, but are limited to situations
in which the prevailing conditions match the critical
parameters on which that relationship are based.
Various authors have identified that 𝛼𝛼 is not only

dependent on undrained shear strength, but also a range
of other physical properties such as clay mineralogy,
Atterberg limits, moisture content, overconsolidation ratio,
pile material, geometry, displacement ratio, and
installation method.
Improved predictions of 𝛼𝛼 are possible if some of the Figure 6. Simulated adhesion relationship for specific
influencing factors are taken into account. Thus Nowacki geology and pile type
et al. (1992) proposed the design chart shown in Figure
5. More sophisticated interpretations of data sets can 2.2 Key findings
lead to more reliable assessments with reduced
uncertainty. To summarize, the key findings (kf) from this review are :
Putting aside whether the 𝛼𝛼 method is the most suitable kf1. Geotechnical design relationships are generally
design method, it is clear that the variability of results will based on data sets with wide scatter
reduce as the number of variables decreases. By kf2. Geotechnical engineers using these
reducing the scope of a relationship to a specific site, or a relationships must understand and account for
small region within the same geological setting, the the inherent uncertainty in their design (with
possibility of developing a more reliable relationship appropriate reduction factors)
increases. Further, if that data is collected for a single pile kf3. Empirical relationships are by nature not
universal, but are specific to the particular
29
conditions existing at the sites used for data 𝐸𝐸𝐸𝐸𝐸𝐸

collection 𝑅𝑅� � (3)
kf4. The reliability of design recommendations can 𝜒𝜒. �𝑠𝑠 � 𝐶𝐶 2�
be improved by reassessment of data sets The empirical factor 𝜒𝜒 (a correction factor which
taking into account critical influencing factors correlates the formula to PDA or Wave Equation
kf5. The reliability can also be improved by estimates of static resistance, 𝑅𝑅� ) is traditionally assumed
developing local relationships for a specific to be constant, however, as will be shown, it can be any
geological formation and particular pile type. empirical function. Although the genesis of Eqn. (3) is
kf6. Statistical analyses can be used to evaluate theoretical, the use of a correlation factor implies a
confidence limits and to determine fundamental reliance on empiricism.
characteristic values for design. This is
fundamental to developing realistic The comparison between Hiley Formula predictions
geotechnical reduction factors (according to Eqn (2)) and static load tests reported by
kf7. Even in the face of significant variability and Fragaszy et al., 1989 are shown in Figure 7. The authors
uncertainty, engineers necessarily synthesize report that of the 103 tests, 38 were rejected because of
the individual test results into a unifying incomplete data and 2 were rejected because the piles
relationship (design method) so that the test were broken. The remaining 63 usable tests included 6
data can be extrapolated to locations not timber, 20 prestressed concrete, 5 H-section, 4 pipe
specifically tested (open and closed), 7 concrete-filled pipe, 5 hollow
kf8. The synthesis of test results is fundamental to concrete, and 16 Raymond step taper piles. Included in
the design process these tests were 41 piles driven in cohesionless soil, 11
in cohesive soil, and 11 where the sub-surface conditions
2.3 Dynamic Formulae and Pile Acceptance consisted of layers of both cohesive and cohesionless
soil.
As noted, the purpose of the previous section was only to
provide an analogy to a process that is well understood In the case of Figure 7, the 63 individual test comparisons
as a basis for argumentation in this section regarding the are shown against reference line of 1:1 correspondence
role of dynamic pile formulae in regard to pile acceptance. – the ‘design’ line
Fragaszy et al., 1989 reviewed the reliability of 10 𝑅𝑅��
common pile driving formulae, including the Hiley �1 (4)
𝑅𝑅��
Formula, in a study of 103 pile tests conducted in Oregon
and Washington states by consultants and their Supplementary lines are also shown for factors of 2 and
departments of transport. 3 overestimation and underestimation for reference.
Because of the traditional use of the Hiley Formula in This design line shown in Eq (4) and Figure 7 is equivalent
Australia, only the Fragaszy results in relation to this to the 𝛼𝛼 design line in Figure 3, and demonstrates the
formula will be discussed here. same key findings kf1, kf2 and kf3 (see Section 2.2).
The Hiley formula is based on a simple theoretical model Although on average the predictions have low bias (the
of pile driving characterized as the inelastic collision with average ratio of Hiley to static capacity is 0.985), the
coefficient of restitution, 𝑛𝑛 of 2 masses (the hammer, 𝑊𝑊� scatter is very wide which is reflected by the coefficient of
which drops from a height ℎ with drop efficiency 𝑒𝑒� and variation of 72.6%.
the pile and helmet, 𝑊𝑊� ). The pile experiences a
permanent set, 𝑠𝑠 , and three components of transient Given that the basic premise of monitoring the installation
movement, 𝐶𝐶. Equation (2) predicts the pile capacity, 𝑅𝑅� . of piles is to reduce the uncertainties inherent in pile
design, such a high coefficient of variation undermines the
𝑒𝑒� 𝑊𝑊� ℎ 𝑊𝑊� � 𝑛𝑛� 𝑊𝑊� very premise on which pile monitoring is based.
𝑅𝑅� � . (2)
𝑠𝑠 � �𝐶𝐶� � 𝐶𝐶� � 𝐶𝐶� � 𝑊𝑊� � 𝑊𝑊�
The basis of this equation is to estimate the striking
hammer energy, determine the energy transferred to the
pile, and then to equate that to the work expended in the
transient and permanent movements of the pile. With the
exception of the Gates Formula (Gates, 1957), pile driving
formulae are based on some simplified energy balance
equation.
That being said, two key parameters in the Hiley Formula

– drop efficiency, 𝑒𝑒� and coefficient of restitution, 𝑛𝑛 can
only be determined empirically1.
Subsequent modifications of the traditional Hiley Formula,

which substitute PDA-computed energy transfer, EMX,
have been proposed by Broms (1989) and Paikowsky and
LaBelle (1994). These modifications were well
established in general practice prior to these publications, Figure 7. Hiley capacity vs Static capacity (after
but with the addition of a correction factor, 𝜒𝜒 as shown in Fragaszy et al. 1989)
Equation (3)
1
And in practice are generally assumed rather than measured
30
Fragaszy et al. do not further analyse the data of Figure 7 Ramey and Johnson (1978) report a data set of 153 piles
in order to determine whether the spread is affected by comprising steel-H, steel pipe, concrete and timber piles.
pile type, hammer type or soil type. Five dynamic formulae, including Hiley Formula, are
compared with static tests. All things being equal, one
The difficulties of comparing end of drive installation data would expect the Hiley/Static capacity ratio to be relatively
(on which driving formulae are based) with static load highest for the piles installed in cohesionless soils (which
tests undertaken some days or weeks later have been are generally not known for setup), and lowest for the
discussed elsewhere (Seidel, 2015a). The phenomenon piles exclusively in cohesive soils (for which end of drive
of pile setup is well documented and is particularly capacity may be significantly less than the capacity
associated with post-installation pile capacity increases in measured later during static load testing).
cohesive soils (Skov and Denver, 1998; Lee et al., 2010).
Figure 8. Hiley capacity vs Static capacity after Ramey and Johnson (1978)
Figure 8 shows that the spread of data is similar to the This resistance comprises two parts – a static component,
Fragaszy data set. However, the data broadly supports 𝑅𝑅� , and a transient dynamic component 𝑅𝑅� . thus
the expected differentiation of test results by soil type,
𝑅𝑅� � 𝑅𝑅� � 𝑅𝑅� (5)
with the spread of cohesive soil data generally being
below the non-cohesive data, and showing the influence As discussed in Seidel (2018a), the relative proportion of
of pile set-up. Key findings kf 4 and kf 5 are demonstrated static and dynamic components of total driving resistance
in Figure 8. vary with the ease of driving. For hard driving, with low
set, the total resistance is dominated by the static
As noted in regard to the design parameter, 𝛼𝛼 insightful component. As driving becomes progressively easier
consideration of the influencing factors (in that case (increasing set), the proportion of dynamic resistance
overconsolidation ratio and plasticity index), can reduce increases as indicated schematically in Figure 9.
uncertainty.
Referring to Equation (3), the correction function, 𝜒𝜒2 will
In order to further reduce the uncertainty in the application be relatively small at low sets, and progressively increase
of dynamic formulae, it is important to understand that as pile sets increase. 𝜒𝜒 can either be evaluated by wave
dynamic formulae do not predict static pile capacity but equation analysis prior to piling, or evaluated empirically
rather predict the total resistance of the soil to driving, 𝑅𝑅� . from dynamic pile testing, particularly at the
commencement of a project.
2
𝜒𝜒 is referred to as Dynamic Reduction Function (DRF) in Seidel (2018)
31
analyzed, which results in a coefficient of variation of only

1%, and a very tight 95% confidence limit band.
Figure 11. Correction factor function 𝜒𝜒 computed by

GRLWEAP analysis
Figure 9. Schematic representation of resistance
components during pile driving It is worth reinforcing that by understanding fundamentally
the factors underlying the variability of Hiley Formula
predictions (see Figure 7) it has been possible to develop
The predicted relationship between Hiley Formula and
a definitive correction factor and to adopt characteristic
static capacity, was evaluated using the Wave Equation
values which are only marginally below the best-fit
program GRLWEAP (Rausche et al., 2004). for a near-
relationship for the data (see Figure11). On the other
shore piling project with a hydraulic hammer and steel
hand, it is noted that the correction function computed
pipe piles driven through bed sediments to refusal on
here is theoretical. In practice, the data spread will
variably weathered rock.
inevitably be greater due to accuracy of set
measurements (Denes et al. 2021), quality of PDA
Figure 10 compares the uncorrected Hiley formula
testing, uncertainty in wave equation matching (Seidel,
capacities inferred from the GRLWEAP predicted values of
2015a) as well as inherent geotechnical variability. The
transferred energy, set and temporary compression with
confidence of any relationship will be reflected in the
the analysed static capacities. The Hiley formula shown
quality of these inputs. Figure 11 demonstrates the same
in Equation (3) was used with 𝜒𝜒 � 1. Hammer strokes of
key findings kf4, kf5 and kf6 (see Section 2.2).
between 0.6m and 1.5m were used in the analyses.
Figure 12. Correction factor function 𝜒𝜒 computed

from project PDA data
Figure 12 presents correction function data for a large
Figure 10. Comparison of Hiley Capacity and Static project. Each data point represents the wave matching
Capacity predicted by GRLWEAP solution for a PDA test. The spread is seen to be larger
than for Figure 11, but a distinct trend is still shown. As 𝜒𝜒
The similarity between Figures 7 and Figure 10 is noted. is on the denominator of Equation (3), the upper bound of
The Hiley formula overpredicts capacity by a factor of up the confidence limit should be adopted as the
to 3 at low capacities. Unlike the traditional Hiley Formula characteristic value line.
of Eqn (2), the modified Hiley Formula of Eqn.(3)
generally does not underpredict capacity. It is evident that The data represented in Figure 12 has a very low
the degree of overprediction progressively decreases with coefficient of variation (about the trendline) of only 4.2%.
capacity for each stroke, and approaches the equality line This was achieved with remote non-contact set
(Eqn. 4) at higher capacities. measurements and with high quality PDA testing and
mindful (non-automatic) wave equation matching. The
It is clear that the ratio of Hiley capacity to Static capacity consequence is the potential for higher capacity reduction
is not constant but varies between 1 and 3 even for a factors (as discussed in the next section).
single project with defined soil model, pile type and
geometry and hammer. Figure 11, plots the correction Figure 12 demonstrates the same key findings kf7 and kf8
factor, 𝜒𝜒 as a function of pile set. There is a compelling (see Section 2.2). The PDA results are only relevant to
dependency of 𝜒𝜒 on set across all hammer strokes the particular piles tested. It is necessary to synthesize
32
these results into a unifying relationship from which a pile 3.1 Pile acceptance based on energy, set and
acceptance criterion can be developed. Locally temporary compression
correlated driving formulae provide the ground truth, and
are fundamental to providing the basis for acceptance of Despite the certainty provided to the evaluation of 𝜒𝜒 ,
any untested pile. there still remain challenges in evaluation of static
capacity, 𝑅𝑅� , in Equation (3) which is reproduced here for
3 PILE ACCEPTANCE convenience.
𝐸𝐸𝐸𝐸𝐸𝐸
The current practices in Australia for driven pile 𝑅𝑅� = (3)
acceptance vary considerably. Generally, piles are driven 𝜒𝜒. (𝑠𝑠 + 𝐶𝐶 2)
to a capacity rather than a penetration criterion. What
might be considered a normal practice is that a There are a range of techniques for measurement of pile
percentage of piles are PDA tested and matched with set and temporary compression. The possibilities, their
wave equation analysis, and this percentage defines the benefits and limitations are discussed in Denes et al.
adopted capacity reduction factor, 𝜙𝜙� in accordance with (2021).
the provisions of Section 4 of AS2159 (2009).
However, the most significant challenge to
AS2159 (2009) introduced the concepts of a basic implementation is 𝐸𝐸𝐸𝐸𝐸𝐸, the energy transferred to the pile.
geotechnical reduction factor, 𝜙𝜙�� , an intrinsic test factor, Although some modern hammers measure and report
𝜙𝜙�� , and a testing benefit factor, 𝐾𝐾, which allowed the kinetic impact energy, 𝐾𝐾𝐾𝐾, these represent only a small
capacity reduction factor, 𝜙𝜙� to vary depending on the percentage of the existing hammer fleet. Even for these
percentage of piles tested, 𝑝𝑝.3. For a typical project with hammers, the challenge still remains to evaluate the loss
low perceived risk category and low redundancy 𝜙𝜙�� = of energy, 𝐾𝐾𝐾𝐾𝐾𝐾𝐾𝐾𝐾𝐾𝐾, through the helmet, particularly for
concrete piles, for which the pile cushion can be highly
0.56, the computed capacity reduction factors are as
variable in thickness and stiffness, especially when the
indicated in Table 1:
cushion properties change during driving.
Table 1: AS2159 (2009) Effect of increasing Flynn and McCabe (2016) compare 𝐸𝐸𝐸𝐸𝐸𝐸 and 𝐾𝐾𝐾𝐾
percentage of PDA tests measurements for hydraulic hammers on 5 projects, and
𝑝𝑝 (%) 3 5 10 25 demonstrate the uncertainties with inferring 𝐸𝐸𝐸𝐸𝐸𝐸 from
𝜙𝜙� (−) 0.69 0.72 0.76 0.80 hammer energy measurements, as shown in Figure 13.
The Flynn and McCabe data indicates that the percentage
Table 1 suggests that by the very act of testing, a reduced of hammer energy delivered to the pile typically varied
risk is imparted to the pile foundation system. The between 75% and 100%, but in isolated instances was as
benefits of testing are manifest, and higher percentages low as 50%. This variability is just the transfer variability.
of testing will increase the likelihood of detection of any The variability of the hammer stroke and kinetic energy is
piles with insufficient capacity. However, there is an not even included in this assessment.
implied assumption in AS2159 (2009) that the installation
of the 75% to 97%4 of piles not tested is informed by the For uninstrumented hammers, transferred energy can
results of the tests. This is generally not true, or is not only be estimated on the basis of estimated ram drop
implemented in a rigorous way. height, assumed hammer drop efficiency, and an
assumed loss of energy in transferring from the hammer
One common practice is to treat the untested piles as to the pile. As just noted, this can be particularly
unrelated to the tested piles. The tested piles must challenging for concrete piles.
demonstrate a capacity in excess of the contractual
requirement. The untested piles must pass a predefined Seidel (2018b) presents the energy measurements for a
acceptance criterion in the specification, such as the project in which 155 PDA tests were undertaken on
traditional Hiley Formula. In this approach, the valuable prestressed concrete piles over the extent of a large
learnings of the PDA tests do not inform acceptance of bridge project. The project is distinguished by the
the untested piles. requirement for all piling and all PDA tests to be
undertaken with a drop height of 0.5m. Despite this
The second common practice is to adopt a maximum set control on drop height, delivered energies averaged
criterion which corresponds to the set measured for a 69.2kJ (88% efficiency), but with a range of 37.5 to 91.0kJ
tested pile with the target capacity, or inferred if the tested (48% to 116%5 efficiency). The coefficient of variation of
capacity exceeded the target. This implicitly assumes EMX was 15.4%. The three projects that the paper details
that delivered energy efficiency is constant and that the demonstrate energy ranges of 36% and 40% for the steel
set-capacity relationship is fixed with no variation. We will pile projects, and 77% for the concrete pile project.
see in the next sections that neither of these assumptions
is correct. Li et. al., (2022) report in regard to another case study
“The driving hammer energy efficiency recorded
It will be shown (see Figure 14) that transferred energy variations of up to 40%, casting doubt on the use of
cannot be assumed to be constant. Figures 10 to 12 achieved set as a means of inferring pile capacities”.
demonstrate that the set-capacity relationship is not
constant. Any acceptance criterion must take into The histogram in Figure 14 presents the sequential EMX
account the demonstrated variability of both the energy values for 51 of the 155 PDA tests.
delivered and of the set-capacity relationship.
3 5
Up to a limit of 𝑝𝑝 = 25(%) of piles PDA tested Efficiency greater than 100% is physically impossible, indicating that the
4 true drop height was significantly in excess of 0.5m
i.e. (100 – 𝑝𝑝)%
33
Figure 13. Relationship between kinetic energy at impact and energy

transferred to pile
Figure 14. Longitudinal study of variability of delivered energy
The only possibility for individual pile energy implied geotechnical reduction factors based on a
measurements6 is from high frequency remote non- transferred energy uncertainty of 5%7.
contact displacement measurements. Peak pile
velocities can be computed from differentiating the It is noted that the 𝜙𝜙� value of 0.70 is derived exclusively
displacement-time signal. It is possible to identify low from a statistical assessment of the pile testing results
energy blows from low velocity measurements. Although and hammer performance and is independent of any
EMX is related to the square of peak velocity VMX, the assessment of the basic geotechnical reduction factor,
relationship is affected by variations in cushion stiffness, 𝜙𝜙�� . This is a logical approach, because assessment of
so that no unique and reliable energy can be computed in the installation process should be independent of any of
most cases. the risk factors considered in developing project average
Figure 15 is a representation of Figure 12, with upper and risk rating, 𝐴𝐴𝐴𝐴𝐴𝐴 and 𝜙𝜙�� in Table 4.3.2(C) of AS2159
lower 95% confidence limits shown. Also shown are the (2009).
6 7
Other than PDA testing or attachment of an accelerometer to each pile, The uncertainty in transferred energy evaluation, either from
both of which have practical issues instrumented hammer or pile velocity measurements should be
determined from each case.
34
Figure 15. Confidence limits and inferred geotechnical reduction factors for Figure 11 data
3.2 Pile acceptance based on energy and set Three key points are shown on the graph at blow counts
of 67, 200 and 870 blows/m (15mm, 5mm and 1.15mm
Pile acceptance methods based on energy and set alone set). What this graph shows is that for any given energy,
are fundamentally premised on use of bearing graphs the capacity at 15mm set will be 50% less than the
correlated to field test results. Bearing graphs represent capacity at 5mm set, and the capacity at 1.15mm set will
the relationship between capacity and pile set, which is of be 50% more than the capacity at 5mm set. This
course dependent on hammer energy. Seidel (2015b) relationship is unique to this data set, but equivalent
and Denes et al. (2021) discuss the importance of relationships will hold for any site.
Energy-Capacity-Movement (ECM) relationships.
Figure 16 compares bearing graphs for 5 different

hammer strokes. The inferred capacities at a blow count
of 200 blows/m (5mm set) are shown for each stroke,
ranging from 3525kN at 0.4m stroke to 5650kN at 0.8m
stroke. Clearly, capacity can only be known if the
delivered energy is known.
Figure 17. Normalized bearing graphs with

normalization to capacity at nominated blow count
The benefit of this approach is that it does not require

measurement of temporary compression.
However, all energy-based methods require

determination of the energy delivered to the pile. The
Figure 16. Traditional bearing graphs for a range of same comments regarding variability of transferred
hammer strokes / delivered energies energy and the reliability of EMX estimates discussed in
Section 3.1 apply equally to methods based on energy
and set alone.
As shown by Seidel (2018b) these bearing graphs can be
normalized by the capacity at a specific blow count (in this 3.3 Pile acceptance based on force and set
case 200 blows/m). All the curves collapse to a near-
unique relationship as shown in Figure 17, which is an Pile driving formulae are almost universally some
interesting finding. expression of a Energy-Capacity-Movement (ECM)
relationship, with assumed or measured energy being a
35
fundamental input. As discussed, this is problematic,

because of the potential variability of hammer
performance, and the uncertainties involved in estimating
energy transferred to the piles, even for instrumented
hammers (Flynn and McCabe, 2016).
The Gates pile driving formula (Gates, 1957) is an

empirical formula which is based not on energy, but the
square root of energy, which leads to an equation which
is not dimensionally correct. Regardless, the Gates
formula, or derivatives of the Gates formula8 continue to
show comparatively the best performance in comparative
studies in the United States where it is in common usage
(Long, 2009).
To better understand why a non-dimensional equation Figure 18. Gates blow-count capacity modifier
performs well, consider Equation (6) for kinetic energy 𝐾𝐾𝐾𝐾
of the pile driving hammer with mass 𝑀𝑀� , which strikes
with velocity 𝑣𝑣�� , The purpose of the parametric study was to investigate
the factors affecting normalized capacity with a view to
�
𝐾𝐾𝐾𝐾 = 12 𝑀𝑀� 𝑣𝑣�� (6) developing reliable predictions of the relationship
between normalized capacity9 and pile set. Only
from which it follows that moderate success was achieved, and a reliable prediction
√𝐾𝐾𝐾𝐾 ∝ 𝑣𝑣�� (7) method is yet to be developed.
Similarly, the energy transferred to the pile, 𝐸𝐸𝐸𝐸𝐸𝐸 is Figure 19 shows the results for 4 of the cases
computed from pile-top PDA measurements of force, 𝐹𝐹(𝑡𝑡) investigated. These cases include concrete and steel
and velocity, 𝑣𝑣(𝑡𝑡) as piles, and span ratios of (a) ram weights of 10.4, (b) pile
impedances of 3.5; (c) damping factors of 3.4 and (d)
𝐸𝐸𝐸𝐸𝐸𝐸 = � 𝐹𝐹. 𝑣𝑣. 𝑑𝑑𝑑𝑑 (8) quakes of 2.1. Despite these large variations in hammer
One dimensional wave mechanics provides the following and pile parameters, the indicated normalized capacity-
relationship which holds between F and v at the set relationships have a limited range. Furthermore, in
commencement of the dynamic event when no reflections practice, the range of this relationship appears to be more
from the soil have reached the pile head. limited than indicated by Figure 19, because the
parametric study included hammer/pile combinations
𝐹𝐹 = 𝑍𝑍. 𝑣𝑣 (9) which could be considered extreme and outside the
bounds of practical experience.
where 𝑍𝑍 = 𝐸𝐸𝐸𝐸𝐸𝐸 is the pile impedance, a function of the
pile modulus, 𝐸𝐸, pile cross-section, 𝐴𝐴 and pile wavespeed, Regardless of whether it will be possible to reliably predict
𝑐𝑐. these relationships in advance with a simple algorithm,
predictions for any project can be made in advance using
From Equations (8) and (9) it follows that there is an wave equation analysis.
(imperfect) relationship between √𝐸𝐸𝐸𝐸𝐸𝐸 and pile-top
velocity, 𝑣𝑣, and pile-top force, 𝐹𝐹.
The Gates formula therefore effectively expresses a

relationship between capacity and force or velocity. The
FHWA Gates formula is
𝑅𝑅� = 1.75�𝑒𝑒𝑒𝑒� log(10𝑁𝑁� ) − 100 (10)
Without detailing all the components, the influence of blow

count, 𝑁𝑁� in blows/inch on modifying the estimate of static
capacity, 𝑅𝑅� , is shown in Figure 18 which has been
shows the relationship in equivalent mm set units.
Seidel (2018c) presented results from a parametric study Figure 19. Wave Equation study of Normalized
of 194 Wave Equation simulations involving 20 different Capacity – Set relationship
piling hammers ranging from 7 to 168 tonne ram weight;
25 concrete or steel pile sections varying from 0.2m
Furthermore, any predicted relationship should be
square concrete to 3m diameter and 80mm wall thickness
calibrated against the results of testing once piling
steel; cushion stiffnesses, short and long piles, and
commences. It is recommended that for the initial pile
varying resistance quakes, damping factors and
installations, PDA testing and wave matching analyses be
resistance distributions. The study therefore ranged
undertaken at a range of pile sets – for instance at 10mm,
across an extreme range of piling scenarios.
5mm and 2mm set per blow.
8
For example FHWA-Gates, modified FHWA-Gates or WSDOT formulas
9
Capacity normalized by the impact force, 𝑅𝑅� 𝐹𝐹��
36
Figure 20 is an example drawn from a project covering practical, however, peak impact velocity, 𝑣𝑣�� , can be
multiple dispersed bridge sites. The data includes both used as a proxy for 𝐹𝐹��
end of drive and restrike tests, but is constrained to a using Equation (11).
single pile geometry (1200mm diameter steel pipe piles)
and hammer system. 𝐹𝐹�� = 𝑍𝑍. 𝑣𝑣�� (11)
Figure 20 also shows the upper and lower 95% It is possible to measure 𝑣𝑣�� with high frequency remote
confidence limits, and the corresponding geotechnical pile measurement devices (see Denes et al., 2021).
reduction factors, 𝜙𝜙� , which vary depending on pile set
and the spread of data. For typical acceptance criteria The basis for the popularity of the Gates formula can be
with sets below 5mm /blow, 𝜙𝜙� is in excess of 0.7 and as seen in the similarity of the capacity-set relationships
much as 0.74. shown in Figures 18, 19 and 20. The empirical success
of the Gates formula reinforces the essential validity of a
Of course, the relationship is premised on measurement force-based, but locally correlated acceptance criterion.
of impact force, 𝐹𝐹�� , for every (untested) pile. This is not
Figure 20. Normalized Capacity – Set relationship for multi-bridge project
In order to undertake this successfully, the PDA tests

4 CONCLUSIONS must be of high quality and the wave matching must be
conducted in a geotechnically relevant and thoughtful
Pile driving formulae have long been considered an approach.
unreliable approach to evaluation of pile capacity.
Although this view may have had historical merit, it The paper has described three alternative approaches to
misunderstands the central role that pile driving formulae implementation of pile acceptance criteria – two based on
have in modern pile driving control. traditional energy-based approaches and a third based on
measurement of pile impact velocity (being a proxy for
Pile driving formulae are not universal truths, but are impact force).
empirical formulations which have relevance only to the
particular set of conditions (piling equipment, pile and soil) It has been shown here and elsewhere that the hammer
on which they are based. energy delivered to piles can be extremely variable. The
greatest challenge to implementation of the energy-based
In the modern context, pile driving formulae should have approaches is the evaluation of delivered energy to the
a central role in being the synthesis of the results of PDA pile for every pile. On the other hand, the alternative
testing and wave matching analysis. The pile driving based on measurement of pile impact velocity (and pile
formula, locally correlated against the test results is an set) is practically attainable.
expression of the local ground truth, and provides the
basis for extension of the individual test results to all piles Finally, it has been shown that these approaches provide
in the foundation system. a logical pathway to determining realistic capacity
reduction factors, based on the reliability of the adopted
37
empirical pile acceptance relationship, the reliability of the capacity. Transp. Res. Record No. 1105: Struct.
energy, force and movement measurements and the Foundations, Transp. Res. Board, pp.32-40.
number of tests on which the relationships are based. Lee, W., Kim, D., Salgado, R. and Zaheer, M., 2010. Setup of
driven piles in layered soil. Soils and foundations, 50(5),
These approaches do not rely on any evaluation of the
pp.585-598.
basic geotechnical reduction factor defined in AS2159 Li, I., Vinod. M. and Hsi, J. 2022. Case Study: Lessons learned
(2009), which is in itself a logical outcome. from large scale pile driving in Waikeria, New Zealand.
Proc. 20th International Conference on Soil Mechanics and
5 ACKNOWLEDGEMENTS Geotechnical Engineering, Sydney 2022 (submitted for
review).
The authors acknowledge the important contributions of Long, J.H., 2009. Comparison of five different methods for
their ex-colleagues at FSG Geotechnics and determining pile bearing capacities. Wisconsin Highway
Research Program.
Foundations, in particular Pearse Casey, Ben Kroenert,
Olson, R.E. and Flaate, K.S., 1967. Pile-driving formulas for
Nick Medley, Alireza Foroughi and Hossein Ahmadi, to friction piles in sand. Journal of the Soil Mechanics and
the development and refinement of the novel pile Foundations Division, 93(6), pp.279-296.
acceptance methods presented here. Oxford University Press. 2021 Available at:
https://www.lexico.com/definition/wake
Paikowsky, S., and LaBelle, V. 1994. Examination of the Energy
Approach for Capacity Evaluation of Driven Piles.
6 REFERENCES Proceedings of the International Conference on Design
and Construction of Deep Foundations, December 6–8,
Orlando, FL. FHWA, Vol. II, pp. 1133–1149
Allin, R., Likins, G. and Honeycutt, J., 2015. Pile driving formulas
Ramey, G.E. and Johnson Jr, R.C., 1979. Relative accuracy and
revisited. In IFCEE 2015 (pp. 1052-1063)
modification of some dynamic pile capacity prediction
AS2159 (1978) SAA Piling Code. Rules for the design and
equations. Ground Engineering, 12(6).
installation of piling. Standards Australia
Rausche, F., Liang, L., Allin, R. and Rancman, D., 2004.
AS 2159-2009 (2009). Piling – Design and Installation.
Applications and correlations of the wave equation
Standards Australia
analysis program GRLWEAP. In Proceedings, VII
Bond, A.J., Schuppener, B., Scarpelli, G., Orr, T.L., Dimova, S.,
Conference on the Application of Stress Wave Theory to
Nikolova, B. and Pinto, A.V., 2013, June. Eurocode 7:
Piles (pp. 107-123).
geotechnical design worked examples. In Workshop
Rausche, F., Moses, F. and Goble, G.G., 1972. Soil resistance
“Eurocode (Vol. 7).
predictions from pile dynamics. Journal of the soil
Broms, B. and Choo, L.P., 1988. A simple pile driving formula
mechanics and foundations division, 98(9), pp.917-937.
based on stress-wave measurements. In Proc. of the 3rd
Seidel, J.P., 2015a. Overview of the Role of Testing and
Intl. Conf. on the Application of Stresswave Theory to Piles
Monitoring in the Verification of Driven Pile Foundations. In
(pp. 591-600).
Proceedings on the 12th Australia New Zealand
Cambridge University Press 2021. Available at
Conference on Geomechanics (ANZ2015).
https://dictionary.cambridge.org/dictionary/english
Seidel, J.P., 2015b. Enhanced Use of Dynamic Pile Testing in
Chellis, R.D. (1961) ‘Pile Foundations’. 2nd (ed). New York:
Foundation Engineering. In Proceedings on the 12th
McGraw-Hill Book Company, Inc.
Australia New Zealand Conference on Geomechanics
Cherubini, C. and Vessia, G., 2007. Reliability approach for the
(ANZ2015).
side resistance of piles by means of the total stress
Seidel, J.P., 2018a. The normalized bearing graph and dynamic
analysis (α Method). Canadian Geotechnical Journal,
reduction function concepts in pile acceptance. DFI-EFFC
44(11), pp.1378-1390.
International Conf. on Deep Foundations and Ground
Coduto, D.P. 1994. Foundation design, principles and practices.
Improvement, Rome, Italy 6-8 June, 2018.
Prentice Hall Inc., Englewood Cliffs, N.Y.
Seidel, J.P., 2018b. The importance of energy evaluation on an
Denes, D., Foroughi, A. and Seidel, J.P. 2021. Pile Testing
individual pile basis. DFI-EFFC International Conf. on
Verification – an Alternative Approach. Australian
Deep Foundations and Ground Improvement, Rome, Italy
Geomechanics Society Victoria Seminar, Oct. 2021.
6-8 June, 2018.
Fellenius, B.H., 2013. Capacity and Load Movement of a CFA
Seidel, J.P., 2018c. Mobilization of Pile Capacity
Pile: A Prediction Event. In Foundation Engineering in the
and Pile Acceptance. Invited presentation to Australian
Face of Uncertainty: Honoring Fred H. Kulhawy (pp. 707-
Geomechanics Society, Victorian Chapter, 8 August, 2018.
719).
Skov, R. and Denver, H., 1988. Time-dependence of bearing
Flynn, K. and McCabe, B.A., 2016, August. Energy transfer ratio
capacity of piles, Proc. 3rd International Conference on
for hydraulic pile driving hammers. In Proc., Civil
Application of Stress-Wave Theory to Piles, Canada, 1–10.
Engineering Research in Ireland Conf.
Fragaszy, R.J., Argo, D. and Higgins, J.D., 1989. Comparison of
formula predictions with pile load tests. Transportation
Research Record, 1219, pp.1-12.
Gates, M., 1957. Empirical Formula for Predicting Pile Bearing
Capacity, Vol 27, No. 3, March : 65-66
Goble, G.G., Likins Jr, G. and Rausche, F., 1975. Bearing
capacity of piles from dynamic measurements (No. OHIO-
DOT-05-75 Final Rpt.).
Groom, D. 2019. An Assessment of Dynamic Pile Driving
Formulae for use by Military Engineers
Hannigan, P.J., Rausche, F., Likins, G.E., Robinson, B.R. and
Becker, M.L., 2016. Geotechnical Engineering Circular No.
12–Volume II Design and Construction of Driven Pile
Foundations (No. FHWA-NHI-16-010).
Hiley, A., 1930. Pile-driving calculations. The Struct. Eng, 8.
Kulhawy, F.H. and Phoon, K.K., 2002. Observations on
geotechnical reliability-based design development in North
America. In Foundation Engineering in the Face of
Uncertainty: Honouring Fred H. Kulhawy (pp. 142-159).
ASCE.
Lawton, E.C., Fragaszy, R.J., Higgins, J.D., Kilian, A.P. and
Peters, A.J., 1986. Review of methods for estimating pile
The Application of 3D Finite Element Method in the Design of Large Piled
Foundation System - Case Study: Melbourne Cement Facility
K. Ranjbar Pouya1, B. Collingwood2 and A. Judi3

1
FSG Geotechnics and Foundations, Unit 12, 71 Victoria Crescent, Abbottsford, VIC, 3067; email: kranjbar@fsg-geotechnics.com.au
2
FSG Geotechnics and Foundations, Unit 12, 71 Victoria Crescent, Abbottsford, VIC, 3067; email: bcollingwood@fsg-geotechnics.com.au
3
Wagstaff Piling Pty Ltd., 33 Nott Street, Port Melbourne, VIC, 3207; email: a.judi@wagstaffpiling.com.au
ABSTRACT
This paper provides an overview of the foundation design and analysis process carried out for the Melbourne Cement
Facility silo located in Port Melbourne, Victoria, Australia. The proposed silo is a cylindrical multi-compartment cement
storage facility supported on a 2.6m thick concrete ring beam with an external diameter of 38.5m. The ring beam is
supported by a piled foundation system comprising 155 CFA piles in an annular pile layout. The site is underlain by
Quaternary Sediments of Yarra Delta which are further underlain by Werribee Formation. This paper describes a detailed
soil-structure interaction analysis performed using the finite element program PLAXIS 3D, which was used to assess the
foundation performance with particular attention to global and differential settlement of the pile group. The study evaluated
the complex load sharing between the piles and the ring beam, and the differences in load mobilisation between piles
within the group. The results of this study highlight the capability of 3D FEM analysis for obtaining an optimised foundation
design solution and understanding and addressing various technical challenges associated with silo foundation systems
of this type.
Keywords: piled raft, piled foundations, foundation design, finite elements, storage facilities, settlement analysis
1 INTRODUCTION 1991) and plate on springs (Poulos 1994; Viggiani 1998;

Russo and Viggiani 1998). Ta and Small (1996)
Piled raft foundations can be a practical and economical developed the Finite Layer Method for the analysis of pile
solution for very tall budlings and heavy industrial raft foundations resting on layered soil stratum. More
facilities where substantial vertical loads will be recent attention has focused on three-dimensional
transferred to the substructure. The proposed cylindrical analysis of piled rafts (Smith and Wang 1998;
multi-compartment cement storage facility weighs up to Katzenbach et al. 1998; Katzenbach et al. 2005; Lee et
922 MN when all compartments are fully loaded. al. 2010). Recent advances in Finite Element Modelling
Considering the size and loads of such large storage (FEM) have made full three-dimensional piled raft
facilities, the settlement of the foundation system analysis a robust tool to capture the realistic soil-
supporting the structure is often a governing factor in the structure interaction and load sharing behaviour among
design. It has been shown that piles can be successfully the piles and the raft. In this study, the FEM program
used as settlement reducers (Burland et al. 1977; PLAXIS 3D was employed to evaluate the performance
Mandolini et al. 2005) with a contribution from the of the foundation system in regard to both stability and
overlying raft. Moreover, in many cases the primary serviceability requirements. A detailed soil-structure
objective of including piles in the design of pile raft interaction analysis was carried out to evaluate and
foundations is to control settlements as raft alone can understand the contribution of the piles and the ring
provide sufficient bearing resistance (Davis and Poulos beam in the design of the foundation system.
1972; Randolph 1994). However, as Viggiani et al.
(2012) identified in the case of a so-called small pile raft, 2 SITE LOCATION, SILO STRUCTURE AND
the un-piled raft could not solely carry the vertical loads LOADING CONDITIONS
and, in contrast to the case of a large piled raft, the key
requirement for including piles is to satisfy an adequate The cement silo facility proposed for construction by
factor of safety against bearing failure. Melbourne Cement Facility (MCF) is a cylindrical
structure with a number of segmented storage
A considerable body of literature has been focused on compartments and a total height of about 71m. The
developing methods for piled raft analysis. Early structure is located in Port Melbourne, Victoria, Australia.
analytical works for the piled raft analysis were carried The site location and the proposed location of the new
out by Davis and Poulos (1972) and Randolph (1983, silo is shown in Figure 1. The structure has an internal
1994). Davis and Poulos (1972) presented a simplified diameter of about 31.4m and comprises circular
analytical approach to evaluate the role of piles as reinforced concrete walls with thicknesses of around
settlement reducers in combined piled raft systems. The 0.5m to 1.2m.
first series of simplified numerical analyses emerged as
in two distinct approaches of strip on springs (Poulos

41
The Application of 3D Finite Element Method in the Design of Large Piled K. Ranjbar Pouya, B. Collingwood and A. Judi
Figure 1. Site location and the proposed location of the new silo - source: Google Earth
The silo structure is supported on a 2.6m thick annular assessment of the contribution of the piles and ring beam
ring beam which is shown in Figure 2. Loads coming to the foundation performance will be discussed in
from the superstructure will be transferred from the subsequent sections. The overall dead load and live load
cylindrical silo walls to the ring beam and the piles of the silo structure is in the order of 900 MN in total. The
constructed underneath. The foundation system is load critical load cases that were adopted in foundation
designed to perform as a piled raft as the ring beam has modelling and design are summarised in Table 1.
a large dimension (and contact area) and can contribute
to the load-bearing of the system. A quantitative
Figure 2. Silo ring beam geometry
Table 1: Summary of design load cases

Load Case Description Load Composition
Serviceability limit state for global settlement check – full
SLS-1 DL+LL
silo loading
Serviceability limit state for differential settlement check – DL across full silo
SLS-2
live loading on one half of silo only LL on one half of silo
Ultimate limit state for maximum vertical loading based on
ULS-1 1.2 DL + 1.5 LL
AS 1170 load combination
Ultimate limit state for earthquake loading based on dead &
ULS-2 DL + LL + EQ
live vertical loads, shear force and overturning moment
Serviceability limit state load cases SLS-1 and SLS-2 foundation system which will be discussed in more detail
were used for settlement assessment. It was assumed in the subsequent sections.
that all silo compartments are fully loaded for SLS-1. It is
understood that some compartments of the silo can be 3 GROUND CONDITIONS
either fully loaded or completely empty at times.
Therefore, SLS-2 was adopted to account for the worst- According to the Geological Survey of Victoria 1: 31, 680,
case scenario in terms of differential settlements when the construction site is located in an area of recent
half of the silo is empty (DL only) and the other half is Quaternary sedimentation which forms part of the well-
fully loaded (DL+LL). Ultimate state limit load cases ULS- known Yarra delta. Yarra delta sediments at the site are
1 and ULS-2 were used to check the generated actions present beneath a fill layer, and comprise Port
in piles against the structural capacity. ULS-1 was also Melbourne Sand (PMS), Coode Island Silt (CIS),
used to check the overall geotechnical stability of the Fishermen’s Bend Silt (FBS) and Moray Street Gravel
(MSG). The Yarra Delta sediments are further underlain
by Werribee Formation and Silurian age Siltstone of the information from both borehole and CPTs. The ground
Melbourne Formation. The groundwater level has been conditions were found to be reasonably consistent with
reported consistently at a depth range of 2.2m to 2.7m our general expectation of the Yarra Delta profile as
below ground surface, consistent with the adjacent river described in many existing references, for example,
water levels. Ground models were developed based on Ervin (1992). A summary of the adopted ground model is
the available site investigation data, which includes presented in Table 2.
Table 2: Summary of geotechnical modelling parameters

Unit Effective Friction Young
Depth BGL (m) Unit Weight
Cohesion Angle ϕ’ Modulus
(kN/m3)
From To c’ (kPa) (deg) E’ (MPa)
Fill 0 3.6 18 2 30 20
Port Melbourne Sand 3.6 6.5 19 2 30 20
Coode Island Silt 6.5 8 17 2 27 5
Fishermen’s Bend Silt 8 29 19 7 27 50
Moray Street Gravel
29 49.5 19 2-7 28 - 35 55 - 180
(interbedded clays and sands)
Werribee Formation (sand) 49.5 63.5 19 2 - 10 28 - 37 60 - 150
Siltstone, EW to HW 63.5 70 20 10 - 25 30 -32 60 - 200
4 FOUNDATION DESIGN The foundation system has the ring beam supported on
155 No. 900mm diameter reinforced concrete CFA piles
The large depth to siltstone rock meant that the piled with pile toes founded in the Moray Street Gravel (MSG)
foundations needed to be founded in the overlying soil unit. The layout for the CFA piles includes five rows of
profile at depths that could be readily achieved by an piles from inside to outside of the ring beam. The pile
available piling plant. The size and significance of the silo spacing varies between rows, but averages about 2.1m.
structure, and the significant loading it imposes on the
founding soils, meant that both geotechnical strength Available site investigation data suggested that the MSG
and serviceability requirements had to be carefully is comprised of interbedded sand and clay layers. To
evaluated. account for the lithological variation of interbedded
granular and cohesive sublayers within the MSG unit and
According to the adopted serviceability design criterion, the possibility of encountering clay lenses at the toe of
calculated foundation settlements under the some of the piles, a series of sensitivity analyses were
serviceability load combinations (SLS-1 and SLS-2) performed to evaluate the foundation performance with
should not exceed the allowable settlements. Allowable an appropriate range of ultimate base resistance values.
settlements for silo were specified as below:
• Total settlement is not to exceed 300mm. Appropriate pile design parameters for shaft friction and
• Differential settlement under Load Case SLS-2 is base resistance were determined in consultation with
not to exceed 60mm from one side of the silo to the Wagstaff Piling, based on previous experience and
other. available load test results from previous tests in similar
ground profiles, ensuring that the selected values were
Design for geotechnical strength was based on the limit reasonably expected to be achievable at the site for CFA
state design principles documented in the Australian piles installed with the available equipment. Table 3
Piling Code (AS 2159-2009). This requires piles to be presents the unit stresses that were adopted in our
designed in such a way that the design geotechnical modelling.
strength (which is the ultimate geotechnical strength
multiplied by an appropriate geotechnical strength Table 3: Summary of pile resistance parameters
reduction factor) will be not less than the design action Depth (m) Ultimate
Ultimate
effect. However, it has been well documented that Shaft
Unit Base Stress
applying such geotechnical criteria to each individual pile Friction
From To (kPa)
within a pile group can result in considerable (kPa)
conservatism in the design (e.g., Poulos, 2017). This is Fill 0 3.5 0 -
acknowledged in AS 2159-2009 in clause 3.2.2, which PMS 3.5 6.5 20 -
allows individual piles within the group to be overloaded CIS 6.5 8 20 -
for the ULS load case provided that the ultimate FBS 8 17 50 to 80 -
geotechnical design strength of the pile group satisfies
MSG 34 40 100 to 150 3000-5000*
the requirements of the code. To this end, the ULS-1 load
* Sensitivity checks were carried out for the range of base
divided by an appropriate Geotechnical Strength stresses presented which cover the expected range of
Reduction Factor (φg), was modelled with the outcomes in practice.
acceptance criteria being that the pile group must
maintain overall stability. This is indirectly a check on the Using the estimated unit stresses, the ultimate
overall factor of safety for the foundation system design geotechnical strengths of piles considered in the
and stability under this load case demonstrates that the modelling process ranged between 9.2MN and 11.3MN,
requirements of AS2159-2009 are met with respect to depending on the various founding assumptions.
strength.
Go to table of contents 39
5 NUMERICAL ANALYSIS showed that the application of embedded beam

elements can be considered a rigorous numerical
Recent advances in computational speed have approach as it results in very similar predictions
facilitated the utilisation of relatively complex and compared to volume elements for pile rafts. Although pile
detailed 3D analyses of piled raft foundations. shaft friction can be calculated automatically by the
Commercially available software packages, such as software (by relating the shaft profile to the strength
PLAXIS3D, can model complex ground conditions, pile properties of the soil), in this exercise the shaft
arrangements and soil-structure interaction. In this study, resistance is manually defined to ensure the modelled
modelling of the silo foundation system was carried out ultimate geotechnical resistance of piles is consistent
using the PLAXIS 3D software. with the values given in Table 3.
Simulation of piles in a sufficiently accurate and realistic A linear elastic perfectly plastic model with Mohr-
manner is a key aspect of a piled foundation analysis. Coulomb failure criteria was adopted for all soil layers.
There are two distinct methods in PLAXIS for modelling Effective stress analyses were carried out using drained
piles in a 3D analysis. One involves modelling piles as material properties to evaluate the long-term settlement
volume elements with a user-defined interface. Although behaviour of the foundation system. Long term
considered to be the more robust approach, the consolidation over the design life of the structure was
application of volume elements can be extremely also considered, by estimating the secondary
demanding in terms of processing time for a large compression of soils over the 50-year design life of the
foundations system with a considerable number of piles. structure. The adopted geotechnical modelling
parameters are summarised in Table 2. To ensure
The alternative option, which is adopted in this study, is loading was accurately distributed to the foundation
the application of embedded beam elements in which the system, loads were converted to equivalent pressures on
piles interact with the surrounding soil by means of top of the ring beam. Figure 3 shows the model geometry
special embedded interface elements. Lee et al. (2010) and pile arrangements.
Figure 3. PLAXIS 3D model geometry for the silo foundation system
6 RESULTS AND DISCUSSION be because a significant proportion of the SLS loads are
being resisted by shaft friction over the length of the
Table 4 summarises the estimated total and differential piles, and shed to the surrounding ground. As a
settlements for the foundation system under SLS load consequence, the load is distributed to a large soil mass
cases, for each of the adopted founding levels and end and small differences in founding level and end bearing
bearing assumptions. It can be inferred from the data in do not have a significant impact on settlements. Figure 4
Table 4 that the settlement performance of the presents the predicted absolute and differential
foundation system is not greatly sensitive to the range of settlement contours for load case SLS-1 and SLS-2 for
assumptions that were used. We consider this is likely to founding condition case 4 specified in Table 4.
Table 4: Summary of calculated settlement from PLAXIS analysis

Maximum calculated Maximum calculated
Founding Condition Case vertical settlement (mm) vertical differential
SLS-1 settlement (mm) SLS-2
1.Toe RL -38.5m, fbu= 5MPa 93 50 (33 to 83)
2.Toe RL -36.5m, fbu= 5MPa 100 56 (35 to 91)
3.Toe RL -38.5m, fbu= 3MPa 94 51 (33 to 84)
4.Toe RL -36.5m, fbu= 3MPa 101 58 (34 to 93)
Figure 4. Absolute and differential settlement contours for load cases SLS-1 and SLS-2
The load sharing behaviour between the piles and the unnecessary over-conservatism in the design by
ring beam is summarised in Table 5, as the ratio of the designing all piles to the highest individual pile load.
sum of the pile head loads over the total load being
applied on the foundation system. This is sometimes
referred to as the combined piled raft foundation
coefficient (Katzenbach et al. 2005). Table 5 indicates
that the superstructure loads are mostly carried by the
piles, with a bearing contribution of about 80%, and the
remaining 20% of the loads are transferred to the subsoil
directly through the ring beam contact area.
Table 5: Pile and ring beam load sharing

Case Qp/Qt
SLS - 1 0.80
ULS - 1 0.78
Figure 5 illustrates the calculated pile head axial loads

under ULS-1 for founding condition case 2. This Figure
shows that the loads are unevenly distributed among
piles, ranging from around 4.5 MN to 10 MN. It is noted
that some of the variations in individual pile loads is Figure 5. Calculated pile head loads, ULS-1
thought to be attributable to the FEM modelling process;
specifically localised meshing effects. For this reason, As mentioned in the preceding sections, an additional
the variability in actual pile loads is expected to be less load case, based on the ULS-1 loading divided by the
in practice than the model outputs would suggest. adopted geotechnical reduction factor (φg), was
analysed for all founding condition cases and a
Nonetheless, a large variation in individual pile loads has satisfactory outcome was achieved. This demonstrated
also been demonstrated in other studies (e.g., Poulos, compliance with AS2159-2009 with respect to the
2017). A highly non-uniform distribution of loads among ultimate geotechnical capacity of the pile group, and an
piles highlights the importance of considering group adequate overall factor of safety for the foundation
stability when assessing the ultimate limit state system design.
geotechnical capacity of the foundation system to avoid
To account for the localised variability in individual pile
loads (inferred to be due to meshing effects) in the
Go to table of contents 41
42
design, our approach was to average the individual pile to develop an optimised design solution for the piling
loads across a selection of adjacent piles for each row of scheme. The nature of the load transfer between the
piles. Table 6 presents the average maximum axial pile piles and the ring beam was able to be assessed, as well
head loads across inside, middle and outer rows of piles as the distribution of loads to individual piles and inner,
under SLS-1 and ULS-1 load cases. It is noteworthy that middle and outer pile rows. The analysis indicated that
the outside row piles attract significantly higher loads the distribution of vertical loads was highly non-uniform
compared to the middle row and inner row piles. among piles in different rows, which can be explained by
the relative displacements experienced by piles in each
Table 6: Summary of calculated pile head loads row relative to the surrounding soil. The study illustrated
the importance of considering the foundation system as
Case Maximum axial pile head load (kN) a group, in order to avoid over-conservatism in the pile
Outside Row = 9413 design.
SLS - 1 Middle Row = 3159
Inside Row = 5765 8 ACKNOWLEDGEMENTS
Outside Row = 10050
The authors wish to acknowledge the contribution and
ULS - 1 Middle Row = 5747
assistance of Melbourne Cement Facility and Fitzgerald
Inside Row = 6591
Construction (Australia) in the execution of the study and
for allowing the publication of this technical paper.
To better understand the load sharing mechanism
between piles in different rows, it is necessary to REFERENCES
consider the differential movements between these piles
and the surrounding soil. Figure 6 shows a cross-section AS 2159-2009 Piling - Design and installation, Standards
of the vertical settlement contour plot through the Australia.
midsection of the silo from the PLAXIS analysis. Burland, J. B., Broms, B. B., and De Mello, V. F. B. (1977).
“Behaviour of foundations and structures.” Proc. 9th int.
This figure needs to be interpreted in conjunction with the Conf. on Soil Mech. & Found. Eng., 2, 495-546.
tabulated values in Table 6. As shown in Figure 6, the Davis, E. H., and Poulos, H. G. (1972). The analysis of piled raft
soil mass enclosed inside the ring of piles settles much systems. Aust. Geomech. J 2, 21-27.
Ervin, M.C., (1992), Engineering properties of Quaternary age
more compared to the soil mass outside the footprint of sediments of the Yarra Delta, Seminar, Engineering
the silo. Due to the relative rigidity of the ring beam, the Geology of Melbourne, 245–260.
absolute settlements of all piles are very similar. Katzenbach, R., Arslan, U., Moorman, C., and Reul, O. (1998).
Therefore, the outer row piles settle more relative to the “Piled raft foundation: interaction between piles and raft.”
surrounding soil mass, which results in mobilisation of Darmstadt Geotechnics (Darmstadt University of
higher shaft and end bearing resistances in these piles. Technology), 4, 279-296.
In contrast, the soil mass inside the silo footprint Katzenbach, R., Schmitt, A., and Turek, J. (2005). “Assessing
Settlement of High-Rise Structures by 3D Simulations.”
undergoes a much larger settlement and the
Computer-Aided Civil and Infrastructure Engineering, 20,
displacement of the inner row piles relative to the soil is 221-229.
correspondingly lower; hence, less resistance is Lee, S. W., Cheang, W. W. L., Swolfs, W. M., and Brinkgreve,
mobilised in the inner row piles. The middle row piles R. B. J. (2010). “Modelling of piled rafts with different pile
mobilise the least resistance, as the soil mass around models.” Numerical Methods in Geotech. Eng. Benz &
these piles is being loaded by adjacent piles in both the Nordal (eds), Taylor & Francis Group, 637-642.
inner and outer rows; hence, middle row piles experience Mandolini, A., Russo, G., and Viggiani, C. (2005). “Pile
the least relative pile/soil displacement. foundations: experimental investigations, analysis, and
design.” Proc. 16th Int. Conf. on Soil Mech. & Geotech.l
Eng., 12–16 Sep, Osaka, Japan, 177-213.
Poulos, H. G. (1991). “In computer methods and advances in
geomechanics (eds Beer et al.), 183-191. Rotterdam:
Balkema.
Poulos, H. G. (1994). “An approximate numerical analysis of
pile-raft interaction.” Int. J. for Numerical and Analytical
Methods in Geomechanics, 18, 73-92.
Poulos, H. G. (2017). “Tall Building Foundation Design.” CRC
Press, Taylor & Francis Group.
Randolph, M. F. (1983). “Design of piled foundations.” Proc. Int.
Symp. on Recent Development in Lab. & Field Tests &
Analysis of Geotech Problems, Bangkok, 525-537.
Randolph, M. F. (1994). “Design methods for pile groups and
piled rafts: state-of-the-art report.” Proc. 13th Int. Conf.
Soil Mech. Found. Eng., New Delhi 5, 61-82.
Russo, G., and Viggiani, C. (1998). “Factors controlling soil-
Figure 6. Vertical settlement contour plot structure interaction for piled rafts.” Intern. Conf. on Soil-
Structure Interact. in Urban Civil Eng., Ed. Katzenbach,
7 CONCLUSION R., and Arslan, U., Darmstadt, 79-102.
Smith, I. M., and Wang, A. (1998). “Analysis of piled rafts.” Int.
This paper presented the analysis undertaken as part of J. Numer. Anal. Meth. Geomech., 22, 777-790.
the design of the foundation system for a large cement Ta, L. D., and Small, J. C. (1996). “Analysis of piled raft systems
in layered soils.” Int. J. for Numerical and Analytical
silo storage facility. A satisfactory outcome was achieved
Methods in Geomech., 20, 57-72.
in terms of both geotechnical strength and serviceability Viggiani, C. (1998). “Pile groups and piled rafts behaviour.”
performance. The study showed that 3D FEM analysis Proc. 3rd Int. Geot. Seminar on Deep Foundations on
can be employed to understand the group performance Bored and Auger Piles, Ghent, 77-94.
of the foundation system and the distribution of load to Viggiani, C., Mandolini, A., Russo, G. (2012). “Piles and Pile
foundation elements. The results were used successfully Foundations.” Spon Press, Taylor & Francis Group.
Pile Testing Verification – an Alternative Approach
D. Denes1, CPEng., A. Foroughi 2, BEng., J. Seidel 3, PhD., CPEng.

1
Golder Associates Pty Ltd, Foundation/Pile Testing Department, Building 7, Botanicca Corporate Park, 570 – 588 Swan Street,
Richmond, Victoria 3121, Australia; email: Ddenes@golder.com.au
2
Golder Associates Pty Ltd, Senior Engineer, Building 7, Botanicca Corporate Park, 570 – 588 Swan Street, Richmond, Victoria 3121,
Australia; email: AForoughi@golder.com.au
3
Foundation QA Pty Ltd, Managing Director, POBox 4280, Croydon Hills, Victoria, Australia, 3136; email: julian@foundationqa.com
ABSTRACT
Dynamic pile testing is undertaken for a number of reasons including: 1. To confirm that the pile meets serviceability and
geotechnical capacity requirements; 2. To assess pile integrity, either during installation (driven piles) or after construction
(cast-insitu piles) and 3. To verify that the piling hammer delivers the energy required to satisfy the design criteria and that
stresses during testing are kept within acceptable limits. In addition, testing allows us to establish and calibrate acceptance
criteria - relationships such as resistance vs set curves, and/or correlated pile driving formulas. These relationships are
premised on the interrelationship of capacity (C), transferred energy (E) and pile movement (M) which is represented
primarily by pile set. These ECM relationships allow capacity to be inferred from measurement of transferred energy and
pile movement and are used to infer the capacity of untested piles. However, for a variety of reasons, transferred energy
can vary significantly between piles which, being undetected, undermines the reliability of ECM relationships. An alternative
approach to using ECM relationships is proposed based on pile set and pile force (F). We demonstrate through parametric
studies and review of project data, that these FCM relationships are reliable alternatives which bypass the problems with
variable energy transfer. Of course, impact force will also vary with hammer performance, but impact force can generally
be accurately determined from the measured impact velocity as a proxy. Pile velocity can be measured by attaching a
single accelerometer to the pile, or by using a high frequency displacement monitoring device. FCM-based acceptance
criteria have the significant advantage that both the necessary force (F) and displacement (M) inputs can be verified by
simple measurements on all untested piles.
Keywords: Pile Testing, Pile Verification, Dynamic Formulae, PDA, Case Study
1 INTRODUCTION constructed. In order to reduce the risk of failure for these

foundation elements, design methods take into account
The intent of this paper is to discuss an alternative this uncertainty. In order to reduce this uncertainty, and
approach to estimate the resistance of untested driven benefit from higher capacity reduction factors, driven piles
piles. The traditional methods of pile verification for these are tested during installation.
piles are almost universally based on Energy-Capacity-
Movement relationships. Pile driving formulae are various Pile testing is undertaken to confirm that the pile meets
forms of ECM relationship (Seidel, 2018, 2021). Without serviceability and geotechnical capacity requirements.
these relationships every pile would need to be tested. Depending on the type of test, the structural integrity of
the pile may also be assessed to ensure that the pile is
An alternative Force-Capacity-Movement (FCM) intact and free of damage. Testing can further provide
approach has been developed which relates pile capacity information on the performance of the driving system. In
to the pile head force generated by the driving hammer summary, testing is a quality control process which helps
and to pile set. In application, this approach involves increase confidence that the test pile meets the design
measuring pile impact velocity (as a proxy for pile impact intent.
force). In all but exceptional cases, pile-top force and pile
velocity are related by pile impedance at the time of The uncertainty of a particular pile capacity evaluation is
impact. The benefit of the FCM approach is that pile-top related to the intrinsic reliability of the test method used.
impact force can be inferred from simple measurements However, the benefit of any test has indirect benefit to the
for each pile, whereas transferred energy (in ECM evaluation of other similarly installed piles. Under
relationships) can only be assumed. The FCM approach AS2159-2009 the applicable geotechnical reduction
is therefore intrinsically more reliable than traditional ECM factor (φg) for untested piles is determined from a risk
approaches. assessment in conjunction with the proposed type of pile
test and percentage of piles being tested. The so-called
This paper will focus on demonstrating the nature of the testing benefit factor increases with the percentage of
relationship between pile head force and pile capacity piles tested.
using data generated from both a GRLWEAP (Goble et al.,
1988) wave equation parametric study and also from Static pile testing is the traditional and direct method of
actual piling data. pile testing where the movement of the pile is monitored
and recorded as the pile is gradually loaded to the desired
2 BACKGROUND capacity. Due to practicality, time and cost implications of
static pile testing, only a limited number of piles are
Deep foundations are constructed to transfer loads from usually tested (Rausche et al., 2008), which is not ideal
a superstructure into the ground. There is an inherent for large projects, especially for sites with significant
uncertainty associated with the interaction between the geological variability as there is a significant risk where
piles and the ground in which they are installed or test piles may not be representative.
Pile Testing Verification – an Alternative Approach D. Denes, A. Foroughi and J. Seidel
The focus of this paper is on driven displacement piles bearing relationship. The bearing graph may be further
which allow a wider range of testing opportunities, calibrated based on observed field results taken during
including high strain dynamic testing (commonly referred end of initial drive or restrike testing. At the end of the test
to as PDA testing). PDA testing has been used in piling process, the aim is to use the pile test results to
Australia and New Zealand since the early 1980’s. PDA produce a relationship that reliably predicts resistance of
testing has substantial cost and time advantages relative the untested piles to the tested piles.
to static testing, and this allows more testing to be
conducted for the same testing budget. PDA testing The result of Wave Equation analysis is a relationship
therefore allows statistically significant, and widely between pile capacity and pile set, which is energy
distributed testing to be conducted – a significant benefit dependent – i.e. a form of ECM relationship. This is
compared to traditional static testing. traditionally presented as one or more curves relating
capacity to blow count (blows/metre) called a bearing
Typically, less than 15% of piles on a project are tested, graph, as shown in Figure 1.
leaving the remaining 85% of piles untested. The
verification of the untested piles can be done in several The problem with this traditional form, is that at very low
ways. For example, by simply comparing installation data pile sets (high blow counts), blow counts become
of untested piles to installation data of tested piles and excessively high and the relationship below 1.0mm set
ensuring that the untested piles have been driven to (1000 blows/m) is difficult to evaluate. Therefore, in the
similar or harder conditions. In practice, dynamic following, the bearing graphs shown will relate capacity to
formulae, such as the Hiley formula (Hiley, 1930) are set, as shown in Figure 2 which represents the same data
often used for this verification. Hiley formula is shown in Figure 1.
recognised in AS2159-2009 and is also in the Auckland
Structural Group Piling Specification (2002). However, 3.1 Set and Hammer Energy Measurements
academic debates from as early as 1941 (American
Society of Civil Engineers, 1941) and to this date (Allin, In order to use the bearing graph for pile capacity
2015) continue to discuss the inaccuracies of the use of verification, measurements of the pile set and hammer
dynamic formulae. Allin recommends using wave energy are required. Different methods can be used to
equation analyses instead of generic dynamic formulae measure pile set, with various degrees of accuracy. The
for the assessment of untested piles. Seidel (2015) traditional method used to measure set involves a person
proposes the use of site-specific dynamic formulae, which (a) standing beside a pile and marking a piece of paper
are to be calibrated based on test data, which has been attached to the pile
further investigated with success (Damen & Denes 2017,
Denes & Kroenert 2019). Nevertheless, the use of
dynamic formulae in projects may lead to mis-guided
results and inaccuracies and should be used with a
correct understanding and informed approach.
Regardless of whether a dynamic formula, a modified

dynamic formula or a wave equation analysis is used for
comparison between tested and untested piles, the
capacity of a pile is related to the applied energy (or
hammer drop height) and the pile movement (or set). This
is discussed further below in detail.
3 PILING RELATIONSHIPS – CURRENT Figure 1. Bearing graph relationship – blow count

PROCEDURE
There is a requirement for most driven piling projects to

provide a driving (or installation) criterion prior to
commencement of pile driving to ensure the required pile
capacity is achieved. This is usually in the form of
graphical correlation between pile capacity, applied
hammer energy and measured permanent movement
(set) from a single (or a specific number of) hammer
stroke(s). This graphical representation is usually
referred to as a bearing graph. Bearing graphs can be
generated from either the traditional dynamic formula
(e.g. Hiley) or Wave Equation Analysis Programs (e.g.
GRLWEAP). Dynamic formula use simplified empirical
relationships relating to energy transfer between two
bodies. Wave Equation Analysis Programs simulate the Figure 2. Bearing graph relationship – pile set
driving sequence and estimate the energy applied to the
pile considering a much wider range of factors including during each hammer strike. Due to safety concerns
stiffness and dimension of the pile and striking weight, associated with standing and working beside a hammer,
hammer and piling cushions, etc. this method has mostly been replaced with one of the
following methods: (b) monitoring using surveying
Once a bearing relationship has been developed for pre- equipment; (c) manually marking the pile prior and
piling, the first installed piles are tested to confirm the
subsequent to the final 10 hammer blows; (d) onboard Using Seidel’s observation, and assuming a 5mm set per
piling hammer instrumentation (such as the iPilier system blow, if we apply a +/- 33% variation to the 44 kNm energy
by Junttan or NDT lazer & computer system); (e) counting example in Figure 2, the resistance acting on the pile can
the number of hammer strikes for the pile to move a range between 2200 kN and 3050 kN which is an
certain distance (blow count); (f) non-contact high unacceptably large range.
frequency displacement monitoring devices, such as the
Pile Driving Monitor (PDM) which has sub-0.1mm 4 AN ALTERNATIVE APPROACH - FORCE
accuracy.
As noted, an alternative approach to using the standard
It can be unreliable to rely on options (a) to (d), especially ECM relationship has been developed using the pile set
when pile set is small. Ideally, more focus towards set and the force applied to the top of the pile. This FCM
measurement using digital equipment should be approach was first discussed and presented by Seidel
embraced, as anything that involves manual pile marking (2018).
and/or counting is likely to introduce an element of error.
In simple terms, the general principle that applies to piling
Set is usually measured over a sequence of 10 blows. It is similar to driving a nail into timber. For a nail to drive
is important that these blows be consistent so that the into timber, the force applied must overcome the force
average set is meaningful. Any ‘warm up’ blows to reach resisting the nail to move. If the nail does not move under
the specified drop height should be discarded from the force of the hammer, one can assume that the force
consideration. In summary, set is a fundamental resisting the nail to movement is equal to or greater than
measurement used in all ECM relationships, so it is the force from the hammer. On the other hand, if the nail
important that set measurement is reliable. moves, then resisting force is expected to be less than the
force applied by the hammer. This is often seen with pile
It is noted that temporary compression measurement is testing. If the pile movement is zero under a hammer
used in the Hiley Formula. This can only be measured strike, one can infer that the resistance acting on the pile
using methods (a) and (f) above. Method (a) is not is equal to or greater than the applied force.
recommended for safety reasons, as discussed.
Based on the principles of one-dimensional wave theory
Hammer energy for an untested pile can be estimated pile force and velocity are related by pile impedance for
from the potential energy of the ram which is a function of as long as there is only a single wave in the pile. Under
the weight and drop height. The drop height of a hammer those conditions, the following holds true.
is usually recorded for the full sequence of pile driving
(referred to as pile driving record) which can be used to F = vZ (1)
estimate the energy applied to the pile. The actual
delivered energy to the top of the pile is often less than where F is the pile-top force, v is the pile-top velocity and
the potential energy of the ram due to frictional losses in Z is a constant of proportionality called the pile
the hammer, energy losses in the hammer and pile impedance, and
cushion, energy losses during impact, and restrictions on
energy transfer that are functions of the pile material, Z = EA/c (2)
cross-section, length and driving resistance. There have
been several papers that document and emphasize the where E is the pile modulus, A is the cross-sectional area,
high variability in energy transfer from impact hammers and c is the pile wave-speed.
and caution for the sensitivity of driving formulae and
wave equation analyses to this parameter (Allin (2015), During impact, and generally up to the time of maximum
Seidel (2015b), Flynn & McCabe (2016)). Given the impact force and velocity, this proportionality between F
sensitivity of driving formulae to this parameter, a and v can be reliably assumed. Therefore, knowing Z, the
calibration using either measurements of actual energy in peak pile-top impact velocity can be used as a proxy for
the test piles or an energy efficiency factor based on measurement of peak impact force.
previous performance of the hammer is usually
recommended for estimation of hammer energy. it is important to distinguish between capacity and
However, the performance of a hammer, including the resistance in relation to pile driving. The resistance to
efficiency, can vary from pile to pile and over time. penetration that the pile experiences when struck by the
Various factors such as hammer alignment, pile-hammer hammer is called the driving resistance. Driving
contact surface and hammer maintenance frequency can resistance comprises a dynamic component due to the
also affect the energy transferred to the pile which in turn velocity of the pile movement, and a static component,
affect the accuracy of the bearing graph correlation. which is the (static) capacity of the pile at the time of
Further to this, there can be errors in identifying the actual impact, and would be the expected capacity if it were
hammer drop if it is undertaken visually when the hammer possible to undertake an instantaneous static load test.
is in operation.
The parametric study and case studies described relate
Seidel (2015) has reported in one comprehensive study to the relationship between pile capacity, C, and peak
for a major project that the measured energy observed impact force, F. The capacity/force (C/F) ratio is
from PDA testing on the same pile type (and hammer) dependent on hammer, soil and pile characteristics, and
varied by +/- 33% about the median energy. The first is also shown to be a function of pile set, being maximum
author’s experience has often seen variation in energy at low set, and progressively decreasing as pile set
ranging between +/- 15 %. This demonstrates the increases (and dynamic resistance represents a
uncertainty associated with estimation of the delivered progressively larger proportion of the driving resistance).
energy to untested piles.
Figures 1 and 2 presented the results of a GRLWEAP
analysis for a 9t hammer hitting a 400mm square concrete
pile from drop heights of 0.4m, 0.6m and 0.8m. The three In each case, the CAPWAP results were normalized using
curves in each Figure relate capacity to blow count or set the PDA-measured peak impact force and the data
for each of the three drops/ energies. The parametric presented as data points on a C/F ratio vs set plot.
study takes this one step further by normalizing the
capacities by peak impact force. The FCM relationship has also been computed using
GRLWEAP based on the site hammer, pile and soil
The same data presented in Figures 1 and 2 is shown in characteristics using representative resistance
Figure 3 in normalized form – i.e. the FCM relationship - distribution, damping and quake values determined in the
C/F ratio vs set. What is striking is that all three curves CAPWAP analyses.
reduce to a single and unique relationship between C/F
ratio and set, regardless of drop height. 5.1 Case Study One – Open Ended Steel Tube
Piles
The project used for case study one is the upgrade of an

existing berth. At this project the soil stratigraphy
consisted of dense clayey sand and very stiff to hard
sandy clay layers. All piles were open ended 610mm steel
tubes with 16mm wall thickness each with a total length
of 35m and an average penetration of 16m (ranging
between 13m to 18m). The end of drive sets were
between 5mm and 22mm using 1200mm drop height. The
piles were driven with a 9t hammer. All piles were restrike
tested between 4 to 18 days after installation using the
same hammer with 1.5m drop height. The restrike sets
ranged between 0.5mm and 10mm and estimated
CAPWAP resistances varied between 3500kN to 6800kN.
Figure 3. Normalized bearing graph (FCM relationship: GRLWEAP analyses were undertaken replicating all
C/F ratio vs pile set) relevant characteristics as noted above.
5 CASE STUDY CORRELATIONS Figure 4 compares the C/F ratio vs set data from PDA
tests and the FCM relationship from GRLWEAP analysis.
Two case studies will be presented for historical piling At pile sets less than 1mm per blow, the inferred capacity
projects. PDA testing equipment from Pile Dynamics is approximately equal (96%) to the peak impact force. At
(PDA-PAX and 8G systems) were used to collect the pile sets of 5mm, the achieved capacity is only 70% of the
force and velocity data. The data was analysed using peak impact force; at 10mm the ratio reduces to 55% and
CAPWAP to determine pile capacity. at 15mm pile set the ratio drops further to 44%.
Figure 4. Case study 1 - data comparison to FCM relationship
It is clear that there is very good correlation between the 2. The FCM approach offers a piling control and
site data and the theoretical prediction. There are two acceptance method which can accommodate
important findings relating to this case study. large variations in hammer efficiency without
impacting reliability. The capacity of all non-PDA
1. If the GRLWEAP parameters can be reasonably tested piles, can be estimated with greatly
estimated in advance, then the FCM relationship improved accuracy provided that the set and pile
can be predicted, and initial target sets can be velocity can be measured accurately.
established in advance of piling.
5.2 Case Study Two – Precast Concrete practicalities of driving precast piles where different drop
height may need to be applied due to varying soil/ground
The project used for case study two is the construction of conditions, controlling driving stresses (especially tension
13 bridge pier foundations for a river crossing. At this stresses) and/or different load requirements.
project the soil stratigraphy consisted of very stiff to hard
sandy clay layers. All piles were 400mm square simply- Figure 5 presents the correlation between the C/F ratio
reinforced precast concrete with total lengths ranging and pile set for PDA/ CAPWAP data compared to the
between 20m to 41m and penetration ranging between GRLWEAP models with 3 different drop heights. Similar to
18m to 36m. The end of drive sets were between 1mm Case 1, the correlation between the C/F ratio and pile set
and 13mm using 300mm to 500mm drop height. The piles can be clearly seen from these results. It is also evident
were driven with a 9t hammer. All piles were restrike that GRLWEAP models can predict the correlation with
tested between 1 to 24 days after installation using the reasonable accuracy if the necessary parameters for the
same hammer with drop heights between 400mm and GRLWEAP models are available.
800mm. The restrike sets ranged between 0.3mm and
9mm and estimated CAPWAP resistances varied Nevertheless, there is some increase in scatter evident
between 1100kN to 4500kN. Timber ply was used as a which may result from minor PDA data issues
pile cushion for all tests. (proportionality), accuracy of the manual set
measurement especially at small sets or variable pile
For the GRLWEAP models, three different drop heights cushion stiffness.
were used for assessment. The reason is related to the
Figure 5. Case study 2 - data comparison to FCM relationship
The general form of the FCM relationship is similar for 1B. Undertake PDA testing and CAPWAP analysis on the
both Case 1 and Case 2, albeit with different specific first pile. Analysis to be undertaken at various different
values, and again, predictable by theoretical analysis pile sets (e.g., 12mm, 6mm and 2mm).
using relevant analysis parameters. 2. Use data from either step 1A, 1B or both, to establish a
site correlated FCM relationship.
For Case 2, at pile sets less than 1mm per blow, the 3. Measure pile velocity and pile set for untested piles.
inferred capacity is slightly less (92%) of the peak impact 4. Estimate pile force using the measured velocity and
force. At sets of 5mm, the achieved capacity is only 63% calculated pile impedance
of the peak impact force; at 10mm the ratio reduces to 5. Use the site correlated FCM curve to find the C/F ratio
42% and at 15mm pile set the ratio drops further to 29%. for the measured set and then calculate resistance using
the force estimated in step 4.
The results also confirm that the use of different drop
heights has little effect on the correlation between C/F Ongoing pile testing with CAPWAP analysis should also
ratio and measured set. be undertaken on a percentage of piles to check the
relationship. Minor adjustments/refinement may be
6 RECOMMENDED FCM RELATIONSHIP required as more data is collected.
PROCEDURES FOR A PROJECT
7 FURTHER DISCUSSION
The FCM relationship can be developed using the
following method: 7.1 Measuring Velocity
1A. Undertake a GRLWEAP analysis with estimated driving There are only two known methods currently undertaken
parameters from site investigation using proposed pile to measure the velocity during pile driving. The first is by
and driving hammer details. using accelerometers recording the acceleration of the
pile which can then be used to obtain the velocity.
Examples of equipment used are accelerometers from the both the pile set and velocity. Pile set is an important
PDA system (PDI / Allnamics). aspect of this force base verification and an accurate
measurement of set should always be a strict
The second is by measuring displacement and requirement.
differentiating with respect to time to calculate velocity.
An example is the Pile Driving Monitor (PDM). PDM is a The use of this force relationship for projects should follow
device that records the displacement of the pile with high the same process used for standard bearing graph
frequency and can therefore accurately measure pile set relationships. That is to undertake a GRLWEAP analysis
and pile velocity, and hence, can be further used to prior to piling works to establish an initial correlation.
estimate pile resistance for all untested piles from a FCM Then once the first pile(s) has(have) been PDA tested,
relationship created from tested piles or GRLWEAP refinement of the relationship is required for application to
modelling. the remaining piles. Ongoing testing should also be
undertaken to identify the need for further refinements as
7.2 Set Limitation the project progresses.
Similar to dynamic formulae, bearing graphs, PDA testing REFERENCES

and CAPWAP analysis, the pile set that is recommended to
be used for the FCM correlation has a range and should AS 2159-2009 (2009). Piling – Design and Installation.
be between 2 mm and 12 mm. Sets outside this range Standards Australia.
may provide results that do not have the required level of Allin, R., Likins, G. and Honeycutt, J. (2015) Pile driving formulas
revisited. Proceedings of the International Foundations
accuracy.
Congress and Equipment Expo 2015. San Antonio, Texas.
Eds. M. Iskander et al.
Small sets (less than 2mm) can be a sign that the full American Society of Civil Engineers (1941) Pile driving formulas.
resistance of the pile may have not been mobilized during Progress report of the committee on the bearing value of
driving which can lead to under-estimation of the static pile foundations. Proceedings of the American Society of
capacity. Civil Engineers. Vol. 67, No. 5, pp853-866.
Damen, R. & Denes, D. (2017) Improving site specific modified
Large sets (greater than 12mm) during driving can driving formulae using high frequency displacement
monitoring. Proc. 20th NZGS Geotechnical Symposium.
generate large dynamic component of resistance which
Eds. GJ Alexander & CY Chin, Napier.
may lead to over-estimation of the static capacity. Denes, D. & Kroenert, B. (2019). A Case Study of Pile Testing
and Verification. Berth 4 Upgrade - Port of Townsville.
7.3 Pile Cushion Australasian Coasts & Ports 2019 Conference – Hobart,
10-13 September 2019.
As discussed earlier, the calculated pile force is a function Flynn, K.N., McCabe, B.A. (2016) Energy Transfer Ratio of
of pile impedance and velocity. Hence, the change in Hydraulic Pile Driving Hammers. Civil Engineering
cushion stiffness does not affect the relation between Research in Ireland (CERI 2016).
Goble, G. G., F. Rausche, and G. E. Likins. "GRLWEAP–Wave
velocity and force.
Equation Analysis of Pile Driving." (1988).
Hiley, A. (1930). Pile-driving calculations with notes on driving
The importance of this is to note that although the pile forces, and ground resistance. The Journal of the
cushion ‘stiffness’ does not affect the pile head force Institution of Structural Engineers. pp 246-259 (part 1, July)
calculation but it may alter the relationship between the and pp278-288 (part 2, August)..
force applied and the pile resistance. Pile Driving Monitor PDM. Website:
https://www.foundationqa.com/pile-driving-monitor.
This will be further researched by the authors and Foundation QA (2019). Pile Driving Monitor, Because
Every Pile Is Important, General Specifications.
presented at a later date.
Melbourne, Australia.
Rausche, F., Nagy,M. and Likins, G. (2008). Mastering the Art of
However, the authors do note that there are cushions Pile Testing. Proc. of the 8th Int’l Conf. on the Application
available in the market which claim to have a very of Stresswave Theory to Piles. Lisbon. J.A.Santos (ed) :
consistent modulus or ‘stiffness’ form the start to the end 19-32.
of the cushion life. This may have significant benefits to Seidel, J.P. (2015) Overview of the Role of Testing and
a project as this will provide a higher level of confidence Monitoring in the Verification of Driven Pile Foundations.
to the FCM relationship. Proceedings on the 12th Australia New Zealand
Conference on Geomechanics (ANZ2015). Wellington. Ed.
G. Ramsay. pp389-396.
8 CONCLUSION Seidel, J.P. (2018) Mobilization of Pile Capacity and Pile
Acceptance. Australian Geomechanics Society Victoria
An alternative approach to pile verification using a force Meeting Presentation August 8, 2018.
base relationship with respect to pile set has been Seidel, J.P. (2021) Ground truth, control and design of driven
generated using GRLWEAP. Two case studies on two very piles – implementing old ways with a new twist. Australian
different pile types show very good correlation between Geomechanics Society Victoria Seminar October 21,
the theoretical GRLWEAP results and the field results. 2021.
The determination of force and consequently capacity

using pile velocity provides more confidence when
untested piles are concerned. The use of this velocity
correlation removes the reliance on hammer energy and
the inaccuracies associated with estimating hammer
energy which are the result of variable hammer
performance from pile to pile and from time to time.
Recording pile velocity using readily available equipment
is quick and easy, with some equipment able to record
SESSION 3
OPTIMIZATION OF RISK
AND SAFETY IN DESIGN
Keynote address
Using geotechnical innovation to reduce project risk

Chris Lyons1, Cameron Cairns2
1
Principal Geotechnical Engineer, Suburban Rail Loop Authority, 80 Collins Street, Melbourne, 3001, Victoria, email:
chris.lyons@srla.vic.gov.au
2
Manager Minerals Geoscience, Geological Survey of Victoria, Level 17, 1 Spring St, Melbourne, Victoria 3000, email:
Cameron.Cairns@ecodev.vic.gov.au
ABSTRACT
Project risk is commonly defined as an uncertain event or condition that, if it occurs, has a positive or negative effect on a
project’s objectives. As the construction industry continues to deliver larger and more complex projects, improving how
we manage project risk is a key priority for clients, contractors and designers. Within this digital age, innovation and change
in our industry will play a key part in how we manage risks and deliver projects. This paper considers how we can define
‘innovation’ and how innovation can be promoted within organisations and on projects. The classification of something as
innovative can vary widely depending on its novelty and context. The journey from creativity (an idea) to a productive
solution (innovation) is difficult and frequently does not materialise. Three examples of innovation are presented to highlight
the range of productive solutions that can be considered as innovation within their context.
1. The development and benefits of public Geotechnical Databases; novel within the context of Victoria
2. Incremental innovation in the use of InSAR satellite monitoring.
3. Radical/modular innovation using automated processed to import piling construction records back into project
models to validate QA and geotechnical ground models.
Keywords: Innovation, InSAR, Geo-database, piles, risk
1 DEFINING INNOVATION concepts (e.g. ability to import ground profiles into a

FEM)
There are many definitions and concepts around • Modular Innovation: innovation that changes only
innovation. The definition of innovation varies according the core design concepts of a technology (e.g.
to the context and field of application such as improved algorithms for resolving FEM)
organisational structure, customer service or technical
fields. Across the many definitions and fields, two main
dimensions of innovation can be considered (Edison, et Each of these fields could have different kinds of
al., 2013): innovation (process, product or service) noted above.
1. the degree of novelty (new to a firm, new to a
market, new to the industry or new to the world), and 1.2 Creativity Vs Innovation
2. the kind of innovation (whether it is a process,
product, or service system innovation) While creativity and innovation are related, they should
not be confused as the same thing. Creativity is the ability
Within this framework, innovation can take many forms to generate new ideas or plans, whereas innovation is the
and in particular, the degree of novelty means the ability to apply the ideas in a productive manner. Early
technology systems which are well developed elsewhere, models of innovation (Utterback, 1971) considered three
can still be considered as innovation providing there is a phases of innovation: 1) creative idea generation 2)
productive outcome. turning the idea into reality (i.e., an invention) and 3)
having an economic impact through implementation.
1.1 Technical innovation
1.3 Delivering innovation
Within technical fields, a commonly adopted framework
divides innovation into four types (Henderson & Kim, Successfully taking creativity and delivering it into
1990): innovation depends greatly on the environment within an
organisation. A culture which both embraces innovation
• Radical innovation: Establishes a new dominant and has a process for delivering innovation is required.
design and, hence, a new set of core design concepts To successfully develop an innovation, first a specific
embodied in components that are linked together in a need must be identified, macro and meso trends must be
new architecture (e.g., the establishment of Finite understood, competency must be developed (people and
Element Modelling (FEM) ). technology) and financial support provided. Goals and
• Incremental innovation: refines and extends an aligned actions must be defined, supported and tracked
using metrics during development.
established design. Improvement occurs in individual
components, but the underlying core design concepts,
Once an innovation is developed, the lifecycle and impact
and the links between them, remain the same (e.g. on the market changes over time as the product is first
improvement in viewing FEM results) established, then widely adopted, incrementally
• Architectural innovation: innovation that changes developed as it matures and eventually declines is usage
only the relationships between them core design (Figure 1). Businesses and industry which rely on
51
Using geotechnical innovation to reduce project risk C. Lyons, C. Cairns
innovation require a succession plan of innovation

projects moving through the product lifecycle.
Figure 1. Innovation life cycle showing with succession planning within a market for continued high performance
(Mazouz, et al., 2019).
1.4 Opportunities for Innovation 2 SHARED GEOTECHNICAL DATABASES
Within this digital age, the opportunities for innovation to Innovation can be classified as the adoption of existing
revolutionise how we operate are emerging into all facets technology or system and applying them within a local
of our industry. Opportunities include; market. Within the Melbourne market, there is a gap in
the collation, management and delivery of geotechnical
• Automated processes ranging from simply eliminating data that supports industry and maximises the return on
repetitive activities, to scripting complex models to public investment. Public geotechnical borehole
parametric design process. databases, and associated geoscience datasets acquired
• The evolution of Digital Twins connected to the as part of geotechnical investigations, are a relatively
common structure in other States within Australia and
Internet of Things (and conversely their Evil Twin,
globally, although the structure and sophistication of the
bloated with excess data) (Hannel, 2019).
databases vary. The Geological Survey of Victoria (GSV)
• Machine Learning and Artificial Intelligence are is working with government partners to scope the
changing our use of data and design processes. development of a Victorian geotechnical database. The
• Robotics and drone automation as our construction aim of the database is to augment existing free public
sites become smarter, with autonomous trains, resources delivered by the GSV such as GeoVic, the
vehicles, and drone ‘swarms’ online mapping application that enables users to build
• Virtual reality and augmented reality their own maps, using an array of datasets, perform
• 3D printing searches and access data, the GSV catalogue, and the
Victorian Drill Core Library located in Werribee.
• Big Data, Blockchain and Cloud technologies
supported by 5G and Wi-Fi 6 networks
2.1 Benefits of public geotechnical databases
For this paper, three examples of innovation have been
selected to show different ranges of activities which could The following are the key benefits generally identified with
be defined as innovative. A key feature of all three are public geotechnical data bases, many of which directly
the degree of novelty, the kind of innovation (process, assit reducing project risk:
product, service) and the generation of productive
outcomes. • The greater the availability of geotechnical
information to inform design and construction, the
1. The development and benefits of public Geotechnical more efficient and less wasteful design and
Databases. construction can be. This logic would be shared by
most geotechnical practitioners.
2. Incremental innovation approaches for the use of • Increased information and knowledge of wider
InSAR satellite monitoring. geotechnical conditions mitigates the risk of project
delays, cost over runs and environmental impacts.
3. Radical innovation in adopting construction monitoring • Databases present geotechnical logs and other
data into digital models, akin to integrating the Internet information in a single accessible location and
of Things into construction. The Smart Piles platform thereby reducing data management inefficiency,
presented is modular application of the broader optimise site investigation planning and support
concept, which is globally novel from a design and QA scoping, early-stage design and feasibility stage
perspective geotechnical risk assessments.
52
• Reduce the cost of projects both by reduced databases. The NZGD has over 130,000 data points
investigations and less conservative designs with and is used by over 6,000 users.
more informed geological (3D) modelling. • The Queensland Geotechnical Database (QGD) was
• Provide a cost-effective way of compiling and sharing launched in 2017 with a view to consolidate primarily
information, opening opportunities for cumulative tax and toll payer subsidised exploratory
crowd sourcing from smaller, private projects, which investigations. The database holds over 1,600
reduces inefficiencies in retrieving achieved files and geotechnical investigation records.
relying on historic knowledge of projects when • The NSW government is currently working on the
planning investigations. Government Geotechnical Repot Database Project
• The data lives on in the public domain following the (GGRD) to collate geotechnical investigation records.
end of the project and/or agency. • Progressive European nations have recognised the
• Facilitate the research into geotechnical properties of value and importance of geoscience (geotechnical
materials, geological profiling, and education. investigation) data and knowledge (urban geology
• The re-purposing of geotechnical investigation data models) in infrastructure planning and development
from projects for other areas of public good for sustainable cities. Sub-Urban group compiles data
geoscience, examples in greater Melbourne include and models from multiple European Geological
improved understanding of sea level rise and fall and Surveys, Cities and Researchers (including
volcanic eruption frequency. international partners) http://sub-
urban.squarespace.com/
It is estimated the total site investigation saving for
residential property from the Christchurch Geotechnical 2.3 Development of a Victorian geotechnical
Database is in the order of $50-$100 million NZD (Scott, database
et al., 2015). This excludes savings in infrastructure and
commercial development project and qualitative benefits There have been recent changes to Victoria’s public
such as the improved confidence in project risk and more construction guidance. When procuring services for
efficient delivery. geotechnical investigations, or Works or Services that
may require geotechnical investigations, Agencies must
2.2 Geotechnical database examples ensure that their contracts provide for the ownership and
custody of geoscience data collected for the project to be
A of review of relevant geotechnical data bases is transferred to the State of Victoria, where:
summarised, demonstrating the recognised benefit to the
construction industry: • ‘geoscience data’ includes geological, geotechnical
and environmental information, reports, maps,
• In the 1970’s the Urban Geotechnical Automated images, recordings, survey results and drill core, drill
Information System (UGAIS) was developed in cutting and associated materials embodied in any
Canada, funding the creation of geotechnical form; and
databases for 27 cities involving records from over • ‘geoscience data collected for the project’ includes
110,000 boreholes. While several of the data bases geoscience data generated, placed, stored,
are no longer in use, many still are. The Ontario processed, retrieved, printed, accessed, or produced
databases includes data from over 90,000 boreholes using data supplied by the Principal, for the purpose
and in 2015 was accessed on average almost twelve of the contract.
times every workday (Thompson, 2016).
• The British National Geotechnical Properties Within Victoria, the GSV is in the process of scoping the
Database was launched by the British Geological compilation and public delivery of all geotechnical
Survey (BGS) in 1992. The database includes over investigation physical (i.e. drill core) and digital
100,000 boreholes and the web viewer portal is geoscience data acquired as part of geotechnical
accessed over 60,000 times a month (Thompson, investigations undertaken as part of State of Victoria
2016). infrastructure projects.
• The Perth, Australia Central Business District (CBD)
The GSV are working through agreements with current
database was created in 2004 with 649 boreholes major projects and stakeholders on the transfer and
from the 1970’s. release of geotechnical data. The Suburban Rail
• The Canterbury Geotechnical Database (CGD) was Authority Loop Authority (SRLA) is coordinating with GSV
created to generate efficiencies in sharing to provide data from 560+ geotechnical investigations
geotechnical information following the earthquakes of already completed for Stage One of the project.
2010 and 2011 around Christchurch. In addition to Discussions are ongoing for the provision of construction
hosting over 26,000 borehole and CPT records, it also stage face mapping and geological records to supplement
includes regional assessments and maps of the project investigation data.
liquefaction susceptibility. The CDG assisted with the
design and construction of residential, commercial 3 DEVELOPMENT OF INSAR MONITORING
and infrastructure rebuilds, as well as city planning
Interferometric Synthetic Aperture Radar (InSAR) is a
and enabling the insurance industry to understand
radical, innovative satellite-based ground movement
and manage development risks. In 2016 the New monitoring system that is increasing become commonly
Zealand Ministry of Business and Innovation launched place on major infrastructure projects and forensic
the New Zealand Geotechnical Database (NZGD), studies. It significantly enhances and complements
which combined the CGD with other regional traditional survey monitoring approaches.
53
Satellite based InSAR has been applied across a range required to assess movements during construction.
of major infrastructure, mining, forensic and geohazard Additionally, data requires processing, which is
assessment projects. Within the sphere of metro rail typically undertaken in batches of 3-to-6-month
tunnel projects, InSAR came to prominence in Europe intervals.
slightly over 10 years ago on multiple projects (Crossrail • The location of monitoring points cannot generally be
in London, Paris Metro Line 4, Barcelona Metro, Warsaw
controlled. The processing requires consistent
Metro). Within the Victorian context it was adopted on the
backscattering to develop a reliable data point for
Metro Tunnel Project in Melbourne and has been
procured by the Suburban Rail Loop project to baseline processing (such as kerbs, outcrops, or structures)
and understand existing historic ground movements. and may not be reliable over vegetated areas. On
Crossrail the X-band data generated one monitoring
InSAR measures ground movements by comparing two point approximately every 40m2 (Garcia, et al., 2016);
or more Synthetic Aperture Radar (SAR) images of an and
area to identify surface movements over time (refer Figure • “Unwrapping” errors if rapid ground movements occur
1). InSAR enables ground movements to be monitored between satellite passes. Where large settlement (i.e.
over spatially extensive areas to a high level of accuracy <10mm) occurs between satellite passes the
(typically 1-3mm). Radically innovative technology comparison of SAR images can assess the data as
considering the data comes from satellite hundreds of heave instead of settlement. Unwrapping errors in
kilometres above the earth. Benefits of InSAR include:
data can be easily re-calibrated with manual
settlement monitoring if required.
• Extensive spatial coverage which is suitable for
assessing regional ground movement trends. While InSAR can’t replace automated systems, it can
• Can assess seasonal shrink/swell effects, thermal successfully complement them, providing historical
effects (e.g., bridges) existing consolidation and/or ground movements, assessing regional
groundwater related settlement. settlement/heave, reduce the extent of conventional
• Satellites have been collecting and storing data over monitoring and provide cost effective long-term
most urban centres in Australia for more than seven monitoring. On the Crossrail project the procurement and
years, enabling historic data to be processed from data management of InSAR was in the order of a
time periods prior to project commencement and magnitude cheaper than maintenance and data
management of automated data monitoring (Gonzalez, et
assist with establishing baseline ground movements;
al., 2017).
and
• Monitoring can cover sensitive urban areas and InSAR’s massive spatial coverage, high accuracy, regular
structures, mitigating the need for access (e.g., rail readings, and ability to model historical movements make
corridors) and approval (e.g., heritage structures). it a radically innovative approach to surveying.
Limitations with InSAR include:
• Satellites typically collect data every 7-14 days and,

therefore, more conventional automated systems are
Figure 2. Two SAR images of the same area are acquired at separate time intervals and the phase shift in the reflected
wavelength is used to assess the differential movement (Geoscience Australia, 2021)
54
3.1 InSAR incremental innovation
Incremental innovative improvement of InSAR is ongoing

to expand its use and application. Two examples are
provided.
3.1.1 Artificial Reflectors to monitor areas of

low coverage
The Suburban Rail Loop Project is collating historic

InSAR data to baseline existing ground movements and
identify areas of existing ground movement risk. Part of
the project includes the construction of stabling yards over
areas which have a history as former mixed landfill sites
(Suburban Rail Loop, 2021). It could be reasonably
expected that historic landfills have an ongoing settlement
risk. These areas would be a perfect application for
InSAR monitoring, however there is very poor spatial
coverage due to the vegetated nature of the area and
ongoing site activities.
Suburban Rail Loop Authority, working with our technical

advisors and Sixense Oceania have established a series
of InSAR artificial reflectors over the site (Figures 3 and
Figure 4). The reflectors generate a strong and reliable
image which can be readily recognised in SAR data. The
adoption of artificial reflectors grassed areas overcoming
one of the traditional limitations of InSAR and facilitates
accurate, frequent and access free monitoring at a
significantly cheaper cost compared with traditional Figure 3. InSAR reflector installed at the SRL stabling
survey approaches. yards (courtesy of Sixence Oceania).
Figure 4. InSAR reflectors installations being picked up clearly within a grassed area (before and after images courtesy
of Sixense Oceania).
3.1.2 Commercialisation of space such as Capella Space are rapidly launching additional
satellites and within three years, by using readings from
The miniaturisation of satellites combined with the multiple satellites the monitoring frequency may reduce to
increasing availability and affordability of commercial daily, or even multiple times per day. Subject to the
space payloads has significantly reduced the cost of InSAR image overlay and batching process becoming
launching satellites. Subsequently the number of more automated, this increased frequency has the
commercial satellites orbiting the earth is quickly potential to significantly alter how InSAR monitoring
increasing, enabling incremental innovation in the use of integrates with traditional monitoring approaches. This
InSAR monitoring. higher frequency of monitoring has the potential to move
InSAR through a subsequent cycle of innovative growth.
Within the context of InSAR, the increased number of
satellites relates to an increased frequency of readings. Within a broader industry context, the commercialisation
Currently InSAR reading using a single satellite are of space with increased frequency and higher resolution
typically recorded on a two-week frequency (8-14 days), of satellite imagery may have large impacts in fields such
which is one of limitations compared with traditional as coastal engineering, hydrology and tracking
survey approaches. However, commercial organisation
55
construction progress, particularly when combined with to provide insights into construction productivity and
advanced in artificial intelligence and data-driven models. quantities. Smart Piles takes the same data and uses
automated processes to provide:
4 CONSTRUCTION DATA TO DIGIAL DATA
1. Automated quality control reviews.
While advances are being made in developing existing 2. Verification and development of project geological
asset data and designs into digital models and digital models,
twins, to complete the data cycle we need to look at 3. Interrogation of data during construction and
integrating construction data directly back into our visualisation of piling data into BIM as-built models.
models. Currently construction data collection and
automation are typically targeted around increased
productive to reduce construction costs and programme. 4.2 Quality control reviews
One of our next steps as an industry must be integrating
these digital systems directly into our digital design Quality reviews of piling records by designers are typically
models to validate geotechnical models, improve quality complete at the end of the project or when zones of the
control systems, manage geotechnical risks and provide work are completed. Designers normally check grout
efficiencies in automated development of as-built records. pressure and embedment depths to confirm that the as-
built records meet the design requirements. Smart Piles
Collection and use of data from sensors, aka the Internet uses automated processes to review piling construction
of Things (IoT), is already rapidly occurring within other records to verify the quality of each pile as it is constructed
industries such as healthcare, manufacturing, agriculture to provide early warning of construction issues (Figure 5).
and energy. The construction industry is not as advanced, The system effectively completes all necessary QA
but the use of automation and machine learning checks, reducing input from engineers reviewing logs and
processes are advancing rapidly. The industry already providing confidence in the works.
has many established systems ranging from data collation
on automated monitoring platforms, robotics in 4.3 Verification of project design models
construction (e.g. drilling for tunnel ferrules), autonomous
mining trucks, UAV surveys, tracking materials/supplies Manual reviews of piling records will often look at changes
and GPS positioning of plant. in drilling penetration/torque to help validate ground
models and design assumptions. This is particularly
Adopting IoT construction solutions has the potential to important in lateral loading designs including lateral
improve workplace safety (e.g. wearable IoT devices, fall spreading associated with liquefaction, where the depth
detection apps), resources management, reporting and to a firm layer is a key design assumption. On sites with
maintenance, reduce insurance premiums, mitigate variable geological profiles, pile layouts may change
wastage and theft. However, the industry also faces across the site according to changes in subsurface
challenges including employee privacy, data security, profiles.
cost vs benefit and collection of all data vs useful data.
For the time being it is likely that advances in IoT Smart Piles uses automated process to review logs during
construction solutions will come from large building and construction, enabling changes in construction conditions
infrastructure projects where the benefit will outweigh the to be assessed immediately once pile records are
effort to implement new systems. completed and integrated back into the geological
models. This is a more proactive approach to identifying
4.1 Smart piles ground changes and managing ground risks, particularly
on complex sites.
Smart Piles is an innovative product which has been
developed by Arup as a tool for collating and using data 4.4 Interrogation of digital pile data
from piling rigs. The tool collects site data from piling rigs
to visualise progress, undertake quality checks and Once pile data is available in a digital format, it opens new
integrate construction record data back into project opportunities for construction reviews. Within the piling
geotechnical models. Research within Arup has played a industry some organisations are using digital piling
fundamental role in identifying and leveraging business records to review construction programmes and assess
opportunities, with a structured, global approach to the rate of piling productivities.
research engaging over 1,900 staff across 500 research
projects annually (Arup, 2021). Smart Piles can take the automated quality control and
design verification process and integrate them spatially
Most modern piling rigs have sensors which record data into digital models. This enables the automated quality
with depth for piling, soil mixing and ground improvement and design verification processes and checks to visually
activities. Examples of data types include penetration communicated across the site, enabling both issues to be
depth, inclination, drilling torque, rate of progress and address early in the construction process and
concrete pressure (see example in Figure 1). opportunities for refinement to be identified early enough
Traditionally these records are provided as hard copy so that they can have a material impact on the works.
PDFs for a review of quality (embedment depth, grout
pressure etc) or as a record of construction. However, Smart Piles also enables digital as-built models of the
this digital information can also be uploaded to a cloud- piling works to be produced in an automated manner,
based server relatively easily akin to an Internet of Things providing cost savings and improved quality to produce
(IoT) approach, opening various avenues of digital LoD 500 BIM models.
manipulation.
Smart Piles is a step towards closing the gap from ‘design’
Within the construction industry, many operators have to ‘as-built’, using modern digital technologies in an
already similar approaches to extract digital piling rig data automated and innovative way.
56
Figure 5. Example of a piling rig installation records where checks can be completed on consistency or variations in the
data.
5 MANAGING PROJECT RISK THROUGH 6 CONCLUSIONS

INNOVATION
There is a range of ways in which innovation can be
Project risk is commonly defined as an uncertain event of implemented to reduce project risk. In defining
condition that, if it occurs, has a positive or negative effect innovation, the degree of novelty is an important concept,
on a projects objectives. As the construction industry ie adopting an established system which is new to its local
continues to deliver larger and more complex projects, context can still be considered innovative. Similarly, we
improving how we manage project risk is a key priority for can differential between radical and progressive
clients, contractors, and designers. Introducing ideas, innovation as we progress through a products lifecycle.
practices or object that are new (aka innovation) are a
central component to improving how we manage project Three examples of innovation have been presented to
risk. highlight the range of ways innovation can be applied on
projects. 1) The development of public geotechnical
The three examples presented show different ways in databases within Victoria is an adoption of an established
which innovation will continue to shape our industry within structure as a novel tool into an existing market. 2) InSAR
Victoria ranging from adopting established systems, satellite monitoring is an established radical innovation for
implementing emerging technology and looking forward ground movement monitoring which becoming increasing
to the development of future tools and processes. The adopted on projects. Innovation with InSAR technology
innovations described all improve our knowledge on is being incrementally developed with the potential for a
project and reduce the potential for errors and uncertainty, subsequent significant growth cycle. 3) Emerging
effectively assisting us in reducing project risk. Added innovation using data-driven automated systems to
benefits of innovation also generally include program and integrate construction data into digital models using tools
cost savings. such as Smart Piles. These innovations improve our
understand of project hazards, reduce uncertainty, and
However, experience shows that the process of taking subsequently enable us to manage and reduce project
creative ideas, developing them into reality and then risks through planning, design and delivery stages.
successfully implementing within the industry is difficult,
requiring focus and effort. The journey from creativity (an Within this digital age, innovation will continue to play a
idea) to a productive solution (innovation) is difficult and key part in how we manage risks and deliver projects.
frequently does not materialise. Typically of the small The journey from creativity (an idea) to a productive
number of innovations which do emerge, very few make solution (innovation) is difficult and frequently does not
a significant impact on our industry. To continue to materialise. Successful continued delivery of innovation
support innovation we need to establish a culture of within organisation and the industry requires a culture of
embracing innovation and processes to support embracing innovation and a commitment to processes to
innovation though its lifecycle. deliver innovation.
57
REFERENCES
Arup, 2021. Research at Arup. [Online]

Available at: https://research.arup.com/
Edison, H., Ali, N. & Torkar, R., 2013. Towards innovation
measurement in the software industry. Journal of Systems
and Software, pp. 1390-1407.
Garcia, J. R., Black, M. & Gomar, B. S., 2016. Correlation Study
Between In-Situ Auscultation and Satellite Interferometry for
the Assessment of Nonlinear Ground Motion on Crossrail
London. [Online]
Available at:
https://learninglegacy.crossrail.co.uk/documents/correlation-
study-situ-auscultation-satellite-interferometry-assessment-
nonlinear-ground-motion-crossrail-london/
Geoscience Australia, 2021. Interferometric Synthetic Aperture
Radar. [Online]
Available at: http://www.ga.gov.au/scientific-topics/positioning-
navigation/geodesy/geodetic-techniques/interferometric-
synthetic-aperture-radar
Gonzalez, J. M., Nevard, S. & Sanchez, J., 2017. The Use of
InSAR (Interferometric synthetic aperture radar) to
complement control of construction and protect third party
assets. [Online]
Available at:
https://learninglegacy.crossrail.co.uk/documents/use-insar-
interferometric-synthetic-aperture-radar-complement-
control-construction-protect-third-party-assets/
Hannel, A., 2019. The Evil Digital Twin. [Online]
Available at: https://hannellbim.com/2019/08/07/the-evil-digital-
twin/
Henderson, R. M. & Kim, C. B., 1990. Architectural Innovaion:
The Reconfiguration of Existign Product Technologies and
the Failure of Established Firms. Administrative Science
Quarterly, pp. 9-30.
Schumpter, J., 1983. The Theory of economic development: an
inquiry inot profits, capital, credit, interest and the business
cycle. New Brunswick, New Jersey: Transaction Publishers.
Scott, J. W., Stannard, M., Van Ballegooy, S. & Lacrosse, V.,
2015. The benefits of a shared geotechnical database in the
recovery of Christchurch following the 2010 - 2011
Canterbury earthquakes and the potential benefits of
expanding it into a national database. Wellington, Australian
Geomechanics Society .
Suburban Rail Loop, 2021. Stage One train stabling. [Online]
Available at: https://suburbanrailloop.vic.gov.au/Planning/Stage-
One-train-stabling
Tarde, G., 1903. The laws of imitation. New York: H. Holt & Co.
Thompson, T., 2016. A 2016 case for public geotechncail
databases. Gold Coast, Queensland, Australia, Australian
Geomechanics Society, pp. 1045-1050.
Utterback, J., 1971. The Process of Technoiological Innovation
Within the Firm. Academy of Management Journal, pp. Vol
14, pp78.
Design methodology and input parameters applicable to foundation design for large
complex towers
Y.Y. Tay1, BEng (Hons), PhD, CEng, FICE, FIEAust, CPEng; A. K. C. Smith2, MA, PhD, MBA, CEng, FICE, CGeol, FGS;
and C. Haberfield3 BSc, BE (Hons), PhD, FIEAust, CPEng, RPEQ
1, 3
Golder Associates Member of WSP, Building 7, Botanicca Corporate Park, 570-588 Swan Street, Richmond, Victoria 3121, Australia;
email: ytay@golder.com.au, chaberfield@golder.com.au
2
Coffey Geotechnics Ltd., Gardner House, Hornbeam Park Avenue, Harrogate, UK, HG5 8NA, email andrew.smith@coffey.com
ABSTRACT
As buildings are progressively getting taller, traditional methods of design that generally relied on considerations of the
vertical load-carrying capacity of the foundation system, assessed by empirical methods and a lumped factor of safety,
have been largely replaced by serviceability-based methods of design which typically result in an optimised foundation
design. Serviceability-based designs typically rely on powerful commercial software packages to enable advanced
numerical analysis of foundation systems. This paper briefly discusses case studies of foundation design processes
including soil-structure interaction analyses adopted for the serviceability design of tall towers. In order to obtain accurate
building movement prediction from complex computer analysis, it is imperative that materials and ground stiffness
properties be accurately characterised and measured. This paper presents ground stiffness properties measured from
various types of tests at different strain levels (i.e., geophysical testing, pile load tests, pressuremeter and laboratory tests)
that have been adopted as input parameters in the numerical analyses. The higher allowable shaft friction values from
serviceability analysis compared to those from traditional methods, are further justified on the basis of bi-directional static
pile testing.
Keywords: serviceability-based design, ground stiffness, mobilised pile skin friction
1 INTRODUCTION swelling tendency of calcisiltite/ siltstone when extruded

from ground at depth.
Serviceability-based methods of design are based on the
performance and movements of building foundations
rather than strength that renders the building unusable.
Therefore, it is imperative that ground stiffness properties
are accurately measured and appropriately adopted for
foundation design. This paper presents the tests carried
out in-situ and in laboratories for ground stiffness
measurements of weak carbonated rock in Dubai, such
as unconfined compressive strength tests with stiffness
measurements, pressuremeter, geophysical and bi-
directional static pile load tests. It attempts to address the
strain levels at which the different tests were undertaken,
amongst other variables, and therefore producing
different measured stiffness range. These test strain
levels were also compared to levels determined from
numerical analysis of pile group foundations of tall towers
when axially loaded.
2 SUBSURFACE CONDITIONS
Bedrock across the emirates of Dubai for engineering

purposes is the Barzaman Formation Miocene Age rocks
overlain by the more recent Quaternary Age Ghayathi
Formations. The ground conditions discussed herein
underlie four plots at a master development in Dubai,
typically comprising superficial deposits, overlying weak
calcarenite/ sandstone (Ghayathi Fm), followed by
calcisiltite/ siltstone and conglomerate (Barzaman Fm).
Only rock properties are discussed in this paper.
Figure 1. qc vs depth at Plot A4, A7, A17, E14/15
The carbonated rocks are typically weak with unconfined
compressive strength (qc) between 1.5 MPa and 4 MPa 3 LABORATORY AND IN-SITU TESTING
as shown in Figure 1. There is little increase in strength
with depth, and no discernable difference in strength 3.1 Ground stiffness measurements
between calcarenite and calcisiltite layers (interface at
around RL -25 m Dubai Municipality Datum, DMD). This In the laboratory, ELAB was determined from the
was likely the result of unconfined compressive strength displacement of rock samples in UCS tests. Similar to the
(UCS) tests performed without confinement and the
59
Design methodology and input parameters applicable to foundation design for large complex towers Y. Y. Tay, A. K. C. Smith, and C. Haberfield
qc profile with depth, ELAB values of the four plots tend not assessment, GHH from pressuremeter presented in the
to increase significantly with depth (Figure 2). following sections have been corrected to GVH assuming
anisotropy is due to the differences in normal stress acting
In-situ Young’s modulus (EPMT-H) was measured by in different directions. Therefore, stiffness in any direction
pressuremeter tests with unload reload cycles. The is a function of K0 as shown by the following equations
subscript “H” used here refers to measurements taken in from first principles:
the horizontal direction. Unload-reload EPMT-H-ur values GHH = EH / [2(1+ νHH)] (1)
from pressuremeter tests were about 4 times higher than
the initial EPMT-H-ini (Figure 2), the main reason being that For undrained expansion,
the latter was an elasto-plastic measurement where part νHH = 1 - K0 / 2 (2)
of the expansion was irrecoverable. where K0 = EH / EV (3)
Intuitively, EPMT-H from pressuremeter tests which test a EPMT-V is conservatively assumed to be EPMT-H / 2 in this
larger rock mass compared to intact rock samples tested paper, as K0 measured from pressuremeter tests was
in laboratory, should be smaller than ELAB because of the greater than 1.5. Correction of anisotropy of rock may
scaling effect and the fact that the bigger rock mass is partly explain higher EPMT-H compared to ELAB.
likely to consist of more imperfections, noting rock mass
in Dubai do not have much fractures. However, Figure 2 3.3 Stiffness and strain relationship
shows initial EPMT-H-ini to be around 2.5 times higher than
ELAB and unload-reload EPMT-H-ur to be around 9 times Cavity strain measured from the expansion of borehole in
higher than ELAB. In addition, EPMT-H increases with depth pressuremeter tests was converted to shear strain. Shear
and the difference between ELAB and EPMT-H is more strain is the constant area ratio as shown in Figure 3,
pronounced at depth. It is likely that ELAB measurements therefore shear strain of the pressuremeter tests is
were underestimated as a result of the lack of determined from (4) below:
confinement in UCS testing and that the calcisiltite
samples extruded from deeper depths were disturbed as Shear strain = 1 – [ 1/ (1+ɛc)2 ], (4)
they swelled. Other explanations of the difference where ɛc is the cavity strain.
between ELAB and EPMT-H, aside from the above and
scaling effects are likely anisotropy of rock and test strain
levels discussed in the following sections.
Figure 3. Pressure and straining around an expanding

cavity
Figure 4 shows initial and unload-reload EPMT-V, corrected

for anisotropy, versus shear strain. The figure also shows
combined in-situ and laboratory E measurements (i.e.
EPMT-V, EDYN from geophysical downhole and crosshole
testing and ELAB) of the four plots versus their respective
shear strain levels. As E values are generally consistent
at strain levels less than 0.001%, maximum and minimum
EDYN measured was plotted at 0.001% strain in Figure 4.
The shear strain of ELAB has been assumed to be the
strain at 50% peak stress.
From Figure 4, E values increase with decreasing shear

strain which partly explains higher EPMT-V-ur than EPMT-V-ini
and ELAB.
Figure 2. ELAB and EPMT vs depth 4 BI-DIRECTIONAL PILE LOAD TESTS
3.2 Correction for anisotropy Large scale bidirectional static load tests to failure, using
Osterberg cell (O-cell), were carried out at the four plots
The rock mass in pressuremeter tests is loaded in the for the assessment of pile-settlement behaviour and pile
horizontal plane, and the tests give horizontal shear ultimate skin friction. Osterberg cell tests are increasing
stiffness (GHH). Whilst this may be directly applicable to in popularity in Dubai. They eliminate many of the safety
the analysis of radial compression or pile lateral risks of static load tests using kentledge or tension piles.
movement assessment, it may not be applicable for They also have the significant advantage of enabling test
axially loaded piles where the piles and rock are deformed piles to be short enough such that full mobilisation of skin
in the vertical plane. In the absence of direct friction between the piles and ground (i.e. failure) can be
measurements of GHH and vertical shear stiffness (GVH) achieved. The full range of the anticipated pile depths can
measured from geophysical testing for anisotropy
60
be tested by the simple expedient of having O-cells at in the model to that measured. ETP and average shear
different depths. strain along the pile, obtained from the model at pile head
settlement of 1% pile diameter are included in Figure 4.
The ETP values fit within the range of the values
determined from other tests carried out.
Figure 4. E Modulus versus shear strain
Each test pile carried out comprised an upper and lower

Figure 5. Diagram showing the pile arrangement
segment. The segments were generally 5 to 10 m long.
The test pile segments were embedded within selected
depths of the rock based on strata and rock strength
profiles, covering the anticipated pile depths between RL
-10 m and -30 m DMD.
The test pile diameters were 1.2m at Plots A7 and A17;

and 1.5m at Plots A4 and E14/15 and maximum test loads
applied were 48.0MN, 53.5MN, 81.6MN and 81MN
respectively. For each test pile, a number of O-cells were
located at a single level separating the upper and lower
test segments. The maximum test load was dependent
on the number of O-cells that could be arranged across
the pile section at a single level, and maximum jacking
load of each O-cell that was used for testing.
A diagram showing the test pile arrangement and

instrumentation is presented in Figure 5, noting that a
“soft toe” had been included in majority of the test piles
such that end bearing is not engaged during the testing
(Figure 6).
4.1 Back analysed ground stiffness
A total of 7 no. pile tests were carried out at the four plots.
Movements of the upper and lower segments of each test
pile were measured as the segments were jacked apart
by the O-cell. These movements were converted into an
equivalent top-loaded pile-settlement curve for each test.
Stiffness ETP from the pile tests were back-analysed by
Figure 6. Photo showing lowering of reinforcement cage
numerical modelling a single isolated pile using Plaxis 2D
with 0.5m thick polystyrene ‘soft toe’ attached to the
(axis-symmetry model) and matching the pile-settlement
base
61
4.2 Measured pile skin friction test values, which would mean that (5) yields conservative
ultimate skin friction and/or qc values measured from UCS
Average mobilised skin friction values along the upper tests were underestimated.
and lower test segments of each test are shown in Figure
7. Test piles at Plots A17 and E14/15 had longer upper 5 DESIGN METHODOLOGY
test segments (10m) than the lower segments (5m), the
average measured values were lower for the upper The design methodology adopted for the four plots follows
segments as they were not fully mobilised. The maximum two slightly different but related means of achieving
measured skin friction (near full mobilisation) at each site optimised foundation solutions by justifying serviceability
ranged between 1250 and 1700 kPa. analysis, also associated with higher stiffness values
derived in part from pile testing; and by justifying the use
Expressions for ultimate skin friction as a function of qc of higher allowable shaft friction values on the basis of pile
have been derived by several authors, some of which are testing to failure.
shown in Table 1. These can typically be normalised, as
shown in (5): 5.1 Serviceability assessment
Ultimate skin friction = 𝑥𝑥𝑥𝑥 √(qc)0.5 (5) For serviceability-based design, soil-structure interaction
Where 𝑥𝑥𝑥𝑥 = 0.25 to 0.4 modelling of foundations is key. Soil-structure interaction
qc = unconfined compressive strength essentially requires compatibility between a structure, its
foundation and surrounding soil/ rock. This requires two
Table 1. Correlations of ultimate skin friction and qc key fundamentals to be achieved, which is load
equilibrium amongst the structure, foundation system and
Correlations Reference
soil; and compatibly amongst displacement of structure,
0·375 (qc)0·515 Rosenberg and Journeaux pile head and ground at the interfaces. The approaches
(1976) to undertake soil-structure interaction assessment are:
0·22 (qc)0·6 Meigh and Wolski (1979)
ab (qc) Williams and Pells (1981) 1. 3D finite element modelling which incorporates the
ground, foundation and structure.
(0·20 to 0·30) (qc)0·5 Horvath et al. (1983) 2. Split the ground, foundation and structure into two
0·45 (qc)0·5 Rowe and Armitage (1987) models (structural and geotechnical models) and
0·15 (qc) Carter and Kulhawy (1987) iterate between the models until compatible
displacements at the interface of the models have
(0·15 to 0·2) × (qc)0·5 Reese and O’Neil (1988)
been achieved.
(0·40 to 0.8) × (qc)0·5 Zhang and Einstein (1998) 3. 3D finite element modelling which incorporates the
ground and foundation system; and column/ wall
loads input as point/ line loads on the foundation
system.
Assessment using Approach 1 and 2, will account for

structure, foundation system and ground stiffnesses.
However, structure stiffness will not be accounted for in
Approach 3. Furthermore, redistribution of column loads
(if any) as a result of foundation displacement will not be
captured. The robustness of design is often inversely
proportional to the duration required to construct the
model for assessment. Therefore, whilst Approach 1
provides the most robust of designs, it is often
impracticable for use given that the project timeframe is
usually short. It will also require both structural and
geotechnical engineers working on a single model to
ensure specific elements are correctly modelled. The
following discusses methodology and input parameters
adopted for Plot A4. A similar process was followed for all
other plots.
A non-linear strain dependent stiffness curve is ideal as

input for assessment by Approach 1. However due to time
constraints, Approach 2 with characteristic EPMT-H-ini
values was adopted for the pile group design, noting that
the geotechnical software used did not allow for non-
linear stiffness analysis. Characteristic EPMT-H-ini values
were not corrected for rock anisotropy, and were
assumed to lie between EPMT-V-ini and EPMT-V-ur as shown
Figure 7. Average measured skin friction along test in Figure 8.
segments
The strain range of axially loaded pile groups determined
Considering a mean qc of 1.83MPa from Figure 1, ultimate from numerical assessment is around 0.2%, which is at
unit skin friction determined from (5) would range between the higher end of typical strains for foundation indicated
450 and 730 kPa with 𝑥𝑥𝑥𝑥 = 0.25 and 0.4, respectively. by Atkinson (2000) as shown in Figure 9.
These values are considerably less than the measured
62
3. Applying a factor of safety of 2.5 to measured skin

friction from pile testing to near failure. The factor of
safety includes the additional design criterion that
attempts to avoid the full mobilisation of shaft friction
along the piles to reduce the risk that cyclic loading
will lead to degradation of shaft capacity for tall tower
foundations (Poulos, 2017), especially those
embedded in weak carbonated rock.
Figure 10 shows allowable pile skin friction determined

from the three methods, and values from Method 1 were
conservative. Serviceability analysis (Method 2) offers a
more economical design, where the lengths of the piles
were shorter with higher allowable skin friction compared
to Method 1. Should higher allowable building settlement
(i.e. > 50 mm) be considered, the pile design may be
further revised. However, allowable friction values from
Method 2 should not exceed values from Method 3, in
which potential pile skin friction degradation is
considered. Adopting factored friction values from pile
tests (Method 3) alone without settlement checks, would
result in movements outside the limiting tolerance as the
piles were tested to failure.
Potential for further optimisation, subject

Figure 8. E Modulus versus shear strain - Plot A4 to building settlement criterion being met
Figure 9. Typical strain ranges for geotechnical

structures (Atkinson, 2000)
Figure 10. Pile skin friction – Plot A4
Considering stiffness E associated with the strain range
of pile group foundations (around 0.2%) and the back- 6 CONCLUSION
analysed ETP from pile tests (see Figure 8), the use of
EPMT-H-ini, (uncorrected for anisotropy) in the design is This paper discusses the design methodology that follows
justified. two slightly different but related means of achieving
optimised foundation solutions by justifying serviceability
5.2 Pile skin friction analysis with the use of high stiffness values and by
justifying the use of higher allowable shaft friction values
Allowable skin friction (fs) values for Plot A4 were on the basis of pile testing.
determined based on the following methods for
comparison purpose: A brief overview of soil-structure interaction assessment
methods is presented, along with ground stiffness
1. Traditional method of applying a lumped factor of 2.5 measurements from various testing at different strain
to ultimate fs determined from (5) with qc from site levels for serviceability analysis. Case studies show that
investigation. allowable skin friction values determined from
2. Back-analysed from serviceability analysis adopting serviceability analysis were higher than those from
EPMT-H-ini and a limiting total building settlement of empirical formula with a lumped factor of safety. The use
50mm. of allowable skin friction from serviceability analyses were
further justified by measured skin friction values, from
63
large scale bi-directional static pile tests to failure, which

were reduced to avoid the full mobilisation of shaft friction
along the piles. This is partly to reduce the risk associated
with cyclic loading that will lead to degradation of shaft
capacity for tall tower foundations embedded in
carbonated weak rock.
7 ACKNOWLEDGEMENTS
The work presented in this paper was carried out by the

authors on projects in Dubai with the support of their
colleagues at Tetra Tech Coffey. They are thankful to
Antone Dabeet and Nabeel Bux who provided expertise
that greatly assisted the projects.
REFERENCES
Atkinson JH (2000), “Non-linear soil stiffness in routine design.”

Géotechnique, 50(5), 487–508.
Carter JP and Kulhawy FH (1987), “Analysis and Design of

Foundations Socketed into Rock”. Geotechnical
Engineering Group, Cornell University, Ithaca, NY, USA,
Res. Rep. 1493-4.
Horvath RG, Kenney TC and Kozicki P (1983), “Methods of

improving the performance of drilled piers in weak rock”.
Canadian Geotechnical Journal 20(4), 758-772.
Meigh AC and Wolski W (1979), “Design parameters for weak

rock”. Proceedings of 7th European Conference on Soil
Mechanics and Foundation Engineering. British
Geotechnical Society, London, UK, vol. 5, pp. 59–79.
Poulos HG (2017), “Tall Building Foundation Design.” CRC

Press, Boca Raton, USA.
Reese LC and O’Neill MW (1988), “Drilled Shafts: Construction

Procedures and Design Methods”. US Department of
Transportation, Dallas, TX, USA, FHWA-HI-88-042.
Rosenberg P and Journeaux NL (1976), “Friction and end

bearing tests on bedrock for high capacity socket design”.
Canadian Geotechnical Journal 13(3), 324–333.
Rowe RK and Armitage HH (1987), “Theoretical solution for axial

deformation of drilled shaft”. Canadian Geotechnical
Journal 24(1), 114–125.
Williams AF and Pells PJN (1981), “Side resistance rock sockets

in sandstone, mudstone, and shale”. Canadian
Geotechnical Journal 18(4), 502–513.
Zhang L and Einstein HH (1998), “End bearing capacity of drilled

shafts in rock”. Journal of Geotechnical and
Geoenvironmental Engineering 124(7), 574–584.
Optimisation of temporary support design for the Northern portal cut & cover tunnel
J. Zeerak1, Dr. M. Wei2, J Roper3, B Clarke4
1
Associate Principal-Geotechnics, EIC Activities Pty Ltd, 567 Collins Street, Melbourne, Australia; email: Jawad.Zeerak@eicactiv.com
2
Technical Principal-Geotechnics, EIC Activities Pty Ltd, 567 Collins Street, Melbourne, Australia; email: Michael.Wei@eicactiv.com
3
Principal-Structures, EIC Activities Pty Ltd, 567 Collins Street, Melbourne, Australia; email: Jeff.Roper@eicactiv.com
4
Engineering Manager, John Holland Group, 180 Flinders Street, Melbourne, Australia; email: Ben.Clarke@jhg.com.au
ABSTRACT
The West Gate Tunnel Project is a city-shaping project that will deliver a vital alternative to the West Gate Bridge, provide
quicker and safer journeys, and remove thousands of trucks off residential streets. Delivery of the WGTP project is currently
underway by a joint venture of CPB Contractors and John Holland Group. Northern Portal cut and cover tunnel is one of
the major structures on the project which requires excavations to a depth of 22 m to allow the launch of the twin 15.6m
diameter Tunnel Boring Machines (TBM). Design of the retention system comprised 900 – 1500 mm diameter secant pile
wall supported with multiple levels of temporary steel struts. A detailed Soil-Structure Interaction (SSI) analysis together
with a review of proposed construction methodology indicated that a two-level propping arrangement as opposed to
propping at three levels, which is what would normally be expected for a structure of this scale, would be adequate. The
opportunity to remove one level of the proposed steel struts with the potential for a significant reduction in materials and
time was considered critical to the completion of this critical path structure. Removal of one level of props would result in
a reduction of steel tonnage in excess of 1,000 tons, in addition to improving constructability, productivity and safety. This
paper discusses technical aspects of the analysis which enabled the development of the above optimised solution. In
addition, the results of the instrumentation and monitoring and performance of the constructed portal structure will be
discussed.
Keywords: WGTP, Cut & Cover Tunnel, Deep Excavation, Secant Pile Wall, Soil Structure Interaction, FEM, Value
Engineering
1 INTRODUCTION the launch of twin TBM machines for the bored tunnels
and forms the final cut & cover tunnel (tunnel portal) after
The West Gate Tunnel Project (WGTP) is a city-shaping completion of the TBM launch and construction of the
project that will deliver a vital alternative to the existing permanent structural lining. The portal structure is over
West Gate Bridge, provide quicker and safer journeys, 330m in length and up to 22.2 m in depth at the interface
and remove thousands of trucks off residential streets. In of the cut & cover tunnel and the TBM tunnel. Temporary
addition to the construction of twin large diameter (15.6 support for the northern portal excavation comprised the
m) tunnels the project will deliver: following retention systems:
• Widening of the West Gate Freeway from 8 to 12 • Secant pile walls supported with heavy steel strutting
lanes & waling
• Multiple crossings across the Maribyrnong River, • Anchor supported sheet pile walls
connected to an elevated viaduct along Footscray Rd, • Base slab
and • Tension piles
• Multiple bridges, entry and exit ramps across the
eastern and western zones of the project. Propped secant pile wall was adopted for the deeper parts
of the cut & cover tunnel at the interface with the TBM
The project is currently in the delivery phase in Melbourne tunnels, transitioning to the anchored sheet pile wall as
by a construction Joint Venture of CPB Contractors and the excavation depth reduces towards the open trough
John Holland Group. structure. Multiple design types were incorporated in the
design to suit excavation depths and ground conditions
2 NORTHERN PORTAL CUT & COVER TUNNEL along with the portal structure. The portal structure was
designed as a tanked structure (undrained) for the full
The Northern Portal (NP) cut and cover tunnel at WGTP length of the portal to minimise disturbance to the regional
is one of the major packages of the works on the project. groundwater regime. Figure 1 presents a general layout
The portal is used in the temporary condition to facilitate of the northern portal.
Figure 1. Northern portal site layout
65
Optimisation of Temporary Support Design for the Northern Portal Cut & Cover Tunnel J. Zeerak, M. Wei, J. Roper and B. Clarke
It is worth noting that the northern portal cut & cover through the foundation rock which was expected to be
tunnel is a major structure comprising propped secant pile minimal.
wall in the temporary and a structural lining wall
constructed on the inside in the long term. The two
structure types combined forms the cut & cover tunnel
with a significant interaction between the two structure
types. Discussions in this paper are limited to the analysis
conducted for the structure in the temporary condition
only before construction of the permanent structural
lining. It is further noted that the discussion presented in
this paper are based on the analysis, which are limited in
nature conducted as part of a value engineering and
optimisation exercise, detailed design of the structure
were undertaken by others.
2.1 Geological Conditions & Geotechnical Figure 2. Geological Survey of Victoria, Melbourne
Parameters Mapsheet, (1:63,360)
Figure 2 presents surface geology of the general project Geotechnical parameters adopted in the finite element
area highlighted on an extract from the Geological Survey analysis were derived from site investigation data,
of Victoria’s Melbourne Mapsheet. laboratory testing, published literature and past
experience. Parameters for the soil units were generally
The geological conditions at the Northern Portal consists derived from laboratory testing, typical correlations, and
of an upper unit comprising man-made Fill overlying a thin past experience. While rock mass parameters for use in
veneer of the soft and compressible Coode Island Silt the design of excavation support were assessed using the
(light green in Figure 2, not shown in Figure 3). These generalised Hoek-Brown strength criterion using the
thinner units are underlain by a more substantial layer of Roclab (RocData) software by Rocscience. Adopted
Quaternary alluvial outwash, and deeper alluvial infill to a parameters are summarised in Table 1.
paleochannel crossing the northern end of the Northern
Portal (green/dark green in Figure 3). Variable
thicknesses of the Brighton group unit (yellow in Figure 3)
are encountered in parts of the northern portal. The older
volcanics rock (pink in Figure. 3) comprises the bedrock
in the area of the portal where most of the piling and
excavation works within the deeper ends of the northern
portal occur.
Variable degrees of weathering of the older volcanic units

were encountered during piling which comprised residual
soils to extremely and highly weathered rock. Tertiary
aged Werribee formation sediments underly the Older
Volcanics rock. A typical geological section at the
Figure 3. Northern Portal Geological Profile
northern portal is shown in Figure 3. Groundwater levels
vary along with the portal structure and ranges typically
In the analysis, both drained and undrained parameters
between RL0 mAHD to RL-4 mAHD. The lower levels
were utilised relevant to each stage of the construction. In
were associated with the long-term depressurisation
general, short term or undrained parameters were used in
effects of local drains e.g. North Yarra Sewer Main. The
the short term i.e. during top down excavation to the final
portal structure was designed as a tanked structure with
excavation level and casting of the base slab after which
the retention structure comprising predominantly a secant
soil parameters were changed to drained parameters
pile wall which is considered watertight. It is worth noting
relevant to long term conditions.
that a row of recharge wells was proposed along the
perimeter of the excavation to minimise the impact of
The above covers two extremes of soil behaviour.
works on the local groundwater regime. Given the secant
wall system, any groundwater inflow was expected to be
Table 1: Adopted Geotechnical Parameters

Unit Unit r Su c’ Phi’ E’ ν'
[kN/m3] [kPa] [kPa] [o] [MPa]
1 Fill – General 17 20 2 26 12 0.3
2 Alluvium Outwash 18 - 1 33 25/20/75 0.2
3 Paleochannel Alluvium 18 120 10 28 25/20/75 0.2
4 Brighton Group-Clays 19 150 10 30 40 0.35
5 Brighton Group-Sands 20 - 1 34 60 0.3
6 Older Volcanics-Residual Soil 20 150 15 28 60 0.2
7 Extremely weathered older volcanics 21 250 30 28 200 0.3
8 Highly weathered Older Volcanics 21 - 50 40 800 0.3
9 Slightly weathered Older Volcanics 28 - 400 55 6000 0.2
10 Werribee Formation 20 200 20 28 50/40/150 0.2
66
Figure 4. Northern Portal strutting layout
In reality, the soil behaves in a lot more complex manner 3 NORTHERN PORTAL RETENTION DESIGN
than that in the modelling. For the purposes of covering
critical cases in terms of forces on the retention structure The adopted retention system for the majority of the cut &
and support elements, various sensitivity checks were cover structure comprised 900 mm to 1500 mm secant
carried out with changing soil properties from undrained piles propped with multiple layers of steel struts. A secant
to drained at various constructing stages. The impact of pile wall option was required to provide a watertight
the above sensitivity checks on the overall design was structure as required by the project technical
found to be negligible in this case. specifications. The larger diameters 1200 mm and
1500mm diameter piles were adopted for the deeper parts
Excavation works are essentially ‘unloading’ problems. of the excavation in a hard-hard pile configuration, while
This behaviour is best captured using the constitutive soil the smaller 900 mm diameter piles were adopted for the
model Hardening Soil (HS) model in a finite element shallower excavation depths with a hard-soft
analysis e.g. PLAXIS. The HS model is an advanced soil arrangement. It is noted that a hard-hard pile arrangement
model that is able to generate a more realistic soil was incorporated in the design for constructability
response in terms of non-linearity, stress dependency and reasons i.e. to reduce reinforcement congestion and ease
inelasticity of soils. The model requires three stiffness of installation.
parameters (E50/Eoed/Eur) as shown in the E’ column in
Table 1. In the analysis, in general an unload/reload The proposed design of the propping for the secant pile
modulus (Eur) of 3 x Secant modulus (E50) was adopted wall comprised 3 levels of steel strutting connected to a
based on literature and common practice. Although waler system against the secant pile walls. The largest
experience and some cases studies in literature indicate steel struts comprised double 1200WB steel sections
that this ratio can be significantly higher than 3 in some connected to a double 1200WB walers system attached
cases. This may also explain the better than predicted to piles. Steel struts were supported in the middle using a
deflection results for the structure which will be discussed row of king posts supported on bored piles socketed to
in later sections. rock. The initial strutting layout at S2 level is shown in
Figure 4 with a typical section shown in Figure 5.
Figure 5. Typical retention design – Northern Portal
67
The assessment criteria relevant to the design of retention walls as shown in Figure 7, discussion on the interaction
system in the temporary condition included: of permanent and temporary structure will be limited to the
- Stability during excavation above.
- Wall deflection and settlement behind the wall
- Crack widths
- Strength checks
Stability, deflection, and settlement checks were

assessed as part of the soil-structure interaction (SSI)
analysis discussed in this paper, while strength and crack
widths along with other criterion relevant to structural
design were assessed under several load combinations
as part of a detailed structural analysis, which are outside
scope of this paper.
Figure 6. Typical PLAXIS 2D Model
4 SOIL STRUCTURE INTERACTION
In the PLAXIS model, all construction stages were
As noted earlier, in the temporary condition the portal simulated based on the proposed sequence until
structure was intended to facilitate the TBM launch for the completion of the excavation and installation of the base
twin bored tunnels. In other words, the portal structure slab and removal of the S3 prop. The above stages can
was on the critical path for the project. Given the required generally be summarised as below:
excavation depths, length, size and type of structure, - Installation of retention piles
there was a considerable scope of piling and strutting - Progressive excavation and installation of
required. Any improvement in the construction program temporary steel props
(and cost savings) was seen critical to the success and - Construction of base slab and casting of upstand
timely completion of the works. The above warranted a walls (partially completed permanent wall)
detailed review of the structure in particular the temporary
support system. The review targeted geotechnical
parameters, improvements in the finite element analysis
(FEA) modelling approach, construction sequencing as
well as piling methodology some of which is described in
brief here.
The portal structure comprised a secant pile (1200

diameter for sections discussed in this paper) wall
retention system in the temporary to retain the soil and
water before a permanent lining structure was
constructed on the inside. The temporary retention piles
and final structural lining was constructed over several
stages. During the top-down construction, steel props
(typically 2x1000WB) were used to support the retention
piles. As the final structural lining was constructed
progressively in a bottom-up approach, permanent
supports comprising a base slab, road deck and roof slab Figure 7. View of northern portal temporary strutting
replaced the temporary props. Given the relatively towards TBM headwall
complex nature of works and associated construction
stag, involving installation and removal of multiple The level for S1 prop was generally fixed at the caping
supports, a detailed Soil-Structure Interaction (SS) model beam level for the portal structure as they are most
was developed using the finite element software program effective in controlling pile head deflections and from a
PLAXIS 2D. A SSI analysis such as one described above, constructability point of view. The aim of the above
is able to capture interaction between structural elements exercise was to eliminate S2 prop from a design where
(retention piles), ground and water loads particularly as feasible. Therefore, the configuration for the S3 (lowest
the loads changes during the staged construction and level prop) was of significant interest in order to optimise
locked in stress are generated at various stages. In the propping design. An optimum arrangement of S3 prop
addition to above, there is also a significant interaction was considered to be one which would allow removal of
between the retention piles and the permanent structural S2, while limiting wall deflections and providing adequate
lining as the permanent lining is progressively installed support to the retention piles in what would be considered
and temporary supports removed. PLAXS analysis a significant unpropped height between the S1 and S3
showed that loads on the retention piles (and supports) (with S2 removed).
were greatest in the short term before installation of the
permanent supports. This was expected given the limited To achieve the above goal, various iterations were run in
support points in the temporary condition in contrast to the the PLAXIS analysis to test and assess the most optimum
final configuration when the structure is propped at configuration for S3 prop. From the iterations of the
multiple levels with road deck, roof slab and base slab. PLAXIS analysis, it became clear that a lowered S3 prop
Once the structural lining and permanent supports were would generally be a lot more effective in providing lateral
installed, there is a load redistribution and load sharing support to the retention piles. However, the extent on how
occurring between the retention piles and the lining much the S3 props could be lowered was constrained by
structure. Given the focus of the current paper is on the the working space required for the base slab and
temporary supports system until excavation of the final construction of the partial upstand walls. Therefore, the
excavation level and installation of base slab and upstand prop was lowered as far as feasible for practical purposes.
68
It is also noted that small section of the permanent • Settlement markers behind the excavation at regular
structural lining was constructed in early stages after intervals
casting of the base slab which was called ‘Upstand Walls’. • Groundwater monitoring wells; and
The analysis also showed that it was critical that the • Strain gauges on selected steel props to monitor
partial upstand walls were constructed prior to removal of forces on the steel struts.
the S3 prop. These upstand walls would prop the
retention piles at a higher level and therefore reduce The I&M requirement was a critical part of the design
additional forces generated in the piles. Analysis allowing the wall and strutting system to be monitored
indicated that this would result in an increase of forces during progressive excavation and installation of strutting
within the base slab and upstand walls slightly, however, system. This was in particular critical where the S2 level
the benefits would outweigh the small increase in forces. props were omitted. The recommendation was that
The above were relatively simple, yet critical modifications retention piles and S1 level props would be monitored
required to enable a two-level strutting system for spans continuously, and wall deflections and prop forces
of over 15 m between the two props which were compared against design predictions.
considered unprecedented in author’s experience.
The monitoring system would allow for early intervention
In addition to the above, various refinements and should the monitoring results indicate a need based on
improvements were also made in the finite element the wall performance during construction. Accordingly,
modelling resulting in improvements of the overall wall deflections were monitored continuously during bulk
analysis outcome. In the recent versions of PLAXIS 2D, excavations and installation of the steel props until base
the software allows the input of flexural stiffness as slab excavation. Based on the PLAXIS analysis, a
‘elastoplastic’ moment-curvature diagram instead of the maximum wall deflection of 55mm was expected in the
more ‘traditional’ way of defining flexural stiffness of the final stage, upon removal of the S3 prop after the base
plate elements as a linear elastic material. With the slab had been constructed. Inclinometer readings
improved moment-curvature (M-K) input feature, it was obtained during construction indicated maximum wall
found that forces generated within the retention piles were deflections of 20mm which was well below the design
more optimised and considerably lower than those which predicted value and set trigger levels. Typical inclinometer
would have been obtained using the traditional ‘EI’ values profiles showing horizontal wall deflections are shown in
reduced by a factor to account for short- and long-term Figure 9.
cracking of the concrete. The combined effects of the
above led to the possibility of the S2 (middle props) being
removed from the design leaving two levels S1 and S3.
Although there were differences in the pile forces at
individual stages, PLAXIS analysis indicated that
retention pile structural actions envelopes generally
remained similar between a two-strut configuration and a
three-strut system. Proposed prop sizes also remained
adequate for the slightly increased loads with the S2
props removed. Figure 8 shows a view of the northern
portal temporary works completed.
Figure 9. Typical inclinometer profile (horizontal wall

Figure 8. View of northern portal temporary strutting movements)
5 INSTRUMENTATION AND MONITORING Similarly strain gauges installed on the steel struts were
RESULTS monitored during construction to obtain prop forces and
comparison against design predictions. Monitoring results
Given the size and scale of the portal structure, a detailed indicated strut forces within the range predicted in the
and comprehensive Instrumentation and Monitoring PLAXIS analysis. There were variations in the peak forces
regime was recommended as part of the design. The as shown in Figure 10. These fluctuations in the forces
following instruments were installed on piles and struts were attributed to the changes in the atmosphere
along the portal: temperature levels causing expansion and contraction of
the steel struts. Progressive and real-time monitoring
• In Place Inclinometers at regular intervals results obtained at each stage of the construction
• Reflectorless prims at multiple levels on the piles provided confidence in the suitability and adequacy of the
design.
69
Figure 10. Strain gauge data
Observations made from the monitoring results were that methodology indicated possibility of removal of one level
significantly lower wall deflections were recorded of props for a major portal structure on the WGTP. The
compared to those predicted in the design and calculated SSI was completed using the finite element software
in PLAXIS analysis. This is in part attributed to some level package PLAXIS 2D. Results of the FE analysis indicated
of prudent conservatism in the design as would expected that a two-level strutting arrangement opposed to
for the scale and size of the excavation described in this propping at three levels, would provide adequate support
paper. However, other observations and lessons learnt to the 22 m deep excavation at the portal structure. Level
relating to better-than-expected performance of the of the lowest strut (along with other improvements
excavation support system, particularly the retention piles discussed in this paper) was found to be critical in
could be attributed to factors stated below: optimising the temporary support requirements for the
• Stiffness parameters for various rock units were portal structure. A thorough and continuous
assessed using the Roclab software using lower instrumentation and monitoring regime prescribed as part
bound and upper bound test results. In the PLAXIS of the design, enabled a close monitoring of the
analysis, typically lower bound parameters were used performance of the structure and provided confidence
given the scale and size of excavation, associated during construction with the suitability and adequacy of
risks and to cover inherent variability and uncertainty the proposed design. Results of the ongoing monitoring
associated with ground conditions. It appears, that indicated a generally better performance of the retention
encountered ground conditions have responded design than expected. This may be attributed to the 3D
considerably better than those assumed in the nature of the structure towards the deeper ends of the
analysis which has resulted in better performance of portal which was not captured in the 2D analysis and
the retention system. Greater levels of confidence can encountering better ground conditions in particular better
be reached with additional targeted geotechnical rock stiffness than those assumed in the design. This
investigations and laboratory testing to optimise underlines the significance of adequate site investigation
parameters further during design stages. and testing to enable detailed assessment and adoption
of refined geotechnical parameters. The adoption of an
• As noted earlier, an unload/reload (Eur) modulus of optimising strutting design at two levels as opposed to
3xE50 was adopted in the design. It is also possible three levels which would normally be seen reasonable for
that in this Eur/E50 ratio is higher than 3, potentially excavation depths such as those required at the northern
greater than 5. Pressuremeter testing can be portal, provided significant cost savings to the project in
undertaken to obtain more accurate estimates of the addition to an efficient, smoother and safer construction
Eur/E50 ratio. program.
• There is a propping effect at the two corners where 7 ACKNOWLEDGEMENTS

the portal structure meets the TBM headwall. Based
on experience from other similar structures, the corner As noted in the introduction, the discussions presented in
effects can be significant due to soil arching in the this paper are based on limited analysis conducted as part
corners, but also the propping effect of the end walls of a value engineering and optmisation exercise. Detailed
(TBM headwall in this case) to the side walls. Given design of the structures are by others. Authors wish to
the limitations associated with a 2D plane strain acknowledge inputs and support provided by the WGTP
analysis, this was not captured in the analysis tunnel zone personnel during design and construction
undertaken for this structure. However, these phases of the northern portal structure. without whose
beneficial effects can be captured well using a 3D type support this paper would not have been possible. Finally,
analysis. authors wish to thank Dr. Jeff Hsi of EIC Activities for a
peer review of this paper.
6 CONCLUSION
REFERENCES
As part of a value engineering exercise, a detailed soil
structure interaction (SSI) and review of construction GSV, (1974), “Melbourne Mapsheet 1:63,360”, Geological
Survey of Victoria.
SESSION 4
THE ROLE OF DESIGN
IN INFRASTRUCTURE
PROJECTS
Session 4 – The Role of Design in Infrastructure Projects
Supporting Innovative Design and Construction
S. R. Taylor1, CPEng, FIEAust.
1
Major Road Projects Victoria, Level 25, 180 Lonsdale Street, Melbourne; email scott.taylor@roadprojects.vic.gov.au
ABSTRACT
Innovation is at the core of the engineering profession. Innovation is driven largely by the need to increase efficiency,
reduce costs, or respond to increasing complexity. In today’s context, these drivers appear to be converging, with
efficiency, cost and increased complexity almost a baseline for all projects, and the impacts of climate change, sustainability
and the circular economy a significant influence on the future of transport infrastructure. To seek the benefits of innovation
on transport infrastructure projects, Major Road Projects Victoria’s (MRPV) aims to facilitate the minimisation and removal
of barriers and obstacles to innovation. The barriers to innovation include a risk-adverse culture, limited capacity and
capability of resources (both within industry and government), leadership, regulatory requirements and a bureaucratic
culture, and rewards and incentives for the implementation of innovation. Through a series of new initiatives, this
presentation will outline how MRPV is supporting innovation in design and construction of major road projects in Victoria.
To address barriers associated with risk-aversion and capacity, MRPV have implemented a new delivery model that
focuses on a program of projects with incentives for innovative solutions. For barriers associated with leadership,
regulatory requirements and bureaucratic culture, MRPV is leading a program of modernising and updating standards and
specifications including trialling intelligent compaction, and creating of a new technical specific for recycled organics.
Keywords: innovation, transport infrastructure, standards, specifications
1 INTRODUCTION Victoria (MRPV) that aim to reduce and remove barriers

to innovation and enable innovation to be realised across
Innovation is a corner stone of engineering. Clear the entire industry. The specific initiatives being
technological and engineering advancements can be undertaken by MRPV include the Program Delivery
associated with human development. Similarly, in the Approach (PDA) used to deliver projects, and a program
transport infrastructure sector, often, innovative design of modernising and updating standards and
and construction facilitates significant step changes specifications.
industry wide. Without innovation, many of the currently
projects being delivered would not be possible or even 2 DRIVERS FOR INNOVATION
conceivable. Therefore, innovation must be fostered and
encouraged by all participants (whether from Scholarly articles relating to “drivers for innovation” in
government or industry) in the delivery of projects. engineering number in the multiple hundreds of
thousands. It is clearly a well-researched and
Innovation in the context of projects, organisations and commented topic. Common themes in the transport
industry can be viewed differently, and has differing infrastructure sector have been identified by a range of
lasting impacts. The transport infrastructure sector is authors including Wipulanusat et al. (2019), Ozorhon
comprised of a number of organisations (including and Oral (2017), and Salter and Torbett (2003). The
government, contractors, suppliers, consultants etc), drivers of innovation include;
which in turn form smaller teams to deliver projects.
Diagrammatically the relationship between projects, • Innovations developed to increase efficiency and
organisations and industry can be represented in Figure effectiveness.
1. When considering the impact of innovation, targeting • Innovations to reduce costs.
the broader industry will obviously result in the greatest • Innovations in response to increasing complexity
outcome as the benefits are realised over the greatest which can be derived from both internal or external
number of projects. sources. The rise of the ‘mega’ project is also a
large source for innovations driven by a complex
environment,
• Innovations in response to an external crisis which
Industry is becoming more commonly related to climate
change, sustainable practices and the circulate
Organisation economy.
Projec When returning to the levels of projects, organisations

and industry, there is no known documented evidence of
the amount of innovation done at the different levels –
possibly due to the fact it would be difficult to
Figure 1. Relationship between projects, organisations categorically assign innovations solely to a project,
and industry. organisation or industry. Yet at the project level, drivers
for innovation would occur more regularly, but the
This paper begins by exploring the drivers and barriers likelihood of the innovations carrying through to the
to innovation in the transport sector, with a focus on organisation or even the industry would be less common.
identifying the barriers to implementation of innovation at That is, an individual project may develop an innovation,
the industry level. The paper then explores two key no matter how small, but fail to pass that innovation onto
initiatives being undertaken at Major Roads Projects other projects within their organisation, or onto other
72
Supporting Innovative Design and Construction S. R. Taylor
projects within different organisations. This means the While this may be a reality of government entities, this is
risk of losing innovation done by a project is significant, not such a barrier for private industry who are more
as once the project finishes and the team is disbanded, willing to invest in innovation without the certainty of
there is little incentive or desire for the project learns to success. Therefore, to effectively remove this barrier,
pass their innovations onto the next project. In the government and industry need to partner in a manner
transport infrastructure sector this is largely driven by the that makes best use of the limited resources, while still
competitive nature of projects, where organisations will allowing individuals the freedom to explore new ideas or
tend to protect innovations from the broader market. concepts.
For the full benefits of any innovation to be realised by 3.3 Failure of Leadership
the broader industry, it requires an industry wide agency
to oversee innovation for the industry. This is a role Wipulanusat et al. (2019) comment that leadership plays
State Governments in Australia have significant part to an important role in fostering a culture of innovation.
play. Governments set the regulatory and policy “Leaders must find mechanisms to encourage the
environment within which industry operates. They are generation, adoption and implementation of
also the major purchaser of civil infrastructure within innovations,” (Wipulanusat et al. 2019). While the
Australia. Therefore, it is within government’s interest to leaders don’t need to be the creative force behind
ensure innovations succeed at the industry level so that innovations, they are critical in supporting and
their benefits can be fully realised. championing innovation to ensure its success.
3 BARRIERS TO INNOVATION In a highly technical area such as transport

infrastructure, this also requires leaders to have a level
Likewise for drivers, barriers to innovation can equally of technical literacy that they are able to identify
have a significant impact on the success of innovations innovation and encourage its growth. At MRPV this has
and the development of an innovative culture. been achieved by establishing a strong leadership team
that consists of all the necessary skills required for
Wipulanusat et al. (2019) provided a framework for successful project delivery in the transport infrastructure
defining barriers and obstacles to innovation. The sector.
context they have defined barriers are those in the public
sector which impede and delay government agencies 3.4 Regulatory Requirements
from adopting and implementing innovations
successfully. They propose barriers to innovation in the Change and approval processes within government are
public sector include a risk adverse culture, limited heavily controlled and regulated (Wipulanusat et al.
resources, failure of leadership, regulatory requirements, 2019). Further, in the engineering context, standards
few rewards and incentives, and a bureaucratic culture. and specifications play a significant role in defining
Each of these barriers is discussed future in the following acceptable minimum requirements. Many of these
sections. standards and specification are prescriptive in nature,
which inherently limit the ability of innovators to use
3.1 Risk Adverse Culture products, systems and processes that lie outside these
defined boundaries.
Wipulanusat et al. (2019) proposed that due to the
negative ramifications of risk-taking in the public sector, An innovation within or for a government project will
such as political damage to the government or public never be successful if it does not meet the wide range of
criticism, the public sector tends to have a risk adverse regulation, guidelines, policies, standards and
culture. In the context of innovation, this results in specifications. Further, where innovations use new
diminishing likelihood by government agencies of taking technologies that lie outside current regulatory
risks through the adoption of new technology. In the requirements there is the risk they will not be accepted
transport infrastructure sector, this is further as they cannot be measured against a set of
compounded by the increasingly complex legal and requirements to determine if they are suitable of use.
commercial environments, where uncertainty in new
technology, systems or processes leads to the A significant factor in much of the regulatory
suppressing of innovation at project, organisation and requirements that exists are they are prescriptive in
industry level. nature, and as a result the product, process or system
needs to fit within the framework and boundaries of the
A symptom of a risk adverse government culture is the requirements very neatly. An alternate approach is to
creation of an overly conservative and prescriptive develop performance-based requirements, where the
regulatory and legal environment. This then means performance requirements are defined and testing
innovations are constrained by the limitations and procedures described, however the parts that constitute
boundaries imposed by technical standards and the final output are not defined. This enables individuals
specifications. to innovate knowing their product, system or process
only needs to achieve the output performance
3.2 Limited Resources requirements.
Capacity can be considered in both the context of people 3.5 Few Rewards or Incentives
and funding. Often innovation will require development
and funding without knowing the outcome of the Wipulanusat et al. (2019) comment that in government
research being undertaken. Government’s, however, innovators rarely received feedback or a reward for their
operate on clear funding cycles and productive use of success, however “if the innovation fails or does not
resources, leaving little space for the ability to innovate. prove to be efficient, the innovators are responsible for
all the costs.” When translated into the transport sector,
73
particularly for traditional D&C delivery models, there is of the project, and then the Project Delivery Phase once
little incentive for contractors and designers to share contracts are executed. A designer is engaged by the
innovations beyond their projects, meaning they can be contractor in a sub-contract agreement similar to a
short lived and not widely adopted. traditional D&C model. The appointment is made after
the award of the project to a contractor and before the
Therefore, a key feature of delivery models that foster a start of the Project Development Phase, such that
culture of innovation are those that contain some MRPV, the contractor and design all work collaboratively
element of reward and recognition for new technology, through both phases of the delivery model.
processes, or systems.
At its core, the PDA approach features elements of both
3.6 Bureaucratic Culture the Alliance, and Design and Construct (D&C) project
delivery models. The initial Project Development Phase
The barrier related to a bureaucratic culture relates is similar to Alliancing, whereby scope, risks and
strongly to leadership and the risk adverse culture within opportunities are shared and resolved with input from
government. Governments tend to be hierarchical in MRPV, contactors and designers. Once the contract is
decision-making, as well as restrained in their promotion executed, the model converts to more of a traditional
of innovation due to their inherent tendency towards D&C, however payment is through an Incentivised
regulation and certainty (Wipulanusat et al. 2019). Target Cost (ITC) payment mechanism, which
reimburses direct costs and includes cost and non-cost
To overcome the barrier of a bureaucratic culture, MRPV incentives through a performance regime. However, a
has partnered with industry to delivery projects through significant difference of the PDA model from current
the Program Delivery Approach (PDA) which has forms of delivery models is the projects are managed as
enabled projects to offer innovative solutions that a program of works enabling a more sustainable
address performance-based regulations, standards, and contractor and designer market.
specifications. This ensures industry suppliers with new
technology, process or systems can successfully The benefits of the PDA model for industry include;
implement their innovative solutions within the
constraints of government requirements. • a more sustainable supply of projects for industry
partners and their supply chain;
4 INITIATIVES TO ENCOURAGE INNOVATION • creation of long-term competition based on
performance rather than short term, cost driven
To increase and encourage new technology, systems, competition;
and processes within road projects, MRPV has been • improved cost certainty through the reimbursement
working on a number of initiatives that aim to remove and of all direct costs;
reduce the barriers to innovation. The next sections • better sharing and allocation of risk to the party best
outline two key initiatives, namely the Program Delivery able to manage the risk;
Approach (PDA) used to deliver projects, and a program • streamlined procurement processes leading to
of modernising and updating standards and reduce procurement time and costs;
specifications. • streamline procurement processes and, in so doing,
maximise efficient engagement of contractors and
5 PROGRAM DELIVERY APPROACH consultants of all tiers to perform works;
• enable more Front End Engineering and Design
In mid-2020, MRPV implemented a new project delivery (FEED), investigations and assessment of project
model called the Program Delivery Approach (PDA). specific risks; and,
The PDA offers the ability for a portfolio approach to the
• better integrated project planning and project
delivery of projects. The objectives of the PDA are to;
delivery through collaboration between government
and industry.
• Create a more sustainable contractor market, by
engaging with the appropriate industry partner When considering barriers to innovation, the PDA model
based on their capacity and capability. has a number of key features that foster and enable
• Improve the efficiency of project procurement creativity and allow new technology to be more easily
resulting in saving time and minimising State and adopted by projects, as well as shared with the broader
contractor costs. industry. The key features of the PDA model to enable
• Improve project outcomes, contractor performance innovation are explored further in the following sections
and optimise value through a more collaborative and include increased collaboration, risk mitigation and
approach to procurement and delivery, by sharing, and incentivisation for innovation.
incentivising contactor performance, both financially
and via future opportunities. 5.1 Collaboration
The PDA model awards projects to pre-qualified A significant feature of the PDA model is the closer
contractors and designers based on their capability, collaboration between industry (both contractors and
capacity, past performance, and ability to deliver value- designers) and government, as well as the supply chain
for-money solutions. These pre-qualified companies partners critical to the projects delivery.
have been assigned to MRPV’s Construction and Design
Panels. The panels divide industry partners into broad By engaging a single team through the project
groups based on their capability. development phase, risk and opportunities can be
rigorously assessed and mitigated, which would not
Once selected, a contractor will follow a two-staged normally be achieved in a more traditional D&C projects.
delivery model under the PDA, entering into a Project Further, the collaboration between government and
Development Phase to determine the scope and value
74
industry means that risks can be assigned to the party of Transport (DoT), who are the owner and operator of
best able to mitigate those risks. It also enables a the assets delivered by MRPV.
greater adoption of innovation, with industry and key
suppliers able to bring forward ideas, with MRPV A significant barrier to new products and new
facilitating and enabling the adoption of the innovation. technologies is that standards and specifications don’t
allow or even consider the application of these emerging
5.2 Risk Mitigation and Sharing materials or concepts. This is a result of standards and
specification being largely prescriptive in nature, where
Risk and risk management is dealt with very differently the minimum requirements define the inputs and
in the PDA model when compared with traditional D&C describe what must be done to achieve the
contracting. Considering only a single contractor and requirements. Prescriptive standards and specifications
designer is taken through the project development phase are best suited to situations where a high level of control
with MRVP to develop a target outturn cost (TOC), it is required and appetite for risk is low. However, under
enables a shared understanding and contribution to risk this framework, changes in technology or methodologies
mitigation. The eventual aim of the TOC development is can not be easily adopted as it is unlikely the prescriptive
to understand and eliminate risk by utilising resources, standards or specification would have accounted for
knowledge and relationships across both government future innovations.
and industry. Traditionally, in a D&C environment a
contractor would assign a cost to items or scope they In an innovative culture, regulations (including standards
could not effectively mitigate or eliminate the risk. and specifications) should be flexible and adaptive to
Whereas, in the PDA model, with input from MRPV, enable changes in technology and new practices. This
project teams work to identify and mitigate or eliminate is achieved using performance-based standards and
all risk during the TOC development. specifications which focus on the requirement outcome.
Under this framework, products much achieve a
When considering the barriers to innovation, this model minimum performance using which ever means they
enables novel problem solving and innovative solutions wish. Typically, the performance-based standard or
to reduce or remove risk during the development phase specification will require a series of tests to be
of the project. Further, MRPV is able to draw on undertaken to verify the product, process or system
innovative solutions from across the industry to facilitate achieves the minimum requirement. The risk with this
solutions for individual projects. approach is that the suite of performance requirements
may not account for a behaviour or outcome that was not
5.3 Incentives for Innovation intended to be achieved.
The final feature of the PDA model that encourages In practice, it is very difficult to write a standard or
innovation is the use of financial and non-financial specification entirely using a performance-based
incentives for those projects that develop and adopt framework. Therefore, what more commonly happens to
innovative solutions. Specifically, there is a KPI (key enable an innovative culture, is that standards and
performance indicator), for both contractors and specifications are written predominantly with
designer who develop innovative initiatives for not only performance-based requirements and use more
the project, but that can be adopted more broadly in the prescriptive-based requirements for elements deemed
industry. The assessment of the innovation KPI is based high risk.
on the type of innovation(s) adopted, and defined as;
Looking back at the barriers to innovation, the move
• Continuous improvement is defined as an towards largely performance-based standards and
initiative which is a continuous improvement specifications enables removal of barriers associated
initiative within the project. This is the lowest level of with risk-aversion, regulatory requirements and
innovation that will trigger the KPI. bureaucratic culture. The basic premise of defining the
• Innovation adoption is an initiative which is performance outcome rather than prescriptive
adopted from another project. This is the starting requirements means the basic elements that have led to
point for more broader industry wide innovation and risk-aversion, regulatory requirements and bureaucratic
encourages teams to look at the practices of other culture are retained, in that risk and uncertainty are
projects to implement a new technology, process or managed, however the performance-based approach
system. enables suppliers to propose new technologies,
• Industry application is the highest level achievable techniques and processes.
within the KPI and is defined as an initiative which
has industry wide application. 6.1 Key Standards and Specifications
Contractors and designers are judged on the number The range of standards and specifications MRPV is
and type of innovations adopted by the project, and given working on with DoT is diverse and covers the full range
short term financial incentives, as well as longer non- of technical disciplines within the road transport sector.
financial incentives related to their ongoing performance. It includes new and updated standards and
specifications for plastic noise walls and slip form
6 MODERNISING STANDARDS AND barriers, as well as guidelines for application of spray
SPECIFICATIONS seal pavements. However, the focus of this paper is a
couple of initiatives associated with geotechnical aspects
The second key initiative MRPV are undertaking to drive of road infrastructure, including trialling the use of
innovation at the industry level is through a program of intelligent compaction for road pavement, and creating a
modernising and updating standards and specifications. new technical specification for recycled organics.
It is acknowledged that much of the work associated with
this initiative is done in conjunction with the Department
75
6.1.1 Intelligent Compaction prescriptive standards and specifications can limit the
type of materials used in civil infrastructure projects. To
Compaction is the process whereby material is placed in conform to a highly prescriptive-based standard or
layers and compressed to increase the density and specification, recycled materials must achieve the same
uniformity of the pavement material. Compaction is inputs as their virgin material counterparts which can be
achieved by a roller and applied to all layers within a road unrealistic.
profile (e.g. subgrade through to asphalt).
In the landscaping discipline, this is particularly relevant
Currently, the most widely used technique used to for recycled organic materials. The Australian Standard
confirm the achievement of compaction is a combination AS 4454 (2012) Composts, Soil Conditioners and
of in-situ spot test with a nuclear gauge density device Mulches specifies the “requirements for organic products
and proof rolling. These methods are limited in that they and mixtures of organic products that are to be used to
only evaluate a small portion of the entire road amend the physical and chemical properties of natural or
composition (e.g. approximately 1 %) and as a result artificial soils and growing media.” The standard is
may miss weak or unqualified compaction areas. largely a prescriptive-based set of requirements, where
physical and chemical requirements of nominated virgin
Intelligent Compaction (IC) is a compaction technology materials are described.
that uses “vibratory rollers equipped with the real-time
kinematic (RTK) Global Positioning System (GPS), Recycled organic materials don’t easily conform with
roller-integrated measurement system (normally requirements of their virgin counterparts. Specifically,
accelerometer-based), feedback controls, and onboard recycled organic materials can vary in quality largely due
real-time display of all IC measurements,” (Chang et al., to the (i) the extent of impurities within the recycled
2011). It is starting to be adopted in Europe, Japan and material (e.g. plastics, paper etc), and (ii) the extent of
North America. processing time allowed to create the compost or soil
conditioner. To enable recycled organic materials to be
The benefits of IC include (Chang et al., 2011); used on civil infrastructure projects, MRPV in
collaboration with DoT and ARRB, are in the process of
• IC allows mapping of each compaction layers to developing a technical specification for recycled
enable real-time identification of weak spots for organics. The intent is to define the performance
corrective actions prior to the compaction of the requirements for a range of uses of recycled organic
upper layers. material, that link back to the classifications used within
• IC provides the means to maintain a consistent AS 4454 (2012). These include for landscaping (e.g.
rolling pattern for 100% coverage of a construction planting, hydro-mulch, turf topdressing, soil conditioning,
area. fertiliser), erosion control (e.g. soil compaction
• IC enables consistent rolling patterns under lower mitigation, filter berms, and compost blankets), and
visibility conditions such as night paving operations. biofiltration (e.g. filter media). By developing a technical
• IC technology enables digital record collection for specification, the innovative use of recycled organics is
future investigation. available for all projects within the industry.
In 2019, DoT developed guidelines for IC trials 7 CONCLUSION

(Papacostas and Walker, 2019). The guidelines
describe how to set-up production trial work lots for Government has a significant role to play in ensuring
placement, compaction, testing and assessment for a innovations flourish and thrive across industry. The risk
single material type. Previously, the requirements within of innovations developed by industry and suppliers being
VicRoads Technical Specification 204 were prescriptive stifled by barriers imposed by government are
in their requirement to undertake some type of nuclear significant. The case of encouraging innovations is
gauge testing and proof rolling. compelling with drivers for innovative solutions including
increased efficiency and effectiveness, reduced costs,
Through support from DoT, MRPV engaged the the ability to respond to increasingly complex
Australian Road Research Board (ARRB) to develop a environments, and the ability to respond to external
set of guidelines for project to successfully implement IC crisis.
monitoring. The guidelines build on DoT’s work and
assist projects in the development of the testing To facilitate innovation, several barriers must be
procedure, including suggesting IC equipment, quality overcome including the public sector risk adverse
control processes, and supporting onsite construction culture, limited resources, failure of leadership,
methodologies. regulatory requirements, few rewards and incentives,
and a bureaucratic culture.
Currently there are a few IC trials underway on MRPV
projects resulting in the innovation being adopted across MRPV has initiated two key programs that aim to reduce
the industry. Further, the flexibility of the guidelines and remove barriers to innovation. The specific
enables projects to utilise as little or as much of the IC initiatives being undertaken by MRPV include the
technology as they wish. Program Delivery Approach (PDA) used to deliver
projects, and a program of modernising and updating
6.1.2 Recycled Organics standards and specifications.
As waste becomes an ever-increasing problem for 8 ACKNOWLEDGEMENTS

governments, the Victorian Government has developed
the Recycled First Policy, whereby projects are The author wishes to acknowledge a number of people
encouraged to optimise the use of recycled and reused and organisations who have been instrumental in the
materials. The obvious weakness of this policy is that development and implementation of both the PDA model
76
and modernising standards and specifications. These

include Allen Garner (MRPV’s CEO) and Steve Cornish
(MRPV’s COO) for their leadership of the MRPV PDA
model, Leon Choo (MRPV’s Quality Manager) for his
leadership of the Intelligent Compaction trials at MRV,
and ARRB and DoT support of both the Intelligent
compaction trials and development of the recycled
organics technical specification.
REFERENCES
Wipulanusat,W., Panuwatwanich, K., Stewart, R. A., and

Sunkpho, J. (2019), “Drivers and barriers to innovation in
the Australian public service: a qualitative thematic
analysis.” Engineering Management in Production and
Services, Vol 11 (1), 7-22.
Ozorhon, B., and Oral, K. (2017) “Drivers of Innovation in
Construction Projects.” Journal of Construction
Engineering and Management, 143 (4).
Chang, G., Xu, Q., Rutledge, J., Horan, B., Michael, L., White,
D. and Vennapusa, P. (2011) “Accelerated
Implementation of Intelligent Compaction Technology for
Embankment Subgrade Soils, Aggregate Base, and
Asphalt Pavement Materials.” Federal Highway
Administration, Report No. FHWA-IF-12-002.
Papacostas, A. and Walker, A. (2019), “Guidelines for
Intelligent Compaction (IC) Construction Trials.”
Department of Transport, Victoria.
VicRoads, (2015) “Technical Specification 204: Earthworks.”
Department of Transport, Victoria.
Australian Standard, (2014) “AS 4454 (2012) Composts, Soil
Conditioners and Mulches.”
77
BIM to Numerical Modelling Interoperability for

Geotechnical Design of Underground Metro Station
M. Q. Huang1, X. L. Chen2, T. J. Sia3 and Q. B. Zhang4
1
Monash University, VIC 3800, Australia, email: Mengqi.Huang@monash.edu
2
Monash University, VIC 3800, Australia, email: Xilin.Chen@monash.edu
3
SMEC, Melbourne, VIC 3008, Australia, email: TongJoo.Sia@smec.com
3
Monash University, VIC 3800, Australia, email: Qianbing.Zhang@monash.edu
ABSTRACT
Building Information Modelling (BIM) is one of the important processes being adopted by the construction industry as it
provides a collaboration platform in conjunction with technical standards for interoperability over the lifecycle of an asset.
However, geotechnical analysis engaging numerical tools has yet to leverage the BIM benefits due to the lack of effective
interoperability means, which not only results in unnecessary remodelling and rework at a cost of labour and computational
waste, but also with possibility of errors, misinterpretation and omission of information. Using a trinocular underground
station as an example, a workflow underpinned by heuristic techniques is proposed to enhance the interoperability between
a BIM design authoring tool (Revit) and a numerical modelling tool (FLAC3D) for geotechnical analysis. A BIM-based
multiple LoD (levels of detail) model framework is proposed to represent different information requirements for different
purposes of BIM use throughout the project lifecycle. Leveraging the associated geometry and semantics, techniques of
parametric modelling, data manipulation via visual and traditional programming are engaged to bridge BIM and numerical
modelling for geotechnical analysis on different design scenarios. The simulated results are visualised through a backward
automation cycle to Revit for design optimisation. The presented solution offers and automates an error-free design-to-
design workflow solution and therefore enables efficient exploration of design scenarios and design optimisation.
Keywords: Building Information Modelling (BIM), numerical modelling, automation, underground, metro station
1 INTRODUCTION information requirements on the exchange scenario of

BIM design to numerical simulation for geotechnical
Building Information Modelling (BIM) is one of the analysis of underground station, the development of a
important processes being adopted by the construction multi-level BIM parametric station model considering the
industry as it provides a collaboration platform in varying richness of information requirements, and the
conjunction with technical standards for interoperability innovation of an intermediary solution based on
over the lifecycle of an asset (Azhar, 2011; Gu and automating information interchange workflow between
London, 2010). While many analyses are now performed design and numerical simulation for geotechnical
in computational environments capable of 3D analysis.
representation and object-based design, there remain
limitations of model-based collaboration and 2 RELATED STUDIES
multidisciplinary integration in the underground
construction processes, such that activities of disciplinary This section provides a brief literature review on aspects
analyses are still performed in non-BIM environment, for of information modelling in underground construction,
example spatial coordination of geological model and multi-arch cavern design, stability analysis and existing
infrastructure model often conducted in geographical interoperability efforts of information and numerical
information system (GIS) and structural analysis in modelling.
numerical modelling programs (Huang et al., 2021). One
important process often being excluded from the BIM Information modelling platforms developed with a
environment while bearing significance to engineering modular or layered structure have been proposed in
success of underground construction is the geotechnical existing research (Huang et al., 2021, Koch et al. 2017,
assessment based on analytical, empirical or numerical Zhu et al., 2017). The importance of geological and
models. Heterogeneous applications are often used to geotechnical information is well understood for
complete this task without referencing to the BIM model underground construction design. However, much of this
that is supposedly the single source of information. This data is either proprietary or stored in formats requiring
inadequately coordinated workflow not only results in the specific software to view. In order to make such
iterative, manual retrofitting with high labour and information more accessible for future project planning
computational cost, but also inherits great risks of and preliminary-stage design, some online infrastructures
misinterpretation and omission of information (Ninic, are built that allow public to interactively search, view and
2021). Therefore, this paper proposes to build a link use borehole data according to their geolocations. Figure
between BIM and numerical modelling to enable the 1 illustrates an example of this type of platforms using the
prediction of geomechanical consequences of design “Data and Information on the Dutch Subsurface
variations via the parametric BIM modelling, parameter (DINOloket Netherlands, 2019)”. With the increasing
visualisation and optimisation, and programming. This accuracy and data type coverage, future engineering
paper focuses on improving interoperability between BIM design for underground construction could better
design authoring and numerical simulation-based leverage the open-source geological and geotechnical
geotechnical analysis for an underground station from data.
several perspectives, including identification of the
BIM to Numerical Modelling Interoperability for Geotechnical Design of Underground Metro Station M. Q. Huang, X. L. Chen, T. J. Sia and Q. B. Zhang
Figure 1. Demonstration of creating synthetic borehole and cross-section (Top: the main interface for user
inputs including selection of model of interest, locating synthetic borehole and viewing custom-defined cross-
section; bottom left: the resulted information of synthetic borehole including relevant geographical and
geological data; bottom right: the cross-sectional geological information). (Huang et al., 2021)
A multi-arch cavern is a common structure adopted for common means to simulate different engineering
underground metro stations with columns and walls scenarios of underground excavation and structure short-
providing support to the wide-span. Examples of stations term to long-term stability. Since BIM models consist of
using such structure include Moscow Mayakovskaya geometry and semantics, BIM could act as a pre-modeller
station (Shilin et al., 2016), Tokyo Kiyosumi-shirakawa and post-viewer for numerical modelling, the BIM-
Station (Konda, 2003), and Beijing Badaling Great Wall numerical modelling interoperability, therefore, allows an
station (Li et al., 2020). Figure 2 illustrates an example of error-free exchange of information between the two
typical triple arched cavern and support systems. The interfaces.
construction methods or philosophies, such as sequential
excavation method (SEM), sprayed concrete lining (SCL),
New Austrian Tunnelling Method (NATM), and the
‘Analysis of the Controlled Deformation in Rocks and
Soils (ADECO-RS)’ emphasised the significance of
excavation and construction with adequate observation
and support measures. The components prescribed in
these methods should be modelled. Their mechanical
properties should be taken care of given the information
modelling approach is undertaken.
The stability analysis of the cavern structure is essential Figure 2. Example of a typical triple-arched cavern and
to geotechnical engineering. With computer software and support systems.
information technology becoming easily accessible,
numerical modelling methods such as the Finite Element BIM encompasses a set of tools, technologies and
Method (FEM), Finite Difference Method (FDM) and processes ranging from parametric modelling (Ninic et al.,
Discrete Element Method (DEM) have been used as
79
2021) to project lifecycle management (Wang and Zhang, setting of meshing parameters, initial and boundary
2021) that are driving the digital transition of the conditions, and excavation and ground support
Architecture, Engineering and Construction (AEC) properties.
industries. Many of the benefits suggested by adopting
BIM and digital information technologies, such as 3.1 Multi-LoD station BIM
improved visualisation and cross-disciplinary
communication and collaboration, are desired for The station model adopts the concept of “Level of Detail
underground construction. For example, the downstream (LoD)”, which specifies what objects are included in the
application of BIM has been investigated for tunnels. In model and the degree of detail with which each object is
Ninić et al.(2018, 2020 a and b), BIM-to-FEM and BIM-to- modelled (Sacks, 2018).
IGA workflows were developed, incorporating machine
learning techniques, to deliver solutions that focus on
automating model generation and simulation for
numerical analysis on shield tunnelling induced
settlement and decision aid on parameter optimisation in
soil condition. Fabozzi et al. (2021) has also studied BIM-
to-FEM and FEM-to-BIM interoperability via the use of
several software tools, including commercially available
rail design modelling package.
The existing research demonstrates the necessity and

achievement to date of BIM uses in underground
construction and solutions of bridging information Figure 3. Workflow proposed in this paper
modelling and numerical modelling. Nevertheless, the
outlined prototypes and modelling methods have yet been The LoD required for 3D modelling varies depending on
generalised to facilitate the information modelling of all the specific BIM goals and the types of model used
underground space usages. Besides, among those throughout the project’s lifecycle. The modelling process
considered the interoperability between the information for any large infrastructure project, such as erecting a
and numerical modelling, the engaged software tools may skyscraper in the urban centre, or constructing a train
not have the built-in constitutive models to characterise tunnel running between opposite ends of a city, often
the materials that make up the model or not necessarily involves structures with widely differing scales, geometric
capable of representing the large deformation of wide- appearances, and spatial arrangements. Therefore, it
span cavern structure. More specially, the workflow of requires the participating engineers and designers to
Ninić et al focuses on circular tunnel in soil conditions, define carefully the LoDs for the to-be-modelled objects in
indicating relatively simple geometrical efforts and order to address this peculiarity of infrastructure design.
singular support structure (lining) involved in both BIM
and numerical modelling process, and the specific Figure 4. demonstrates examples of a triple-arch station
meshing tool (GiD) and the research-oriented open- model with different LoD emphasising the variation of
source simulation software program (Kratos) are model content in terms of excavation support. At a low
engaged. Whereas the solution proposed by Fabozzi et level of detail, it mainly includes the spatial illustration of
al. (2021) is based on use of a commercialised software excavation support means such as the major primary and
suite designed for railway, which could be leveraged for secondary support while the high LoD-model includes
analyses in similar condition but may lack the versatility physical elements of the primary (e.g. forepoling and rock
given that the required design or analysis exceeds the bolts) and secondary support (e.g. cast-in-place concrete
capability of the software packages. lining). In this study a LoD300 model is adopted for the
use case of geotechnical analysis.
This paper is therefore founded on the need of examining
solutions to enhance interoperability between BIM 3.2 Parametric modelling
authoring tool and numerical modelling software codes
that are particularly suitable for cavern stability problems. From a historical perspective, BIM model generation and
design technology are evolved and matured based on 3D
3 METHODOLOGY solid modelling, which represents the ability to generate
and revise arbitrary 3D solid, and reaching its climax in
Using a tri-arch underground station as an example, the object-oriented parametric geometry modelling. The then-
information exchange between two main interfaces state-of-art method integrated two forms of 3D solid
investigated in this paper are Revit and FLAC3D. A tri- modelling techniques, namely the boundary
arch station BIM model will be established under the representation (B-rep) and the Constructive Solid
multiple levels-of-detail (LoD) framework that intents to Geometry (CSG), to realise functions of editing,
represent the different information requirements for the visualising, measuring, clash detection as well as other
BIM model throughout the lifecycle of the construction. non-editing uses (Sacks, 2018). Thereafter, the solid
Leveraging the associated semantics, data mapping and modelling computer-aided design (CAD) systems were
parameterisation is achieved via visual programming, improved by recognising the connectivity of shapes
through sharing parameters and building links.
which also facilitates visualisation of numerical results
Eventually, the realisation of automatic update and rebuilt
back in the BIM environment. The two-way workflow was
of shapes via parametric relations marked the era of
automated via the use of Python interface. Figure 3 parametric geometry modelling.
illustrates the proposed workflow, which incorporates
BIM-based alignment selection, visual programming Parametric modelling is adopted for reusability and
generating parametric station models along with other extensibility, which refers to establishing intelligent
project data to feed the exchange enabler and to allow objects and object assemblies described with
80
parameters. The to-be proposed methodology largely parametric families are created for this study, mainly
relies on computational design concepts, which have concerning the geotechnical analysis that will be
been in use for architectural designs since emerging in performed. Example of an outer shell and rock bolts as
the 60s, and has more or less influenced the development forms of excavation support are illustrated in Figure 5.
of BIM tools. Both parametric design and generative The triple arched primary lining as the outer shell is
design are popular terms considered in the scope of essentially created by extrusion, which is a simple
computational design (Caetano et al., 2020). Establishing parametric modelling procedure.
a component (referred to as a family in the selected BIM
software), for example a bolt, based on constraints of
parameters and rules denotes parametric design for a
single element.
(a) Primary Lining
(a) LoD 100
(b) Rock Bolts

Figure 5. (a) Primary lining family as a type of
support and(b) assembly of rock bolt family
instances.
(b) LoD 200
Then, we propose to use computational design approach
in conjunction with parametric modelling and visual
programming to account for assembly and construction
constraints. This approach holistically considers the
elements and variations of elements implying regional
geological conditions, excavation sequence and support
installation, so that design flexibility and convenience of
modification and retrofit can be achieved. Figure 6
(c) LoD 300 illustrates the layout of visual programming procedures
engaged to create a station model ready for geotechnical
Figure 4. A triple-arch station model with LoD 100 to analysis.
LoD 300 (construction sequence and support
systems are modelled)
Parametric modelling is adopted for reusability and

extensibility, which refers to establishing intelligent
objects and object assemblies described with
parameters. The to-be proposed methodology largely
relies on computational design concepts, which have
been in use for architectural designs since emerging in
the 60s, and has more or less influenced the development
of BIM tools. Both parametric design and generative
design are popular terms considered in the scope of
computational design (Caetano et al., 2020). Establishing
a component (referred to as a family in the selected BIM
software), for example a bolt, based on constraints of Figure 6. visual programming layout of instantiating
parameters and rules denotes parametric design for a lining, ground and rock bolts (controlled by alignment
single element. and faces) and generation of batch files for numerical
simulation.
BIM applications and BIM enabled platforms usually
provide an extensive set of predefined parametric object 3.3 Model semantic enrichment
classes and families while allowing users to customise
when a desired parametric object does not exist in the As aforementioned, geological and geotechnical
specific BIM tool. However, underground infrastructure quantification is of paramount importance to numerical
such as the metro station often do not reflect the target simulation for ground stability analysis. The process of
functionality of most BIM tools designed for architectural assigning or embedding related attributes in the
and building modelling, constructing custom parametric geometries is regarded as semantic enrichment in BIM. In
objects and families is inevitable. Several design-based this study, we dedicated another visual programming
81
algorithm to achieve this goal. The stored parameters for

a rock bolt as an example could include density, grout-
cohesion, grout-friction, grout-stiffness, yield-
compression, yield-tension, and young’s modulus. These
properties are then read and input by Python script into
numerical simulation program to perform calculation
concerning geotechnical problems such as ground
settlement, structural deformation and yielding.
3.4 Information exchange and process execution
The interface selected for numerical simulation for

geotechnical analysis is FLAC3D, which supports a
command-driven program execution and allows the use
of data files (.dat format) to describe the model (Itasca,
2019). This supposedly outdated program design is
actually favoured in information exchange scenarios as it (a)
makes the complete automation of program execution
possible.
An exchange enabler is proposed to facilitate exchange

of BIM model and automate the execution of numerical
simulation. The enabler is realised by a set of Python
scripts under a modular structure to allow future
extensions and convenient modifications. The modules
are constructed in accordance with the general solution
procedure of the numerical modelling program. The
processes of data file generation and process execution
is illustrated in Figure 7.
(b)
(b) simulation execution
(a) exchanger (c) simulation result

(c)
Figure 7. (a) batch file generation enabled by the
exchanger; (b) program execution and (c) simulation Figure 8. Visualisation with (a) contour coloured map
result plotted in the numerical simulation interface. in visual programming interface, (b)mesh distortion
and (c) contour coloured map illustrating deformation
3.5 Results Visualisation of a cross-section via cutting plane perpendicular to
the triple-arch cavern.
A back-analysis loop is enabled also through a dedicated
algorithm to help visualise the impacts of underground 4 DISCUSSION
construction in the BIM native environment. By combining
the settlement effect with existing structures such as The workflow consisted of three parts: (1) generation of
buildings, existing tunnels, or geological structures, structural model of underground metro station
design variations could be proposed and tested following encompassing both geometries and semantics via
the same procedure. parametric modelling and visual programming; (2)
automation of model exchange and calculation
Both coloured contour and mesh distortion are achieved commanding by python scripts; and (3) development of a
to visualise the settlement results obtained from reverse loop to visualise numerical calculation results in
numerical simulation. Since parametric modelling is used, BIM environment for further analysis, parameter
parametric study could be easily facilitated by adapting, optimisation and scenario refitting. The workflow is
for example, the overburden thickness, station cavern marked by its modularity and extensibility that could be
shape, lining thickness and installation pattern of ground easily adapted for further development. Each part is valid
support systems. as a single system while creating an efficient BIM-to-
82
numerical modelling loop for geotechnical analysis. Tunnelling and Underground Space Technology 108,
Currently, the exchange module is developed for the 103677.
specific interfaces that it is connecting, this means that Koch, C., Vonthron, A., König, M., 2017. A tunnel information
modelling framework to support management, simulations
certain adaptations are required if other software tools are
and visualisations in mechanised tunnelling projects.
employed for BIM and numerical simulation. Another Automation in Construction 83, 78-90.
limitation the current exchange module may suffer from is Konda, T., 2003. Reclaiming the underground space of large
the level of complexity in meshing. The success of FEM cities in Japan, Proceedings of the ITA World Tunnelling
or FDM simulation largely relies on the quality of meshes Congress, pp. 1-13.
while examples used to testify the workflow have only Li, R., Zhang, D., Fang, Q., Liu, D., Luo, J., Fang, H., 2020.
involved relatively simple geometrical structures without Mechanical responses of closely spaced large span triple
comprehensive set of interfaces. A potential approach tunnels. Tunnelling and Underground Space Technology
105, 103574.
that would be investigated in the next stage is to engage
Ninic, J., Alsahly, A., Vonthron, A., Bui, H.-G., Koch, C., König,
a meshing program as a middle-ware or to experiment M., Meschke, G., 2021. From digital models to numerical
with the octree mesh in FLAC3D. Ultimately, proper analysis for mechanised tunnelling: A fully automated
translators could be developed based on BIM data design-through-analysis workflow. Tunnelling and
schema, such as the industry foundation class (IFC). Underground Space Technology 107, 103622.
There is yet standard for underground construction based Ninić, J., Bui, H.G., Meschke, G., 2020a. BIM-to-IGA: A fully
on IFC while research and development efforts are automatic design-through-analysis workflow for
ongoing by buildingSMART, an international industry segmented tunnel linings. Advanced Engineering
Informatics 46, 101137.
body aiming at improving interoperability between
Ninic, J., Giang Bui, H., Meschke, G., 2018. Parametric Design
software applications used in the construction industry, and Isogeometric Analysis of Tunnel Linings within the
cooperating with other organisations focusing on Building Information Modelling Framework. CEUR
subsurface utilisation, such as the International Tunneling Workshop Proceedings.
and Underground Space Association (ITA-AITES). Ninić, J., Koch, C., Vonthron, A., Tizani, W., König, M., 2020b.
Integrated parametric multi-level information and
5 CONCLUSION numerical modelling of mechanised tunnelling projects.
Advanced Engineering Informatics 43, 101011.
Sacks, R., Eastman, C., Lee, G., Teicholz, P., 2018. BIM
This study aims to provide a solution to tackle the problem
handbook: A guide to building information modeling for
of lacking open standard data models that cover the owners, designers, engineers, contractors, and facility
context-specific knowledge required for underground managers. John Wiley & Sons.
construction. By leveraging the parametric modelling Shilin, A., Kirilenko, A., Znajchenko, P., 2016. Complex
technique of BIM, the workflow could mitigate the efforts reconstruction project of Mayakovskaya metro station in
of numerically retrofitting and re-modelling at feasibility the centre of Moscow. Structural Analysis of Historical
and earlier design stage. The enabled a design-to-design Constructions—Anamnesis, Diagnosis, Therapy, Controls;
workflow with the reversed visualisation loop creates Van Balen, K., Verstrynge, E., Eds, 1736-1741.
Wang, G., Zhang, Z., 2021. BIM implementation in handover
opportunity for backanalysis and design optimisation.
management for underground rail transit project: A case
study approach. Tunnelling and Underground Space
6 ACKNOWLEDGEMENTS Technology 108, 103684.
Zhu, H., Wu, W., Li, X., Chen, J.Q., Huang, X.B., 2017. High-
This work was financially supported by the ARC precision acquisition, analysis and service of rock tunnel
Nanocomm Hub (IH150100006). The first author would information based on iS3 platform. Chin. J. Rock Mech.
like to acknowledge the Australian International Eng. 36. 2350–2364.
Postgraduate Research Scholarship and Monash
Graduate Scholarship. The technical information and
photos used in this paper are from the Metro Tunnel
Project in Melbourne (https://metrotunnel.vic.gov.au), and
the models are partially supported by the Government
and Industries.
REFERENCES
Azhar, S., 2011. Building Information Modeling (BIM): Trends,

Benefits, Risks, and Challenges for the AEC Industry.
Leadership and Management in Engineering 11, 241-252.
Caetano, I., Santos, L., Leitão, A., 2020. Computational design
in architecture: Defining parametric, generative, and
algorithmic design. Frontiers of Architectural Research 9,
287-300.
DINOloket Netherlands, 2019. Data and Information on the
Dutch Subsurface, Data en Informatie van de Nederlandse
Ondergrond (DINO), p. 10.
https://www.dinoloket.nl/en/subsurface-data
Fabozzi, S., Biancardo, S.A., Veropalumbo, R., Bilotta, E., 2021.
I-BIM based approach for geotechnical and numerical
modelling of a conventional tunnel excavation. Tunnelling
and Underground Space Technology 108, 103723.
Gu, N., London, K., 2010. Understanding and facilitating BIM
adoption in the AEC industry. Automation in Construction
19, 988-999.
Huang, M.Q., Ninić, J., Zhang, Q.B., 2021. BIM, machine
learning and computer vision techniques in underground
construction: Current status and future perspectives.
83
On best practices for trackbed design

N. Yousefpour, PhD, PE1
1
Department of Infrastructure Engineering, The University of Melbourne; Arup, Melbourne, VIC, Australia; email:
negin.yousefpour@unimelb.edu.au
ABSTRACT
This paper reviews the best practices of trackbed design for railway projects. Various existing methods have been studied
and recommendations for more economical design are provided. The analytical/empirical methods from various standards
such as UIC, AREMA, British Rail, and Australian standards, as well as other commonly used methods such as Raymond
and Li-Selig are compared based on a typical track embankment cross section. The outcome was then evaluated against
2D and 3D numerical models. Incorporating numerical methods is shown to render considerable reductions in the required
prepared subgrade/structural fill materials and allow for assessment of long-term design issues, such as subgrade shear
failure due to excessive plastic deformations.
Keywords: Trackbed Design, 3D Numerical Modelling, Best Practices, Railway Standards
1 INTRODUCTION Figure 2. Table 2 provides the method of determination

of subgrade bearing capacity class and required
With major ongoing and emerging rail projects across the prepared subgrade class and thickness.
world and in Australia, the rail infrastructure industry
seeks innovative solutions to provide sustainable and
cost-effective design solutions. Proper engineering
design of track-bed and track formations is one of the key
components for effective design and operation of
railways and minimizing the construction and
maintenance costs.
This paper presents a review on various common

analytical and empirical methods and standards for
track-bed and track formation design. The variations in
the design outcomes using these methods are
investigated. In addition, the application of innovative
solutions including advanced nonlinear finite element
method (FEM) analysis and subgrade stabilisation for
optimizing the track formation design is discussed.
2 TRACKBED AND TRACK FORMATION
A typical track embankment cross-section is presented

in Figure 1. The main function, typical dimensions, and
typical material properties for all the required layers are
defined in Table 1. Ballast and subballast are normally
defined as “trackbed” laying on top of the ‘track
formation,” which is the natural subgrade [1].
Figure 2. UIC method of calculating the minimum

thickness of trackbed layers [1]
3.2 AREMA method

Figure 1. Trackbed and formation typical section
AREMA [2] recommends the Talbot equation for
3 ANALYTICAL AND EMPIRICAL DESIGN estimating the thickness of the granular trackbed (i.e.
METHODS AND STANDARDS ballast + sub-ballast). This method was developed based
on field tests conducted in 1910s and 1920s. Subgrade
3.1 UIC method conditions, heavier axle loads, dynamic effects, and
granular material quality are factors that have not been
UIC [1] requires the following items to be considered in considered in this method.
dimensioning the track bed layers:
• Problems of frost protection H=0.24 (Pm/Pc)0.8
• Bearing capacity of subgrade
• Type and spacing of sleepers Where,
• Traffic characteristics H is the granular material thickness (Ha + Hs),
Pc is the allowable subgrade stress (138 kPa
The UIC method of estimating the suggested minimum recommended by AREMA), and Pm is the vertical stress
thickness of the trackbed layers including is depicted in applied on the ballast surface.
On best practices for trackbed design N. Yousefpour
Table 1: Trackbed and formation layers

Track layer Main functions Typical thickness Typical material
Ballast -Transfers and distributes loading 300 – 450 mm Crushed rock/aggregates with maximum
from the ties to the underlying size of 63.5 mm (2.5 in)
subballast or subgrade of the track
structure at a tolerable level.
-Provides drainage through the
support structure and away from the
track
Subballast/Capping -Providing a filter/buffer between 150 mm Crushed rock/aggregates the maximum
subballast and subgrade grain size of the subballast shall not exceed
-Drainage the maximum grain size of the track ballast
and no more than 5% of the subballast shall
pass the No. 200 sieve.
Prepared An additional layer on top of the fill 0-500 mm Varies with different standards: well-graded
Subgrade/Structural or natural subgrade to further reduce (varies depending coarse-grained material with low to medium
Fill the stress on subgrade and also to on the subgrade plasticity and maximum 15% to 30% fine-
provide sufficient overall stiffness strength) grained
Embankment Fill Provide the required track level Depends on the Should be free of unsuitable material:
track level design -Materials containing stumps, weeds, leaves,
wrt the existing coals, peats, ashes, grasses, and other
ground organic materials.
-Contaminated soil
-Material with soaked CBR1 <2% and swell
index2 >3 %
Natural Subgrade Foundation to the trackbed NA Must have a minimum of CBR=2 %; Should
be free of unsuitable material as per above
1. 95% Standard compaction – 9 kg Surcharge
2. ASTM D4546 – 14e1 “Standard Test Methods for One-Dimensional Swell or Collapse of Soils”, Test Method A.
Table 2: Determination of subgrade bearing capacity and prepared subgrade thickness

Subgrade/ Fill Soil Min. CBR2 (%) Req. Bearing Prep Subgrade Prep Subgrade Min. thickness:
Quality1 Class for Quality Class Min. CBR3 “ef” (m)
1
Subgrade
QS1 2-3 P2 QS2 5 0.5
P2 QS3 10-17 0.35
P3 QS3 10-17 0.50
QS2 5 P3 QS3 10-17 0.35
QS3 10-17 P3 QS3 10-17 NA
1. QS0: “Unsuitable”, QS1/P1: “Poor”, QS2/P2: “Average”, QS3/P3: “Good”. Refer to [1] for details
2. CBR of the in-situ (soaked)
3. CBR of remoulded samples (soaked)
3.3 Raymond Method

• Ballast, sub-ballast + subgrade act together as a
Raymond (1985) [3] modified the design method single homogenous layer
recommended by AREMA. Figure 3 shows the design • Uses a single value of axle load disregarding the
chart developed by Raymond for vehicles weighing 70- amount of cumulative tonnage
125 tons. The following assumptions have been made in
the development of this method: 3.4 British Rail Method
Heath et al. (1972) [4] developed a threshold stress

design method for selecting the granular layer thickness.
The threshold stress is the limit stress in the subgrade to
protect the subgrade from progressive shear failure (See
Figure 4).
Figure 4. Subgrade threshold stress for selecting

Figure 3. Raymond design chart granular layer thickness [4]
85
3.5 Li-Selig Method granular layer thickness. This normally causes a heave
at trackside which can block the proper drainage. Water
The Li-Selig Method (1998) [5,6] can be used based on coming from the granular layer can be trapped in the
the allowable stress at the subgrade surface (allowable subgrade depression areas which can aggravate the
cumulative plastic strain), as well as the allowable failure.
subgrade deformation. Figure 5 provides examples of
design charts. In these charts, H is granular material 2) Excessive Plastic Deformation due to Repeated
thickness. Iԑ and Ip are referred to as the strain influence Loading: This can cause ballast pockets under the track
factor and the deformation influence factor, respectively: (Figure 7) and happens predominantly in soft soil
subgrades.
Other types of failure include mud pumping, excessive

consolidation settlement, and slope stability failure. The
granular layer thickness does not influence these [5].
Where, 𝜎𝜎𝜎𝜎𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 is the allowable deviator stress at the
subgrade surface, 𝜌𝜌𝜌𝜌𝑑𝑑𝑑𝑑 is the allowable total subgrade
plastic deformation for the design period, N is the total
equivalent number of load repetitions during the design
period, Pd is the design dynamic wheel load, 𝜎𝜎𝜎𝜎𝑠𝑠𝑠𝑠 is the soil
compressive strength, a, m, and b are the material
parameters dependent on soil type, A is the area factor
selected to make deformation influence factor
dimensionless, (0.64m2, 1000in2), and L is a factor to
make charts dimensionless (6in, 0.152m).
Figure 6. Subgrade progressive shear failure
Figure 7. Excessive plastic deformation due to repeated

loading [5]
3.6 Australian Standards
ARTC and ASA [7,8], VLine [9] and MTM [10] standards
provide relatively similar guidelines for trackbed design. A
minimum of 0.3m ballast and 0.15m subballast or capping
is recommended (total granular layer thickness of 0.45m).
Typical thickness/depth for the prepared
subgrade/structural fill layer is provided based on the
CBR of the natural subgrade as presented in Table 3.
Table 3: Structural Fill/Prepared Subgrade Depth (m) per

Australian Standards
Subgrade/Fill
Material Soaked ARTC/MTM1 VLine
CBR (%)
1-2 1 NA2
Figure 5. Li-Selig example design charts [5] 2-3 1 1
3-5 0.5 0.5
Two main failure mechanisms were considered for the 5-7 0.5 0.3
development of this approach: ≥8 0.5 0/0.2
1) Subgrade progressive shear failure (Figure 6): This 1. Minimum subgrade CBR is 3% by MTM, 2% by VLine, 1% by
has been reported by the European railways and in the ARTC
United States for track under heavy axle loads 2. Minimum fill CBR is 2% by VLine
especially on fine-grained soils without sufficient 3. 0mm for subgrade, 200mm for fill
3.7 Comparison of Methods 4.3 Material Model
An example is worked based on the discussed methods An Elasto-Plastic, Mohr-Coulomb constitutive model was
in this section to compare the required thickness of the implemented, adopting material parameters as per Table
granular layer (H=Hb+Hs) from different methods. The 5. The water table was not considered within the model,
example is based on an axle load of 160 kN. It is assumed assuming a dry subgrade (zero pore water pressure).
that granular layer will lay on an underlaying layer with
QS2 soil quality with CBR~5% (~150kPa). Table 4
provides the estimated required thickness ranging
between 0.4m to 0.5m.
Table 4: Granular layer thickness from various

methods/standards
Method Ballast+Sub-ballast
Thickness (m)
UIC 0.5
AREMA 0.4
Raymond 0.4
British Rail 0.45
Li-Selig 0.45
Australian Standards 0.45
The required prepared subgrade thickness is 0.35m

based on UIC specifications. Based on Australian
standards a thickness of 0.5m is required.
4 FINITE ELEMENTS METHOD
4.1 Finite Element Method
In the previous sections, it was shown how various

methods and standards can result in different thickness
values for trackbed layers.
With the current computational capabilities, 3D finite

element methods (FEM) can help engineers with more Figure 8. Isometric and cross section view of the track
economic and more reliable design. Following sections formation
provides the result of a 3D train embankment model using
LS-Dyna in comparison with a 2D Plaxis model. The
impacts of cyclic degradation in subgrade has not been
considered.
4.2 Geometry
Figure 8 presents the isometric and cross section view the

3D model for a track embankment cross section from a
project in Victoria. The length of the model is about 25m
equal to the length of a train car. The trackbed includes a
0.3m ballast, underlain by a 0.15m subballast/capping
layer over a prepared subgrade/structural fill. The natural
subgrade is assumed to have a CBR=3.
As shown in the analytical example above, this condition Figure 9. Train loads
renders a structural fill of 0.5m according to Australian
standards. However, in this case, a 0.3m of prepared 4.4 Results
subgrade from stabilised subgrade material with CBR=15
has been proposed to replace the structural fill to save on A summary of model outputs, including stress and strain
cost. Using a Ls-Dyna model, the proposed design is distribution, as well as deformations and plastic strains
being validated by checking the stress and strain are provided below (see Figures 10 to 14). The shear
developed in the subgrade and by checking for bearing stress within the subgrade is below 15 kPa and the
failure. vertical stress is below 50 kPa, the subgrade allowable
stress, hence no plastic stain is observed. The
The train loading assumes a 23 tons axle load (two deformation at the sleeper level shows to be less than 5
bogies, four axles per car), resulting in 450kN/m linear mm, which is below the common acceptable criteria of 10
load for each wheel applied on top of the rail track over mm (1/2 in). This validates the proposed design of 0.3m
the length of a sleeper (0.25m) (see Figure 9). prepared subgrade with higher CBR, instead of the 0.5m
structural fill with CBR=8 required by Australian
standards.
87
Results of LS-Dyna model was then compared with an maximum shear stress is around 15 kPa in the subgrade
equivalent Plaxis 2D model of the longitudinal cross and around 27 kPa in the prepared subgrade layer,
section. As shown in Figure 15 and 16, the 2D model whereas Plaxis shows maximum of 12 kPa in subgrade
shows a maximum stress of 140 kPa in ballast, about 20% and 40kPa in the prepared subgrade. The 2D model
lower stress as compared with 3D LS-Dyna. Similarly, underestimated the maximum stress, hence the
looking at the shear stress distribution, from Ls-Dyna the deformations, in the trackbed layers.
Table 5: Track formation material properties

Layer Friction Cohesion Undrained Unit Young's Poisson's
Angle Shear Weight Modulus Ratio
Strength
Deg kPa kPa kN3/m MPa
Ballast 40 0 0 23 280 0.3
Sub-ballast 35 0 0 23 150 0.3
Prepared Subgrade (CBR=15) 0 0 250 20 200 0.3
Natural Subgrade (CBR=3) 0 0 50 19 30 0.3
Figure 13. Von-Mises stress (=2 × shear stress)

Figure 10. Stress under the sleepers
Figure 14. Plastic strains
5 CONCLUSIONS
Figure 11. Vertical deformations Among the discussed analytical design methods, Li-
Selig is the only method that considers both allowable
stress and deformation at subgrade with respect to the
train loading cycles, providing a more reliable
trackbed design. Incorporating 3D numerical
simulations, can further improve the design, by

providing a more accurate assessment of stress
distribution and deformations in the subgrade,
resulting in potential reduction of the required depth
for the structural fill/prepared subgrade layer on soft
natural subgrades.
Figure 12. Vertical stress
Figure 15. Vertical stress – Plaxis 2D
Figure 16. Shear stress – Plaxis 2D
These methods also allow for identifying potential long- Li, D. and Selig, E., Method for Railroad Track Foundation
term issues with bearing failure or excessive plastic strain Design. I: Development, Journal of Geotechnical and
due to cyclic loading. Although it is common for clients to Geoenvironmental Engineering, ASCE, 316-322, 1998.
Selig, E. and Waters, J. M., Track Geo-technology and
demand compliance with standards, there is often
Substructure Management, Thomas Telford, 1995.
provisions for applying more rigorous design methods ARTC (Australien Rail Track Corporation), Track & Civil Code of
beyond the standards by getting client approvals for Practice, Railway Earthworks - Supplementary Appendix,
derogations in early phases of a project. ETC-08-02, 2006.
ASA (Asset Standards Authority) Earthworks and Formation
6 ACKNOWLEDGEMENTS Standard, Version 1.0, 2018. Supplementary Appendix to
ARTC.
This study was sponsored by Arup University. VLine, Network Infrastructure Standard (NIST) 2659, Earthworks
and Drainage Standard, 2019.
MTM (Metro Train Melbourne), Earthwork and Formation
REFERENCES Specification, L1-CHE-SPE-178, Version 2.0, 2018.
UIC (International Union of Railways), Earthworks and Trackbed

for Railway Lines, Code 719, 3rd Edition, 2008.
AREMA (American Railway Engineering and Maintenance-of-
Way Association), 2015.
Raymond, G. P., design of railroad ballast and subgrade support,
Journal of the Geotechnical Engineering Division, ASCE,
Vol. 104, No. GT1, 45-59, 1985.
Heath, D. L., Shenton, M. J., Sparrow, R. W., and Waters, J. M.,
Design of Conventional Rail Track Foundations,
Proceeding of Institution of Civil Engineering Vol. 15, 251-
267, 1972.
NOTES
2021 AUSTRALIAN GEOMECHANICS SOCIETY
VICTORIAN SYMPOSIUM
Innovations in Geotechnical Design
VICTORIA CHAPTER

Innovations in Geotechnical Design: 2021 Australian Geomechanics Society Victorian Symposium

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Innovations in Geotechnical Design: 2021 Australian Geomechanics Society Victorian Symposium

Uploaded by

Copyright:

Available Formats

PROCEEDINGS

2021 AUSTRALIAN GEOMECHANICS SOCIETY

SYMPOSIUM ORGANISING TECHNICAL REVIEWERS

2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN i

SESSION 1 1 Design methodology and input 63

The Application of 3D Finite Element Method in 41

Pile Testing Verification – an Alternative Approach 47

ii 2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN

JULIAN SEIDEL SCOTT TAYLOR

2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN iii

11:30 am Presentation 4:10 pm Presentation

iv 2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN

2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN v

Key capabilities include:

Engineering • Laboratory • Drilling • Instrumentation

Global Synthetics is a 100% Australian-owned company,

Stay on top of the latest Global Synthetics information and

Use of probabilistic methods in geotechnical engineering

1 INTRODUCTION rocks suggests that geotechnical systems are highly

If we were to plot the multitude of data from the

2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN 3

φ=′ 32° 3 HOW CAN WE USE PROBABILISTIC METHODS

4 2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN

Suppose n pile capacity tests have been performed and

4 HOW MUCH FIELD/TEST DATA DO WE NEED?

Suppose we want to do an excavation on a site as shown

2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN 5

Figure 5. Conditional Monte Carlo simulations based on

Figure 3. Unconditional Monte Carlo simulations based

Figure 6. Probability of failure verse number of CPT tests

. Generally speaking the approaches that we use to assess

Figure 4. Conditional Monte Carlo simulations based on

6 2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN

Figure 7. Geotechnical CPT test and geophysical multi

geotechnical information for proper design, construction 2

study of existing geophysical data; and 2) a dedicated in 0.1

A more scientific integration of geophysical data with

Huang et al. (2018) developed a Bayesian updating

2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN 7

Bayesian updating can use prior knowledge (e.g., lab and

8 2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN

A. L. E. Lochaden1 and S. Buttling2

Keywords: geotechnical design, working stress, limit state, reliability.

1 INTRODUCTION 2 THE STATE OF GEOTECHNICAL DESIGN IN

2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN 9

By plotting both the load and resistance frequency

10 2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN

Effective friction angle of retained material, φ' (°)

Mean margin of safety = R - S

Loads (S) Resistance (R) 15 Failure

2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN 11

4 UNDERTAKING RBD IN PRACTICE Many geo-engineering professionals will reasonably

12 2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN

bridge would probably be defined as a Class 3 structure

4.5 Design examples – retaining walls

Low (2005) presents reliability analyses of two types of

Figure 6. Reliability analysis of anchored retaining wall (from Low, 2005)

2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN 13

14 2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN

1 INTRODUCTION enhance the work efficiency by removing encapsulation

2021 AGS VICTORIAN SYMPOSIUM INNOVATIONS IN GEOTECHNICAL DESIGN 15

western/eastern end of the soil nail wall indicate that the

2.2 Laboratory soil and rock testing and rock