New Generation Design Codes For Geotechnical Engineering Practice - Taipei 2006

Proceedings of the International Symposium on
New Generation Design Codes for

Geotechnical Engineering Practice -Taipei 2006
(with CD-ROM)
editors
Meei-Ling Lin, Chung
Yusuke Honjo and

(with CD-ROM)

(with CD-ROM)
National Taiwan University of Science and Technology,

Taipei, Taiwan 2 - 3 November 2006
editors
Meei-Ling Lin
(National Taiwan University, Taiwan),
Chung-Tien Chin
(Moh & Associates Inc., Taiwan),
Horn-Da Lin
(National Taiwan University of Science and Technology, Taiwan),
Yusuke Honjo
(Gifu University, Japan) &
Kok-Kwang Phoon
(National University of Singapore, Singapore)
\fc World Scientific

NEW JERSEY • LONDON • SINGAPORE • BEIJING • SHANGHAI • HONG KONG • TAIPEI • CHENNAI
Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE
British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.

NEW GENERATION DESIGN CODES FOR GEOTECHNICAL ENGINEERING
PRACTICE — TAIPEI 2006 (with CD-ROM)
Copyright © 2006 by World Scientific Publishing Co. Pte. Ltd.
All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the Publisher.
For photocopying of material in this volume, please pay a copying fee through the Copyright
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to
photocopy is not required from the publisher.
ISBN 981 -270-382-9 (pbk)
Printed in Singapore by World Scientific Printers (S) Pte Ltd

PREFACE
Communication of design risk within a transparent and rational framework is necessary

in view of increasing interest in code harmonization, public involvement in defining
acceptable risk levels, and risk-sharing among client, consultant, insurer, and financier.
Activities in code harmonization in particular are noteworthy. The advent of the World
Trade Organization (WTO) has added impetus to the formation of trading groups that
result in multilateral free trade areas or bilateral free trade agreements. Traditionally,
geotechnical engineering practice has always been viewed as a localized activity under
the purview of the relevant federal and/or state authorities. However, the move towards
greater economic cooperation and integration will require the elimination of some
technical obstacles that exist as a consequence of differences in national codes and
standards, and harmonization of technical specifications.
For the geotechnical engineering profession, there is added pressure to undergo

significant revamp because structural and geotechnical design are increasingly
incompatible. The structural engineering design community has adopted limit state
design and probability-based design since the seventies and appears to be gradually
evolving towards a performance-based design philosophy. The structural engineering
community is also the main driving force behind international standardization activities,
such as IS02394 on "General Principles on Reliability for Structures".
Engineers and regulators in many countries are struggling to accommodate the complex
and multi-faceted changes occurring at the international scene. The status of local design
codes in view of globalization and their compatibility in view of evolving design
philosophies, are issues of major concern that do not admit simple solutions. A large
number of countries do not have the scale of economy, organizational structures, political
support, and perhaps financial resources to solve these complex problems on their own.
This conference intends to follow the spirit of IWS Kamakura (2002) and LSD2003 to
promote greater awareness, to facilitate debate and information exchange, and to
accelerate research and practice on important issues relating to new generation
geotechnical design codes. The bottom-line is to move geotechnical engineers forward
together as a community in response to significant changes occurring globally.
The idea behind this symposium grew out of a discussion between Chung-Tien Chin, Jie-
Ru Chen, Yusuke Honjo, and Kok-Kwang Phoon during the 16th ICSMGE in Osaka last
year. Subsequent discussion between Meei-Ling Lin and Kok-Kwang Phoon during the
GEDMR05 conference in Singapore helped set the path in commencing the organization
of this event. Given the gathering pace of geotechnical design code developments, there
is a compelling reason to consider a follow-up symposium to LSD2003. It is also timely
to discuss the possibilities of establishing a more regular series of symposiums and a joint
working group to coordinate these activities.
Thirty-five abstracts from thirteen countries were received during the initial call for paper.
Thirty-one papers were accepted for publication after review. Topics covered include
v
geohazards, geotechnical uncertainty and variability, probabilistic and reliability analysis,
design code concept and harmonization, and performance-based engineering practice. In
addition to the submitted papers, special invitations were extended for contribution as
keynote lectures, invited lectures, and Taiwan special project lectures. A total of 11
papers were obtained for these lectures. This publication contains extended summaries
of 42 papers. Complete contributions are available in the accompanying CD-ROM.
This symposium is jointly organized by the Taiwanese Geotechnical Society and TC23 of
ISSMGE. It is supported by the National Taiwan University, National Taiwan University
of Science and Technology, Taiwan Construction Research Institute, ASCE Taiwan
Chapter, JWG-DMR, ASCE Geo-Institute, TC39 of ISSMGE, and Southeast Asian
Geotechnical Society.
The publication of this proceedings will not be possible without the considerable efforts
invested by a committed editorial committee that include Jie-Ru Chen, Jian-Ye Ching,
Yo-Ming Hsieh, Chih-Ping Lin, and C.H. Wang. Papers appearing in this proceeding are
subjected to technical and editorial reviews. We are also grateful for the constant support
and timely assistance given by the technical reviewers (C. Hsien Juang, Kok-Kwang
Phoon, Robert S.R. Lo, Liming Zhang, Kenichi Horikoshi) and editorial reviewers
(Hsiang-Ju Chen, Ting-Rong Chen, Te-Wei Chen, Yu-Hua Hsieh, Wei-Nan Jian, Yueh-
Ting Lai, Jing-Hang Lin, and Mei-Ling Liu), and the secretariat (Ms Wei-Ling Lin and
Tsui-Hui Chiang). The significant assistance rendered by Prof. Der-Wen Chang in the
arrangement of travel visas for speakers is deeply appreciated. Lastly we would like to
thank Rhaimie Wahap and his team at World Scientific for working patiently with us
under a very tight publication schedule. His professional assistance is greatly appreciated.
Editors
Meei-Ling Lin
Chung-Tien Chin
Horn-Da Lin
Yusuke Honjo
Kok-Kwang Phoon
vi
Organising committee
Prof. Meei-Ling Lin, Symposium Chairperson

Dr. C.H. Wang, Secretary General
Dr. Jie-Ru Chen
Dr. Chung-Tien Chin
Prof. Jian-Ye Ching
Prof. Jia-Jyun Dong
Prof. Yo-Ming Hsieh
Prof. Chih-Ping Lin
Prof. Horn-Da Lin
Advisory committee
Prof. Yusuke Honjo

Prof. C. Hsein Juang
Prof. Chien-Chung Li
Dr. Za-Chieh Moh
Dr. Chin-Der Ou
Prof. Kok-Kwang Phoon
Dr. Ming-Teh Wang
Editorial committee
Dr. Jie-Ru Chen

Prof. Jian-Ye Ching
Prof. Yo-Ming Hsieh
Prof. Chih-Ping Lin
Dr. C.H. Wang
vn
TABLE OF CONTENTS
Preface v
Organization vii
Keynote Lectures 1
Limit states design based codes for geotechnical aspects of foundations in Canada 3
D. E. Becker
Risk assessment in rock engineering 5

H. H. Einstein
Some movements toward establishing comprehensive structural design codes

based on performance-based specification concept in Japan 7
Y. Honjo
Development and implementation of Eurocode 7 9
T. L. L. On-
Invited Lectures 11
New trend toward performance-based design in the construction industry 13

K. Horikoshi, Y. Honjo, A. Iizuka
Risk analysis of lining structure in large-diameter shield tunnel 15
H. W. Huang, Q. F. Hu, Y. Y. Yang
Energy approach to earthquake-induced slope failures for performance-based

design 17
T. Kokusho
A preliminary study on load and resistance factors for foundation piles in Taiwan 19
H. D. Lin
Evaluating probability of seismic landslide based on the Chi Chi's events, Taiwan 21
M. L. Lin, C. J. Chung, M. H. Ho
Serviceability limit state reliability-based design 23

K. K. Phoon
Eurocode 7 for geotechnical design — basic principles and implementation in

the European member states 27
B. Schuppener, R. Frank
Session I: Code Concept and Harmonization 29
The study and revision of the probabilistic seismic hazard map and dam safety
code of Taiwan 31
C. T. Cheng, S. J. Chiou, C. T. Li, P. S. Lin, Y. B. Tsai
A simple reliability assessment of pile design: resolving some Hong Kong

challenges 33
S. R Lo, K. S. Li, J. Lam
Geotechnical standards in Hong Kong 35

W. K. Pun, W. M. Cheung, L. S. Lui
Implementation of Eurocode 7-1 geotechnical design in Germany 37

N. Vogt, B. Schuppener
Introduction to international joint study ofreliability-based design for port and

harbor structure 39
G. Yoon, T. Nagao, W. Lu, K. Lee, H. Kim
Session II: Performance Oriented Geotechnical Analysis 41
Effect of lateral cyclic load on axial capacity of pile group in loose sand 43
S. Basak
Evaluation of design methods for large-diameter bored piles 45

Florence L. F. Chu and L. M. Zhang
Engineering problems for performance-based design of earth structures 47

Y. Honjo, M. Honda, K. Ogawa,Y Wakatuki
Performance-oriented risk assessment and retrofitting strategy for electricity

towers on slopes 49
C. H. Wang, M. H. Chang, C. F. Chang, D. C. Wu, K. P. Hsiung
Session III: Geotechnical Reliability Analysis 51
Equivalence between reliability and factor of safety 53

J. Y. Ching, T. R. Chen
Bearing capacity of open ended piles in port construction in Japan 55

Y. Kikuchi
Variance of soil parameters: some common misconceptions 57

K. S. Li, S. R. Lo
Reliability analysis of excavation-induced building damage 59
M. J. Schuster, C. H. Juang, E. C. L. Hsiao, M. J. S. Roth, G. T. C. Kung
Session IV: Geohazards 61
The assessment and prediction of landslides and debris flows in Ta-Chia river
after Taiwan Chi-Chi earthquake 63
C. T. Cheng, Y. L. Chang, S. J. Chiu, Y. S. Lin, C. Y. Ku, S. M. Shu,
J. C. Chern, S. H. Yu, S. D. Yang, C. F. Wang, C. H. Chiao, L. T. Hwang
Predicting landslides probabilities along mountain road in Taiwan 65

J. Y. Ching, H. J. Liao
Optimal design of sand compaction pile based on liquefaction hazard analysis 67

J. H. Hwang, C. W. Yang, C. C. Lu
Verifications and physical interpretations of the discriminant model for

evaluating liquefaction potential on SPT-N value 69
S. Y. Lai, M. J. Hsieh, W. J. Chang, P. S. Lin
Seismic performance-based design for canal embankment 71

Y. Otake, T. Hara, T. Horikawa, Y. Ito, T. Kato, M. Hosoyamada, Y. Kasai
Session V: Engineering Practice and Challenges 73
Observational design approaches for safe and economical deep basement

construction in the urban environment 75
I. Askew, J. A. Frame, D. Sein
The performance of laterally loaded single pile in reclaimed land 77

C. S. Chen
Settlement calculation of large-area thick raft foundation under irregular

high-rise buildings 79
J. F. Gong, X. L. Huang, D. H. Di
Geotechnical risk assessment and performance-based evaluation of a deep

excavation in the Kaohsiung MRT system project 81
B. C. Hsiung, H. Y. Chuay
An essay on typification of verification methods used in the design procedure of

geotechnical structures 83
S. Kobayashi, K. Aita, T. Fujiyama, M. Honda, T. Kaneko, A. Morikage,
A. Murakami, M. Nabetani, M. Nozu
xi
Session VI: Geotechnical Uncertainties and Variabilities 85
Reducing performance uncertainties with monitoring data 87

J. Y. Ching, Y. H. Hsieh
Evaluation of spatial variability of weathered rock for pile design 89

S. M Dasaka, L. M. Zhang
Analysis of uncertainties in analytical pile design methods in South Africa 91

M. Dithinde, K. K. Phoon, M. de Wet, J. V. Retief
Probabilistic uncertainties in estimating the vertical bearing resistance of piles 95

M. Suzuki, M. Shirato, S. Nakatani, K. Matsui
Soil parameters used in the new design code of port facilities in Japan 97
Y. Watabe, M. Tanaka, Y. Kikuchi
Taiwan Special Project Series 99
Performance design of Taipei 101 foundation 101

D. S. Chen
Design and construction issues of deep foundations for the Taiwan high
speed rail 103
S. W. Duann, J. R. Chen, T. C. Su, C. T. Chin
Experiences from Hsuehshan tunnel constructions 105
L. P. Shi, Y. S. Hsieh
Author index 107
List of past ISSMGE TC23 proceedings 109

Keynote Lectures
TAIPEI2006 International Symposium on New Generation Design Codes
for Geotechnical Engineering Practice
Nov. 2-3, 2006, Taipei, Taiwan
Limit States Design Based Codes for Geotechnical Aspects of

Foundations in Canada
Dennis E. Becker
Golder Associates Ltd, Calgary, Alberta, Canada
EXTENDED ABSTRACT
The geotechnical engineering profession in Canada, and elsewhere throughout the world, is in the
process of incorporating limit states design into codes of practice for geotechnical design aspects of
foundation engineering. Primary benefits of the use of limit states design are that it provides a
consistent design approach between structural and geotechnical engineers, as well as providing a
rational and consistent framework for design and risk management of design uncertainty. This paper
describes the needs and objectives for limit states design in Canada, and its development in codes;
identifies and describes the primary Canadian Codes; discusses the role of the Canadian Foundation
Engineering Manual and other authoritative references related to these Codes; discusses some of the
experiences and challenges encountered in practice during implementation and application of limit
states design; and outlines ongoing and proposed code development work, and associated future
directions and research needs. The importance of understanding fundamental principles, effective
communications between structural and geotechnical engineers, education and training is emphasized.
All of these components will be required for successful implementation and acceptance of limit states
design for geotechnical aspects of foundation engineering.
Limit states design, based on a factored strength approach similar to that of the European practice,
for geotechnical aspects of foundations was first introduced into Canadian engineering practice in the
early 1980s. However, this initial introduction did not get off to a good start because factored strength
concepts were not well accepted by geotechnical engineers; it also generated a fair amount of
confusion and controversy because the promised economy of design was not achieved. Canadian
geotechnical practitioners felt that it was not logical or rational for strength parameters to be reduced
(factored) to reflect weaker "artificial" soils and then use them directly in the same equations for
calculating design resistances. In the early 1990s, an overall factored resistance approach, based on a
Load and Resistance Factor Design (LRFD) format, was proposed for limit states design based codes.
Subsequently, a LRFD format for foundations became a mandatory requirement in the 2000 edition of
the Canadian Highway Bridge Design Code (CHBDC) and in the 2005 edition of the National
Building Code of Canada (NBCC). Nevertheless, confusion continues to exist concerning the
objectives of limit states design as engineering practitioners in Canada struggle to undergo the
transition from traditional working (allowable) stress design to design based on limit states (LRFD)
concepts.
The primary structural codes in Canada are the NBCC, the CHBDC and the Canadian Offshore
Structures Code. These codes involve the interaction of structural and geotechnical engineers; they
generally apply to the design and construction of foundations, retaining walls and other buried
structures. There is no national code document for aspects in which geotechnical engineers do not
normally interact with structural engineers. The current geotechnical state-of-practice in Canada does
not use limit states design concepts to design slopes, earth embankments, dams and other earth
structures. The code requirements are normally written as performance requirements and are based on
scientific or technical principles. The codes avoid standardizing certain methods or procedures of
design and construction. For example, the NBCC (2005) is published in an Objective-Based Code
format where each code requirement is linked to the four basic objectives of Safety, Health,
Accessibility (in particular for persons with disabilities), and Fire and Structural Protection of
Buildings. Although some countries are striving to establish Performance-Based Codes, the NBCC
code developers are of the opinion that current building science knowledge is inadequate to write a
"true" (as per their perspective) Performance-Based Code, and that the measures to verify performance
3 Full paper in TAIPEI-2006 CD-ROM

are not yet adequately in place. It is anticipated that it will be many years before a true Performance-
Based Code format exists in the NBCC and other Canadian Codes. The current implementation of an
Objective-Based Code format in NBCC (2005) is considered to be an initial step in this regard.
There appears to be a general lack of understanding, communication, education and training
concerning the fundamental principles and intent of limit states (LRFD) design. In the LRFD format,
it is important to note that the load and resistance factors are interrelated to each other. The values of
the load and resistance factors depend on the target reliability index that the design is to achieve, the
variability of the parameters that affect loads and resistances, and the statistical definition of then-
characteristic values. For consistent and rational design in practice, the selection of a given
characteristic value for geotechnical resistance needs to be made in the same manner as that used to
derive the specified geotechnical resistance factor. The mean or a "cautious estimate" of the mean
value for the affected volume of ground (zone of influence) is generally considered to be appropriate
for the characteristic value and the basis of the load/resistance factors derivation (calibration). The
quantification of "cautious estimate" has not been formalized completely; there may be a need to
establish an unambiguous quantitative definition for it.
In general, practicing geotechnical engineers who have completed limit states design for
foundations do not object to the use of the NBCC and CHBDC specified geotechnical resistance
factors for shallow foundations. However, some of the specified resistance factors for deep
foundations are considered to be too low. In particular, it is felt that static pile load tests are being
unduly penalized by the specified resistance factor of 0.6. There appears to be support for the use of a
value of 0.7, which is also under consideration by the AASHTO Bridge Code. A review of the
geotechnical resistance factors is anticipated to be part of new code development work, including an
assessment of the influence of class (level of detail) of geotechnical site characterization. In addition,
effects such as subsurface variability, construction quality control, and previous site and construction
experience would be interrogated to account for specific knowledge that an engineer has and can be
utilized in design. Although it is generally recognized that site investigation, test dependent and
knowledge-based resistance factors have merit, the approach for both the CHBDC and NBCC was to
keep the design process simple, at least during the initial stages of transition between working
(allowable) stress design and limit states design. It was felt that it is more important that the
fundamental principles of limit states design for foundations be conveyed to and understood by
geotechnical practitioners. The initial transition should be as gradual and smooth as possible.
Providing a myriad of partial factors that cover a large range of methods used in practice may not be
conducive to better understanding and acceptance of limit states design (LRFD) by geotechnical
engineers. Refinements and level of sophistication and details can come later when more experience
with limit states design for foundations has been gained. Without the "test" of designs in practice,
there can be no substantive verification of appropriate numerical values of geotechnical resistance
factors.
Assessment of appropriate partial factors for serviceability limit states has not received the same
kind of attention and scrutiny as applied to ultimate limit states. Currently the specified factor is 1.0 in
the NBCC and CHBDC. The effects of sampling disturbance and other effects will need to be
considered carefully. It is anticipated that partial factor values of both less than and greater than 1.0
may be a result of the assessment of partial factors for serviceability.
4
TAIPE12006 International Symposium on New Generation Design Codes
Risk Assessment in Rock Engineering

H. H. Einstein
Massachusetts Institute of Technology
SUMMARY
Risk assessment in rock engineering is based on the formal identification of uncertainties and
on their assessment and possible modification in the context of risk analysis and
management. The best way to include uncertainty in the engineering design process is
through the use of the basic structure of decision making under uncertainty which progresses
from information collection, to deterministic and probabilistic modeling to end up with risk
assessment and related decisions. These decisions, i.e. risk management range from
accepting the risk as is to modifying it.
Before applying this decision making process to rock engineering, it is necessary to be clear
as to what criteria engineering structures have to fulfill: safety, susceptibility, economics and
aesthetics and, particularly, to identify the relevant sources of uncertainty. In rock
engineering, the most important sources are inherent spatial variability,
measurement/estimation errors and model uncertainties.
In information collection, one needs to determine the relevant parameters and associated
uncertainties (distributions) through appropriate sampling procedures. Specifically, potential
biases have to be avoided and corrected for. Also, one needs to relate the sample to the
sample population and, most importantly, to the target population, the latter usually requiring
judgement. The result of information collection are state-of-nature models, which express
the natural variability. Stochastic fracture pattern models are examples.
In the deterministic modeling, phase one relates parameters to outcomes, i.e. predicted
performance. The performance can be related to stability, deformation, flow or economic
aspects (or combinations). In rock engineering, such performance is related to the typical
problems of slope stability, foundation performance, flow and tunneling. An important
aspect of the deterministic phase is the concluding sensitivity analysis, which is used to
identify the parameters having the greatest effect on the results. Usually only these
parameters will be varied in the probabilistic phase.
Probabilistic modeling is entirely analogous to the deterministic one but now the relevant
parameters and their uncertainties (distributions) are propagated through the model. Hence,
the state of nature models mentioned earlier provide the required input. An important issue
specifically related to rock engineering is the treatment of fracture persistence, i.e. the fact
that fractures and intact rock are interspersed; which has a significant effect on rock mass
performance. The probabilistic approach allows one to rationally solve "the persistence
problem". Probabilistic models are also well suited to deal with uncertainties affecting
economics such as the cost and time to build a tunnel.

In the final phase, one combines the uncertain performance from the probabilistic phase with
its consequences; this combination is the risk. When doing this one has to be aware of the
fact that a particular performance does not always have a consequence, another uncertain
aspect usually called vulnerability. Also, consequences can be expressed in form of cost or,
better, in form of utilities. Risk management can then be used to modify the risk through
active actions which change the probability of unsatisfactory performance, or passive actions
which change the consequences or the vulnerabilities.
Determining and using uncertainties in predictions have a long tradition in rock engineering.
Hence, quite a few procedures and models are available. It is, however, most important to
put all this in the context of the decision making structure as was done here.
6
Some Movements Toward Establishing Comprehensive Structural

Design Codes Based on Performance-Based Specification Concept
in Japan
Y. Honjo
Gifu University, Gifu, Japan
SUMMARY
Introduction
There have been movements in Japan to develop a serious of comprehensive structural design codes
which can harmonize all the major Japanese structural design codes. This movement is much
motivated by the rapid development and popularisation of international and regional structural design
codes such as IS02394 and Structural Eurocodes, as well as of the performance based design concept
especially after the conclusion of WTO/TBT agreement in 1995. In proposing such efforts, it is much
contemplated to propose a concept that can harmonize all the major Japanese structural design codes
that have been developed rather separated way due to many historical reasons. The performance based
design (PBD) (or the performance based specification (PBS)) and the limit state design (LSD) are the
two concepts we introduced to achieve this aim. One of the final aims of this activity is to propose a
new framework of structural design codes for harmonizing structural codes in regional and
international levels. Two of such efforts, namely development of 'Principles for Foundation Designs
Grounded on a Performance-based Design Concept' (nick name 'Geo-code 21') by JGS (Japanese
Geotechnical Society) and 'code PLATFORM ver.l' by JSCE (Japan Society of Civil Engineers) are
presented in this paper.
The relationships among WTO/TBT, PBD/PBS and LSD in the current design framework are
illustrated in Figure. 1. It is our belief that the specifications of performance of the structures
would be described based on the concept of PBD/PBS, whereas the verification of design would
be based on LSD/RBD for all the major design codes in the world.
In order to cope with the situations explained in the previous section, movements to establish a
series of comprehensive design codes have been started in Japan. One of the initial works of this kind
of movements started in 1997 at JGS (Japanese
Geotechnical Society) as drafting of 'Geocode WTO/TBT
21'*, a proto type comprehensive foundation PBD
(Performance Ni
design code that can harmonize all the major specifications Based Design) >
foundation design codes in Japan. The by
performance
comprehensive design codes stand at the top
hierarchy level in all the structural design codes
RBD/
in Japan to give concepts, framework and Respect
LSD/ ; other
International
terminologies for structural design codes as Standards LRFD ' design
(IS02394etc.) methods
indicated in Figure. 2. It is not intended to be
legally enforced but as agreements among the
professions (more specifically, the code writers)
to draft structural codes based on the rules,
terminologies and concepts established by the Figure 1 WTO/TBT agreement, PBD and LSD/RBD
comprehensive codes. Therefore, it is thought
that it is most appropriate for professional societies such as JSCE (Japan Society of Civil Engineers)
and JGS (Japanese Geotechnical Society) to publish such codes.
* 'Geocode 21' is a nick name of this design code. This name has been used from the beginning of the
project. The final official name of the code is Principles for Foundation Designs Grounded on a
Performance-based Design Concept and the official number of the code is JGS-4001-2004.

Geocode 21
Geocode 21 is drafted pursuing for an ideal

foundation design code at present time in Japan.
That is to say, the code is aiming at systematizing
and harmonizing the major foundation design codes
in Japan that have been developed rather
independently due to some historical and legal
reasons.
In proposing such code, it is neither meaningful
nor successful to try to develop a code at the same
level to the existing major design codes: An
advanced concept is definitely required in
proposing such a code. The PBD/PBS concept
is employed as the backbone of this code, and Figure 2 Concept of the comprehensive design code
is used to harmonize the major design codes on
a ground that is different from that of the present major design codes are based. The
comprehensive design code is fully performance based design code; but at the same time, it can be
looked at as 'a code for code writers'.
The aims of this code are as follows:
• To define means to specify the structure performances.
• Unification of terminologies.
• Methods and formats to introduce the safety margin to various limit states in design.
• Standardize characteristic value determination in geotechnical design.
• Standardize information flow (i.e. documents preparation) among owner, designer,
constructor, geotechnical investigator and others.
• The limit state design (LSD) concept is introduced for design verification.
For all the major design codes in Japan, it is principal that the design changes from the next
day a revised code is enforced for the category of structures under the control of that code because
of the legal background. It is too strong constraint for a code to introduce new concepts. For this
reason, it is our experience that all the new concepts introduced to the codes are creepingly
deformed, stripped of its essential contents in the process of drafting, and finally enforced with no
substances.
It is not expected that Geo-code 21 is to be used in the actual design from the day it is issued;
it is rather pursuing an ideal code which all the code should finally merge into it in the near future.
It is expected that various foundation design codes in Japan to accept the concepts and the formats
etc. proposed in this code, and finally mildly harmonized to this code in a certain time interval.
Final Remarks
Some of the activities on harmonizing Japanese major civil engineering structural design codes
are introduced in this paper. The authors are hoping this kind of activities are extended to Asian
region so that we can cooperate together to develop our own regional codes system to promote
construction industries within this region by unifying the market, and strengthen the
competitiveness of our construction industry to the outside.
Development and Implementation of Eurocode 7

Trevor L.L. Orr
Trinity College, Dublin University, Dublin, Ireland
EXTENDED SUMMARY
Eurocode 7 for geotechnical design is one of the set of Eurocodes for structural design using
different materials that are about to be implemented in Europe. The Eurocodes are all based on the
same limit state design method, set out in Eurocode EN 1990, with partial factors applied to
characteristic parameter values. In this paper, the development of Eurocode 7 from the initial work in
1981 to prepare a model limit state code for geotechnical design, through the preparation of the pre-
standard, ENV version of Eurocode 7, to the publication in 2004 of the of the full European standard,
EN 1997-1, Eurocode 7 Geotechnical Design - Part 1: General rules, is outlined. The issues that
arose in developing Eurocode 7 as a code that was consistent with EN 1990, took account of the
special features of soil and geotechnical design, and was acceptable to the European geotechnical
community were:
• The scope of Eurocode 7
• The definition of the characteristic value of a geotechnical parameter
• The value of the partial factor on permanent loads
• The application of partial factors to material parameters or resistances
• The treatment of water pressures and forces
• The accommodation of national design practices.
The nature of these issues and how they were overcome is discussed in this paper.
Regarding the scope of Eurocode 7, it was accepted by CEN TC 250, the management committee
for the Eurocodes, that the requirements for ground investigations and determining parameters from
field and laboratory tests are part of the design process and should be included within the scope of
Eurocode 7.
The definition of the characteristic value of manufactured structural materials as the 5% fractile of
an unlimited series of test results is shown to be not appropriate for geotechnical design. The
principal reason for this is because the geotechnical parameter controlling, for example, a failure in
the ground, is the mean strength over the failure surface, not the strength of an individual test
element. Hence it is the 5% fractile of the mean strength along the failure surface that is required, not
the 5% fractile of the test results. Another reason is because, in geotechnical design, only a limited
number of test results are normally available and hence statistics need to be used with caution.
Eurocode 7 states that the characteristic value "shall be selected as a cautious estimate of the value
affecting the occurrence of the limit state". This definition is an important innovation in Eurocode 7
and some guidance on the selection of the characteristic value is provided in the paper.
Since the Eurocodes are for structural design, the partial factor chosen for permanent loads in EN
1990 was 1.35. This value caused a problem for Eurocode 7 because in geotechnical designs, for
example in slope stability analyses, the permanent actions due to the weight of soil are not normally
factored. If they are factored, then illogical situations can arise; for example, in the case of a circular
vertical failure surface below horizontal ground, if the unfavourable soil weight, treated as a
permanent load, is factored by 1.35 while the favourable soil weight is not factored, then analysis of
this situation can predict failure of the horizontal ground when it is not loaded. This is not logical and
therefore the Eurocode 7 drafting panel successfully resolved this issue by getting TC 250 to accept a
partial factor of unity for permanent actions in geotechnical design when factors greater than unity
are applied to the soil strength.
In the ENV version of Eurocode 7, partial material factors are applied to the soil strength
parameters c', tan<j>' and c„, in the same way as partial material factors are applied to the strength
parameters in the other Eurocodes. The partial factors for geotechnical ultimate limit states adopted
in the ENV version are the sets of partial factors referred to as Cases A, B and C. However, many
geotechnical engineers in Europe were not happy with these three Cases, with partial factors applied
to the soil strength parameters. The inclusion of partial factors applied to soil resistances was sought,

after these had been calculated using unfactored soil parameter values. Hence, in the EN version,
three Design Approaches, DAI, DA2 and DA3, were introduced that allow for partial factors to be
applied either to soil resistances or to the soil strength parameters.
Pressures and forces due to groundwater are treated as permanent actions in Eurocode 7 and
hence, in ultimate limit states involving failure in the ground, the appropriate partial factors in
Eurocode 7 for permanent actions are applied to water pressures. In the ENV version, the partial
factors in Case A are for use in the case of buoyancy ultimate limit states and no specific guidance or
partial factors are given for ultimate limit states involving seepage. Hence, when the EN version was
being prepared, it as decided to include a new section on hydraulic failure, which includes design
rules and partial factors for design against seepage failure as well as buoyancy.
Since Eurocode 7 is not a prescriptive codes but a code with the general rules for geotechnical
design, providing the principles and only a few calculation rules in informative annexes, and since
different national practices have developed in the European countries, reflecting different geologies
and soil conditions, TC 250 accepted that the valuable experience embedded in these practices may
be accommodated by supplementing Eurocode 7 with non-conflicting national standards.
Two examples, a spread foundation with a vertical central load and an embedded retaining wall,
are presented to demonstrate the effect of using the different Design Approaches. In the case of these
examples, it was found that for low 0' values, DA2 is more conservative than DAI, while for high <]>'
values, it is less conservative, as shown by the foundation widths in Figure 3 below from the paper.
0.5 -
0 -I 1 , 1
20 25 30 35 40
Friction angle <])' (degrees)
Figure 3: Design widths for the square pad foundation
National Annexes with values for the Nationally Determined parameters (NDPs), e.g. partial
factor values, are due to be completed by November 2006 so that each member state can implement
Eurocode 7 as a national code and usher in the Eurocode era. The advantages of Eurocode 7 are that,
in providing the limit state principles for geotechnical design and not being prescriptive, it
harmonises geotechnical design with structural design, it takes account of the special features of soil
and geotechnical design, and it allows national design practices to be accommodated. Hence
Eurocode 7 is applicable, not only in Europe, but also worldwide.
10
Invited Lectures
New Trend Toward Performance-Based Design in the

Construction Industry
K. Horikoshi
Taisei Corporation, Kanagawa, Japan
Y. Honjo
A. Iizuka
Kobe University, Hyogo, Japan
SUMMARY
More than 10 years have passed since performance-based design was introduced into the
construction industry along with a set of expectations in Japan. Nowadays, Japanese engineers are
very familiar with the term 'performance-based design'. Designers and contractors expected more
flexible application of new design and construction methods for public works. The public sector
expected more cost-effective construction by setting the required performance of individual
structures based on importance. Despite these expectations, however, the application of
performance-based design in public works is still limited.
At present, some of the large Japanese projects are based on design-build contracts. Design-
builders must take more risks to achieve the required performance. Risk sharing related to
geotechnical conditions is always a difficult issue between the client and the design-builder.
Insurance systems that cover such risks are not fully developed due to difficulty in determining
the insurance premium. It seems that there is still a number of unsolved issues to make much use
of the advantages of performance-based designs in public works.
A significant portion of the present paper is based on the activities of the JSCE (Japan Society
of Civil Engineers) Committee on 'Performance-based design of soil structures' chaired by Prof.
Atsushi Iizuka of Kobe University. The committee was organized in 2004 to discuss and clarify
the applicability of performance-based design to soil structures such as embankments, reclaimed
islands, and dams. Investigation of European systems to support performance-based design was
conducted as one of the activities.
This paper describes
1) The current situation of the Japanese construction industry;
2) New contracting and bidding systems employed in Japan that are oriented more toward
performance and quality without compromising cost; and
3) Results of investigation on performance-based design in European countries.
Following trends and findings are included through the study described in this paper.
1) The Japanese domestic construction market has become increasingly competitive. Severe price
wars have led to client concerns over maintaining the quality of public works, which has led to
the application of new contracting and awarding systems.
2) Some of the top contractors have expanded their markets to overseas. The increase in overseas
projects has resulted in designers, consultants, and contractors shifting toward performance-
based design in order to avoid or reduce potential risks in the target projects.
3) New contracting systems such as design-build system, value engineering system, technical
proposal integrated evaluation system, performance requirement ordering system, and PFI
(Private Finance Initiative) system has been applied in Japan. These systems enhance the
implementation of better performance in public works through the private sector's proposals.

4) Expectations for performance-based design differ depending on professional affiliation.
Designers and contractors tend to expect more flexibility in performance-based design,
whereas public-sector clients tend to expect better performance and quality at reduced cost.
5) Although the actual term 'performance-based design' may not be popular in European
countries, the performance of structures is always considered by designers in their contracting
systems, such as design-build and PFI.
6) Consultants probably play a more important role in Europe than in Japan. It seems that
lawsuits are also more popular in European countries. Again, consultants play an important
role in such lawsuits.
7) Contracting systems, insurance systems, relationships among clients, designers, and
contractors, all differ from one country to another, although design codes such as Eurocode
have been developed in harmony. Although the difference in systems will become smaller
with time, Eurocode operation will progress with different supporting systems. The unification
of these systems may take a longer time.
8) As illustrated in the figure, performance-oriented code development, as well as balanced
development of other systems, is inevitable for enhancing the dissemination of performance-
based design, and maximizing its advantages. It is also important that these systems be
considered based on global standards.
Performance oriented
Code development
Bidding/Awarding systems Verification Systems of
(Technical Proposal Integrated Design/Construction
Evaluation System, etc)
method
v
Insurance Systems
M
Contracting Systems
(PI insurance etc.) (Design-Build, CM, PFI etc.)
Systems promoting performance-based design
14
Risk Analysis of Lining Structure in Large-Diameter Shield

Tunnel
Hongwei Huang
Department of Geotechnical Engineering, Tongji University, Shanghai, P.R.China
Qunfang Hu
Department of Geotechnical Engineering, Tongji University, Shanghai, P.R.China
Yuanyuan Yang
COWI Consulting Ltd. Co., Shanghai, P.R.China
SUMMARY
In general, the lining structure of shield tunnel is composed of segments and joints. According to
the difference of joints among segments, the assembly of lining structure is defined as segment
seams in sequence and segment seams in stagger. The service life of shield tunnel is usually
designed as 100 years, the lining structure is below the ground level with carrying the load of soil,
water and vehicle in its life cycle. So the risk condition of the lining structure will have great
effects on the safety during tunnel construction and operation, and risk analysis is much
significant.
In this paper, risk accidents during lining structure construction are analyzed, and various
potential risk factors are identified, and risk analysis model is established based on the faulty tree
theory. Considering the functional requirement and mechanical behaviour of lining structure, the
possible damage types mainly include:
(1) ULS: circumferential or longitudinal damage due to insufficient strength of bolts or
segments, circumferential or longitudinal joint expansion, segment crack, etc.
(2) SLS: Lining structure damages due to too large diameter deformation, longitudinal
differential settlement, durability failure, operation factors, etc.
Combining the two damage types as above and considering the uncertainty of geotechnical
engineering, the main risk factors that affect the shield tunnel lining structure are as follows in the
view of engineering risk identification:
(1) Internal factors mainly are from the structure itself, including the design model of lining
structure, material, prefabrication and curing method, assembly tolerance of segments, bolt
strength, joint tolerance, circumferential and longitudinal variation of segments and joints, which
will cause risk accidents directly.
(2) External factors mainly are the loads and environmental action on the lining structure. Due
to the constraints during geological investigation, the investigation results are usually not the same
with actual condition. Furthermore, large variation appears in the random distribution of soil, the
physical and mechanical parameter, and underground water. Therefore, the uncertain factors such
as site construction loads and environmental medium (e.g.Cr'.SOj ,Mg2 ,HC0 3 ,C0 2 ) and motors
may also cause risk accidents of lining structure.
Based on the analysis of the risk factors and accidents for shield tunnel lining structure, the
fault tree theory could be used to establish the structural risk analysis model to quantify the risk
evaluation.
Considering the variation of stratum and lining structure, the risk occurrence probability for the
events in the model is calculated by the means of engineering reliability theory and finite element
method analysis. And then the above method is used to analyze the lining structure risk in

Shanghai Yangtze River Tunnel. Simultaneity, the key degree of each risk-induced factor was also
discussed. In the end, there conclusions were drawn from the research work.
(1) The 3 main failure types of lining structure during construction mainly include segment
damage, joint damage and sealing pad damage. As there are many factors that would cause lining
structure damage, it is very essential to consider engineering risk analysis and evaluation during
construction which could provide guidance for decision-making.
(2) Systemic risk analysis model for lining structure during construction is established based on
the fault tree theory. To evaluate the risk quantitatively, the mechanical behaviour and
deformation condition of lining structure are selected as control parameters, of which the risk
occurrence probability is calculated on the basis of reliability theory and numeric analysis.
Systemic risk evaluation method is proposed and the risk of lining structure is analyzed, and the
significance of each risk factor is discussed. Therefore, the proposed model and method could be
applied into real projects.
(3) The risk analysis of lining structure in Shanghai Yangtze River Tunnel shows that the
reliability of the tunnel lining meets the design requirement in general. The risk at K2+373 where
the depth is maximum and K7+396 where the cover thickness is maximum is large, so special
management, pre-warning, prediction and control measures during construction and operation
shall be taken. Meanwhile, to reduce the risk occurrence of lining structure, construction load
shall be paid attention to, especially the risk factors during segment transport, grouting pressure
and erection assembly, which shall be avoided during construction.
16
Energy Approach to Earthquake-Induced Slope Failures for

Performance-Based Design
T. Kokusho
Chuo University, Tokyo, Japan
SUMMARY
Seismically induced slope failures have normally been evaluated based on equilibrium of forces acting on
a potentially sliding soil mass. This force approach can evaluate the safety factor against the slope
failure or displacement not exceeding about 1 m normally by Newmark method but cannot predict slide
deformations, once large failure occurs. There exists no simple method available for performance-based
design in which different levels of slope performance including ultimate flow-type failure can be
evaluated at this moment.
In this research, an energy approach is proposed for performance-based design for predicting
residual displacement including long run-out distance in seismically induced slope failure. Shake table
tests of dry sand slopes are carried out to examine associated energy balance by comparing test results
with a Newmark-type rigid block model. The energy approach is then applied to a hypothetical slope
under extremely large earthquake energy to show its potential for evaluating different levels of slope
performance including ultimate collapse. Major findings obtained in this research are:
1) In shake table tests of dry sand slopes with different slope inclinations and different input frequencies,
earthquake energy used for slope failure could be successfully measured, quantifying the energy
balance involved in the failure of the model slope.
2) The model tests yielded a unique relationship for each slope inclination between the energy EEQ and
residual slope displacement Sr, which is independent of input frequency, indicating that energy can
serve as a better parameter for evaluating residual slope displacement than acceleration.
3) The above mentioned EEg versus 5r relationship shows clear thresholds of EEQ below that 5r =0,
which is again independent of input frequency. This implies that not only residual displacement but
also initiation of slope failure is determined not by acceleration but by energy.
4) Comparison of the test results with energy balance for a Newmark-type rigid block indicated that the
rigid block model can almost perfectly emulate continuously deforming sand slope provided that an
appropriate friction coefficient in the rigid block model can be estimated.
5) In order to back-calculate the friction coefficient, the energy-based method was applied to a recent
case history during the 2004 Niigata Chuetsu earthquake. The equivalent friction angle 12.6° was
back-calculated, much lower than the inclination of the slip plane 20°, indicating that the failed soil
mass accelerated and piled up on the opposite side of the valley.
6) A case study on a simple slope demonstrates that even under an extremely large earthquake, a
flow-type slope failure of a long run-out distance cannot occur unless equivalent friction angle
approaches to slope angle. Also indicated is that even the strongest earthquake energy is actually
far smaller than potential energy in long run-out distance failures.
7) Thus, the case study indicates that the energy-based method can serve as a powerful tool to predict
different levels of slope performance corresponding to serviceability, reparability and ultimate
collapse with long run-out distance.

A Preliminary Study on Load and Resistance Factors for

Foundation Piles in Taiwan
H. D. Lin
National Taiwan University of Science and Technology, Taipei, Taiwan
SUMMARY
This paper describes the current pile design code of Taiwan, presents a rational approach for the
determination of the resistance and load factors for foundation piles subjected to axial load, and then
compares the results with some LRFD design codes. Results show that the resistance and load factors
may be significantly affected by the target reliability, the ratio of live load to dead load, the pile
failure criterion, and the design method. Therefore, these influencing factors should be carefully
considered when developing the load and resistance factors for an LRFD code. This paper also
presents a preliminary set of resistance and load factors for bored piles and driven piles in Taiwan.
Pile design code in Taiwan

The Foundation Design Code for Building developed by the Taiwan Geotechnical Society (TGS) is
the most comprehensive and the most widely used foundation design code in Taiwan (TGS, 2001).
The WSD method was adopted in the TGS code and remains the mainstream design philosophy.
However, some initiatives have been taken by the TGS toward the development of a new
performance-based geotechnical design code. Axial capacity of piles is a summation of the skin
resistance along the pile shaft and the point resistance at the pile tip. Both the static equations and the
SPT-N formulas are suggested for axial pile capacity estimation. The ultimate axial capacity is then
divided by a safety factor to obtain the allowable design capacity. In this study the LRFD factors are
developed mainly based on the pile capacity design equations recommended in this TGS code.
Reliability analysis of partial factors

The performance function defined for the analysis is as follows: G = R-QD-QL , in which G =
safety margin of the pile, R- pile resistance (log-normal), QD= dead load (normal), and QL= live
load (Type I). The point on the failure surface (G=0) with minimum distance to the origin is the
most probable failure point. In the reduced normal space this distance is the reliability index of the
system. After finding the reliability index, an iterative procedure, the "second moment formulation"
(Ann and Tang, 1984), can then be applied to obtain a set of load and resistance factors such that the
reliability index of the system meets the target reliability (3.0). Before one can perform this
reliability analysis the uncertainties of the pile resistance and pile loads need to be determined.
For the evaluation of the pile resistance uncertainty, a number of sources are considered, including
soil properties, design equations for axial pile capacity, and pile load failure criteria. Axial capacity
of piles can be denoted as follows: R = NR R, in which R = actual pile resistance interpreted from
the pile load test, R = pile resistance estimated using the design equation, and NR = the correction
factor. Pile load test results can then be analyzed together with the statistics of the soil stratum to
evaluate the uncertainty of the pile resistance.
Regarding the pile load, only dead load and live load were considered in this study. At present,
there is no reliable load distribution data for pile design in Taiwan. Therefore, the load criteria
recommended by the American National Standard A58 was used. The load ratio (QLIQD) was varied
in the analysis to study its effects on the LRFD factors.
Load and resistance factors of bored piles and driven piles

For bored pile analysis, 14 full-scale load test results of bored piles embedded in the Taipei
metropolitan area were used for the analysis of pile resistance uncertainty. Four design equation sets
were used to estimate the axial pile capacity. For pile load interpretation, the Davisson criterion was
used. For driven pile analysis, 33 load tests of PC piles embedded in a reclaimed industrial park in
south-western Taiwan were selected for the analysis. Three design equations and four failure criteria
were considered. Highlights of the analytical results can be summarized as follows:
(1) Load and resistance factors are summarized in Table 1 below. They are consistent with values
documented in other LRFD codes (Eurocode, the CGS code, and the ASSHTO bridge code).
(2) The resistance factor and the live load factor increase significantly with increasing load ratio.
However, the dead load factor decreases only slightly. In addition, all the partial factors exhibit a
tendency to level off after the load ratio increases to 2.0, as shown in Figure 1.
(3) The pile failure criterion as well as the design method may have a significant impact on partial
factors. The influence on the resistance factor is the most pronounced. At present, the individual
effects of failure criterion and design method seem to be coupled together. Further study may be
warranted to clarify this issue.
(4) The target reliability also has a significant influence on the resistance factor and the live load
factor for driven piles but the effect on the dead load factor is negligible. When determining the
partial factors for pile design code, appropriate target reliability is essential.
Table 1. Pile load and resistance factors.

Pile Type Design Method <t>YD YL
SPT-N 0.33-0.50 1.02- 1.07 1.28- 1.52
Driven Pile
Static Method 0.27-0.55 1.02- 1.07 1.25- 1.54
Bored Pile SPT-N or Static 0.49 - 0.62 1.04- 1.08 1.35- 1.53
2 -
Reliability Index=3.0 - Davisson
1.8 - + + Resistance Factor Fuller&Hoy
— • — • Dead Load Factor Terzaghi
# # Live Load Factor - Chin
1.6 -
1.4 -
I-
! -
0.8 -
0.6 - tsgnk?^,-.--^:-
0.4 -
0.2 -
i ' i ' i ' i ' r T
0 0.5 1 1.5 2 2.5 3 3.5
Live Load / Dead Load
Figure 1. Load and resistance factors of driven pile (design method I).
Conclusions
The load and resistance factor design method (the LRFD method) that is based on reliability analysis
offers a very rational framework to account for all aspects of the uncertainty related to pile
performance. The reliability approach also offers more flexibility for future adjustment of the safety
factors. Studies conducted by the author are in strong support of the above statements. As a result, a
set of partial factors for bored piles and driven piles in Taiwan is suggested. The suggested partial
factors seem reasonable and are consistent with available LRFD codes. However, they should be
used with caution because further study using a broader data base is warranted.
20
TAIPE12006 International Symposium on New Generation Design Codes
Evaluating Probability of Seismic Landslide Based on the Chi

Chi's Events, Taiwan
M. L. Lin
National Taiwan University, Taipei, Taiwan
C. J. Chung
Geological Survey of Canada, Ottawa, Canada
M. H. Ho
National Taiwan University, Taipei, Taiwan
SUMMARY
Locating in the circum-pacific seismic zone with frequent earthquake activities, Taiwan is
susceptible to hazard induced by earthquakes. In September 21 st , 1999, the Chi-Chi earthquake
with a magnitude of 7.3 struck central Taiwan, and caused severe damages to human lives and
properties. Among all the damages, more than 20000 cases of landslides were identified with a
total area of more than 8600 hectares by Council of Agriculture (2000). In order to mitigate the
hazard and plan for the land use, it is essential to evaluate probability and develop prediction map
for potential landslide induced by the earthquake. The evaluation of the probability of future
landslide can be performed by applying the conditional probability concept to the documented
events from the Chi-Chi earthquake in this case.
Quantitative model for conditional probability

The quantitative prediction models for landslide hazard can be developed based on a spatial
database consisting of causal factors of the landslide occurrences using the conditional probability
function to represent a quantitative measure of future landslide hazard (Chung, and Fabbri, 1999).
The procedures suggested by Chung and Fabbri (2003) for estimating probability of occurrences
of future landslides under the assumptions of a given scenario are used in this study. A prediction
model of future landslide can be generated based on the likelihood ratio function of the
multivariate frequency functions of the causal factors. With assumption of conditional
independence of the factors, the likelihood ratio function can be estimated as a multiple of
separate univariate likelihood ratio functions. The overall likelihood ratio function can be
estimated using:
Hxi xk,yl,...yh) = \{xi)-...-X{xk)-X(yi)-...-\(yh)

(1)
with xl,...,xk, y,,...y h denoting categorical and continuous data.
Study area and database

The research area locates in the Nantou County along the Cherlungpu fault close to the central
ridge mountain area with a width of 25.3 km and a length of 27.5 km. The study area is divided
into the upper half and the lower half subareas for construction of model and cross validation. The
ground variation data by Bureau of Soil and Water Conservation, Council of Agricultural (2000)
were used to provide the landslide scar information, which were based on the pixel variation of
the SPOT satellite images taken before and after the earthquake. After verification, a total of 1061
landslide scars for the whole study area are identified. The causal factors related to the geo-
environmental characteristics of landslide are used, which include: elevation, slope angle, aspect,
and geological formation of the study area. The factors related to the characteristics of the

landslides and geo-environment required extraction from the basic information, pre-treatment, and
quantification, which were performed using the GIS system.
Prediction map and validation

The likelihood ratio of each pixel of the study area is calculated according to Eq. (1) using data set
of each causal factor, and the hazard map is generated based on the scar information. The
prediction map of the future landslide could be generated using the final fitting model, and
verification is required to check the effectiveness of the prediction model The final prediction
model established from triggering area of landslide scar and causal factors of slope angle, slope
aspect, and geological formation data is used, which yields a prediction rate with 10% of the
mapped highest hazard area containing about 56% of the scar data. The prediction map shown in
Fig. 1 is generated by combining the prediction map from both the lower subarea and the upper
subarea. The uncertainty of the prediction can be estimated from the likelihood ratio calculated for
each pixel.
Conclusions
Based on the previous discussions, a prediction model is constructed using likelihood ratio method
following the theory of conditional probability. From the results of the analysis, the geological
formation, slope angle, and slope aspect factors appeared to be effective causal factors and the
prediction model established accordingly yielded satisfactory results, and the uncertainty of the
prediction results can be estimated.
Figure 1. Combined prediction map of the study area.
22
Serviceability Limit State Reliability-Based Design

K. K. Phoon
National University of Singapore, Singapore
SUMMARY
In foundation design, the serviceability limit state (SLS) often is the governing criterion,
particularly for the cases of large-diameter piles and shallow foundations. In addition, it is
widely accepted that foundation movements are difficult to predict accurately. This paper
discusses the application of a probabilistic hyperbolic model for performing reliability-based
design checks at the serviceability limit state. The important nonlinear feature of the load-
displacement curve is captured by a two-parameter hyperbolic curve-fitting equation:
Q = y
Qm a + by
in which Q = applied load, Q m = interpreted capacity, "a" and " b " = curve-fitting parameters, and
y = pile butt displacement. The uncertainty in the entire load-displacement curve is represented
by a relatively simple bivariate random vector containing the hyperbolic parameters as its
components. It is important to highlight that these hyperbolic parameters typically exhibit a
strong negative correlation. A translation-based Hermite expansion can be used to capture this
correlation aspect correctly in a fairly general way:
a = a 1 0 H 0 ( X 1 ) + a 1 1 H 1 (X 1 ) + a 1 2 H 2 ( X 1 ) + a 1 3H 3 (X 1 ) + a 1 4 H 4 ( X 1 ) + a 1 5 H 5 ( X 1 ) + ---
b = a20H0(X2) + a21H1(X2) + a22H2(X2) + a23H3(X2) + a24H4(X2) + a 2 5 H 5 ( X 2 ) + - -
in which Xi and X 2 are standard normal variables and the Hermite polynomials Hj(.) are given by:
H„(X) = 1
H,(X) = X
H k + 1 (X) = X H k ( X ) - k H k _ , ( X )
The correlation between " a " and " b " (p s , b ) is related to the correlation between Xi and X 2 (pxix2):
^ k ! a l k a 2 k p X ix2
: k=l
Pa,b
f>!a?Jjrk!aL

3.0 -i 3.0 -i
0 10 20 30 40 50 60 70 80 0 10 20 30 40 SO 60 70 80
Displacement (mm) Displacement (mm)
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Displacement (mm) Displacement (mm)
Figure 1. Normalized hyperbolic curves for ACIP piles:

(a) measured, (b) pXix2 = -0.8, (c) pXiX2 = -0.4, and (d) pXix2 = 0.
The common assumption of statistical independence can circumvent the additional complexity
associated with a translation model, because the bivariate probability distribution reduces to two
significantly simpler uni-variate distributions in this special case. However, the scatter in the
measured load-displacement curves cannot be properly reproduced by simulation under this
assumption as shown in Figure Id. On the other hand, Figure lb looks more realistic because the
proper negative correlation is included in the bivariate probability model for "a" and "b".
The SLS is defined as that in which the vertical or lateral displacement (y) is equal to the
allowable limit (ya) imposed by the structure. The foundation is considered unserviceable if the
displacement is greater than the allowable limit. Conversely, the foundation is considered
satisfactory if the displacement is less than the allowable limit. These three situations can be
described concisely by a performance function:
P = y - y a = y(Q)-y«
An alternate performance function is:
P = Q . - Q = Q.(y.)-Q
Figure 2 illustrates the uncertainties associated with these performance functions. In Figure 2a,
the applied load Q is assumed to be deterministic to simplify the visualization. It is clear that the
displacement follows a distribution even if Q is deterministic because the load-displacement
24
(a) /! (b)
1.6-1 Random displacement caused by uncertainty in load- 1.6 - /[ Random allowable
displacement curve M\ . load caused by — —
/ \ uncertainty in load-
1.2 1.2 - f Jf*** displacement curve
0.8 -
1 ^f^*^ Random applied load
0.4 -
1 ^
J > Deterministic allowable
1 displacement
00 -
20 40 60 20 40 60 80
y (mm) y (mm)
Figure 2. Serviceability limit state reliability-based design.
curve y(Q) is uncertain. The allowable displacement may follow a distribution as well. In Figure
2b, the allowable displacement is assumed to be deterministic. In this alternate version of the
performance function, the allowable load Qa follows a distribution even if y„ is deterministic
because of the uncertainty in the load-displacement curve. The effect of a random load Q and the
possibility of upper tail values falling on the nonlinear portion of the load-displacement curve are
illustrated in this figure.
The probability of failure (pf) at the serviceability limit state can be computed easily using the
first-order reliability method (FORM) once the probabilistic hyperbolic model is established:
p f =Prob(Q a <Q) = Prob -Qm<Q °r

a + by,
mQQ Ya , 1 Q '
p f =Prob = Prob
a + bya "QmVm
a + bya F S Q ;
The interpreted capacity Qm, allowable displacement ya, and applied load Q are assumed to be
lognormally distributed with a coefficient of variation (COV) = 0.5, 0.6, and 0.2, respectively.
The hyperbolic parameters are negatively correlated with an equivalent normal correlation of-0.8
and are lognormally distributed with mean and COV of "a" = 5.15 mm and 0.6 and mean and
COV of "b" = 0.62 and 0.26. The mean of y, is assumed to be 25 mm. The other parameters are
defined as: mQm, mQ = mean of Qm and Q, FS = mQm/mQ, and Qm*, Q* = unit-mean lognormal
random variable with COV = 0.5, 0.2. Note that FS refers to the ultimate limit state (ULS) global
factor of safety commonly adopted in prevailing practice.
For a typical FS = 3, the SLS reliability index for this problem is 2.21, corresponding to a
probability of failure of 0.0134. The probabilistic hyperbolic model can be simplified to a linear
model by setting b = 0 (physically, this is equivalent to an infinitely large asymptotic limit). For
the above example, the FORM solution under this linear assumption is p = 2.76 (pf = 0.00288).
Note that the solution is significantly unconservative — the probability of failure is about 4.7
times smaller than that computed for the correct nonlinear hyperbolic model. Hence, it is crucial
to model the nonlinearity in the load-displacement curve in reliability analysis. The reason is that
the applied load follows a probability distribution and it is possible for the load-displacement
behavior to be nonlinear at the upper tail of the distribution.
The allowable load can be evaluated from the interpreted capacity using a SLS model factor
(Ms) as follows:
25
Qa=-Ir-Qm=MsQm
a + bya
If M s follows a lognormal distribution, then a closed-form solution exists for the reliability index
(P):
m
l n fmMs Qm"|| 1 + COV^ ~
m
[I Q J l J ^ C O V ^ J l + COV^j
V^ll1 + C O V M S J(l + COV^, j(l + COV* )J
This procedure is similar to the bias/model factor approach used in ULS reliability-based design.
The catch is that the distribution of M s has to be evaluated for each allowable displacement. In
contrast, the foundation capacity is a single number (not a curve) and only one model factor
distribution is needed. Another practical disadvantage of the SLS model factor approach is that
M s cannot be easily evaluated from a load test database if y, follows a distribution.
Reliability analysis for the more general case of a random allowable displacement can be
easily carried out using an implementation of the first-order reliability method in EXCEL. It can
be seen in Figure 3 that a target reliability index = 2.6 is not achievable at a mean factor of safety
= 3 for a mean allowable displacement of 25 mm. However, it is achievable for mean factors of
safety larger than about 4. Note that the 50-year return period load Q50 « 1.5 mQ for a lognormal
distribution with COV = 0.2. Hence, a mean factor of safety of 4 is equivalent to a nominal
factor of safety = mQm/Q50 = 2.7. A nominal factor of safety of 3 is probably closer to prevailing
practice than a mean factor of safety of 3. It is also debatable if the allowable displacement
prescribed in practice is a mean value or some lower bound value. For a lognormal distribution
with a COV = 0.6, a mean allowable displacement of 50 mm produces a mean less one standard
deviation value of 25 mm. Using this interpretation, a target reliability index of 2.6 is achievable
at a mean factor of safety of about 3 or a nominal factor of safety of about 2. Overall, the EPRI
target reliability index for SLS is consistent with foundation designs produced from our
traditional global factor of safety of between 2 and 3.
The proposed approach is shown to be more general and more convenient to use than the
bias/model factor approach because the distributions of the hyperbolic parameters do not depend
on the allowable displacement. It can be argued that a random vector containing two curve-
fitting parameters is the simplest probabilistic model for characterizing the uncertainty in a
nonlinear load-displacement curve and should be the recommended first-order approach for SLS
reliability-based design.
mean COV
- Hyperbolic parameters: a 5.15 mm 0.6
b 0.62 0.26
Capacity: Qm mom 0.5
Applied load: Q mQ 0.2
Allowable displacement: ya mya ag^^"
•*— EN1990
•—EPRI
m
ya j / \
(mm) J^[/^'^
50-V^V^^
2 5 ' ' \ s ^
1 5 ^
1 2 3 4 5
Mean factor of safety = mQm/mQ
Figure 3. Relationship between reliability index and mean factor of safety for ACIP piles.
26
Eurocode 7 for Geotechnical Design - Basic Principles and

Implementation in the European Member States
Bernd Schuppener
Federal Waterways Engineering and Research Institute, Karlsruhe, Germany
Roger Frank
ENPC (National School of Bridges and Roads), Paris, France
SUMMARY
Eurocode 7 on geotechnical design consists of two Parts: General rules (Part 1, EC 7-1) and Ground
investigation and testing (Part 2, EC 7-2). Their conversion into foil European Standards (ENs) has
now been completed. EC 7-1 is a rather general document giving only the principles for geotechnical
design within the general framework of Limit State Design. These principles are relevant to the calcu-
lation of the geotechnical actions on structures and to the design of the structural elements themselves
in contact with the ground. EC 7-1 includes fundamental sections on the basis of geotechnical design
and geotechnical data, sections on the design of various types of geotechnical structures such as spread
foundations, piles, anchorages, retaining structures and embankments and sections on geotechnical
verifications such as hydraulic failure and overall stability. Detailed design rules or calculation mod-
els are only given in informative Annexes. Annex A is important as it specifies the partial factors for
verifications of the Ultimate Limit State in persistent and transient design situations as well as correla-
tion factors for the characteristic values of pile bearing capacity. However, the numerical values of the
partial or correlation factors given in Annex A are only recommended values. All the other Annexes in
which detailed design rules or calculation models are given are informative annexes.
EC 7-2 is devoted to laboratory and field testing and states the essential requirements for the
equipment and test procedures, for the reporting and the presentation of results, for their interpretation
and, finally, for the derivation of values of geotechnical parameters for design. It consists of Section 1
General, Section 2 Planning of ground investigations, Section 3 Soil and rock sampling and ground-
water measurements, Section 4 Field tests in soils and rocks, Section 5 Laboratory tests on soils and
rocks and Section 6 Ground investigation report and contains informative annexes dealing with the
most important laboratory and field tests.
The discussions about verifications of geotechnical design usually focus on approaches performed
by calculations. Nevertheless, it should be stressed that calculations are not the only means of check-
ing that the basic requirements are fulfilled. EC 7-1 also offers the adoption of prescriptive measures,
experimental models and load tests, as well as the observational method. Thanks to the Eurocodes, a
single format will be used for the mathematical analysis of the ultimate limit states throughout the
construction sector in Europe in future. Accordingly, for any section in a structure, structure-soil inter-
face or the soil, it will have to be verified that the design value of the effects of actions, Ed, never ex-
ceeds the design bearing capacity or the design resistances, Rj, i.e.:
Ed < Rd (1)
There has to be a clear-cut distinction between the effects of actions and resistances in order for the
general limit state equation to be applied. However, in geotechnical engineering, there are many cases
in which it is not possible to make a clear-cut distinction between the effects of actions and the resis-
tances. Additional problems concerning the application of equation (1) are caused by the fact that there
are two entirely different ways of introducing the partial safety factors in geotechnical engineering. On
the one hand, the design values, Ed and Rd, of the geotechnical effects of actions and resistances can be
determined by what is known as the method of factored shear parameters. In this method, the partial
factors are applied to the characteristic shear parameters, <p\ and c\. Thus the design value of the ef-

fective coefficient of friction, tan q>d, is determined by dividing the characteristic coefficient of fric-
tion, tan tp'k, by the partial factor for friction, y r Similarly, the design cohesion, c' d , is obtained by
dividing the characteristic cohesion, c' k , by the partial factor for cohesion, yc, i.e.: tan cp'd = tan cp'k / yv
and c'd = c'k/ yc. The design values of the geotechnical actions and resistances, Ed and Rd, to be used
in the limit state equation (1) are then determined with the design values of the shear parameters, cp'd
and c' d . On the other hand, there is the method of factored actions and resistances. In this method, the
characteristic values of the actions, effects of actions and resistance of the soil, Fk, Ek and Rk respec-
tively, are determined using the characteristic values of the shear parameters, cp'k and c' k . The design
values of the geotechnical effects of actions, Ed, (stresses, internal forces and moments) and the resis-
tances are then obtained by applying the partial factors for the geotechnical effects of actions and re-
sistances, YE and YR, to the characteristic values, i.e.: Ed = Ek • YE and Rd = Rk / YR-
The different ways of introducing the partial factors into the calculation are the principle reason
why EC 7-1 offers three different methods of verifying ultimate limit states for persistent and transient
situations. The numerical values of the partial factors to be applied in a given design procedure can
also be determined nationally and must be specified in the National Annex to EC 7-1.
After the publication of the three language version of EC 7-1 in 2004 the European Member States
have a period of two years in which to prepare the national versions of EC 7-1. These will comprise
- a national title page and national foreword,
- the full text of the Eurocode with all annexes and
- a National Annex.
The National Annex is needed to make EC 7-1 operational in the Member States and will therefore
- define the values of the partial safety factors,
- select the national design approaches and
- draw up specifications on the use of the informative annexes of EC 7-1.
There are three basic principles that have to be adhered to when harmonising European standards:
- Eurocodes must be introduced by the national standards bodies and applied in all Member
States of the EU.
- National standards in the technical fields in which European standards exist must be with-
drawn after a transitional period.
- National standards in the technical fields not covered by European standards are permitted as
long as they do not conflict with the Eurocodes.
There are a great number of geotechnical standards and recommendations containing valuable ex-
perience in geotechnical design in many European countries. It goes without saying that such experi-
ence must be preserved. This is possible as long as there is no conflict with the principles and applica-
tion rules of EC 7-1. Over the next five years, the national standards bodies in Europe will have not
only to introduce the Eurocodes and adapt their national standards but also to communicate the new
concept of limit states and partial factors to practising civil engineers in their respective countries. The
paper will show how the process of introducing the Eurocodes and adapting national standards is pro-
gressing in Germany and France.
Eurocode 7 is an umbrella code as analytical geotechnical models are contained in informative an-
nexes instead of the normative core text. Moreover, Eurocode gives a number of options which have to
be decided by the national standards bodies, such as three design approaches for the verification of
geotechnical ultimate limit states and the values of the partial factors. On the one hand, this is of
course a shortcoming for a code but, on the other hand, it constitutes an openness which makes the
adoption and the implementation of the code attractive not only for Europe but also world-wide as a
gradual evolution of national traditions of design procedures is possible. Whatever the precise legal
status of Eurocode 7 in the various countries, it will prove to be very important for the whole construc-
tion industry. It is meant to be a tool to help European geotechnical engineers speak the same technical
language and also a necessary aid to dialogue between geotechnical engineers and structural engineers.
It is the authors' belief that it will also be very useful to many geotechnical and structural engineers all
over the world, not only in Europe.
28
Session I
—«s®**—
Code Concept and Harmonization

The Study and Revision of the Probabilistic Seismic Hazard Map

and Dam Safety Code of Taiwan
Chin-Tung Cheng
Sinotech Eng. Consultants, INC., Taiwan
Shian-Jin Chiou
Sinotech Eng. Consultants, INC., Taiwan
Chyi-Tyi Lee
National Central University, Taiwan
Po-Shen Lin
National Central University, Taiwan
Yi-Ben Tsai
Pacific Gas and Electric Company, USA
SUMMARY
Before Chi-Chi Taiwan earthquake, the results of Probabilistic Seismic Hazard Analysis (PSHA)
produced by different agencies and different researchers totally underestimated earthquake hazards in
Central Taiwan. Therefore, there were some impacts on PSHA from the Chi-Chi Earthquake
experience, such as: (a) Saturation of local magnitude, (b) Significant hangingwall effect,(c) The need
to emphasize the fault source, (d) The need to use the close-distance to fault-plane. After the
earthquake, the need to emphasize the fault source was recognized, and the need to adopt a proper
attenuation relationship with considering the strong-motion data recorded during the Chi-Chi
earthquake was urged for PSHA usage. In this study, we conducted a review of readily available
information on attenuation of peak ground acceleration, tectonic setting, geology, active faults, and
seismicity of Taiwan for PSHA. However, we will consider fault activity and adopt proper attenuation
relationship in this study.
Taiwan is situated on the boundary between the Eurasian Plate (EP) and the Philippine Sea Plate
(PSP) where active oblique collisions and subduction are taking place. Therefore, Taiwan's orogenic
belt has high rates of seismicity and high fault-slip rates. We adopted available information of geology
and seismology to revise the probabilistic of Taiwan. New Seismic hazard maps were done for 10%
and 2% probability of exceedance in a 50 year period individually. In order to construct the maps we
defined four major seismic sources in Taiwan which were: (a) fault source, (b) areal/regional source,
(c) subduction plate interface and (d)subduction plate intraslab. We used the mainshocks from the
earthquake catalog of 1900 to 1999 to evaluate the earthquake recurrence rate for the regional zones
and subduction-intraslab sources by Truncated-Exponential model. We also used the fault-slip rate to
estimate the earthquake recurrence rate of faults and subduction-interface sources by Characteristic-
Earthquake model. For the first time in Taiwan our revised PSHA took into consideration the fact that
subduction plate sources induce higher ground-motion levels than crustal sources, and active faults
induce the hanging-wall effect in attenuation relationships. We also used the logic tree method to deal
with the uncertainties of each parameter in the PSHA.
The two highest hazard levels in Taiwan were shown in the areas of eastern longitudinal valley and
from the western foothills to the coastal plain. These two areas are separated by the central mountain
range which has a decidedly lower hazard level. After considering the fault activity in our revised
PSHA, we found that the PGA level of near-fields in Taiwan always exceeds 0.4g in 475-year return
period. However, in previous studies the proper hazard could not be obtained because fault sources
were not considered in the PSHA, especially in long return period. This situation was very obvious in
the central part of Taiwan (Chi-Chi earthquake disaster region) and in the HsinChu-MiaLi region. It

has now been realized that Northern Taiwan also has a much higher hazard level than previously
estimated. This is due to the discovery of active faults in the vicinity and also due to the current
realization that subduction plate sources induce higher ground-motion levels than crustal sources.
Taiwan Seismic Hazatd Map
DSHA and PSHA of Dam Safety Code

4 Before CM-CM « Fte«vahJfcBiA8erai-Ct* -DamMettan
VMatuatf-1 sfgma ~»™KS».2f75Yr -ISJASOOOYr
~&--BSHAlO00OYr
32
A Simple Reliability Assessment of Pile Design: Resolving some

Hong Kong Challenges
S.R. Lo
University of New South Wales, ADFA Campus, Canberra, Australia
K.S. Li & J. Lam

Victor Li and Associates, Hong Kong, China
SUMMARY
This paper demonstrates that simple reliability analysis can effectively provide an objective basis
for making engineering decisions. The equations for a reliability evaluation based on reliability
index as defined in Hasofer and Lind are presented. The background of the challenge in the
design of large diameter bored piles and high capacity driven H-piles are presented. The
necessary databases are established so that real and relevant statistical attributes are available for
reliability assessment. The results of such an investigation showed that:
• Current bored pile design practice for side resistance in saprolite stratum is neither excessively
conservative, nor overly-optimistic.
• The current practice for high capacity steel H-pile is rather conservative.
• Both the CAPWAP and the HKCA methods perform equally well and one cannot conclude
which method is superior.
Reliability assessment
The pertinent equations for reliability assessment are based on Li, Lee and Lo (1993). It is
recognised that engineering design are based on Rpu, ultimate capacity as predicted by a certain
design model; and Rpu in general differs from (XR, the mean resistance. To bridge the "gap"
between Rpu and \lR and to take into account the range of design models available in pile deign, we
define a bias factor, B, for a given design model as:
actual pile capacity (as measured by a load test)

*~
Therefore B is a random variable that is dependent on the prediction model. From the above
definition, flR = flBRplland COV(R) = COV(B). Defining fR as the partial factor on Rpu, the limit
state design equation can be rewritten as:
/ JR
where ys, y& are partial factors on the mean values of action and resistance respectively. The
partial factor on calculated pile capacity, Rpu, is given by:
fs =
XrA)
The required overall factor of safety, FOS, is simply given by: FOS = JsfR = 7SI^B7R
Bored pile design

The ultimate side resistance in kPa unit, fsu, is normally estimated based on the empirical
equation:

f
" 100 r ~fsi
where p a is atmospheric pressure in consistent unit (taken as 100 kPa), N is the standard
penetration test blow count, and fsi is the upper limit of side resistance. We take fsi = 150 kPa
based on a detailed survey of the database. The Building Department will also required site and
project specific pile load test to "prove" the design as experience suggest the side resistance can
be sensitive to detailed construction method. There has been debate on whether such an approach
is too conservative. Zhang et al (2002) examined 33 load test results for a range of foundation
consitions and construction methods, and conclude that that current practice is in fact optimistic.
The authors examined pile load test results from 10 sites (Lo and Li 1999, 2000, 2002). Only
test data from granitic saprolites are included, and piles constructed with the inappropriate
technique of using a permanent undersized liner are excluded. From these data pairs we can
calculate 19 data points for bias factor of bored pile design. This data based for granitic saprolite
yields the statistical attributes for the bias factor. The outcome of the reliability analysis showed
that current practice is neither too conservative as some might claim, nor too optimistic.
Driven H-Pile
Two methods for predicting the ultimate pile capacity from dynamic measurements have been
proposed and debated. The first method is the CAPWAP analysis, and the second method is a
simplified approach proposed by HKCA (Hong Kong Construction Association). The latter is
based on Broms and Lim (1988) but with minor adjustment to suit local conditions. We define
two Bias factors, BCAP and BHK:
The authors have established a database of 29 H-piles load tested to failure from two different
sources. From this database, the statistical attributes of for both the CAPWAP and HKCA method
were calculated. The KS goodness-pf-fit test (Ang and Tang 1975) was used to test two assumed
distribution, normal and lognormal. The results of the KS test indicate that either distribution is
acceptable. Therefore, normal distribution is assumed for sake of simplicity. We still assume Vs
= 0.15. However a target P of 3 was assumed because the resistance along the complete
embedded pile length was accounted for during dynamic measurements and pile load tests. It is
then a matter of substituting the statistical attributes of BCAP and B H K and the partial factors and
required overall FOS were calculated. The outcome of the reliability calculations
unambiguously indicated that both methods are equally reliable, and that the current Hong Kong
practice for friven H-pile may be too conservative.
Conclusions
Simple reliability analysis provides an objective basis for making engineering decisions on
debatable issues. This is illustrated by examining the design and construction of bored piles and
H-piles in Hong Kong. However, like any analysis, suitable input parameters are needed. In
reliability analysis, the inputs are statistical attributes of database compatible with the problem
under investigation. This is not a new concept, but analogous to requiring representative soil
parameters for any meaningful deterministic modelling. This requires a compatible marriage
between engineering know-how and reliability analysis. Reliability analysis is not a replacement
of our deterministic deign tool or engineering know-how. Rather it is an effective supplementary
tool.
34
Geotechnical Standards in Hong Kong

W.K. Pun
Geotechnical Engineering Office, Civil Engineering and Development Department, HKSAR
Government, Hong Kong
W.M. Cheung
L.S. Lui
SUMMARY
Setting Standards
Prior to the 1970s, slope design and construction were generally based on rules of thumb, such as 10:6
cutting with 1 to 2 m wide berms at about 7.5 m intervals giving an average slope angle of about 50°.
Following the occurrence of disastrous landslides in the 1970s, the Hong Kong Government established
the Geotechnical Control Office in 1977 (renamed as the Geotechnical Engineering Office, GEO, in
1992). An important function of the GEO is setting geotechnical standards. Since its establishment, the
GEO has produced many publications covering a wide range of geotechnical engineering topics. The
more comprehensive ones are called Manuals, Geoguides and Geospecs (Table 1). The main objective of
publishing these documents is to allow the profession to use a series of common, up-to-date and
comprehensive geotechnical standards which are appropriate to Hong Kong conditions. The documents
present recommended standard of good practice for various geotechnical activities.
Table 1: List of Manuals, Geoguides and Geospecs

Manuals
• Geotechnical
Gei Manual for Slopes, 2nd edition
• Highway Slope Manual
Geoguides:
• Geoguide 1: Guide to Retaining Wall Design, 2nd edition
• Geoguide 2: Guide to Site Investigation
• Geoguide 3: Guide to Rock and Soil Descriptions
• Geoguide 4: Guide to Cavern Engineering
• Geoguide 5: Guide to Slope Maintenance, 3 rd edition
• Geoguide 6: Guide to Reinforced Fill Structure and Slope Design
Geospecs:
• Geospec 1: Model Specification for Prestressed Ground Anchors
• Geospec 3: Model Specification for Soil Testing
Apart from these three categories of publications, the GEO published other related geotechnical
publications under the series of GEO Publications, GEO Reports, Technical Guidance Notes and
Geological Memoirs. GEO Publications document results of comprehensive literature reviews. GEO
Reports generally present results of applied researches and studies or reviews of aspects of geotechnical
engineering. GEO Technical Guidance Note is to satisfy the need to promulgate new or revised
technical guidance from time to time where the introduction of a new standard or updating of an existing
standard is not justified. When the technical guidance in a Technical Guidance Note has been widely
accepted by the geotechnical profession as standards of good practice, the practice will be incorporated

into the relevant GEO standards document at the time when the particular standards document is
updated.
Up to mid-2006, the GEO has released some 300 publications. A full list of GEO publications are
available from the CEDD website (www.cedd.gov.hk>.
Other parties including professional institutions also produced publications on geotechnical standard.
Status of the Publications

For public development projects, the prevailing government policy is that the details of all
permanent geotechnical works for man-made slopes and retaining walls shall be submitted to the
GEO for checking. The policy also stipulates that related activities, including investigations,
designs and works, shall be carried out in accordance with the prevailing standards. Some
documents, including Manuals, Geoguides, and some other publications, are adopted as local
geotechnical standards by the government through administrative means by the issue of Technical
Circulars.
The standards adopted for public development projects are generally also adopted for private
building and civil engineering developments in Hong Kong. This is achieved through the Buildings
Ordinance (Law of Hong Kong - Chapter 123) and its related Regulations and Practice Notes.
Process of Production
The GEO benchmarks against international standards and adapts the standards for local use as
appropriate in the course of producing geotechnical guidance documents. New specifications and
guidelines are prepared as needed to suit the specific nature of the local geological condition, works
practice, and legal and environmental requirements. Stakeholders are always consulted in the setting of
geotechnical standards. For Manuals, Geoguides and Geospecs, extensive consultation with consulting
engineers, contractors, academics, professional bodies and other government departments are carried
out. All comments are duly considered to ensure that the document would be considered a consensus
document by interested parties in Hong Kong.
Conclusion
Numerous geotechnical guidance documents in the form of Manuals, Geoguides, Geospecs and other
publications and reports are available in Hong Kong. These documents aim to promote good practice in
geotechnical engineering. Some of the guidelines are adopted as the local standards by the Government
through Technical Circulars for public development projects. The same standards are generally adopted
for private buildings and civil engineering projects through the Buildings Ordinance and its related
Regulations and Practice Notes. These standards have been benchmarked against international ones and
are adapted to suit local conditions.
36
Implementation of Eurocode 7-1 Geotechnical Design in

Germany
N. Vogt
Technical University of Munich, Centre for Geotechnical Engineering, Germany
B. Schuppener
Federal Waterways Engineering and Research Institute, Karlsruhe, Germany
SUMMARY
When Eurocode 7: Geotechnical Design, Part 1: General Rules (EC 7-1) is implemented in the European
Member States, each state will need to make two important decisions concerning the design of geotechni-
cal structures. Three design approaches are described in the code and each state can select the one that best
suits its national design traditions and stipulate its use in geotechnical design. Furthermore, the Member
States must establish the values of the partial factors in accordance with national safety requirements.
Both, the choice of design approach and the selection of the partial factors, must be seen as a single unit as
they are interdependent.
The selection of the design approach and the numerical values of the partial factors in Germany was
based on the principle that the safety level of the global safety concept that has been used successfully for
decades and should be maintained as far as possible. I.e. a geotechnical design in accordance with EC 7-1
should result in more or less the same dimensions as the former global safety concept. A comparative de-
sign, in which each of the three Design Approaches in EC 7-1 is applied to a strip footing, is used in the
paper to illustrate the option that has been selected for Germany. It shows that the Design Approach DA
2*, in which the partial factors are applied at the end of the calculation when the limit state equation is
checked, not only best fits the tried and tested safety level of the former global safety concept but also re-
sults in the most economic design.
Thanks to the Eurocodes, a single format will be used for the mathematical analysis of the ultimate limit
states throughout the construction sector in Europe in future. It has to be verified that the design value of
the effects of actions, Ed, never exceeds the design bearing capacity or the design resistances, Rd, i.e.:
E d <R,,
However, in geotechnical engineering, there are many cases in which it is not possible to make a clear-cut
distinction between the effects of actions and the bearing capacity. For instance, the action of the active
earth pressure also depends on the shearing resistance or the shear strength in the failure surface of the
active sliding wedge. In other cases, the resistance of the soil depends on the magnitude of the action. For
instance, the sliding resistance is governed by the magnitude of the effect of the action due to the vertical
component of the bearing pressure resultant. Last but not least the ground bearing resistance is dependent
on the eccentricity and the inclination of the acting forces.
Additionally, there are two entirely different ways of introducing the partial safety factors in geotech-
nical engineering:
- On the one hand, the design values, Ed and Rd, of the geotechnical effects of actions and resistances can
be determined by what is known as the method of factored shear parameters:
tan<p'd = tanip'k/Y<p c'd = c'k/Yc
Ed = f (<p'd, c' d ); Rd = f (<p'd, c'd)

- On the other hand, there is the method of factored actions and resistances. In this method, the charac-
teristic values of the actions, effects of actions and resistance of the soil, Ft, Ek and Rk respectively, are
determined using the characteristic values of the shear parameters, cp'k and c'k and then multiplied or
divided by partial factors to get design values:
Ek = f (q>'k, c'k); Rk = f (<p'k, c'k)
E<i = Ek-YE Rd = Rk/YR
The different ways of introducing the partial factors into the calculation are the principle reason why EC
7-1 offers three different methods of verifying geotechnical limit states. They differ in the way in which
they distribute the partial factors between geotechnical actions and resistances (see Table 1 in the paper).
In Design Approach DA 1, two combinations of partial factors have to be investigated. In Combination
1, uncertainties concerning the loads are checked. Partial factors greater than 1.0 are applied only to the
permanent and variable actions from the structure and the ground. Combination 2 of Design Approach
DA 1 checks the uncertainties of the soil applying partial factors greater than 1.0 mainly to the ground
strength parameters.
In Design Approach DA 2, the method of factored actions and resistances is applied and there are two
ways of performing the verifications. In the one referred to as "DA 2", the partial factors are applied to the
characteristic actions right at the start of the calculation and the entire calculation is subsequently per-
formed with design values. By contrast, in the design approach referred to as "DA 2*", the entire calcula-
tion is performed with characteristic values and the partial factors are not introduced until the end when
the ultimate limit state condition is checked. As characteristic internal forces and moments are obtained in
the calculation, the results can generally also be used as a basis for the verification of serviceability.
In Design Approach DA 3 the method of factored shear parameters is applied and additionally partial
factors are applied to the actions on the structure or coming from the structure. The entire calculation is
performed with design values of the actions and the design shear strength.
Germany took the choice of DA 2* for all verifications of shallow foundations, retaining structures, piles
and anchors and of DA 3 for verifications of slope stability. In order to maintain the safety level of the
former global safety concept (using a global safety factor, T|) in Design Approach DA2*, partial factors
were chosen fulfilling the equation yR • YG/Q « r|g,obal. Using the same partial factors for the permanent
and variable effects of actions in geotechnical engineering as in other fields of structural engineering (YG =
1.35, YQ = 1.50) with a weighted value YG/Q ~ 1A this comes out to YR ~ Tleiobai / YO/Q- For example, apply-
ing this to the partial factor for the ground bearing resistance with a former global safety factor, r)giobal, of
2.00, this yields to YR,V « 2.00/1.40 « 1.40.
In the paper, for a simple strip footing a comparative design is shown using the three design approaches of
EC 7-1 and also pointing out the difference between DA 2 and DA 2* which come out in great differences
in the footing width, B. In procedure DA 2*, in which the partial factors are applied at the end of the veri-
fication, the characteristic ground bearing resistance is determined on the basis of the characteristic values
of the effects of actions in the base of the foundation (Figure 6 in the paper). I.e. the characteristic inclina-
tion, 8k, and the characteristic eccentricity, ek, are used to determine the characteristic ground bearing re-
sistance. In procedure DA 2, in which the partial factors are applied to the actions at the beginning of the
calculation, the characteristic ground bearing resistance is determined on the basis of the design values of
the effects of actions in the base of the foundation. This means that the design value, 8<i, of the inclination
and the design value, ed, of the eccentricity are used to determine the characteristic ground bearing resis-
tance. This is incompatible with a logically structured safety philosophy. Furthermore, as the partial factor
for the (horizontal loads as) variable actions, YQ, is greater than the partial factor for the permanent (verti-
cal) actions, YG, the eccentricity and the tangent of the inclination of the resultant effects of actions in the
base of the foundation obtained by DA 2 will always exceed those obtained for DA 2*. The effect is even
greater if the vertical load acts favourably and the design value, Vd, has to be determined with the partial
factor, YG, mi, set at unity.
38
Introduction to International Joint Study of Reliability-Based

Design for Port and Harbor Structure
G. Yoon
Korea Ocean Research & Development Institute, Ansan, Korea, glyoon® kordi.re.kr
T. Nagao
National Institute for Land and Infrastructure Management, Japan
W.Lu
CHEC Guangzhou Port Construction Company, Guangzhou, China
K.Lee
Konyang University, Nonsan, Korea
H.Kim
Korea Ocean Research & Development Institute, Ansan, Korea
SUMMARY
International Organization for Standardization (ISO) has adopted the applicability of the reliability
based technical standards based on ISO 2394, which may affect to various structures including port
and harbor facilities. Eurocode discussed among European countries might most probably be adopted
as ISO standards. It surely has some negative implications for Northeast Asian countries' economies
such as Japan, Korea, and China which have completely different types of technical standards,
especially, since the ISO is compulsory for member economies of the WTO. A new round of revision
for technical standards of port & harbor facilities are being carried out in China, reliability based
design methods are the fundamental of technical standards. At the same time, the design codes based
on reliability are being prepared in Japan, and will be completed by end of the year 2006 or later. Also,
in Korea, reliability based design methods are currently hot issue among code developers, professors
and government officers to adopt as national standards. Therefore, joint studies are necessary for the
three countries in order that the technical standards of three countries are more scientific and
reasonable. Moreover a certain compatibility among the technical standards of three countries are
critical to promote the cooperation of technology and economy. To improve the effectiveness of joint
study, the representatives of three countries discussed the subjects of joint study on September 1999.
And the work of joint study was divided among three countries and the following arrangement was
approved by the representatives of three countries. Three countries carried out their roles for
international joint study of reliability for port & harbor structure from March 2003 to September 2005.
This paper summaries some results of the joint study performed during the first and second year of the
joint study among China, Japan, and Korea.
In Conclusion
1) Korea reviewed the uncertainties of soil properties for port and harbor structure which is a vital
part in determination of partial safety factor. However, quantification of their uncertainties to design
port structure is most difficult and challenging. The Korea reviewed certain soil uncertainties and
evaluation methods of uncertainties which are closely related with design variables for port and harbor
structures with following conclusions; a) Enough information is not available in the prototype cases for
the stochastic variables of relevant soil parameters to evaluate the uncertainty of the parameters. The
variability of soil properties such as the effective friction angle, undrained shear strength of the clay,
cohesion and angle of dilation are relevant parameters in determining the bearing capacity of the
39 Full paper in TAIPE1-2006 CD-ROM

foundation. In addition, the stress history is seldom known and samples extracted from site may not be
homogeneous and isotropic. Some methodologies of quantifying the uncertainty are introduced, b)
Korea introduced how to define the characteristic values in Eurocode 7. When interpreted using
Schneider's method, it provides a rational system for selecting parameter values for use in design and
one that results in values similar to the values that have traditionally been selected on the basis of
experience. Selection of the appropriate design cases and partial factors on loads, ground properties
and geometrical data is shown to be important in order to ensure the reliability of geotechnical designs
as Eurocode 7
2) Japan analyzed safety index using life cycle cost analysis, and estimated sensitivity index based
on FORM for port and harbor structure focusing on the external stability problems (sliding,
overturning) of caisson-type breakwaters. Japan concluded as:, a) The dominant parameters having a
large influence on the safety index are the wave force/the uplift pressure, the coefficient of friction,
and the surcharge. The tide level is not dominant with respect to the external stability of breakwaters.
Since these dominant parameters change their sensitivity factors depending on the safety indices, their
values should be carefully evaluated, b) The same design parameter takes different values of
sensitivity factors depending on performance functions and structural types. When setting the values
of partial safety factors in the Level 1 reliability-based design method, it is necessary to use proper
values of sensitivity factors in accordance with performance functions and structural types, c) The
sliding failure mode is dominant in the present design section. The failure modes are positively
correlated. The correlation becomes stronger as the sliding safety decreases. These facts indicate that
overturning failure or foundation failure unaccompanied by sliding rarely occurs, which agrees well
with the past disaster cases.
3) China has studied how to determine partial safety factors for gravity type quay wall, which usually requires
a large sample of statistical data for probabilistic code calibration, actions that are used to derive action effects
and resistance. The reliability indices for structures of port facilities implied by use of previous deterministic
design codes have been derived by regressive calculations on some structural members by the calibration
method. Due to incomplete statistical analysis of some uncertain parameters, the derived reliability indices are
not totally accurate, but can be regarded as an indicator of the real possible reliability indices. China also
determined some results of partial safety factors for different failure modes of gravity quay wall, however,
they are not general. More comprehensive code calibration study with as more samples data and
construction cases as possible should be conducted in order to derive appropriate partial safety factors for
gravity type quay wall.
40
Session II
—*s*s*—
Performance-Oriented
Geotechnical Analysis
Effect of Lateral Cyclic Load on Axial Capacity of Pile Group in

Loose Sand
S. Basak
Applied Mechanics Department, Bengal Engineering & Science University, Shibpur, India
SUMMARY
Offshore structures, namely, oil drilling platforms, jetties, Quay & Harbour structures, tension leg
platforms, etc. are supported on pile foundations. Besides usual load from super-structure (dead load,
live load, etc.), these piles are subjected to continuous lateral cyclic loading due to ocean waves. The
quasi-static nature of such loading induces progressive alteration in the bearing capacity and head
displacement of the foundation.
A comprehensive review of literature indicates that limited research works have been done in the
related areas. The relevant contributions are made by Poulos (1981), Purkayastha & Dey (1990),
Narasimha Rao & Prasad (1992), Narasimha Rao et al (1993), Gupta & Dutta (1996), Purkayastha et al
(1997) and Basak & Purkayastha (1999 & 2003). Some of the works are theoretical while the others
are experimental (laboratory and field).
As reported by Poulos (1981), basically three reasons against alteration of strength and stiffness of
pile foundation under lateral cyclic loading can be identified. These are :
• Development of excess pore water pressure generated during cyclic loading in progress.
• Gradual accumulation of irrecoverable plastic deformation of soil surrounding the pile surface.
• Rearrangement and realignment of soil particles surrounding the pile surface.
Understandably, the first two reasons are adequate for cohesive soil, whereas for cohesionless soil, the
third reason is primarily valid.
It has also been found that the alteration in pile capacity due to transient loading depends upon the
following parameters (generally known as cyclic load parameters) : number of cycles, frequency and
amplitude.
Since no standard apparatus for imparting lateral cyclic load on piles is available, a new set up is
designed and fabricated. By means of this apparatus, a symmetrical two way lateral cyclic loading can
be applied on model piles for a specified no. of cycles at a specified frequency and with a specified
amplitude with the help of motor and other mechanically and electrically controlled units. After
completion of the specified no. of cycles, the axial post-cyclic pile capacities are determined.
Experiments are carried out in loose sandy soil bed at a relative density of 35%. Stainless steel
shafts are used as piles using 2 x 2 group, the L/d ratio being 18 and the c/c distance between the piles
are 3d.
During experiment in progress, it is observed that a basin-like depression is formed around the pile
group in the vicinity of soil surface. This may be partially due to gradual shifting of soil particle away
from the pile surface as well as partially because of compaction of the sand mass surrounding the pile
group.
The alteration in the axial capacity of the piles due to cyclic loading has been expressed by a term
degradation factor which is defined as the ratio of the post-cyclic to pre-cyclic static axial capacities of
the pile group (Poulos, 1981). It is observed that some improvement in the axial capacity of the pile
group has occurred which is indicated by the values of the degradation factor to be greater than unity.
The load deflection curves are observed to be hyperbolic in nature. The ultimate capacities are
estimated by double-tangent method. It is observed that for all the tests conducted, the degradation
factor increases with no. of cycles and also there is a tendency of asymptotic stabilization. Also, the
degradation factor increases with frequency and also there is a tendency of asymptotic stabilization.

The degradation factor increases with cyclic load amplitude in a curvilinear manner. However, no
definite pattern can be concluded.
Sand beds are prepared at various relative densities. For densities of 35% and 48%, rainfall
technique is adopted, whereas for higher densities, the ramming technique is used, i.e., filling in the
tank in five equal layers of sand, each layer being compacted with several blows of C.B.R. rammer.
The no. of blows applied for each layer was 3, 5 and 10 to obtain the relative densities of 65%, 80%
and 88% respectively.
For each density, the ultimate axial and lateral static capacity of pile group was initially determined
so as to determine the degradation factor and the load amplitude to be applied for cycling. Thereafter,
the bed is gain prepared at the same density and the cyclic test is conducted. Each of the cyclic tests is
conducted at a frequency of 50 c.p.m. and amplitude of 50%. The no. of cycles applied were 100, 250
and 500.
It is observed that the degradation factor decreases with increase of relative density non-linearly but
no definite pattern can be concluded.
44
Evaluation of Design Methods for Large-Diameter Bored Piles

Florence L.F. Chu and Limin Zhang
The Hong Kong University of Science and Technology
1 SUMMARY
Large-diameter bored piles (LDBPs) are a popular foundation type in Hong Kong. Depending on the
geology, the pile capacity may consist of shaft resistances in soil and rock socket, and toe resistance on
rock or soil. Several design methods are available for the estimation of these resistances, and one or
several resistance components may be considered in design. The design methods include those outlined
by Buildings Department (2004) in PNAP141 and O'Neill and Reese (1999) (for comparison purpose),
and those based on empirical correlations with blow counts from the standard penetration test (SPT-N),
effective overburden stress ( a , ' ) or unconfined compressive strength (UCS) of rocks. The estimated
capacity values using these methods vary widely and the corresponding reliability levels of these methods
are not clearly known to date. In this paper, the design methods for large-diameter bored piles are
reviewed. A database containing information of load tests on 60 large-diameter bored piles in Hong Kong
is developed. The response of the test piles is studied and the resistance components of the piles are then
used as a basis to evaluate the performance of the individual design equations and combinations of these
design equations. The performance of a specific method is measured by a model bias factor and its
coefficient of variation.
2 CORRELATION STUDY
Information of 60 load tests on large-diameter bored piles in Hong Kong is collected for study. A
correlation study was performed to obtain empirical relations for calculating the shaft and end bearing
resistances of piles. For the shaft resistance in the saprolites, the measured average maximum skin friction
in a soil layer is correlated with the mean standard penetration test value (N) or the mean vertical
effective stress (^) at the centre of the layer. Correlations are established for different types of soils
including fill, alluvium, decomposed volcanics, decomposed granites. For piles socketed in rock, the
correlation study was preformed to establish the relations between the mobilized shaft resistance in the
socket or the end bearing resistance and the unconfined compressive strength of rock.
3 EVALUATION OF DESIGN METHODS FOR RESISTANCE COMPONENTS
The accuracy of a design equation can be measured by a model factor that is defined as the ratio of
measured resistance to calculated resistance. Using the results in the load test database, the model factors
for the O'Neill and Reese method and two empirical correlations for the calculation of shaft resistances in
various soils are studied. The degree of dispersion for these design methods is quite large. The COV
ranges from 0.48 to 1.99.
Similar to design methods for piles in soils, the accuracy of three design methods for piles in rocks is
also evaluated. The degree of dispersion of the three design methods is also quite large. The COV ranges
from 0.54 to 1.34 for both shaft resistance and end bearing. In particular the O'Neill and Reese (1999)
method for predicting the end bearing of the piles is associated with a very large model bias factor and
severely underestimates the end bearing.

The total pile capacity may be calculated using a combination of various design equations for estimating
each pile resistance component. Each combination is called a 'design method'. Six combinations for piles
in soils and three combinations for piles on rocks are considered. The calculated total capacity from each
'design method' is compared with the measured pile capacity from the pile load test. The ratio between
the measured ultimate pile capacity and the estimated total capacity using a particular design method, the
bias factor, measures the degree of the deviation of the method from the measured capacity. The mean
bias factor (XR) and the coefficient of variation (COVR) can be calculated for each design method.
Piles in saprolites
The estimated capacity based on the method involving presumed bearing values shows a large bias factor
of 11.28 with a fairly low COVR. This means that the scatter of the model bias factor is not large
compared with other calculation methods but the pile capacity is systematically underestimated. The
predicted capacity values based on the O'Neill and Reese (1999) methods and other correlations are
associated with relatively low XR but moderately high COVR values. However, the accuracy of these
methods is still comparable to the correlations derived for driven piles in sand or clay. The COVR
predicted by the Meyerhof (1976) correlation for driven piles in sand is 0.5 while the COVR of the
correlations established in this study ranges form 0.39 to 0.53.
Piles in rocks
The estimated capacity based on the O'Neill and Reese (1999) correlations shows a large bias factor of
14.97 with a fairly high COVR, which means the predictions from this method are highly scattered. The
predicted capacity values based on the correlations with UCS established in this paper have a mean model
bias factor closer to 1 and a lower COVR of 0.40 compared with that based on the PANA 141 method. The
accuracy of the two methods is still comparable to the correlations derived for the driven piles in sand.
46
TA1PE12006 International Symposium on New Generation Design Codes
Engineering Problems for Performance-Based Design of

Earth Structures
Y. Honjo
M. Honda
Nikken Sekkei Civil Engineering Ltd., Tokyo, Japan
K. Ogawa
Pacific Consultants Co., Ltd., Tokyo, Japan
Y. Wakatuki
Fukken Co., Ltd., Hiroshima, Japan
SUMMARY
Introduction
There is a trend in Japan for the structural design codes to move towards the performance based
design/specification (PBD/PBS) corresponding to the conclusion of WTO/TBT agreement.
Furthermore, there is also some expectation in the background that, since PBD/PBS scheme can
evaluate the performances of structures more directly compared to the conventional descriptive
design scheme, structures can be designed more rationally.
To design geotechnical structures, it is necessary to evaluate the interaction between structures,
e.g. foundations, and soil which may exhibit very complex behaviour. It is more so in earth
structures where the mechanical characteristic of soil need to be understood more precisely.
Moreover, grounds are always heterogeneous and amount of information is mostly not sufficient
to estimate all spatial variation of them. Consequently, some uncertainties are always included in
the evaluation of mechanical behaviour of geotechnical structures even if an external load is fixed.
The traditional descriptive type design specifications have supplied a large amount of
infrastructures successfully. Although some performances which are required to satisfy objectives
are evaluated indirectly in the traditional design scheme, the reliability of the design has been
confirmed by experience, especially those which caused damages.
To introduce PBD/PBS scheme, it is anticipated to cause some problems, e.g. difficulty of
design, evaluation of reliability etc., because it is expected to introduce more sophisticated new
design methods to evaluate performances of structures directly. However, further rationalization
of structure design can be expected by introducing such methodologies, and more flexible
responses to the required performances that corresponds to peculiar conditions at site. It is also an
important task to take into account the long-term maintenance cost and optimise the life cycle cost
of a structure.
This paper discusses technical issues in introducing the PBD/PBS scheme to design codes of
earth structures. The direct aim of this study was to add two chapters to "JGS4001-2004
Principles for Foundation Design Grounded on a Performance-based Design Concept" (whose
nickname is "Geo-code21 (JGS(2004))") on embankments and cut slopes. The basic concept of
PBD/PBS based design code is well presented in "code PLATFORM (JSCE(2003))". The
guidelines like IS02394 and "Basis of Structural Design for Buildings and Public Works
(JSCE(2002))" are also taken into consideration, which recommend LSD (Limit State Design) as a
design verification method in design code drafting.

Objective
According to the code PLATFORM, the objectives of a structure should be an explanation of
necessity of the structure in plane language.
For an example, objectives of an embankment, an representative earth structures, can be given
as follows.
The road or railway fills to give firm ground to run vehicles or trains safety. Water storage
dams that, through reservoirs, stable water supply is achieved in case of water shortage. The river
dike to prevent the flood to inland in case of flooding, the flood control dams or the regulating
reservoirs to prevent the breaking dikes by controlling the outflow in case of heavy rain, the
coastal levee to defend the coastal area from storm surge or Tsunami.
Performance requirement
According to the code PLATFORM, the performance requirements are statements expressed in
plain language describing the performance of the structure with respect to the given objectives.
Usually, serviceability to keep functions of the structure for given combinations of actions;
safety to protect lives of inside and outside of the structure from given combinations of actions;
restorability to enable the use of a structure by rehabilitation that is technically possible and
economically feasible; are required to structures.
Among the three requirements, the restorability is newly invented performance requirement in
Japan based on the experiences of Kobe earthquake of 1995. If the buildings are kept with in this
performance requirement, property value of a building is restored, which was much concern of
many apartment room owners.
Performance criterion
Performance criteria are more quantitative and engineers oriented specifications of given
performance requirements. The performance requirements are described in more general
terminologies, whereas the performance criteria are more qualitative so that they can be used for
design verification.
The difficult aspects of performance criteria for earth structures are that most of the traditional
design method only deals with the stability of earth structures, whereas some performance
requirements are given by displacement or deformation terms. In order to overcome this
difficulty, either new design method, for example FEM, need to be introduced or the traditional
design method need to be newly interpreted and approved by a deemed to satisfy solution.
It is also felt that critical displacement/deformation values for earth structures are very difficult
to determine mainly due to lack of background data; for example, what would be the maximum
differential settlement for drivers to feel uncomfortable when driving a car on a highway?
Verification
It is the stance of Geo-code 21 that the limit state design, which clearly define the limit states and
limit the occurrence of the calculate probability of the defined limit states within the allowable
amount, serves basic framework in PBD/PBS scheme.
The most of traditional design verification methods are just calculating the stability aspects of
the earth structures, and really estimate the displacements and deformations. It becomes
necessary to apply the numerical analysis technique, like FEM, to design. Due to lack in the
performance evaluation of these sophisticated methods, many aspects of these numerical based
techniques are discussed. The conclusion is yet not reached. However, it is necessary to
introduce these modern techniques to routine design of earth structures in the near future.
Final remarks
The final conclusions are not yet reached, so as the draft. However, it is felt that the design code
of the earth structures can be put in the framework of PBD/PBS scheme.
48
Performance-Oriented Risk Assessment and Retrofitting

Strategy for Electricity Towers on Slopes
C.H. Wang, Mei-Hsun Chang, Chia-Feng Chang
Taiwan Construction Research Institute, Taiwan
Deng-Chiou Wu, Kai-Ping Hsiung

Taiwan Power Company, Taiwan
SUMMARY
In the modern new era, the society demands for increasing of mega facilities and structures are
threatened by several challenges, such as the natural hazard, human crisis, and engineering
difficulty. In order to solve the challenges with sufficient safety and reasonable budge, a
performance-oriented problem-solving has been immerged among the geotechnical engineering.
This article presents the risk assessment and retrofitting strategy for the steel-grillage-anchored
foundation of electricity tower on slope. The electricity power network system is an essential
industry that provides electric power for the entire country. The Taiwan Power Company is
responsible for this duty to safely deliver electricity to all necessary corners among the island of
Taiwan. Besides, to maintain the electricity power facility in a safety level to against all possible
threats from hazards is an important issue among the emergency response and maintenance works.
In order to ensure the safety of these transmission towers, the main focus of this article aims to
evaluate the earthquake and landslide risk assessments of the electricity towers on slope for the
No.l north-south direction 345kV line. Furthermore, the adequate retrofitting strategy is
established to ensure the safety in the future service life.
This article first presents a performance-oriented procedure for the risk assessment for the
general geotechnical structures, especially for the existing structures. This procedure emphasized
on four important stages (shown in Figure 1), including the initial condition assessment, the risk
assessment, the impact and retrofitting assessment, and the long term performance maintaining
strategy.
Stage One StageTwo Stage Three Stage Four

Iritial CcrdtionAssessment RiskAssessmert Impact and Retrofitting Long-Termed Performance
• Irwertary • Hazard Scenarios Evaluation Assessment MaintErenceSlretagc
• Sitelrvesbgation • Hazard Analysis • ImpactConsequerce^css • Retrofitting Program
• Current Performance Status • Risk Analysis Study Optimization
Evaluation • PricritizatiGn • Performance Upgradng Plan
• Warmation System
—v • Performance Standard • Lcng-tErmed Maintenance
• Retrofitting Feasibility Plan
• Structure Mairtenance and
M anegementltformatjcn
System
Figure 1: Proposed Performance-oriented Risk Assessment Procedure
In the first stage, in order to completely understand the initial design purpose, the past
maintenance and hazard history, and the current performance of the structure, the basic
information collecting on design, construction records, and the history on the requirement and
mitigation records are fully desired. Besides, a proper investigation and monitoring approach can
be great helps to understand the current safety level of the structure itself. Furthermore, with the
help on the modern technology, a proper database system should also be established to store and
maintain the accumulated information. Based on the information from the first stage, the second
stage emphasizes on analyzing the possible threats that producing the various hazard scenarios. In
the beginning, all possible factors that may affect the safety of the structure should be analyzed.
For example, potential factors include earthquake, rainfall, geology, tomography, previous hazard

region, etc. By assembling factors to potential hazard scenarios, the possible failure modes and its
extend on the geotechnical design can then be estimated, which also includes the various risk level.
In the third stage, the possible impact or consequence produced by various levels of risk is
estimated, impact or consequence including the damages to human lives, economy loss, society
disaster, etc. Before the final retrofitting program can be finalized, two major criteria should be
further examined. The first criteria is the prioritization. This prioritization represents the
importance of components within the structure, also the structure itself among the authority
decision-making process. This prioritization will decided the retrofitting or re-construction
sequence of the component or the structure. Another important criteria is the performance
requirement of the component and the overall structure serviceability. Based on different failure
types, risk, and performance requirement, the retrofitting feasibility study can then conclude
several possible retrofitting techniques. Combining the overall results of the previous 3 stages,
the fourth stage not only generates the final optimized retrofitting program, but also provides the
plan for the future maintenance plain. The retrofitting program optimization process evaluates all
the possible retrofitting techniques, as well as to evaluate the possible performance upgrade.
Finally, the long-termed maintenance plan provides the regulation of continuing maintenance
schedule, hazard response procedure, and structure safety monitoring system.
The second part of this article is to evaluate the risk assessment and retrofitting plan for the
steel-grillage-anchored foundation of the electricity tower for the No.l north-south direction
345kV line, in total of 566 towers. The main concerns are to characterized the risk levels with
their corresponded retrofitting strategy. By following the risk assessment and retrofitting strategy
stated in the first part of this article, the possible risk produced by slope instability was first
characterized into various levels. Then, by combined with the current condition assessment, the
corresponded retrofitting strategy is discussed. Finally, in order to decide the most appropriate
retrofitting method and schedule, the retrofitting plan is proposed for the future works in 4 various
levels, as shown in Figure 2.
Current Condition
Assessment ,s ,
Low * *"*•»«' i^cveJHi'
Medium Low I eve 1 2 (1) * •»>J «* S
Medium High tt (sreU $>$ '' I evel 2 (16 ^-. ^ i.. '<
s
High ; S «YJ3 I *#?> W e i 5 <S?S , L evel 2(1) i U*e* 3 i >J
w J i Mtdium ,, ,
Low Medium' ow T High
hign
Risk Level
Note: Number shown in () represents the amount of tower
Figure 2: Retrofitting Plan for the steel-griliage-anchored foundations of electricity towers on slope.
In which,
Level 4: The foundations of towers classified in this level require immediate retrofitting
attention. The possible actions (with site investigation) include re-locating tower,
foundation retrofitting (e.g. strengthening or replacing the existing foundation), and
slope instability protection (e.g. retaining structure system).
Level 3: The foundations of towers classified in this level require retrofitting attention.
However, this attention can be reasonably postponed by establishing monitoring
system for emergence. The possible actions (with site investigation) include
foundation retrofitting (e.g. strengthening or replacing the existing foundation) and
slope instability protection (e.g. retaining structure system).
Level 2: The foundations of towers classified in this level require regular inspection with
moderate retrofitting program after proper site investigation.
Level 1: The foundations of towers classified in this level require only regular inspection.
50
Session III
—-mam-—
Geotechnical Reliability Analysis

Equivalence between Reliability and Factor of Safety

Jianye Ching
Ting-Rong Chen
SUMMARY
Uncertainties are abundant in civil engineering systems. Traditionally, geotechnical engineers have been
used to design approaches based on safety factors to account for uncertainties due to their simplicity and
convenience. However, safety-factor approaches are not rigorous for the purpose. Recently, reliability
analysis is popular for quantifying these uncertainties. Intuitively, reliability increases with factor of
safety, implying that there exists a monotonic increasing functional relationship between them. However,
it is not clear whether this relationship exists for any problems and if it exists, how to determine the
relationship. This research aims to answer these questions.
Let Z and 6 be uncertain variables and design variables of the target system. Also let D be the al-
lowable design region in the 6 space. Let F denotes the failure event: F= {/?[Z,G]>1), where R[Z,Q] is
called the limit-state function. This function does not necessarily define the complete collapse of the
system but the performance of the system, e.g. serviceability and ultimate capacity. Let us further define
the nominal limit-state function R„(6), a positive function of 9. An example of R„(Q) is to take R[Z,Q] but
fix Z at certain chosen nominal values, e.g. their mean values. The safety-factor approach is to enforce
the following constraint during the design process:
7j'Rn{0)<\ (1.)
where 7] = 1 is the designated safety factor. On the other hand, the reliability-based design approach is to
enforce the following constraint during the design process:
P(R[Z,0]>\\0)=$p(Z\0)-l(R(Z,0)>l)dZ<Pr (2)
where PF* is the desirable failure probability; p(ZB) is the probability density function (PDF) of Z given
6; /(.) is the indicator function, i.e. it is equal to 1 if the inside statement is true; otherwise, it is zero.
If the equivalence between the two approaches can be proved and established, it will be significant
in the following sense:
(a) One can then achieve a reliability-based design by using a safety-factor approach, which is much
simpler and more convenient than the former.
(b) Practical geotechnical engineers who are not familiar with reliability concept can easily achieve
reliability-based design by using the equivalence.
A theorem is derived to show that under certain condition, safety-factor designs and reliabil-
ity-based designs are equivalent:
Theorem: If the distribution of R[Z,Q]/R„(Q) is invariant over the entire allowable design region D, there
exists pairs of [TJ ,PF] such that the following constraints are equivalent:
T)'Rn{0)<! (3.)

and
P{R[Z,0]>P)<P; (4.)
Moreover, the functional relationship between the pair [TJ*,PF*] is as follows:
P(R[Z,0]-1]'Rn(0)>O) = P^ (5.)
where 9 is treated as random and uniformly distributed over D.

According to the above theorem, the reliability constraint (4.) can be transformed into the
safety-factor constraint (3.) if the distribution of R[Z,Q]/R„(Q) is invariant over D, i.e.: let us denote £ R
i {0:/,(/}[Z,6]>ll6)S Pf} be the design region that satisfies the reliability constraint that failure prob-
ability is less or equal to the target failure probability PF'. From the theorem, if the distribution of
R[Z,9]//?„(8) is indeed invariant over D, it is assured that the region £R is identical to the following re-
gion: E s = {Q:rfR„(6)S 1}, where rf can be found by solving (5.).
A simple procedure based on Monte Carlo simulation (MCS) can be used to estimate the rela-
tionship between rf and PF '• Draw N samples of [Z,6], denoted by ([Z^.O0']:i'= 1,... ,N}, where Z samples
are drawn from the PDF of Z, and 6 samples are drawn from the uniform PDF over D. For a chosen rf
value,
K^%l{G^\e^)>rf)^±l{G^>rf) (6.)
where GP=G(Z<''\ 9 W ). By changing the rf value, one can find the corresponding PFMCS* values by re-
petitively applying (6.), i.e. the entire functional relationship between TJ and PF is obtained. Note that
the same N MCS samples can be repetitively used to estimate the entire functional relationship.
When Pr' is small, Subset Simulation (SubSim) can be used to estimate the entire functional re-
lationship more efficiently than MCS:
I ' ( G < : W ) + I £'(<%,•, »n-ni (7.)

P> — + 2'(G£_>7'HIl N
* N
where CPJ-IJ denotes the i-th sample obtained in thej'-th SubSim stage. By changing the rf value, one
can find the corresponding Pr estimates by repetitively applying (7.). The performance of SubSim in
estimating the rf and PF relationship is also independent of system dimension, complexity, and the
uncertainty dimension.
Recall that the theorem is based on the premise that the distribution of R[Z,B]/Rn(Q) is invariant
over D. This can be achieved if the nominal limit-state function 7?„(0) is chosen carefully. However,
finding such a perfect choice is itself a difficult problem. Although finding the perfect choice is hard,
finding a nominal function Rn(&) such that the distribution of R[Zfi]/R„(fi) is roughly invariant is a
relatively easy task. As seen in the examples in the next section, /?„(6)=R(.E(Z),6) and R„(Q)=E[R(Z,6)IQ]
are usually acceptable choices. This is because although the distribution of R[Z,Q] may change dra-
matically with 9, the distribution of R[Z,G]/R„(&) usually does not due to the cancellation effect between
R[Zfi] and J?„(8).
54
Bearing Capacity Estimation of Open Ended Piles in Port

Construction in Japan
Y. Kikuchi
Port & Airport Research Institute, Kanagawa, Japan
In recent years, large diameter steel pipe piles embedded deeply are often used in port
construction in Japan. This situation is greatly different from former situation. However, the
estimation procedure of the bearing capacity of a pile has seldom changed. Comparison of the
bearing capacity estimation technique of a pile internationally and examination of the bearing
capacity mechanism of an open ended pile are conducted.
Conclusion is as follows;
In this paper, the change of the pile used for the port facilities in Japan for about 40 years was
introduced. Although the bearing capacity estimating equation of the pile hardly changed in the
meantime, it is needed to be changed.
Then, feature of the bearing capacity estimation equation of a pile used in the port facility
construction standard of Japan is compared with which used in Europe and the United States. On
these standards, a big difference is in the estimation method of a toe bearing capacity. But it is
difficult to evaluate the superiority or inferiority of bearing capacity equations only comparing
the measured results of the total bearing capacity of open ended piles with calculation results.
Then, the loading test of the pile of the large diameter which embedded to large depth was done.
The toe bearing capacity of the pile was evaluated by dividing into two, the toe resistance of a real
part and the resistance by inner surface friction. The resistance of a real part is being able to
estimated by the proposals already presented. For the inner surface friction, that working in the
section of ID to 2D from the bottom can be taken into consideration. The toe bearing capacity of
an open ended pile can be evaluated in this way.

Variance of Soil Parameters: Some Common Misconceptions

K.S. Li
Victor Li and Associates, Hong Kong, China
S.R. Lo
University of New South Wales, ADFA Campus, Canberra, Australia
SUMMARY
The use of probabilistic method in geotechnical engineering has been advocated by many
researchers, probably following the pioneering publications of Lumb (1966, 1967, 1968)
since the 1960s. Reliability Based Design (RBD) is a current hot-topic in geotechnics,
either as a design methodology in its own right or for underpinning Limit state Design
(LSD) codes. The variances of soil parameters are needed for any form of reliability
analysis, simplified or complicated. The determination of this simple statistical
parameter for real engineering problems, often considered as pedestrian in reliability
research, can be challenging. The author have the experience of coming across
unrealistic "reliability predictions" because of overlooking a number of essential issues (of
practical and theoretical origins) that affect the variance of soil parameters. Having
variances that are representative of the actual engineering problem is analogous to having
representative soil parameters for any meaningful deterministic modelling. This paper
consolidates these issues and use the title of "common misconceptions" in order to
stimulate discussions.
Importance of engineering knowledge or judgement

Sometimes, there is a lack of linkage between geotechnical practitioners (and indeed deterministic
researchers) and the reliability camp. This leads to the perception that reliability-based design is a
"blind" application of statistical analysis without appropriate input of engineering judgement.
The concern about the apparent lack of engineering judgement has been echoed many times by
both researchers in the deterministic camp and practising engineers. However, this is a
misconception. Engineering judgement is in fact enshrined in reliability analysis, either
theoretically in the form of Bayesian probability or practically in the form of judgemental values.
Point value versus spatial average

A soil property is a spatially random variable as the actual values (i.e. realizations) of the soil
properties vary from one point to another within an apparently homogenous soil. This spatial
variation can be modelled by the Random Field Model (Varmarcke, 1977). The relevant soil
parameter is often the average soil parameter along the failure surface (or of the failure domain);
and the variance relevant to the limit state considered should be that of the spatially-averaged
property. The variance determined from sample data is that of point variability. These two
variances are different but inter-related by the Variance Reduction Factor. As pointed out by Li
and co-workers in the early 90s, the variance of a spatially-averaged soil property is significantly
less than that of the point value. This difference can be taken into account to avoid having an
unrealistically large variance for any reliability analysis.

Uncertainties in the estimation of Population Mean
In practice, the true mean value of a soil property is never known and it can only be estimated by,
say the sample mean value, m, determined from limited test data. This leads to an additional
variability in the system. The variance of a spatially-averaged soil property estimated by the
sample mean value is (Li and Lumb, 1987; Li, 1989).
Var{x}= Var{m}+<T2T2(L)
where Var{.) denotes variance of a random variable and X is the spatially-averaged soil property
estimated by the sample mean. The first term on the right hand side of the above equation is
associated with the sampling uncertainty while the second term is associated with the spatial
variability of the soil property. The first term is affected by the number of soil measurements and
will approach a negligible value if the number of measurements are adequate large. In most
geotechnical designs, soil data is usually limited and the contribution of sampling is normally
significant.
Conversion uncertainties
The soil parameter relevant for a given design model may be calculated from a linear conversion
of the measured soil property. For example, a conversion factor of 0.7 may be applied to the
undrained cohesion of triaxial compression to "derive" the undrained cohesion relevant to a
particular design scenario. This is referred to as derived value in EC-7. Irrespective of the
reason behind the conversion, this will introduced additional variability. The methodology of
dealing with conversion uncertainties is addressed.
Importance of testing Errors

Most soil testing is error prone. If the error can be corrected by a bias factor, then it can be
handled by the equations for conversion uncertainties. However, significant random testing error
is a norm rather than an exception. In such a case, the variance as calculated from test results,
without adjustment, will be elevated.
58
Reliability Analysis of Excavation-Induced Building Damage

Matt J. Schuster, C. Hsein Juang, Evan C.L. Hsiao
Clemson University, Clemson, South Carolina, USA
Mary J.S. Roth

Lafayette College, Easton, Pennsylvania, USA
Gordon T.C. Kung

National Cheng Kung University, Tainan, Taiwan
SUMMARY
Concerns often arise during braced excavations due to the resulting excavation-induced ground
movements. When buildings are in close proximity to the excavation, building damage (referred to herein
as the violation of serviceability requirements) may occur as a result of the excavation-induced ground
movements. The application of reliability analysis can provide a valuable assessment of the building
damage potential in a braced excavation.
The purpose of this paper is to establish a framework for the simplified reliability analysis of
excavation-induced building damage potential through the development of an unbiased limit state. The
Bayesian mapping functions are established empirically based on the distributions of principal strains in a
database of cases compiled and examined by Son and Cording (2003), and subsequently used to calibrate
the model uncertainty (jici and COVci) of the assumed limit state. Based on the calibration of limit state
model, an unbiased limit state (jtci = 1.00) is established.
To perform the initial reliability analysis, the cases in the database collected by Son and Cording are
split into two categories, tolerable (undamaged) and intolerable (damaged), based on the classification of
damage by Burland where any damage levels less than or equal to "Slight damage" are considered to be
undamaged.
In addition, the limit state is established with the principal strain, a value derived from angular
distortion and lateral strain, as the critical criterion. Previously, Son and Cording correlated the damage
levels with principal strain and the limiting principal strain is empirically assigned based on classification
levels. Any principal strain associated with a damage level greater than "Slight damage" is considered
intolerable, and thus, the limiting principal strain is assumed to be 1.67xl0"3. The limit state with a
loading principal strain (epL) is subsequently equated as:
i S = /I(epi) = 1.67xlO-3-e„l = 0 (1)
In a deterministic analysis, LS < 0 would indicate that building damage will occur.
In order to evaluate the model uncertainty and establish an unbiased limit state equation, the
probability of damage for a given principal strain must be determined. This probability may be interpreted
based on the distributions of the principal strains of the groups of intolerable and tolerable cases using
Bayes' Theorem. Using the method developed by Juang et al. (2000), the following mapping function
relating EP to PD can be established:
P(s
K pl
\D)P(D)
PD = P(D | ejp = ' K ' (2)
' P{£p\D)P(D) + P(Ep\ND)P{ND)

where P(Z)|Sp) = conditional probability of damage for a given principal strain ep; P(ep\D) = probability of
ep given that damage did occur; P(sp\ND) = probability of ep given that damage did not occur; P(D) = prior
probability of damage; P(ND) = prior probability of no damage.
The primary purpose for developing the Po-ep mapping function is to provide a basis to estimate the
probability of damage for a given Ep, which in turn, provides a reference for back-figuring the uncertainty
of the limit state model (Equation 2). Subsequently, an unbiased limit state can be established based on
the fact that the limit state is unbiased when the mean model uncertainty (jicj) is 1.00. To account for
model uncertainty in the limit state equation, a model uncertainty (or bias) factor, c1; is introduced so that
the limit state model becomes:
LS = A(c,,e pL ) = c,(1.67x 1(T 3 )- EpL = 0 (3)
Using the method developed by et al. (2004), the model bias factor (pcl and COVcj) is calibrated until
the probability of damage calculated from the limit state (Equation 3) matches the probability of damage
calculated from the Bayesian mapping functions. Based on the calibration with each of the 41 mapping
functions (and thus, the results of 41 sets of [icl and COVci values), the mean values and standard
deviations of JXCI and COVcl are determined for the initial limit state: fici = 0.71 with apcl = 0.03 and COVci
- 0.33 with Bcovci ~ 0.
In addition, the effect of the limiting principal strain is examined by redefining the limiting principal
strain so that any damage greater than "Very Slight" is considered intolerable (limiting principal strain is
equal to 0.75x10"3). With the new limit state, the model uncertainty is recalibrated, and the mean values
and standard deviations of /xci and COVci respectively are determined: /xcl = 1.59 with apci = 0.07 and
COVcI = 0.33 with Ocovd ~ 0. It should be emphasized that 0.71 x (1.67 x 10"3) for the initial limit state
and 1.59x(0.75xl0" 3 )forthe new limit state equate to the same value of 1.19xl0~ 3 . This implies that
the limiting principal strain of the unbiased limit state should be equal to 1.19 x 10~3.
To more accurately formulate the serviceability limit state, the limiting principal strain is recalibrated
using a trial-and-error procedure so that the mean model uncertainty is unbiased ( ficl = 1.00). From this
analysis, the limiting principal strain is determined to b e l . l 9 x l 0 ~ 3 , and the mean values and standard
deviations of picl and COVc, respectively are determined: fici = 1.00 with apcl = 0.04 and COVct = 0.33 with
ocovd ~ 0. The uncertainty of the model bias can be combined and is equated to o c / = 0.37. Therefore,
the unbiased serviceability limit state can be expressed as:
I S = A ( c , , f p I ) = C l ( 1 . 1 9 x l 0 - 3 ) - e p , i =0 (4)
where the model uncertainty, Ci, is characterized as /xcl =1.00 and cct' =0.37.
In summary, using the database collected by Son and Cording, the framework for a
probabilistic analysis of excavation-induced building damage has been developed. The Po-ep
Bayesian mapping functions have been established based on the distributions of principal strains
for the group of intolerable cases and the group of tolerable cases. The mapping functions
provide a basis for calibrating the uncertainty of the limit state model. Based on an extensive
calibration process, the limiting principal strain has been determined to be i . i 9 x i o - 3 for an
unbiased limit state with the model uncertainty, in terms of bias factor, characterized as fici - 1.00
and c c i' = 0.37. Reliability analysis of the excavation-induced building damage potential can be
analyzed using the unbiased limit state as expressed in Equation 4.
60
Session IV
— * * ® « * —
Geohazards
Nov. 2—3, 2006, Taipei, Taiwan
The Assessment and Prediction of the Landslides and Debris

Flows in Ta-Chia River after Taiwan Chi-Chi Earthquake
C.T. Cheng, Y.L. Chang, S.J. Chiou, Y.S. Lin, C.Y. Ku, S.M. Shu, J.C. Chern
Sinotech Engineering Consultants, INC., Taiwan
S.H. Yu, S.D. Yang, C.F. Wang, C.H. Chiao, L.T. Hwang
Taipower Company, Taiwan
SUMMARY
Ta-Chia river is the one of abundant water resource in central Taiwan, and there are seven branch
hydro power plants of TPC (Taipower Company). After Chi-Chi earthquake took place, the
follow-up typhoons also caused damages along Ta-Chia river. The disastrous typhoons for Ta-
Chia river were Toraji typhoon in 2001, Mindulle typhoon and Aere typhoon in 2004, and Haitang
typhoon in 2005, which triggered new landslides and the debris flows were flushed into riverbed,
the events caused the river channel silted up and the flood level raised. Those geohazards
destroyed most infrastructures and villages nearby the river, especially the hydropower facility. In
order to assess the impact of sediment yields from landslides and debris flows and to investigate
the strategies of mitigating the geohazards, quantitative assessment was conducted by using aerial
photos and satellite images obtained at 6 stages of major earthquake and typhoon events. In order
not only to estimate the volume of the sediment yields from landslides and debris flows, but also
to establish the relationships between the volumes of sediment yields, the rainfalls intensity, and
the discharge.
Elevation changes in DTM

Figure presents the total elevation changes in the DTMs From Chi-Chi Earthquake to Hi-Tang
Typhoon( 1999-2005). The width of the almost whole riverbed became wider and wider through
those heavy rainfalls and the width of the shrunk sections were changed slightly. After Mindulle
typhoon, maximum elevation change of the riverbed was situated at the front of outlet of
ventilation tunnel, the change of elevation was more than 20 m. The most significant changed area
is situated at the vicinities of Chingshan switchyard and office.
0 1000 2000 3000 4000 5000 6000 7O00 8000 9000 10000 11O00 12000
Distance from Kukuan Reservoir (m)
The elevation changes from Chi-Chi earthquake to Hi-Tang typhoon(1999-2005).
New added landslide areas

The increased landslide areas of each event, and the summation of the increased landslide areas
triggered by difference events is totally 24 million m2. Furthermore, the total area of the increased

landslides became smaller and smaller, it seems that the geological condition would be more
stable in the future.
Volume and thickness of landslides

The volume of landslide triggered by Chi-Chi earthquake is 5 times the volume of the landslide
triggered by whole typhoon events. The average thickness of landslides of each event became
smaller and smaller, the thickness was between 1.5-3 m. When the rocks fall down and deposit as
taluses, the sliding material would be inflated. If the inflated rate was suggested to be 20%, Ta-
Chia main river and its branch river had generated the volume of landslide in 17 million m3 and 32
million m3 (lower bound estimation), respectively. Furthermore, if we suggest the inflated rate to
be 33% and consider the error of DTM, Ta-Chia main river and its branch rivers had generated the
volume of landslide in 24.3 million m3 and 46.7 million m3, respectively. In conclusion, the total
volume of landslides within the watershed between Techi dam and Kukuan dam was approximated
from 50 - 70 million m3.
Estimate new landslide areas in the future

In order to estimate the loss of sediment yields cased by heavy rainfall in each sub-watershed of
Ta-Chia river, Uchihugi's empirical model was adopted for predicting the new landslide rate of
each sub-watershed. The volume of Landslide will trigger by next heavy rainfall (200 year return
period rainfall) are total 6 million m3 in this study region. The results show that new landslide
areas of Pi-Ya-Sun river nearby Chingshan office is the largest quantity.
Long-term prediction of riverbed erosion and deposition

In order to predict the future trend of sediment yields transportation in Ta-Chia river, the
computing program HEC-6 was adopted for simulation the situation of that. Actuality, there have
no any real data for developing the discharge and the total load relationship for our study.
Therefore, The change of landslides after 1923 great Kanto earthquake (Mw7.9) from 1896 to
1980 in Japan caused a lot of landslides and the number of the landslides kept on the high peak
more than 15 years. It was took more than 40 years for the number of landslide becoming stable. It
was a very good reference for assuming the discharge and the total load relationship after Chi-Chi
earthquake, because the landslide area larger the total load would be more larger.
In the condition that great earthquake would no more happen in study area, the annual average
height of Sediment could decrease with exposure time increasing, and the result after almost 30
years would be flushing in the Ta-Chia main river. Ta-Chia main river Between Chingshan
switchyard and Chingshan Office of TPC Chingshan plant will be still deposit in 50 years and
400 m upstream of Kukuan dam will become flushing from deposit after 10 years.
After Chi-Chi earthquake, the geologic condition have became more vulnerable, the huge
sediment yields would transport to downstream after each heavy rainfall, and landslides and debris
flows will also be occurred in the next heavy rainfall. In order to predict the situation of
geohazards and to mitigate the geohazards before it happens, the statistics and analysis of the
information collected from the monitoring stations are very important to verify the expediency of
proposed method. The results show the highest level of riverbed around the Chin-Shan area would
raise more than 20 m in addition. Among the branch rivers of Ta-Chia main river, Ji-Ler river and
Pi-Ya-Sun river brought the most sediment yields from landslides in the sub-watershed. In
conclusion, there are at least 40% of the total sediment yields from landslides still remain in the
study area. Therefore, the sediment will transported out in the near future, and monitoring should
be conducted continually to mitigate the hazards.
64
Predicting Landslides Probabilities along Mountain Road

in Taiwan
Jianye Ching
Hung-Jiun Liao
SUMMARY
The total length of roads with elevation above 100m in Taiwan is more than 67,000 km. Some of
them were built with high engineering standards, but a large number of them were built with low engi-
neering standards. Therefore, landslides in different failure types are not unusual along mountain roads
when the slopes are experiencing long period of rainfalls or torrential rain accompanied with typhoons.
In this study, Route T-18 in central Taiwan is chosen to demonstrate the suitability of landslide predic-
tion using Gaussian Process model. Two main questions are of concern: (a) Given the historical land-
slide data along the demonstrative mountain roads in Taiwan, where are the locations along the roads
with high landslide potential in the future? (b) Given the historical landslide and rainfall data, what are
the landslide probabilities of the slopes along the roads in a future heavy rainfall? The former mainly
concerns with the locations of future landslides, while the latter concerns with the time of landslide
occurrence in future rainfalls.
Landslides along the mileage between 21.5km and 83.5km of Route T-18 are documented. In
total, 55 failed unprotected slopes along T-18 were extracted from the road maintenance records.
Among them, 12 slopes failed during Typhoon Herb, 18 during Toraji, 9 during Nari, and 16 during
Mindulle due to heavy rainfalls. To match the number of failed slopes, 54 not-failed unprotected
slopes were chosen. Note that the not-failed slopes in the database are roughly uniformly distributed
over the chosen Route T-18 section. The data format is as follows: D = {(*,-,/,•): i = 1 109}, where
Xj£ R15 contains the values of the 15 landslide features of the i-th slope in the database; r,- = 1 if that
slope failed, otherwise U = 0; p is the total number of slopes in the database. Fifteen landslide features
are categorized into natural features and man-made features. Among them, thirteen are natural fea-
tures, and two are man-made features. The natural features cover the aspects of topography (4 features:
slope direction, slope angle, slope height, and road curvature), geologic conditions (1 feature: outcrop
strata age), bedrock structure (2 features: slope and dip direction difference and slope and dip angle
difference), weathering & fracturing (2 features: block size and rock volume percentage), vegetation
cover (2 features: area percentage of vegetative cover and thickness of canopy cover), drainage condi-
tion (1 feature: catchment area size), and seismicity (1 feature: peak ground acceleration). The man-
made features (2 features: excavation height and change of slope grade due to toe cutting) quantify the
impacts induced by road construction.
A single index is used to capture the landslide potential: P(t=lbc,D), i.e. the probability that t = 1
given x and D. The Gaussian Process analysis is implemented to estimate the landslide poten-
tial. The discriminant function analysis is also implemented to compare with the Gaussian
Process analysis. Common practice of examining the performance of the adopted model/analysis is to
quantify the so-called training errors. For fair calculation of prediction errors, the so-called leave-one-
out (LOO) prediction errors of the adopted model are adopted here. The LOO prediction error is an
unbiased estimate of prediction error on unseen slopes of the trained model. The basic idea of LOO
prediction error is to mimic the prediction process by removing one data point out of the training da-
taset and use the removed data point for prediction testing.
Table 1 shows the traditional training error rates and the LOO prediction error rates induced by
the Gaussian Process analysis and the discriminant function analysis. Note that both training errors are
quite small (one of them is zero), but these are not realistic estimates for the actual prediction errors on

unseen slopes. The LOO prediction errors, which more realistically reflect the actual prediction error
rates, are always larger than the training error rates. It is also clear that the Gaussian Process analysis
results in smaller LOO prediction error rates than the discriminant function analysis, indicating that the
performance of the former is superior.
Table 1 Training errors and LOO prediction errors for the landslide location analysis
Training # of False LOO Predictions LOO
Methodology
errors (out of 109) Prediction Error Rate
Discriminant Function Analysis 7.3% 13 11.9%
Gaussian Process Analysis 0% 6 5.5%
The analysis shows that slope height, catchment area, height of toe cutting, block size, and
change of slope angle are the dominant features, among them are the two man-made features. This re-
sult implies that the slope stability along mountain roads is noticeably affected by road construction.
Also note that catchment area is among the dominant features. This result is consistent with the sense
that slope stability should be sensitive to the amount of seepage and surface water.
Besides predicting potential landslide locations, it is desirable to predict "when" (i.e. during
which typhoon) the dangerous slopes will fail. In Taiwan, landslides are mostly triggered by heavy
rainfalls during typhoons. As a consequence, we propose a second stage of analysis (landslide poten-
tial is the first stage) to predict the occurrence times of landslides for the dangerous slopes. In the oc-
currence time analysis, the size of catchment area and "effective" rainfall amount are the two features
studied. The latter is treated as the triggering feature of landslides. The goal is to develop a methodol-
ogy that determines the relationship between landslide probability and the two features for a dangerous
slope given past landslide and rainfall data. This relationship is called the rainfall fragility graph.
Figure 1 shows the results of the Gaussian Process analysis and is called the rainfall fragility
graph. In the figure, the crosses "+" indicate the failed dangerous slopes in the database (43 data
points), while the circles "o" are the not-failed dangerous slopes (33 data points). Note that in the
lower-left region, i.e. small rainfall and small catchment area, most slopes did not fail, while most
slopes failed in the upper-right region, i.e. large rainfall and large catchment area. This observation
agrees with our intuition. The contour values indicate the value of the predicted landslide probability
P(t=l\x,D) by the Gaussian Process analysis based on 43 failed dangerous slopes and 33 not-failed
dangerous slopes. This rainfall fragility graph can be used to predict landslide probability of a danger-
ous slope due to future typhoon.
£Bif«w8itflffotf Amnsm* #nro$ +'j'«' \"f^'J1*
Figure 1 Rainfall fragility graph for dangerous slopes along Route T-18
66
Optimal Design of Sand Compaction Pile Based on Liquefaction Hazard

Analysis
J.H. Hwang", C.W. Yang2' & C.C. Lu' ;
"National Central University, Chung-Li, Taiwan
2>
Moh And Associates, Inc., Taipei, Taiwan
SUMMARY
In this paper, a fully probabilistic method for liquefaction evaluation is first proposed, in which
both the uncertainties associated with the earthquake loading and the cyclic strength of soils are
considered. Then, a liquefaction hazard model is established for computing the total cost of a
building including the probable liquefaction loss during its service life. They are combined to
form a decision framework to determine the optimal design solution of ground improvement
method. A case of a damaged building in Wufeng, Taichung County that caused by soil
liquefaction in the Chi-Chi earthquake was chosen to demonstrate the feasibility of the proposed
methodology. The results show that the proposed methodology is workable and provides a logical
and reasonable way to design the ground improvement for preventing the liquefaction hazard.
Fully Probabilistic Method (FPM)

The framework of the proposed FPM contains two calculation loops are implemented to calculate the
probability of liquefaction for a soil layer at a site. The inner loop calculates the probability density
function (PDF) of liquefaction by using a newly developed reliability model, and the outer loop is to
sum the contributions from all potential seismic sources that are capable of producing significant
ground motion at the site by Monte Carlo Simulation.
For engineering purposes, an index is necessary to be proposed for characterizing the liquefaction
severity of the "whole" ground that might consist of several layers of liquefiable soils. In this study,
p
the liquefaction potential index L defined by Iwasaki et al. (1982) and the post-liquefaction
settlement ^ defined by Ishihara and Yoshimine (1992) are adopted to characterize the liquefaction
hazard of the whole ground.
Strategy of ground improvement

The ground improvement method used in the paper is sand compaction pile(SCP). By this way,
loose sandy soils can be compacted by squeezing sand columns and vibrating the natural soil
around the columns. The design of sand compaction pile is based on the empirical relationship of
the replacement ratio of sand piles, the SPT-N value of the subsoil before and after compaction
for different fines content. The relationship was established by collecting reliable Japanese case
history data of SCP in the past (JSSMFE, 1988).
Case Study
The proposed liquefaction hazard analysis model is similar to that suggested by Nishimura and
Shimizu (2005) with omitting the constant initial construction cost of the structure.
The case used for the study is a residential district, named the Prince's Castle, where many
buildings were damaged by soil liquefaction during the 1999 Chi-Chi earthquake. The Prince's
Castle residential district is in Wufeng, Taichung County and surrounded by the Kehniaokengchi,
which is a 5 meters wide creek. The representative geological profile of the site is shown in Table.
Based on the proposed fully probabilistic method for analyzing the liquefaction potential, the
annual exceedance probability (AEP) of liquefaction potential index PL and post-liquefaction

settlement S at this site can be calculated.
The result of the hazard analysis, shown in Figure, indicated that the probable overall cost due
to liquefaction hazard CL is a concave curve of the replacement ratio rs and has a minimum value
of 300,000 NT where the corresponding replacement ratio rs =0.115. In the proposed hazard
model, the value of rs that minimizes the CL function is defined as the optimal solution rs a, of
SCP design. Thus, the optimal solution of the SCP design is rs a, = 0.115 in this case.
Table: The geological profile of the Prince's Castle site

soil layer depth unit weight FC
SPT-Ar soil classification
No. (m) «/m3) (%)
1 0-3.5 1.90 4 46 SM
2 3.5-8.0 2.02 8 25 SM
3 8.0-12.5 2.10 23 40 SM
4 12.5-14.5 2.15 26 53 ML
5 14.5-17.0 1.94 30 32 SM
Ground water table»G.L. -2.5m
0.1 0.15
Replacement Ratio, r,
Figure. Optimal analysis of sand compaction pile
Conclusion
This paper proposes a fully probabilistic method for analyzing liquefaction potential of a ground and a
liquefaction hazard model for structures. They are combined to form a decision framework to determine
the optimal design solution of ground improvement method. From the case study, the feasibility of the
proposed methodology has been demonstrated.
68
for Geotechnical Engineering Practice.
Verifications and Physical Interpretations of the Discriminant

Model for Evaluating Liquefaction Potential on SPT-N Value
S.Y. Lai
Harbor and Marine Technology Center, Institute of Transportation, Taichung, Taiwan
M.J. Hsieh
Harbor and Marine Technology Center, Institute of Transportation, Taichung, Taiwan
W.J. Chang
Department of Civil Engineering, National Chi Nan University, Nantou, Taiwan
P.S. Lin
Department of Civil Engineering, National Chung-Hsing University, Taichung, Taiwan
SUMMARY
Development of statistical models from after-earthquake investigations has been an important

subject in geotechnical earthquake engineering. Discriminant method, which is a multivariate
statistical method, has been employed to analyze binary data related to multiple parameters. In
liquefaction study, discriminant models have been developed to evaluate the liquefaction potential
from in situ testing data, such as blow counts in standard penetration test (SPT-N value), cone tip
resistance (CPT-qc), and Shear wave velocity (Vs). Discriminant models based on 592 cases of
both liquefaction and non-liquefaction occurrences, including 288 cases from the Chi-Chi
Earthquake of 1999 in Taiwan, are developed to correlate the cyclic resistance ratios with
corrected SPT-N values and fine contents.
Discriminant curves of different probabilities of misclassification for cases with fines content
less than 10% are shown in Fig. 1. The correlations of cyclic resistance ratios with corrected blow
counts (Ni)so for different fine contents are developed based on the occurrence of C(P)=0, in
which the probability of misclassification is equal for both liquefaction and non-liquefaction
situations. Models for 10%= FC= 20%, 20%= FC= 30%, and 30%= FCS 40% can be referred in
Lai et al (2005). The evaluation models for cases with various fine contents are summarized and
illustrated in Fig. 2. The curves in Fig. 2 are further simplified by regression analyses and
expressed as:
CISR7J=exp(A-V(N1)60-fl) (1)
where 4 = 0.3865548+0.0072398 FC , fl = -(3.3597395 + 0.0186297FC-0.0001093FC 2 ) and the
square of the correlation coefficient (R2) is 0.99. Equation (1) provides a simple way to calculate
the cyclic resistance ratio using SPT data and can be easily implemented in computer code. And
the factor of safety against liquefaction (F s ) can be computed by:
FS=CRR1$ICSR15 (2)
where CSR7.5 represents the cyclic stress ratio induced by a earthquake.
The Fig. 2 shows that the liquefaction resistance CRR75 increases with an increase in fines
content for the same (Ni)6o, which agrees with the liquefaction phenomena observed in the field.
Moreover, the liquefaction resistance of different soils for (N!)60= 5 is almost the same.
To support the discriminant model, physical meanings of the generated discriminant curves are
discussed in view of the soil mechanics and the physical mechanism of liquefaction. Probability
examinations are performed on both parameters to determine statistically satisfied explanatory
variables. The cyclic disturbance of a soil induced by an earthquake is proportional to the
logarithm of CSR1S instead of CSRTS.The logarithmic transform of the induced cyclic stress ratio
is consistent with the definition of seismic intensity, which is linearly proportional to the

logarithmic transform of the maximum ground acceleration amplitude ( ln(a m „) ) of a site.
Analyses that correlate the properly transformed CRR7.5 and ( N ^ o are shown in Fig. 3. The
results reveal that the logarithm of liquefaction resistance ln(CRR15) of a soil is proportional to
JlNja instead of (#,)«, , implying that the logarithm of liquefaction resistance of a soil is
proportional to the relative density Dr. It is consistent with previous laboratory results.
The developed model is implemented to map the liquefaction potential of Taichung harbour for
Chi Chi earthquake event and the results are shown in Fig. 4. The map shows that the most severe
liquefied area occurred between the North Terminal and the North Pier, where the black area
located. The mediate liquefied areas are located in the West Terminal and the South Terminal. The
results are compatible with the damage condition at Taichung harbour during Chi-Chi earthquake.
REFERENCES
Lai, S.Y., Lin, P.S., Hsieh, M.J., and Jim, H.F. (2005), " Regression Model for Evaluating
Liquefaction Potential by Discriminant Analysis of the SPT N value." Canadian Geotechnical
Journal. Vol. 42, No. 3, p.856-875.
O'/.SFCSlO'/. V*0X
0.5
C(B=0 / P^0.1i
0 0.4
* ^*/ ,
'15
% 0.3 P-005
P=0JU
t °2
ju^yXy4C4.
0.1 ~W%^ 9o°
• Liquefaction
0^s^ ONonliquefactbn
Conectai Bbw Count (N >* Square root of N values (^JV) )
Fig.l Discriminant curves with 0%£ FCs 10% Fig.3 Physical meanings of the developed models
0.6
pc=m
0 5 10 15 20 25 30 35 40 45 50
Conected Blow Count, (Ni)u
Fig. 4 The result of liquefaction potential
Fig. 2 Comparison of discriminant curves with fines evaluation by Discriminant's model at
content of 5%, 15%, 25% and 35% for C(P)=0 Taichung Harbor area
70
Seismic Performance-Based Design for Canal Embankment

Y. OTAKE, T. HARA & T. HORIKAWA
CTl Engineering Co., Ltd., Japan
Y. ITO, T. KATO, M. HOSOYAMADA & Y. KASAI

Japan Water Agency, Japan
SUMMARY
Development of reasonable seismic verification and design of the countermeasure for existing
structures, such as buildings, bridges, dams and etc., is an important issue in Japan because the
occurrence of big earthquakes in the near future has been predicted. A type of canal, which is built on
embankment, Figure 1 shows typical transverse section of the canal, is also one of the structures
required the reasonable seismic treatment from the viewpoints as follows;
there are many canals of such type in Japan
considerable effects of the canal damage during the earthquake on human life and social
economy are concerned
expensive seismic countermeasures will be needed for almost all the canal if they are verified
with respect to the present seismic design specification, satisfying safety factor, 1.0, of land
slide analysis, thus tremendous amounts of money and time will be required for the project
Therefore, a study on the proposal of quantitative seismic performance based design for the canal as
a reasonable treatment of the issue has been conducted.
In this paper, contents of the proposed seismic performance based design code, seismic
performance, limit states, quantitative criteria for verification, and prediction methods of seismic
response of the canal embankment, are introduced.
Figure 1 Typical transverse section of canals built on embankment
Study on the proposal of quantitative seismic performance based design for the canal
In order to verify the canal embankment quantitatively to satisfy seismic performance of the canal,
both quantitative criteria for verification and prediction method of seismic response of the
embankment have to be stipulated along with seismic performance and its corresponding limit states.
Therefore, firstly, seismic performance, limit states and criteria for verification of the canal
embankment are newly proposed from the viewpoints of soil strain for safety and reparability of the
embankment, and subsidence of top of the embankment for serviceability of the canal. And then, the
prediction method that can reproduce well actual seismic response of embankments was proposed
from the comparison with centrifuge tests. Furthermore, because simple prediction method is valuable
for practical design, applicability of several prediction methods, which are the method using static
elastic FEM, the method based on Newmark proposed method (N.M. Newmark, 1965) and more
simple methods adopting charts or equations as well as dynamic elasto-plastic FEM analysis, are
studied.
Procedure of seismic verification of canal embankments and design of the countermeasure

Figure 2 shows concrete procedure of seismic verification of canal embankments and design of the

countermeasures based on the performance criteria and the prediction methods proposed in this study.
It has been expected that large amount of canal embankments would be able to be verified and treated
reasonably with respect to big earthquakes, which has been predicted to occur in the near future, by the
procedure of verification and design method.
1 st verification of embankments / " ' "\
by using simple chart or equation
2nd verification of embankments
and design of countermeasure j
' "
3£
Prediction of behavior
by using simple prediction method (
3rd verification of embankments and

design of countermeasure by using
more detailed analysis such as UWLC
[Design of seismic]/
[ countermeasure y
Figure 2 Procedure of seismic verification and design of the countermeasure for canal embankment
72
Session V
Engineering Practice
and Challenges
Observational Design Approaches for Safe and Economical

Deep Basement Construction in the Urban Environment
I. Askew
Lambeth Associates, Hong Kong
J.A. Frame
Gammon Construction, Hong Kong
D. Sein
Lambeth Associates, Hong Kong
SUMMARY
This paper presents the author's experiences of applying Observational Design (OD) approaches to
substructure design and construction in the South East Asia. To illustrate the potential benefits that can
be achieved, a case history is described for a deep basement excavation in Singapore. The design
objectives, methodology and performance data are presented.
There is an ever-growing push to gain further economies in the construction industry. Construction
projects, particularly in congested urban areas, are becoming increasingly complex but with shorter
expected construction periods. At the same time regulators and the public are increasingly intolerant to
any form of risk. The OD approach can provide a means by which construction safety and efficiency
are improved for deep excavations in the urban environment.
Modern instrumentation has become more accurate and reliable and can provide real time
automated monitoring which directly benefits the processes in OD. Instrumentation data loggers can
remotely connect to Internet based instrumentation databases, which makes data accessible and
transparent to all stakeholders in a project.
The construction industry is also undergoing change in the way contracts are procured, which has
lead to greater interest in OD. Design and build contracts and greater emphasis on partnering amongst
stakeholders has made OD more viable compared to the adversarial nature of historical construction
contracts.
The excavation at a site in Singapore Central Business District involved a 20m deep basement
structure. The base design required 5 layers of struts and imposed constraints on the phasing of the
construction of the pile caps. Objectives of the OD were to delete the lowest level of struts and avoid
the phasing constraints.
The OD design approach involved a series of sensitivity analyses covering a range of design
parameters so that the predicted behaviour could be benchmarked against the measured performance
on site. The range of design parameters covered "most probable soil parameters" and "original design
soil parameters".
A monitoring control framework was developed to track the predicted against measured wall
deflections during basement excavation. A "traffic light" system was implemented which reflected the
status of the wall deflections at different excavation depths (green as below predicted response, amber
as alert and red as action). A similar traffic light system was used in the control framework for the strut
loads.
Decisions to proceed with deletion of struts or implement contingency measures were made at
predefined stages of the excavation based on the monitoring records. A decision flowchart was
included on the construction drawings to clearly identify the key decisions.
All of the stated objectives of the OD were achieved on the project resulting in safety, cost and
programme benefits. Site safety benefits included reduced handling of heavy steel struts and
providing more working space for manoeuvring excavation equipment. Valuable information on
the importance of temperature on strut forces was also obtained. There was a daily fluctuation in
strut force of about 25% recorded.

There are a number of critical success factors. A design-and-construct project procurement
approach, in which the designer and contractor are fully integrated, is best suited for OD. "Traditional"
engineer designed and tendered contracts will be difficult to evaluate and administer effectively since
there will be potentially a number of different outcomes depending upon the field monitoring results
and speed of decision making.
There must also be a clear decision framework at the outset that has been agreed by all of the
stakeholders. The regulatory framework must be sufficiently flexible to allow a design approach
leading to a number of potential work sequences and outcomes rather than a single design solution. The
decision framework must be clearly stated on the construction drawings.
Designers must have adequate experience and resources to be able to track measured field
behaviour closely, report regularly, recognise anomalies and act on them if necessary. In the
contractual relationship between the contractor and designer, there needs to be an acceptance on
the part of the contractor that design spend will continue for the duration of the works covered by
the OD in addition to costs for specific design deliverables.
There must be adequate instrumentation in terms of quantum and reliability. Critical observations
should be made in "real time" with all data fed into a GEOMON system or similar. It is imperative that
the database is transparent and readily accessible to all stakeholders to alleviate possible concerns on
manipulation of the monitoring records.
Stakeholders with a partnering mindset are important. An adversarial environment without openness
will not lend itself to the successful implementation of OD. There needs to be effective communication
between all of the parties. In particular, the construction team must communicate timely, detailed
information on the works progress to the designer to allow field observations to be evaluated.
The wider application of OD approaches will improve understanding and generally result in more
efficient design solutions for substructure construction provided all the necessary steps are taken.
Whilst Regulators are currently exploring possibilities of general use of OD, more needs to be done to
ensure robust practical frameworks are put in place to ensure maximum benefit can be obtained from
OD without any compromise in safety.
In addition to potential direct benefits in cost and time from OD, there may also be safety
benefits from reducing the quantity of support members that need to be handled and increasing
working space. Experience also suggests that because of the high level of interaction between the
design and site teams general communication improves and construction issues are more readily
resolved.
Whilst our experience, particularly in Singapore, shows that significant cost and time benefits can
be achieved as a result of applying the OD approach, it must be emphasised that OD requires the
implementation of strict review and risk management processes.
76
The Performance of Laterally Loaded Single Pile in

Reclaimed Land
C.S. Chen
SSP Geotechnics Sdn Bhd, Kuala Lumpur, Malaysia
SUMMARY
Piles are normally designed to resist axial compression and uplift forces for most of the building
foundations. However, for deep foundations of structures such as towers, bridge abutments,
jetties, mooring dolphins, high-rises etc, the piles are often subject to significant lateral loads in
addition to the axial loads. The induced lateral deflection and bending moment not only will affect
the design pile capacity but also the performance of the structure. The ultimate lateral resistance
of pile depends very much on the subsoil surrounding the pile. It can be evaluated using
approximate solution as proposed by Broms (1965). Extensive researches on the deformation of a
single pile subject to lateral load had also been carried out and some analytical techniques have
been developed too. In general, the analytical model can be grouped into two main categories: the
Elastic Continuum model and the Spring Idealisation soil model (Winkler soil model or p - y
curve method). Both models had been used extensively today for the prediction of pile
deformation when subject to lateral load.
A recent development on a tidal land required pile foundation not only to resist axial load but
also to resist lateral load. The site originally was submersed at most of the time and the ground
surface level was at about 0.75 mLSD (Land Survey Datum). Reclamation had been carried out by
using hydraulic sandfill. The design platform level for the proposed development is about 4
mLSD. Due o the thick soft clay, long term post construction settlement was expected. In order to
expedite the minimise the post construction consolidation settlement of the soft marine clay,
surcharge method with the installation of prefabricated vertical drains had been adopted and more
than 90% consolidation has been completed at the time of pile tests. Soil investigation carried out
at the test piles locations revealed that the subsoil composed of 7.5m thick of sand below the
design platform level. Underlying is the treated soft clay with thickness of about 18m. Liquid
Limit and Plasticity Index of the soft clay vary from 80% to 120% and 40% to 80% respectively.
A thin layer of loose silty sand was found below the treated soft clay layer follow by medium stiff
to hard soil layer.
In order to evaluate the pile performance when loaded laterally, in-situ full-scale lateral load
tests were planned and conducted on two single piles. Test pile LTP-1 is a 600 mm diameter pre-
stressed concrete spun pile with thickness of 100 mm. The concrete strength is about 78.5 MPa.
The pile was installed using hydraulic hammer from the design platform level of 4 mLSD. The
penetration depth is about 30 m, i.e pile toe level at -26 mLSD. Inclinometer of 70 mm outer
diameter was installed into the spun pile and grouted. Test pile LTP-2 is a 750mm diameter cast-
in-situ bored pile. The pile was constructed from the design platform level to 34m depth, i.e. pile
toe at about -30 mLSD. After the boring depth had reached to the design depth, a steel cage
consisted of 10 numbers of high yield strength steel reinforcement of 32mm diameter was
installed. A steel pipe of 150mm diameter was welded to the steel cage for the installation of
inclinometer in later stage. The concrete strength is 30 MPa and it was placed by tremie pipe
method. Inclinometer with outer diameter of 70mm was installed into the pre-installed 150mm
diameter steel pipe. The annular space between the steel pipe and the inclinometer was filled with
cement grout. For both test piles, about 2m excavation was carried out for the setting up of the
test. The applied load was at the level of about 2 mLSD.
Reaction piles with similar diameter to the test piles were adopted as the reaction system. The
test load apply to the test pile was using a hydraulic jack together with a load cell. The pile head
deformation was measured by dial gauges or transducers. In addition to the dial gauges or
transducers, inclinometer was installed in the test pile to measure the pile deflection at different
depths.

Test Pile LTP-1 was loaded to 225 kN and the measured pile deflection at the ground surface
(at level of 2 mLSD) was about 39mm. LTP-2 was loaded to 400 kN and the pile head deflection
was about 61mm. Preliminary analysis had been carried out prior to the lateral pile load tests for
the prediction of the pile head deformation. The assessment methods were based on the guidance
from NAVFAC DM-7.2 and using commercial available computer software. Differences between
the estimated and measured results are expected. Nevertheless, for preliminary assessment
purpose, guidance from some of the Standards or using computer software is still very useful.
Lateral pile load test may be carried out later to verify the assumptions and to refine the analysis.
In addition to the pile head deflection, the measured results on the deformation of pile at
various depths indicate that the pile deformations are confined to the upper part of the piles only.
Based on the results of the two tests, the deformations are confined within 6 to 7 pile diameters
from the ground surface.
78
Settlement Calculation of Large-Area Thick Raft Foundation

under Irregular High-Rise Buildings
J.F. Gong, X.L. Huang & D.H. Di
China Academy of Building Research, Beijing, China
SUMMARY
With the development and utilization of the underground space, large-area frame structure with
thick raft foundation under a single or multiple high-rise buildings is widely used in China.
Settlement calculation method for large-area thick raft foundation under regular high-rise
buildings has been already proved feasible and reasonable in studies and in-situ settlement
monitoring in recent years. In fact, with the variety and complexity of high-rise buildings, large-
area frame structure with thick raft foundation under irregular multi-tall buildings is increasing,
for this foundation form, existing approach cannot accurately calculate the settlement of the whole
range of the raft. To solve the above-mentioned problem, based on the mechanical concept of
large-area thick raft foundation, a deformation analysis method is developed in this work, which
can be employed in the settlement calculation of large-area thick raft foundation with irregular
multi-tall buildings, and it has been successfully applied in the projects of China Petroleum
Mansion and Beijing HengFu Garden Mansion.
Design Principle for Large-Area Thick Raft Foundation

The large scale model test proves that the rigidity of frame structure with large-area thick raft
foundation as approximately equal to the rigidity of box foundation. Tests show that the
dispersion through the thick raft foundation for the tall building load is limited to a certain extent
under service ability limit state. When the ratio of thickness to span of the raft h/L > 1/6, if the soil
is uniform, rigidity of superstructure is good, load distribution is comparatively uniform, for high-
rise building at the circumference connecting one-span podium symmetry, the contact pressure of
foundation distributes linearly, raft design can only consider local flexure and its distribution rule
of the internal force is as same as that of the global analysis method (with the consideration of
superstructure). When the circumference of high-rise building exceeds one-span podium, the
contact pressure under high-rise building still distributes linearly, and the value of contact
pressure under high-rise building can be calculated on average with the area of further adding one
span, The raft design of high-rise building may only consider local flexure and the raft design of
podium needs to consider total flexure.
The basic mechanical characteristics of large-area thick raft foundation require the following
considerations for design:
In case many high-rise and low-rise buildings are built on the same large-area integral thick raft
foundation, by taking each building as the centre, and the raft thickness within the one and half
span from the side of the tall building remains unchanged. When the deflection of the raft meets
the requirements, and the variable-thickness raft design is needed, the raft thickness change points
shall be within the second span from the side of the tall building and the gradating way is adopted
for transition. If the difference in settlement between tower and the podium attached is not
acceptable and it is necessary for delay poured strip, it should be within the second span next to
the tower, and the raft of the podium linked with high-rise building should be as thick as that of
the tower. The settlement of large-area thick raft foundation with multi-tall buildings can be
calculated based on superimposition method.
Calculation method of settlement for irregular multi-tall buildings

Based on the mechanical characteristic of large-area thick raft foundation and deformation
controlling principle of foundation designing, under serviceability limit state, settlement

calculating method for large-area thick raft foundation under irregular high-rise buildings can be
drawn as follows:
Basic assumptions: The raft is limited, its deformation is irregular and continuous; Foundation
is unlimited, and its deformation is continuous and elastic; Load transferred through the raft is
limited.
Applicable conditions: The soil is uniform, rigidity of superstructure is good, and load
distribution is comparatively uniform. The ratio of thickness to span of the raft h/L>l/6. The ratio
of length to height of tower is not more than 1.5.
Rectangular cutting and superimposition method: According to structural characteristic, each
irregular high-rise building is divided into several rectangle parts on the projected plan of
superstructure. The settlement of divided portions of tower is calculated on elastic theory
respectively, and then superimposition method is used for the settlement of high-rise building.
Under serviceability limit state, the settlement of large-area thick raft foundation under multiple
irregular high-rise buildings can be calculated by superimposition method based on the settlement
of large-area thick raft foundation under each irregular high-rise building.
Practical application
Optimized designs using the aforementioned method for the large-area thick raft foundation of
China Petroleum Mansion, which is of four high-rise buildings of L-shape, and Beijing HengFu
Garden Mansion, which is of single high-rise building of Z-shape, have been carried out, and in-
situ settlement monitoring has been made during the construction stage. In comparison with
theoretical calculation, measured settlements agree well with numerical results. From engineering
applications, it is demonstrated that thick raft are capable to effectively spread the intensive high-
rise loading outwards to podium area, as a result to adjust the differential settlement in between.
In comparison with flexible raft/plate, rigid thick raft can significantly reduce the unacceptable
differential settlement occurred normally at the location of 1 to 2 span outside tower perimeter.
The concept of adopting thick raft to replace the conventional delay poured strip is proved to be
feasible and practical.
Conclusion
Under the interactions of superstructures, raft and foundation soil, the measured deformation of
entire raft is irregular and continuous, the rigidity of raft foundation is limited, and the raft
foundation under high-rise building is still limited rigidity. Based on the 'rectangular cutting and
superimposition' method, the settlement of irregular high-rise building can be calculated by elastic
theory. The settlement of large-area thick raft foundation under multiple irregular high-rise
buildings can be calculated based on elastic theory and superimposition method under
serviceability limit state. It is necessary to extend the thick raft of irregular high-rise building
outside one-span podium to decrease the additional stress under the foundation of high-rise
building, and to increase the foundation's stability of irregular high-rise building.
80
Geotechnical Risk Assessment and Performance-Based

Evaluation of a Deep Excavation in the Kaohsiung MRT System
Project
B. C. Benson Hsiung
National Kaohsiung University of Applied Sciences, Kaohsiung City, Taiwan
H. Y. Chuay
Mott MacDonald Ltd, Kaohsiung City, Taiwan
SUMMARY
Kaohsiung is the largest city in southern Taiwan as well as the economic and political centre. To
fulfil the need of fast development of the city, the construction has started since 2000 for the
Phase 1 of Kaohsiung mass rapid transit system. The system is expected to be in full operation
from 2007. There are two lines in Phase 1, Red Line in N-S direction and Orange Line in E-W
direction. A risk assessment was carried out in the design stage for the deep excavation of a
cofferdam without any lateral strutting system, 140 m in diameter and 27 m deep, in silty sand, in
order to ensure the construction safety and programme of the only interchange station in the Red
& Orange lines of Kaohsiung mass rapid transit system (KMRTS). The main risks associated
with this type of excavation are unbalanced ground water pressure and soil pressure outside the
cofferdam before the completion of the concrete structure, which are critical to the development
of the arching effect.
A risk assessment was carried out for the excavation at O5/R10 station and it identifies that
possible risks during the cofferdam excavation include: (1) excavation error, (2) over/under-
digging, (3) collapse of trench wall, (4) design load on ground level, (5) ground conditions, (6)
groundwater level, (7) direction and velocity of groundwater, (8) temperature, (9) earthquake, (10)
structure buckling, (11) field measurement, (12) water pumping and draw-down of groundwater,
(13) uplift failure and (14) leakage of diaphragm wall.
After the identification of hazards, a performance-based evaluation of the excavation is
considered. Based on the field measurements, a deflection path of O5/R10 was made and also be
compared with reference envelope determined from excavations having different ground
conditions and wall thicknesses in Taipei. It was found that the ratio of maximum lateral wall
movement (5hmax) to the excavation depth (D) varies from 0.07% to 0.27% at O5/R10 and this
ratio is much smaller for excavations in Taipei with D greater than 4m. This might be connected
with a thicker wall, different ground conditions and excavation shape at O5/R10.
The observational method was considered to be applied in the project. In the observation
method, the design is reviewed from time to time during the construction in response to the
monitored performance of the structure. It was suggested that the observation method should be
taken for a project where a precise prediction of the geotechnical behaviour is difficult.
The management of geotechnical works at O5/R10 is reviewed and it was recommended that
the advantage and flexibility associated with a turnkey contract should be properly considered in
geotechnical works. In addition, having one consultant to carry out independent checking
services might reduce the risk in geotechnical design successfully.
The efficiency of using circular cofferdam excavation is explored in this study. It was found
that the use of circular cofferdam excavation could dramatically reduce the construction cost,
even though it increased design complexity.

Further, a two-dimensional analysis was carried out for the excavation at O5/R10 using
computer software PLAXIS. Two sets of soil stiffness were used here: one is determined based
on an empirical estimation of SPT-N values; the other is defined by the shear wave velocity of
soils measured from the site. It indicated that the analysis using soil stiffness interpreted from
shear wave velocity measured in the ground could have closer results with field observations.
At the end, it was recommended that partial factors should be applied to actions, soil
parameters and resistances and a sensitivity study can be conducted by the use of numerical
simulation in this paper to explore influences from the partial factor.
82
An Essay on Typification of Verification Methods Used in the

Design Procedure of Geotechnical Structures
S. Kobayashi
Kyoto University, Kyoto, Japan
K. Aita
Sato Kogyo Co., Ltd., Tokyo, Japan
T. Fujiyama
Central Research Institute of Electric Power Industry, Abiko, Japan
M. Honda
Nikken Sekkei Civil Engineering Ltd., Tokyo, Japan
T. Kaneko
Fukken Co., Ltd., Hiroshima, Japan
A. Morikage
Chubu Chishitsu Co., Ltd., Kanazawa, Japan
A. Murakami
Okayama University, Okayama, Japan
M. Nabetani
Sato Kogyo Co., Ltd., Tokyo, Japan
M. Nozu
Fudo Construction Co., Ltd., Tokyo, Japan
SUMMARY
Technical committee on the performance-based design for geotechnical structures in Japan Society of
Civil Engineers was founded in 2004 and has actively exchanged the ideas among the administrators,
practitioners and academia. Working group 2 (WG2) of this committee has collected and investigated
many verification methods used for previous practices to establish a new methodology of verification
methods in the context of the performance-based design.
We should notice that the preciseness of calculated results depends on not only a verification
method itself but also other data qualities such as boundary conditions, initial conditions and modeled
geometrical configurations. We discuss on the potential abilities of each verification methods with the
unified framework. However, a verification method itself is only one factor to a solution. As a nature
of a geotechnical problem is that of an initial boundary value problem, we can point out three major
factors of a problem. One factor is expressed as a term " material" which includes a constitutive
model and its material parameters. Another factor is expressed as a term "ground profile" including
the constitution of soil strata and initial conditions of ground. External forces such as seismic motions,
wind forces and tidal forces are also included in ground profile. The other factor is expressed as a term
"solver" which describes how to formulate and solve a problem.
In a performance-based design procedure, choice of verification method is one of the key issues for
the reliability of an obtained result. From the view point of cognitive science, three typical behavior
can be observed as pointed out by Rasmussen (1986) shown in figure 1.; i.e., skill-based behavior
(SBB), rule-based behavior (RBB) and knowledge-based behavior (KBB). This model is
convenient for our research purpose to consider qualitative classification of verification method. By

combining three major factors and Rasmussen's model, we derive a matrix expression of qualitative
levels in the verification. A desirable choice for the quality / cost ratio is that these three factors are
suitably balanced in the similar level. Possible errors which will occur in the verification are also
discussed based on the cognitive engineering. The obtained results in this paper indicate an important
view point for the establishment of regulations and implementation of technical matters in a social
system.
Goal
Knowledeg-based
level
Rule-based level
Skill-based level
Figure 1. Operator's three behavior modes by Rasmussen (Yoshikawa, 2003)
Material Ground Solver Frequency Importance of

profile structure
Knowledge base rare very high
Rule base few high
Skill base lot low
Figure 2. Matrix expression of three levels of modes and three factors in geotechnical design
84
Session VI
—<&&*&*—
Geotechnical Uncertainties
and Variabilities
Reducing Performance Uncertainties with Monitoring Data
Jianye Ching
Yi-Hung Hsieh
SUMMARY
Uncertainties are abundant in geotechnical engineering. It is sometimes the case that the amount of
uncertainties associated with geotechnical systems is so significant that the resulting failure probability
is quite large. How to reduce the uncertainties in geotechnical systems can be an important research
topic. This research focuses on the following subject: how to reduce uncertainties and update reliability
by using monitoring data.
In this research, a new method is proposed to update reliability of general systems without the
dimensionality and linearity constraints. This new method is based on Bayesian analysis and Monte
Carlo simulation (MCS), and update reliability of an instrumented system by using its monitoring data.
In fact, as long as the probability distribution of the uncertainties is given, the functional relationship
between the updated failure probability and the monitoring value can be obtained prior to the monitoring
process. This means in real applications, it is not necessary to conduct the new algorithm in an online
manner. Instead, the relationship can be calculated a priori so that the reliability update can be achieved
right away once the monitoring data is obtained.
According to the Bayes' rule, we know
s f(<p\F)P(F)
P(F$p)=J^$ K
> (1)
JVP)
where F is the failure event; q> is the monitoring value; P{F]ip) is the failure probability given that the
actual monitoring value is f. In fact, P(F\<p) is the updated failure probability;fi<p\F)is the probability
density function (PDF) of the monitoring value conditioned on the failure event; P{F) is the failure
probability without the monitoring information; aadfifp) is the marginal PDF of the monitoring value.
For our purpose, the goal is to find P(F\<p). According to (1), iff(tp\F), P(F) and fly) are all available,
P(F]tp) can be readily obtained.
The failure probability P(F) can be estimated with MCS. Let Z denote the uncertain variables of
the target system. First, N sets of Z samples {Z*' : i = 1.. .N} are drawn from the prescribed PDF of Z.
According to the Law of Large Number, we have
1
* ;=1
where R denotes the limit-state function that defines failure event F, i.e. failure event is defined as
R[Z]>\.
Please note that in the process of MCS, samples distributed asf{(p\F) snAfij^F0) can be obtained
(F° denotes the non-failure event): Corresponding to the N sample sets {Z*' : /' = I...N) are the JV
samples of the monitoring value {#>*' : i = 1.. .JV }. Assuming that among the N samples, there are NF
failure samples, i.e. samples satisfying J?[Z"]>1, so the corresponding <p samples are distributed &sj{(p\F).

On the other hand, there are N-NF non-failure samples, so the corresponding <p samples are distributed as
ftjptf^). With the samples, the unknown WFsfljp) and./fy>IF) can be estimated by using the maximum
entropy theory.
The maximum entropy theory is employed to estimate f(<p\F) and /(pIF 0 ) based on the sample
moments of the samples obtained from f((p\F) and ftpiF0), i.e. to solve the following optimization
problem:
m a x - J l o g g ( ( p ) g(<p)d<p subject to \g(<p)d<p = \ jVg(<p)d(p = nt i = \..A (3)
Please note that in the optimization problem, the variable is the entire g(f) function, where g(cp) can be
either ftjp\F) or/[^IF c ); L and U are the upper and lower bounds, respectively, of the monitoring value <p
(if they exist, otherwise, they are -co and oo); the quantity -lUi}og(g(<p)). g(<p)d<p is the entropy of g(<p), and
Hi, k=\..A are the first four sample moments. Furthermore, j\q>) can be estimated with the following
equation:
f((p) = P(F)xf{(p\F) + P{Fc)xf{<p\Fc) (4)
where P(F) has been estimated by MCS, and P{FC)=\-P(F).

Please note that the approach proposed in this research for reliability updating is based on MCS and
the maximum entropy theory. The former is applicable to general linear or nonlinear systems whose
uncertainty dimension can be arbitrarily large, while the latter is also applicable to general PDF.
Let us take an example, where the monitoring value is the height of the water table in an infinite
slope, and failure is defined as the sliding of the slope. Using our method, it is possible to estimate the
functional relationship between the updated failure probability and the height (see the left-hand-side
figure in Figure 1). Suppose the height is monitored, and the daily data is shown in the upper-right figure
in Figure 1, we can use the functional relationship to compute the updated daily failure probability of the
slope, as shown in the lower-right plot in Figure 1. The lower-right plot has significant application:
based on the plot, decisions can be made to maintain the slope to prevent possible failure.
20
Monitoring value (m) Days
Figure 1. An infinite slope. The left-hand-side plot is the estimated P(F\<p) function. The upper-right plot
is the monitoring value, while the lower-right plot is the resulting failure probability time history.
88
TA1PEI2006 International Symposium on New Generation Design Codes
Evaluation of Spatial Variability of Weathered Rock for

Pile Design
Satyanarayana Murthy Dasaka
Limin Zhang
SUMMARY
It is well recognized that the geotechnical properties of soil and rock show considerable spatial
characteristics in all directions, though in different degrees. In this paper the spatial variability of
a weathered rock at a site in Hong Kong is quantified using the statistical models based on random
field theory and geostatistics. The borehole information available at the site is used in the
analyses. Depth of Grade-III weathered rock at a construction site is an important indicator in
design to evaluate the preliminary founding depth of the bored piles. The theory of geostatistics is
made use of to enhance the available subsoil information. Additional data at unsampled locations
are obtained using the spatial interpolation technique, called Kriging. The spatial variability
characteristics of Grade-III rock in the horizontal direction are evaluated in terms of scale of
fluctuation, using random field theory. The advantage of getting such information is that it helps
the engineer in charge of nearby sites or sites of similar geological origin for an effective and
efficient planning of the future site investigation. An effort has also been made in this study to
evaluate the scale effect of sampling size on the autocorrelation characteristics of Grade-III
profile. The results clearly demonstrate that the scale of fluctuation increases with the sampling
domain; the highest value is obtained for the data from the whole site and the least value when the
borehole information from only one building block is considered.
In general, the Grade-III weathered rock is deemed to support the loads of the structure
effectively by producing distortions that are within permissible limits as specified in the codal
provisions. Hence, for the preliminary designs, depth of Grade-III weathered rock is an important
reference for the determination of bored pile lengths.
Similar studies have been reported for the evaluation of spatial characteristics for soil and rock
properties for various applications using either geostatistics or random field modelling by many
researchers over past 3 decades (Vanmarcke 1977, Kulatilake 1989, De Groot and Baecher 1993,
De Groot 1996, Fenton 1999, Jaksa et al. 1999, Murakami et al. 2006, Liu and Chen 2006).
The whole site is divided into 7 blocks for site exploration purpose, which are referred to as
blocks 1 to 7. Figure 1 shows the locations of borehole data, which are irregularly spaced over the
areas of each block and whole site. The spatial correlation characteristics are evaluated for three
data sets with varying sampling size, as shown in Table 1. The first data set contains the depth of
Grade-III information obtained from the boreholes within block 1. The second data set contains
the depth of Grade-III information from the boreholes within blocks 1, 2, and 3 together.
Similarly, the Grade-III data obtained from all the boreholes within the whole site comprise the
third data set.
A clear trend is observed in the depth of Grade-III profile for this site, and it is found that the
slope of the Grade-III surface is as high as 1V:3H.

1050
E- 1000
a>
13
c
S. 950
o
o
o
V,
iS 900
'700 650 600 550 500 450 400 350

North coordinate (m)
Figure 1. Plan of borehole locations within the whole site
Table 1. Size of exploration area and number of boreholes for individual blocks and the whole site
Size of exploration area No. of boreholes
Whole site 160.2 m x 252.4 m 174
Blocks 1,2, and 3 together 122 m x 137.8 m 87
Block 1 alone 55 m x 59.3 m 49
CONCLUSIONS
Depth of Grade-Ill weathered rock is an important indicator used in Hong Kong design practice to
evaluate the preliminary founding depth of bored piles.
To ease the evaluation of semivariogram using irregularly spaced data, the separation distance
of all pairs of data are grouped into different classes of 1 m interval, and semivariance
corresponding to each class is calculated considering all the pairs of data grouped within that
class. This approximation could result in slightly inappropriate experimental semivariogram, and
might have affected the reliability of the predicted depths of Grade-Ill at unsampled locations.
N/4 pairs of data are chosen as minimum number for evaluating the experimental
semivariogram reliably, where 'N' is the number of measured data points. The results obtained
from the study reveal that the maximum lag of N/4 for constructing the experimental
semivariogram suggested in case of regularly spaced data may not be effective when the data is
randomly spaced.
The detrending process reduces the sill and range of semivariogram. The range obtained using
the observed data (without trend removal) from the whole site is many scales higher than the
sampling size. This demonstrates that the data set is non-stationary and has to be de-trended.
The horizontal scale of fluctuation of depth of Grade-Ill surface using the linearly detrended
data from the whole site is 137 m. This value is well comparable with horizontal scales of
fluctuation of some of the geotechnical parameters available in the literature.
The horizontal scale of fluctuation of depth of Grade-Ill depends on the sample size. Higher
scale of fluctuation is obtained from the analysis when the data from the whole site is used, as
shown in Table 2.
Table 2. Horizontal autocorrelation distances and scales of fluctuation of depth of Grade-Ill
Block 1 Blocks 1,2, and 3 Whole site
Size of exploration area 55 m x 59.3 m 122 m x 137.8 m 160.2 m x 252.4 m
Correlation distance (m) 13.5 45.4 77.3
Scale of fluctuation (m) 24.0 80.5 137
90
Analysis of Uncertainties in Analytical Pile Design Methods

in South Africa
M. Dithinde
University of Stellenbosch, Department of Civil Engineering
K.K. Phoon
National University of Singapore, Singapore
M. De Wet
J.V. Retief
ABSTRACT
The importance of uncertainty quantification is increasingly recognised in geotechnical

engineering as reliability based methods are assuming a prominent role in the calibration of
new generation design codes for geotechnical engineering practice. Accordingly model uncertainty in
the classical static pile design method was quantified in terms of model factors. The realisations of the
model factor were subjected to statistical analysis to facilitate the extrapolation of the results beyond
the current database.
Database
A database of 172 compression pile load tests was compiled in this study. The test loads were
generally carried out in accordance with SABS 1200F (1983) reaching a maximum load of 1.5 to 3
times the design load. In both cohesive and non-cohesive materials, the available geotechnical data
was limited to: soil profiles, SPT, Dynamic Probe Super Heavy (DPSH), and CPT (very few). For
evaluation of geotechnical properties, the DPSH and CPT measurements were converted to equivalent
SPT values. The geotechnical properties required for the calculation of ultimate capacity were then
obtained from correlations with SPT N-values.
Model factor for pile resistance
The collected data was processed and used to determine the predicted capacity and the
interpreted capacity. The predicted ultimate capacity was obtained from the basic static
formulae while the measured capacity was interpreted from load settlement curves. The model
factor for each case was calculated as the ratio of the interpreted capacity to predicted
capacity, thereby generating a model factor dataset of 172 individual values. A statistical analysis of
the model factor realisations was carried out to reduce the data to manageable form as well as to
facilitate the extrapolation of the results beyond the current database.
Graphical display
Histograms for the model factors in cohesionless materials, cohesive materials and all soil types were
constructed. Visual examination of the histograms showed similar characteristics for the three cases.
These characteristics were: most of the data points were clustered around the mean value; the
histograms were unimodal, i.e. they had one point of concentration or a single peak indicating that the

data was statistically homogeneous; and the data are not symmetrical about the peak frequency,
suggesting that the underlying distribution for the data is not normal.
Numerical summary
Although graphical presentation condenses a set of data for easy visual comprehension of its general
characteristics, numerical sample characteristics are required for calculations, statistical testing, and
inferring the population parameters. These are quantities used to describe the salient features of the
sample. The key statistics are the mean (jtM) and the standard deviation (<rM). These key statistics
together with the coefficient of variation (COV) are given in table 1. In addition to the measure of
centrality and dispersion, in this study the mean and standard deviation of the model factor are
considered as indicators of the accuracy and precision of the predication method. An accurate and
precise method gives HM = 1 and OM = 0 respectively, which means that for each pile case, the
predicted pile capacity equals the measured capacity (an ideal case). However, due to uncertainties the
results of an ideal case can not be attained in practice. Therefore in reality, the method is better when
fiM is close to 1 and aM is close to 0. When /JM > 1, the predicted capacity is less than the interpreted
capacity, which is conservative and safe. When HM < 1, the predicted capacity is greater than the
interpreted capacity, which is unconservative and unsafe.
Table 1: Summary statistics for the bias factor random variable
Case Summary statistics

Non-cohesive materials N mean Std. Dev. COV
(a) Driven piles 29 0.92 0.31 0.34
(b) Bored piles 31 1.30 0.34 0.26
(c) All pile types 60 1.11 0.38 0.34
Cohesive materials
(a) Driven piles 59 1.10 0.32 0.29
(b) Bored piles 53 1.20 0.30 0.25
(c) All pile types 112 1.14 0.32 0.28
Overall 172 1.13 0.34 0.30
Sensitivity of the bias factor to pile design parameters
The correlation between the bias factor and the various pile design parameters was accessed on the
basis of the p-value associated with a null hypothesis of zero rank correlation. The results of the
correlation analysis indicated that:
• Cohesive materials: The bias factor was not significantly correlated to pile diameter, pile length,
and undrained shear strength for both driven and bored piles. However, the model factor appears
to be correlated to the predicted capacity in driven piles but not in bored piles.
• Cohesionless materials: The bias factor was not significantly correlated to pile length. However, it
was significantly correlated to the pile base diameter for both driven and bored piles; angle of
friction in driven piles but not bored piles; the predicted capacity in both driven and bored piles.
Theoretical distribution for the bias factor
A lognormal distribution was assumed to be the most appropriate theoretical distribution for the model
factor on the basis of theoretical and historical considerations. This assumption was verified using
goodness of fit statistical tests. When a random variable X is lognormally distributed, its natural
logarithm, ln(X) is normally distributed. Accordingly the natural logarithms of bias factors were
determined and some normality tests supported by the available software were applied to the resulting
data set. The normality tests included Kolmogorov-Smirnov (K-S) test, Lilliefors test and Shapiro-
Wilk's W test. The P-values for all the three tests are greater than 0.05 and therefore there is no
92
evidence to reject the null hypothesis of normal distribution for the logarithms of bias factors of all the
cases considered. Since the distribution of the natural logarithms of the bias factors is normal, it
follows that the distribution of the bias factors is lognormal.
Conclusions
It has been shown that the model factor varies with individual observation and therefore it is best
modelled as a random variable. In this regard the lognormal distribution was found to be the most
appropriate theoretical model for the model factor. It was also established that the bias factor was
significantly correlated to pile design parameters in non-cohesive materials while such correlation in
cohesive materials was not significant.
93
Probabilistic Uncertainties in Estimating the Vertical Bearing

Resistance of Piles
M. Suzuki
Shimizu Corporation, Tokyo, Japan
M. Shirato, S. Nakataui & K. Matsui

Public Works Research Institute, Tsukuba, Japan
SUMMARY
We describe how to draft a formula for estimating the side resistance of a pile by reviewing a great
number of in situ pile loading tests conducted in the past. Design values of the unit side resistance by
an average SVT-N value (fj/N) provided as estimation formulae in the Specifications were determined
by selecting respective peak values of unit side resistances in each layer from the relation of the unit
side resistance, /, and displacement, S, during a test. Consequently, it has been pointed out that
estimated side resistances using fd /N tend to overestimate the resistances observed in the tests.
Furthermore, the magnitude of the safety margin excluding the safety factor in the current
Specifications is unclear.
We examine two SVT-N value normalized unit side resistances in each layer: (1) fy/N at the yield
point of a P0-S0 curve and (2)fp/N at the peak side resistances at pile tops based on the in situ pile test
results. To determine the yield point of the P0-S0 curve, a Weibull curve was used here. The resistances
observed in the test piles based on past tests have collected, in which SPP means driven steel piles and
CCP means cast-in-place bored piles. These tests all had measurement devices for dividing the
resistance into the side and base, although their loading sizes were different.
Figure 1 shows the relations of total resistance, Pa, at the pile top, base resistance, Pp, and side
resistance Pf - displacement S0 at the pile top for pile A-4, in which dotted lines mean yield points of
the P0-S0 curves and chain lines mean peak values of side resistances. This figure reveals the
following: (1) peak side resistances are mobilized at the displacement of around 2 to 5% of the pile
diameter, (2) base and total resistances increase in proportion to the increased load at the pile top with
displacement of more than 10% of the pile diameter, and (3) total resistances tend to yield at around
the points where the peak side resistances are fully mobilized.
Figure 2 show the relations of unit side resistance by normalized by SVT-N value (f/N) - pile top
displacement normalized by pile diameter (S0 /D) in each layer for pile A-4, in which yield points
(dotted lines) of P0-S0 curves and peak values of side resistances (chain lines) are also shown. In this
figure, the symbols As, Ac, Ds, and Dc mean alluvium sand, alluvium clay, diluvium sand and
diluvium clay, respectively. As shown in these figures, we can obtain the yield point (fy/D) but it is
hard to find the peak value (fp/D) of the side resistances compared to fy/D.
We examined the reliability of two SVT-N value normalized unit side resistances in each layer, (1)
fy/N at the yield point of a Po-S0 curve and (2)fp/N at the peak side resistance by using the reliability
index, (3. Statistical values offy/D and fp/D for each pile type and ground condition are shown in Table
1, in which 95% reliability lower levels of mean values (hereafter, characteristic values) from the mean
values are also provided. Figures 3 show fy /D and fp /D histograms of pile types SPP based on the
values provided in Table 1. In this figure, sample mean values of ;t^, D and x^^ with full lines and
characteristic values of fi ID and ll^,ID with dotted lines are also marked. With regard to the reduction
ratio of the characteristic to mean values, those of fp /D are greater than those of / , /D because the
number of the former is remarkably smaller than that of the latter.
In order to estimate the reliability of the resistances fy/D andfp/D, we first set up (fd/D)d = 0.667 by
using current design values fa/D = 2 and a safety factor (= 3 for the present situation) in the case of
SPP sand, for example. The safety margin can be estimated with the reliability index, P, considering
the characteristic values and variation of fy/D,fp/D and (fd/D)d-

The main findings were as follows:
1) In the case of SPP, the mean value of fp /N is larger than that of fy /N by 90% whereas the
characteristic value of fp/N is larger than that of fy/N by only 35%. In the case of CCP, on the
contrary, reductions were 20 to 30% and 25 to 35% respectively.
2) With regard to the coefficient of variation, V, Vjy/N was smaller than Vjp/N.
3) When comparing the reliability of fy/N to fp/N by using reliability index p, the former has a more
stable safety margin than the latter because the B values of fy/N are almost constant at 0.4 to 0.6 for
every pile type (SPP and CCP) and soil type (sand and clay), whereas the P values offp/N are small
in the case of SPP but large in the case of CCP.
4) In principle, it is desirable to select fp /N rather than fy /N as an index for estimating the side
resistance of a pile, unless it is generally hard to obtain information on fp/N in the pile loading tests.
Thus, overall it is practical to use fy/D instead of fp/N.
P P r max
1400 Pf-max
1200 Po
1000
p 800
P
(kN) 600 / - • —
'
400
// P„
200
1/ 2.0 10.6
0.0 5.0 10.0 15.0 15.0

S„/D(%)
Figure 1. Total, base and side resistances of pile A-4 Figure 2. Unit side resistances of pile A-4
Hfy/N SPP(sand) Wp/N SPP(sand)

=2.251 =3.080
*IWN —
= 2.)1
y
•
Xfp/N
=5. i72
| MtL
6 8 10 12 14
•• •5 10 15 20 25 30
fy/N fp/N
Figure 3./,/D and fp/D histograms of pile type SPP
Table 1. Statistics and reliability index of f/D

pile soil No. ave. st. dev. cov Characteristic
value (95%) P
sand 44 2.917 2.629 0.901 2.251 0.603
SPP
clay 41 10.084 9.861 0.978 7.524 0.425
f,/D sand 73 4.426 4.266 0.694 3.594 0.452
CCP
clay 53 20.370 23.535 1.155 14.956 0.494
sand 23 5.572 6.962 1.249 3.080 0.347
SPP
clay 14 19.187 19.115 0.996 10.140 0.356
fp/D
sand 11 3.788 1.470 0.520 2.711 0.530
CCP
clay 5 13.707 4.276 0.312 9.630 1.472
96
Soil Parameters Used in the New Design Code of Port Facilities

in Japan
Y. Watabe, M. Tanaka and Y. Kikuchi
Port and Airport Research Institute, Yokosuka, Japan
ABSTRACT
This paper proposes a new practical method, which will be adopted in the annex of the Japanese
design code of port facilities, to evaluate soil parameters for performance based design. The
characteristic value of the soil parameter is the expected value in principle, but it is not the mere
average in consideration of statistical errors in association with the testing method, the
inhomogeneity of the soil itself, and the limited number of the test data. The new method uses a
correction factor corresponding to the coefficient of variation.
PROPOSAL OF CHARACTERISTIC VALUE DETERMINATION METHOD
/ Modeling of the depth profile

It is required that a sufficient number of test results have been obtained in order to statistically
estimate the characteristic value. For the soil parameter a, the estimated soil parameter a* has to
be modeled as uniform depth profile (a* = C\); linear depth profile (a* = ci Z + C2); etc. Where z is
the depth; c\ and c% are constants. The estimated value a* is the value on the approximated depth
profile of the soil parameter. In modeling the profile in a certain depth range, the more data is the
more desirable; however, it is realistic to be said that 10 data is practically sufficient, because the
number of the test results is generally very limited.
The variation of some soil parameters is reasonable to examine on the logarithmic axis: e.g.
consolidation yield stress p c ; coefficient of consolidation cv; coefficient of volume compressibility
mv, and so on. In dealing these parameters, it is ideal to model the depth profiles on the
logarithmic axis (note here that the depth is on the arithmetic axis). However, in the case of
normally consolidated clay with slight skeletal structure, the depth profile of the consolidation
yield stress pc is expressed as a linear function with depth, characterized by a constant OCR and a
constant effective unit weight y'.
2 Concept of the characteristic value

The basic concept of the proposed method to determine the characteristic value from the derived
value is briefly described below. Because the derived value is influenced by sampling method,
laboratory testing method, sounding method, empirical/theoretical equation, etc., the design values
must reflect these influences. For example, it has been well known that the reliability of the
undrained shear strength obtained by the unconfined compression test is much lower than that by
the recompression triaxial test; however, it is very difficult to take account of this fact in design.
The new method proposed in this study adopts the concept, in which the characteristic value
must have been corrected corresponding to the reliability of the testing method. The coefficient of
variability (COV) is introduced to represent the data variation. To reflect the data reliability to the
characteristic value, the estimated value is to be corrected corresponding to the COV.
Consequently, we can establish the design standard with the common safety factor by using the
characteristic value determined by the proposed method, even the derived value of the soil
parameter has been obtained with different soil test.
3 Correction factor
In order to evaluate the variation of the derived values a, it is convenient to use the standard
deviation (SD) of a/a*, where a* is the estimated parameter obtained by modeling the depth

Table 1. Value of correction factor b\.
Coefficient of variation Correction factor 6,
COV Parameter for safety side Parameter for unsafe side
CO V< 0.1 1.00 1.00
0.1 COV< 0.15 0.95 1.05
0.15 CO V< 0.25 0.90 1.10
0.25 CO V< 0.4 0.85 1.15
0.4 COV < 0.6 0.75 1.25
0.6 COV Reexamination of the data / Reexecution of the soil test
profile of the derived values. Note here that the profile of a* must be estimated as either the
expected constant value or the regression function to minimize the errors. This is the major
premise based on Table 1, in which the characteristic value is generally defined as the expected
value of the derived values. The SD of a/a* is equivalent to the COV of a.
It has been known that the COV of the derived values obtained by soil tests carried out
carefully by a well skilled technician becomes less than 0.1 (Watabe et al., 2004). In other words,
the variation in this level is inevitable because of the ground heterogeneity and the laboratory test
errors. Ground heterogeneity, sample disturbance, inappropriate soil test, bad modeling of the
depth profile, and so on result a larger COV. In such cases, it is reasonable to conservatively
determine the characteristic value taking account of the uncertainties.
In order to calculate the characteristic value at from the estimated value a*, correction factor b\
is introduced as a function of the COV of a, then a k is defined as b\Xa*. When the soil parameter
a contributes to either resistance in the performance verification (e.g. the shear strength in the
stability analysis) or safety margin in the prediction (e.g. the consolidation yield stress pc; the
coefficient of consolidation cv in the consolidation calculation), the correction factor is defined as
bi-l-(COV/2). On the other hand, when it contributes to either sliding force in the performance
verification (e.g. the unit weight of the earth fill in the stability analysis) or safety dearth in the
prediction (the compression index Cc; coefficient of volume compressibility mv in the
consolidation calculation), the correction factor is defined as b\ = 1 +(COV/2). In these definitions,
the characteristic values correspond to either 30% or 70% fractyle value. Because a simplified
method is aimed in this proposal, the values listed in Table 1 are to be used instead of the
correction factors with detailed fractions. When the COV is larger than 0.6, it is judged that the
reliability of the soil parameter is too low for design. In this case, the test results are reexamined;
i.e. the depth profile is remodeled if necessary. In some cases, the ground investigation may be
reexecuted.
As mentioned above, some soil parameters, such as pc; cv; mv are examined on the logarithmic
axis. It is, thus, reasonable to evaluate the variation with the logarithmic normal distribution. For
the soil parameter a, the SD of (log a)/(log a*) is the COV of log a. Using this relationship, the
values listed in Table 1 can be used as the correction factor b\ on the logarithmic axis. In this
case, since ak = a* b'' is obtained from log ak = b\ log a* = log a* b\ the correction factor b\
corresponds to the exponent on the arithmetic scale.
If the number of the data is not sufficient for statistic treatment, another correction factor b2 is
introduced to correct the b\. Since about 10 or more data in the depth profile concerned can be
thought sufficient to reliably calculate the COV. In the case with less than 10 data, when the soil
parameter contributes to either resistance in the performance verification or safety margin in the
prediction, the correction factor is defined as bi = 1 - (0.5 / n). On the other hand, when it
contributes to either sliding force in the performance verification or safety dearth in the
prediction, the correction factor is defined as b2= 1 +(0.5/n). Here, b2 for the case with only one
data is set to 0.5 or 1.5, respectively, and the reliability is assumed to rapidly increase with the
data number. In this regard, however, the correction factor b\ cannot be obtained in the case of n =
1, because the COV cannot be calculated. This indicates that more than two data are required in
this proposed method. Note here that b\ = 1 and b2 = 1 can be used for soil parameters that
contribute equivalently to both action and counteraction.
98
Taiwan Special
Project Series
Performance Design of Taipei 101 Foundation

Dunstan D S Chen
Sino Geotechnology, Inc.
SUMMARY
The capacities and behavior of a cost-in-place reinforced concrete pile is largely determined by the
way it is installed. The design of the pile foundation for Taipei 101 was based on a series of full scale
pile trial installation and a series of comprehensive instrumented pile load tests with downward and
pull-out loads up to 4000 tons and 2200 tons, respectively.
The load transfer within the pile and the characteristic t-z curves for each soil strata were evaluated
and used in the pile length determination for each of the 380 piles, in accordance with the soil
stratification each pile is located and the maximum anticipated load to be acted on the pile during
service.
Base on results of pile trial installation and load tests, pile installation specifications, including
equipments to be used and installation procedures were finalized, and were strictly enforced under the
full time supervision of geotechnical engineers.
The structural design of the foundation take into consideration not only the pile-mat system, but
also the 5 levels of basements and the surrounding slurry wall retaining structures. The combined
effect on, and the behavior of the foundation system under loadings from the tower and the podium
were also investigated and evaluated.
The estimation of the foundation behavior under various load combinations were conducted using
the above mentioned structure model and the estimated distribution of sub-grade reactions under the
foundation mat. The estimated distribution of sub-grade reactions takes into consideration the
settlements of both the compression of each pile under load and the compression of the bearing
soil/rock strata under pile group loads. The high and low ground water conditions were also considered
in the foundation behavior estimates.

Design and Construction Issues of Deep Foundations for the

Taiwan High Speed Rail
S.W. Duann, J.R. Chen, T.C. Su, C.T. Chin
MAA Group Consulting Engineers, Taipei, Taiwan, R.O.C.
SUMMARY
The Taiwan High Speed Rail Project is considered one of the largest BOT (Build-Operate-Transfer)
projects in the world with an estimated construction cost of US$ 15 billions. The THSR runs through
populated west coast of the Taiwan Island, and its 345-km guideway links major cities from Taipei to
the southern city of Kaohsiung. The maximum design speed of the THSR is 350 km/hr with a capacity
of carrying 300,000 passengers per day, the Rail is expected to contribute significantly in alleviating
the overcrowded traffic of the west corridor. In this paper, an overview of the THSR Project is first
provided, and then important design requirements and considerations are discussed, which include
design requirement, seismic effect, near/crossing fault effect, ground subsidence and foundation
scouring etc. This is followed by discussions of load test programs implemented during the
installation of piles. Finally, several issues related to design and analysis of foundation systems were
discussed from a retrospective perspective. Extensive technical matters regarding bored piles have
been learned through this project.
The Taiwan Island is located on the western edge of the Pacific-rim earthquake belt and sits over
the juncture of the Eurasian and Philippine Sea Plates. Complex geological structures of the island
were formed by the active tectonic activity, which also created active faults islandwide. As a result,
seismic forces and earthquake related effects are major concerns for the design of the THSR. The
geological conditions along the route vary substantially, having mountainous terrain in the north to
thick sedimentary deposits in the south. As a result, the northern section is constructed mostly by cut-
and-fill, with bridges and tunnels, while the southern half is mostly elevated supported by large bored
piles. More than 30,000 piles were installed because of the presence of large earthquake forces, thick
alluvium soil deposit, and limited right of way. Due to significant variations in ground conditions,
different construction methods also were adopted for the installation of piles along the THSR route.
The design of the THSR structures followed the basic concept of performance-based design, in
which different performance requirements were considered for different earthquake levels. However,
the performance requirements were not implemented in an explicit form of performance matrix. Table
1 attempts to summarize the loosely specified design requirements in the context of the performance
matrix. Although the basic performance-based design was attempted in the THSR design of civil
works, some redundancy may still exist because performance requirements and verification methods
were not addressed in a systematic manner. With experiences of the THSR design, it seems important
considering the essence of performance-based design concept and making critical review on current
practice.
Given the large number of piles to be constructed, there is a great benefit to optimize the pile
design. The full-scale pile load tests were executed to determine the actual shearing resistance of local
soils. The test piles were instrumented with rebar stress transducers installed at different depths. The
results of the pile load tests in compression, tension and lateral direction were evaluated. Some
correlations between the unit skin friction and the SPT N values were established for various types of
soils.

Table 1 Performance Matrix Attempted to Summarize Foundation Design Requirements
Limit State Serviceability Ultimate (Repairable)

Long Term Condition Case 5 Cases 1,2,3
*specified for
Type II Earthquake
structural design
Type I Earthquake Case 4
In this series of pile load tests, the construction methods of bored pile installation were improved to
a great extent through the use of proper stabilizing agent (polymer), shorter construction time, multi-
stage toe grouting etc. Based on some of the available results, the ratio of ultimate tension capacities
to compression capacities (without end bearing) ranges from 62 to 105%, with a mean of 81±16 %.
Lateral pile load tests were conducted at locations with different soil conditions for determining
horizontal modulus of subsoil reaction and to verify the design parameters. Test results revealed that
the responses of test piles are quite similar to those predicted by the "p-y" curve method; hence this
method was adopted by some of the designers for predicting response of piles in design
With the benefit of hindsight, it can be concluded that improved foundation engineering may be
achieved with more elaborated studies of the load test data. The project also revealed the need for
sound and realistic design concepts. Performance-based design with due considerations of explicit
reliability levels should be the framework for implementation in future geotechnical seismic design.
The geotechnical community is way behind the structural community in this subject area. It calls for
serious efforts in conducting relevant investigations and developments such that a harmonized design
on both structural and geotechnical aspects may be achieved.
104
Experiences from Hsuehshan Tunnel Constructions

L. P. Shi
RSEA Engineering Co., Taiwan
Y. S. Hsieh
RSEA Engineering Co., Taiwan
SUMMARY
The main structures of Hsuehshan Tunnel are composed of one pilot tunnel and two main tunnels
(southbound and northbound). These 12.9 km long tunnels are the longest traffic tunnel in Taiwan.
There are three ventilation stations each including twin shafts for fresh and exhaust air
interchanging respectively, their depths vary from 250 m to 500 m. The experiences from the
above tunnelling works are discussed. These tunnels go through the Hsuehshan Ranges in the
northern Taiwan. There are 6 major faults and 98 shear zones in the field of construction site. The
maximum width of these weak zones is about 80 m. The tunnels were bored by tunnel boring
machines (TBM) from the south portals due to the environment constrains of the Feitsui
Reservoir.
The difficult problems during construction mainly came from TBM excavation stuck in the
Szeleng Sandstone. Difficulties occurred in the excavation of the pilot tunnel, main tunnel and
ventilation shafts are described. The rock mass properties of the Szeleng Sandstone can be
categorized as high strength, joint well developed to highly fractured with many faults and shear
zones, and with large quantities of groundwater behind them. The distributions of the geological
structure and groundwater are so complicated that they cannot be exactly predicted before
excavation, especially for the irregular high or low angle of the shear zone that left great
quantities of groundwater behind. These interactive effects create a lot of problems in probing
from excavation faces and TBM workings. The difficulties encountered include collapses of rock
masses, obstructions of borings and TBM excavations, wears and breakages of probing bits and
TBM disc cutters, cut and indent of casings and shields, and huge inflows of pressurized water
discharge..
The strategies for solving the problems proposed by the team are described as the follows:
a. Top heading method
By using this protection method, the southbound TBM worked well in the Szeleng Sandstone
and then make a greater excavation rate after passing that difficult ground condition.
b. Bypass tunnel for rescuing stuck TBM
The bypass tunnel whose excavation area almost same as the pilot tunnel was successfully used
to solve the TBM stuck problems.
c. Prolongation of tunnelling from northern portal
The excavation length from the north portal working face is more than 1/3 both in pilot and
main tunnel. This plan really increased the excavation rate.
d. Additional tunnelling faces from ventilation shaft No. 2
The total length excavated by these additional working faces was about 5,454 m. It actually
promoted the construction rate for this Project.
e. Raise boring applied for shaft No. 1
By using this sophisticated directional drilling system, it took about 5 months to complete the
reaming work of the fresh air shafts and 2 months for exhaust air shaft.
f. Application of long distant boring
It is worth to note that long-distance probing investigations are very good for the TBM
excavation, especially in the difficult geological condition.
g. Reformation of boring machine on the pilot TBM

After reformation of boring machine on the pilot TBM, it became really easier and much more
time saving than that of long-distance boring. By using the different boring rates, the insitu
geological condition can be quickly judged.
h. Study projects
Many fruitful and helpful results obtained from these study projects. It is important to note that
the proposed strategies make quite good results for the completion of the Project.
The experiences from these constructions included:
a. Contract execution problems
It is suggested that the suitable contract and reasonable insurance condition for the long
tunnelling must be carefully taken into consideration.
b. TBM usage
Using the open type TBM may more appropriate in Taiwan and the supporting segments must
be closed to the excavation face as near as possible.
c. Additional excavation faces
It must be kept in mind that increasing any possible working faces is the most important thing
in the long tunnel construction.
d. Applications of grouting techniques for weak zones
The best policy to deal with the faults or shear zones on tunnelling is applying sophisticated
grouting techniques to overcome seepage inflow and unstable collapsed problems.
106
AUTHOR INDEX
Aita, K. 83 Hsiung, B. B. C. 81
Askew, I. 75 Hsiung, K. P. 49
Basak, S. 43 Hu,Q. F. 15
Becker, D. E. 3 Huang, H. W. 15
Chang, C. F. 49 Huang, X. L. 79
Chang, M. H. 49 Hwang, J. H. 67
Chang, W.J. 69 Hwang, L. T. 63
Chang, Y. L. 63 Iizuka, A. 13
Chen,C. S. 77 Ito, Y. 71
Chen, D. D. S. 101 Juang, C. H. 59
Chen, J. R. 103 Kaneko, T. 83
Chen,T. R. 53 Kasai, Y 71
Cheng, C. T. 31,63 Kato, T. 71
Chern, J. C. 63 Kikuchi, Y. 55,97
Cheung, W. M. 35 Kim, H. 39
Chiao, C. H. 63 Kobayashi, S. 83
Chin, C. T. 103 Kokusho.T. 17
Ching,J. Y 53,65,87 Ku,C. Y 63
Chiou, S. J. 31,63 Kung,G. T. C. 59
Chu,F. L. F. 45 Lai, S. Y 69
Chuay,H. Y 81 Lam, J. 33
Chung, C.J. 21 Lee, C. T. 31
Dasaka, S. M. 89 Lee, K. 39
deWet, M. 91 Li,K. S. 33,57
Di,D. H. 79 Liao,H.J. 65
Dithinde, M. 91 Lin, H. D. 19
Duann, S. W. 103 Lin, M. L. 21
Einstein, H. H. 5 Lin, P. S. 31,69
Frame, J. A. 75 Lin, Y S. 63
Frank, R. 27 Lo, S. R. 33,57
Fujiyama, T. 83 Lu, C. C. 67
Gong, J. F. 79 Lu, W. 39
Hara,T. 71 Lui, L. S. 35
Ho, M. H. 21 Matsui, K. 95
Honda, M. 47,83 Morikage, A. 83
Honjo, Y. 7,13,47 Murakami, A. 83
Horikawa, T. 71 Nabetani, M. 83
Horikoshi, K. 13 Nagao, T. 39
Hosoyamada, M. 71 Nakatani, S. 95
Hsiao, E. C. L. 59 Nozu, M. 83
Hsieh,M. J. 69 Ogawa, K. 47
Hsieh, Y H. 87 Orr, T. L. L. 9
Hsieh,Y S. 105 Otake, Y 71
107
Phoon, K. K. 23,91 Tsai, Y. B. 31
Pun, W. K. 35 Vogt,N. 37
Retief, J. V. 91 Wakatuki, Y. 47
Roth, M.J. S. 59 Wang, C. F. 63
Schuppener, B. 27,37 Wang, C. H. 49
Schuster, M. J. 59 Watabe, Y 97
Sein, D. 75 Wu, D. C. 49
Shi, L. P. 105 Yang, C. W. 67
Shirato, M. 95 Yang, S. D. 63
Shu, S.M. 63 Yang, Y. Y. 15
Su,T. C. 103 Yoon, G. 39
Suzuki, M. 95 Yu, S. H. 63
Tanaka, M. 97 Zhang, L. M. 45,89
108
Listing of ISSMGE TC23 proceedings and other related publications
Ovesen, N. K., Mortensen, K. & Steenfelt, J. S. (Editors), Proceedings, International

Symposium on Limit State Design in Geotechnical Engineering, Copenhagen, Denmark,
26-28 May 1993. Danish Geotechnical Society 1993, 748 pp., ISBN 87-983058-9-1.
Simpson, B. & Driscoll, R., Eurocode 7 — a commentary, Construction Research

Communications Ltd. 1998,179 pp., ISBN 1-86081-226-0.
Orr T. L. L. & Farrell E. R., Geotechnical Design to Eurocode 7, Springer, London, 1999,
166 pp., ISBN 1852330384.
Ovesen, N. K. & Day, P. (Editors), Proceedings, International Workshop on Limit State

Design in Geotechnical Engineering (LSD 2000), Melbourne, Australia, 18 November
2000, CDROM.
Honjo, Y, Kusakabe, O., Matsui, K., Kouda, M. & Pokharel, G. (Editors), Proceedings,
International Workshop on Foundation Design Codes and Soil Investigation in View of
International Harmonization and Performance Based Design (IWS Kamakura 2002),
Tokyo, Japan, 10-12 April 2002. Balkema 2002, 459 pp., ISBN 90-5809-381-6.
Phoon K. K., Honjo, Y. & Gilbert, R. B. (Editors), Proceedings, International Workshop

on Limit State Design in Geotechnical Engineering Practice (LSD 2003), Massachusetts
Institute of Technology, Cambridge, 26 June 2003. World Scientific 2003, 54 pp., ISBN
981-238-784-6 (pbk).
Orr, T. L. L. (Editor), Proceedings, International Workshop on the Evaluation of

Eurocode 7, Trinity College Dublin, Ireland, 31 March - 1 April 2005. Department of
Civil, Structural and Environmental Engineering, Trinity College, Ireland, 2005, 320 pp.,
ISBN 1-873279-23-X.
Lin, M. L., Chin, C. T, Lin, H. D., Honjo, Y. & Phoon, K. K. (Editors), Proceedings
of the International Symposium on New Generation Design Codes — Taipei 2006 (With
CD-ROM), Taipei, Taiwan, 2-3 November 2006. World Scientific 2006, 122 pp., ISBN
981-270-382-9 (pbk).
109
(with CD-ROM)
Communication of design risk within a transparent and rational
framework is necessary in view of the increasing interest in code
harmonization, public involvement in defining acceptable risk levels,
and risk-sharing among client, consultant, insurer, and financier.
Activities in code harmonization in particular are noteworthy. For the
geotechnical engineering profession, there is added pressure for it to
undergo a significant revamp because structural and geotechnical
design are increasingly incompatible. The contributions in this volume
tackle the important issues relating to new generation geotechnical
design codes.in a bid to move geotechnical engineers forward together
with the significant changes occurring globally.
Cover art design by Kuo-Hsing Chang

Following the style of Chinese landscape painting, geometrical lines are used to
delineate the relationship between geotechnical engineering and the environment.
Circular patterns symbolized the ground and the water conserved in it. The harmony
between these elements represents the spirit of fullness — an ideal pursued by the
Chinese. The rectangles correspond to buildings, highlighting the goal towards more a
harmonized super- and sub-structure design as part of sustainable development.
631 Osc
ISBN 981 -270-382-9 (pbk)
i 9 8 i - 2 o ~o—6 www.worldscientitic.com

New Generation Design Codes For Geotechnical Engineering Practice - Taipei 2006

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

New Generation Design Codes For Geotechnical Engineering Practice - Taipei 2006

Transféré par

Droits d'auteur :

Formats disponibles

Proceedings of the International Symposium on

New Generation Design Codes for

Meei-Ling Lin, Chung

Yusuke Honjo and

New Generation Design Codes for

National Taiwan University of Science and Technology,

\fc World Scientific

British Library Cataloguing-in-Publication Data

Proceedings of the International Symposium on

ISBN 981 -270-382-9 (pbk)

Printed in Singapore by World Scientific Printers (S) Pte Ltd

Communication of design risk within a transparent and rational framework is necessary

For the geotechnical engineering profession, there is added pressure to undergo

Prof. Meei-Ling Lin, Symposium Chairperson

Prof. Yusuke Honjo

Dr. Jie-Ru Chen

Risk assessment in rock engineering 5

Some movements toward establishing comprehensive structural design codes

Development and implementation of Eurocode 7 9

New trend toward performance-based design in the construction industry 13

Energy approach to earthquake-induced slope failures for performance-based

Serviceability limit state reliability-based design 23

Eurocode 7 for geotechnical design — basic principles and implementation in

A simple reliability assessment of pile design: resolving some Hong Kong

Geotechnical standards in Hong Kong 35

Implementation of Eurocode 7-1 geotechnical design in Germany 37

Introduction to international joint study ofreliability-based design for port and

Session II: Performance Oriented Geotechnical Analysis 41

Evaluation of design methods for large-diameter bored piles 45

Engineering problems for performance-based design of earth structures 47

Performance-oriented risk assessment and retrofitting strategy for electricity

Session III: Geotechnical Reliability Analysis 51

Equivalence between reliability and factor of safety 53

Bearing capacity of open ended piles in port construction in Japan 55

Variance of soil parameters: some common misconceptions 57

Session IV: Geohazards 61

Predicting landslides probabilities along mountain road in Taiwan 65

Optimal design of sand compaction pile based on liquefaction hazard analysis 67

Verifications and physical interpretations of the discriminant model for

Seismic performance-based design for canal embankment 71

Session V: Engineering Practice and Challenges 73

Observational design approaches for safe and economical deep basement

The performance of laterally loaded single pile in reclaimed land 77

Settlement calculation of large-area thick raft foundation under irregular

Geotechnical risk assessment and performance-based evaluation of a deep

An essay on typification of verification methods used in the design procedure of

Reducing performance uncertainties with monitoring data 87

Evaluation of spatial variability of weathered rock for pile design 89

Analysis of uncertainties in analytical pile design methods in South Africa 91

Probabilistic uncertainties in estimating the vertical bearing resistance of piles 95

Y. Watabe, M. Tanaka, Y. Kikuchi

Taiwan Special Project Series 99

Performance design of Taipei 101 foundation 101

Experiences from Hsuehshan tunnel constructions 105

Author index 107

List of past ISSMGE TC23 proceedings 109

Limit States Design Based Codes for Geotechnical Aspects of

3 Full paper in TAIPEI-2006 CD-ROM

Risk Assessment in Rock Engineering

5 Full paper in TAIPEI-2006 CD-ROM

Some Movements Toward Establishing Comprehensive Structural

Gifu University, Gifu, Japan

7 Full paper in TAIPEI-2006 CD-ROM

Moreover, the functional relationship between the pair [TJ,PF] is as follows: