Experience Seamless Full Ebook Downloads for Every Genre at textbookfull.

com

Practice of Bayesian Probability Theory in

Geotechnical Engineering Wan-Huan Zhou

https://textbookfull.com/product/practice-of-bayesian-
probability-theory-in-geotechnical-engineering-wan-huan-
zhou/

OR CLICK BUTTON

DOWNLOAD NOW

Explore and download more ebook at https://textbookfull.com

 Recommended digital products (PDF, EPUB, MOBI) that
you can download immediately if you are interested.

Geotechnical and Foundation Engineering Practice in

Industrial Projects 1st Edition Mihir Roy

https://textbookfull.com/product/geotechnical-and-foundation-
engineering-practice-in-industrial-projects-1st-edition-mihir-roy/

textboxfull.com

Principles of Geotechnical Engineering Braja M. Das

https://textbookfull.com/product/principles-of-geotechnical-
engineering-braja-m-das/

textboxfull.com

The Theory and Practice of China s Tourism Economy 1978

2017 Peng Zhou

https://textbookfull.com/product/the-theory-and-practice-of-china-s-
tourism-economy-1978-2017-peng-zhou/

textboxfull.com

Advances in Numerical Methods in Geotechnical Engineering

Hany Shehata

https://textbookfull.com/product/advances-in-numerical-methods-in-
geotechnical-engineering-hany-shehata/

textboxfull.com
New Prospects in Geotechnical Engineering Aspects of Civil
Infrastructures Hadi Khabbaz

https://textbookfull.com/product/new-prospects-in-geotechnical-
engineering-aspects-of-civil-infrastructures-hadi-khabbaz/

textboxfull.com

Mathematical Theory of Bayesian Statistics First Edition

Watanabe

https://textbookfull.com/product/mathematical-theory-of-bayesian-
statistics-first-edition-watanabe/

textboxfull.com

Probability and Mathematical Statistics Theory

Applications and Practice in R 1st Edition Mary C. Meyer

https://textbookfull.com/product/probability-and-mathematical-
statistics-theory-applications-and-practice-in-r-1st-edition-mary-c-
meyer/
textboxfull.com

Mathematical Theory of Bayesian Statistics 1st Edition

Sumio Watanabe

https://textbookfull.com/product/mathematical-theory-of-bayesian-
statistics-1st-edition-sumio-watanabe/

textboxfull.com

Statistical independence in probability analysis number

theory Kac

https://textbookfull.com/product/statistical-independence-in-
probability-analysis-number-theory-kac/

textboxfull.com
Wan-Huan Zhou
Zhen-Yu Yin
Ka-Veng Yuen

Practice
of Bayesian
Probability Theory
in Geotechnical
Engineering
Practice of Bayesian Probability Theory
in Geotechnical Engineering
Wan-Huan Zhou Zhen-Yu Yin
• •

Ka-Veng Yuen

Practice of Bayesian
Probability Theory
in Geotechnical Engineering

123
Wan-Huan Zhou Zhen-Yu Yin
State Key Laboratory of Internet of Things Department of Civil and Environmental
for Smart City, Department of Civil and Engineering
Environmental Engineering The Hong Kong Polytechnic University
University of Macau Hong Kong, China
Macau, China

Ka-Veng Yuen
State Key Laboratory of Internet of Things
for Smart City, Department of Civil and
Environmental Engineering
University of Macau
Macau, China

ISBN 978-981-15-9104-4 ISBN 978-981-15-9105-1 (eBook)

https://doi.org/10.1007/978-981-15-9105-1
Jointly published with Tongji University Press
The print edition is not for sale in China (Mainland). Customers from China (Mainland) please order the
print book from: Tongji University Press.

© Tongji University Press 2021

This work is subject to copyright. All rights are reserved by the Publishers, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publishers, the authors, and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publishers nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publishers remain neutral with regard to
jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Dedicated to our families, teachers
and students.
Foreword

The development of numerical simulations in geomechanics has promoted the

construction of advanced constitutive models for soils. However, the limitations of
available methods for model class selection and parameter identification make it
difficult to efficiently apply these advanced models in practice. The selection of
constitutive model affects simulation results, leading to different engineering
decisions, which raises the problem of safety, economy, and risk in geotechnical
engineering. Therefore, the influence of the choice of constitutive model on
decision-making and its related consequences should be fully considered.
Currently, the existing studies focus on how to develop more accurate constitutive
modeling approaches, but there is still no systematic research on the evaluation
of their applicability to engineering problems. The selection of a given constitutive
model is often determined based on user preference and past experience, both of
which tend to be subjective. Neglect in the selection of constitutive models has
become one of the most important sources of risk for geotechnical engineering
constructions. Therefore, constitutive model class selection and parameter identi-
fication methods based on optimization and probabilistic theory have important
theoretical and practical significance for accurately predicting soil deformation,
controlling engineering cost and preventing engineering disasters. In this regard, a
clear identification of the model parameters has become central to the successful
practice of soil mechanics and to the development of artificial intelligence in
geotechnics.
I have known Dr. Wan-Huan Zhou since the years of her Ph.D. study at the
Hong Kong Polytechnic University. Her research project focused on soil nail
reinforcement mechanisms for landslide prevention. Her supervisor was Prof.
Jianhua Yin, who is well known for his expertise in constitutive modeling and
advanced physical experiments in geotechnical engineering. As one of her Ph.D.
examiners, I was impressed by her enthusiasm in research and knowledge in the
physical and numerical modeling of soil nail pullout resistance. After receiving her
Ph.D. degree, Dr. Zhou joined the University of Macau as a young scholar in 2009.
Since then, she has collaborated with Prof. Kelvin Yuen on the research of Bayesian
methods in geotechnical engineering.

vii
viii Foreword

I came to know Dr. Zhen-Yu Yin during the Winter of 2011. At that time, he
was at HKUST as a Visiting Scholar from Shanghai Jiao Tong University. We
began academic discussions and collaboration from then on. His rigorous research
approach, deep understanding of soil behavior and solid analytical skills have left a
deep impression on me. He has led a dynamic research team and has published
many well-cited papers. He also had 5 years of experience as an engineer before
returning to academia, which prompted him to develop engineering application
methods.
I have known Prof. Kelvin Yuen for more than 20 years since he was a Master’s
student at HKUST. Although his major research is on the development of Bayesian
system identification for structural systems, many of his methods have been applied
to the geotechnical engineering field, including those introduced in this book. I am
sure that this book will raise a lot of attention because uncertainty quantification is
very important in geotechnical engineering.
This book consists of a summary and a refinement of results obtained by the
authors on the topic of model class selection and parameter identification. It gathers
the authors’ original results in optimization algorithms, model class selection
methods, parameter identification processes, constitutive modeling approaches, and
engineering applications. The development of the optimization theory in geotech-
nics is clearly presented based upon rigorous analyses. The illustrative examples
allow the reader to quickly understand the concept of optimization-based parameter
identification and its application to various practical cases. The authors have
deliberately provided relevant source codes for optimization algorithms and con-
stitutive models so that the reader can make use of this work for his or her own
applications.
In summary, this book provides an informative exposition of the current
advances in the formulation of numerical simulations in geomechanics and will
benefit both scientists and practitioners. Students will also find in this book a
stimulating introduction to the mechanics of geomaterials.

Limin Zhang
Ph.D., FASCE, FHKIE
Chair Professor of Geotechnical Engineering
Department of Civil and Environmental Engineering
The Hong Kong University of Science and Technology
Clear Water Bay, Kowloon, Hong Kong
Preface

The soil as the research object of soil mechanics has very complex property, and
how to correctly understand or describe the soil physical or mechanical behaviors
under different cases is crucial to the development of the soil mechanics. In
geotechnical engineering, the laboratory and field tests are important ways to study
and determine a certain soil physical property or parameters. Some soil properties
can be measured directly, e.g., water content, density, specific gravity, etc. But
some soil properties can only be estimated through their relationships with other
soil properties, such as void ratio, porosity, and degree of saturation. There are still
some cases in which the soil properties must be estimated through the complex or
empirical formulae. In these situations, some unknown parameters in the model are
required to be determined to describe the soil property. Back analysis is usually
conducted based on the relevant observed responses, e.g., the determination of the
compression index and the swelling index involved in the deformation predictions
of one-dimensional compression problem. Therefore, it is worthwhile to evaluate
the performance of models on the estimation of the real soil behavior.
It is well known that the soil has the complicated variability and uncertainty, and
its state is affected by many factors, e.g., stress level and history, rainfall conditions,
and physical and chemical changes. Therefore, the soil properties change in space
and time even for the homogeneous soil, and this is considered to be the inherent
variability of soil. The quantification of inherent variability is necessary for the
appropriate evaluation of soil properties. To quantify the uncertainties of soil
properties has great engineering significance. Moreover, due to the uncertainty,
complexity, and time-consuming in the measurement of the soil properties, indirect
methods are developed for predicting soil properties. Based on a large amount of
experiments and engineering practices, the researchers have proposed the empirical
or numerical formulae to estimate certain soil properties or mechanics index with
appropriate assumptions, e.g., numerous empirical formulae for predicting SWCC
and various soil constitutive models for describing the time-dependent behavior of

ix
x Preface

soft soil or critical state behavior of granular soils, etc. These models usually
contain unknown parameters, which are generally determined using the regression
analysis based on test data. Many of empirical models may not reflect the mech-
anism of soil properties due to the complex mechanism of soil behavior, and the
unknown parameters in the model may have no physical meanings.
Notably, most of empirical models for describing soil properties are obtained
from data fitting, and this will introduce the uncertainty to the model performance.
Even for the theoretical models, the uncertainty is also inevitable due to the ide-
alizations and simplifications in the theories. Three primary sources of uncertainty,
i.e., inherent soil variability, measurement error, and transformation uncertainty,
contribute to the uncertainty in the model predictions. These components should be
combined consistently by using a second-moment probabilistic approach. Besides
the evaluation of model uncertainty, the efficiency of models for reflecting the
actual soil properties is another valuable research issue. The investigation on
the model efficiency can provide engineers with a straightforward suggestion for the
engineering application of models.
Hence, research works in this book have been conducted to investigate different
geotechnical problems with the use of Bayesian probabilistic approach. In the study
of soil-water retention property, the effect of initial void ratio on the SWCC of same
textured soils was investigated and new models were introduced in Chaps. 2 and 3.
The involved predictive model was determined by using the Bayesian linear model
class selection. Chapters 3 and 4 are dedicated to evaluating the model uncertainty
for the creep behavior of soft soil and the effect of load duration while Chaps. 6 and 7
focus on shear behavior of granular soils. The Bayesian probabilistic method was
adopted to determine all model parameters and evaluate the uncertainty of model
parameters and model predictions. In Chap. 8, Bayesian linear model class selection
is employed again for the estimation of pullout shear strength of grouted soil nails.
Chapter 9 is dedicated to the study of the unconfined compression strength of natural
fiber reinforced soil. The Bayesian nonparametric general regression method was
introduced to select the most significant features from a set of experimental data. In
Chap. 10, we introduce an efficient model updating method under the Bayesian
framework for the deformation prediction of braced excavations.
This book begins with the most basic theoretical knowledge and combines
geotechnical experiments to realize how Bayesian probabilistic method can be
applied in practice, which can help readers quickly and accurately understand the
constitutive simulation of soil characteristics and methods to evaluate the associated
uncertainty in the predictions. In addition, this book provides source codes and a
free software platform from the authors’ work that can be directly used by readers
for analysis and training. We also hope to enrich the developed application methods
Preface xi

in the future by sharing with readers the latest scientific research results in the
geotechnical field and promoting the application of these methods to achieve the
purpose of geotechnical engineering innovation.

Macao, Hong Kong Wan-Huan Zhou

May 2020

Zhen-Yu Yin

Ka-Veng Yuen
Acknowledgments

In the process of writing this book, we received much sincere guidance and help
from experts and colleagues. Here, I would like to express my sincere gratitude to
Prof. Pierre-Yves Hicher from Ecole Centrale de Nantes in France, Prof. Jian-Hua
Yin from the Hong Kong Polytechnic University, Prof. Limin Zhang from Hong
Kong University of Science and Technology, Prof. Shuilong Shen from Shantou
University, and Prof. Hongwei Huang from Tongji University for their careful
guidance and help during the book-writing process. And also we thank Dr. Zhao
Linshuang, Dr. Tan Fang, Dr. Qi Xiaohui, and Ms. Tian Chen currently or formerly
from the University of Macau; Dr. Yinfu Jin, Dr. Jie Yang, Dr. Pei Wang, and
Dr. Hanlin Wang from the Hong Kong Polytechnic University for their participa-
tion in the research projects and hard work in the process of organizing and
reviewing the book. We are also grateful to Tongji University Press and Springer
for performing detailed work on the publication of this book.
Some of the results and publications in this book have been funded by the
Science and Technology Development Fund, Macau SAR (File no.
0193/2017/A3, 0125/2014/A3, and SKL-IOTSC-2018-2020) and the University of
Macau (MYRG2018-00173-FST). We would like to express our heartfelt thanks to
National Natural Science Foundation of China (52022001, 51508585, 41372285,
51579179) and the Research Grants Council (RGC) of Hong Kong Special
Administrative Region Government (HKSARG) of China (Grant No.:
15209119, 15217220, R5037-18F) for their funding.

xiii
About This Book

This book systematically introduces the application of Bayesian probabilistic

approach in soil mechanics and geotechnical engineering. Four typical problems are
analyzed using Bayesian probabilistic approach, i.e., to model the effect of initial
void ratio on the soil-water characteristic curve (SWCC) of unsaturated soil, to
select the optimal model for the prediction of the creep behavior of soft soil under
one-dimensional straining, and to identify model parameters of soils and to select
constitutive model of soils considering critical state concept. This book selects the
simple and easy-to-understand Bayesian probabilistic algorithm, so that readers can
master the Bayesian method to analyze and solve the problem in a short time. In
addition, this book provides MATLAB codes for various algorithms and source
codes for constitutive models so that readers can directly analyze and practice.
This book can be used as a postgraduate textbook for civil engineering,
hydraulic engineering, transportation, railway, engineering geology, and other
majors in colleges and universities, and as an elective course for senior under-
graduates. It can also be used as a reference for relevant professional scientific
researchers and engineers.

xv
Contents

1 Problem of Uncertainties in Geotechnical Engineering . . . . . . . . . 1

1.1 Uncertainties in Geotechnical Engineering . . . . . . . . . . . . . . . . 1
1.1.1 Uncertainties in Soil Properties . . . . . . . . . . . . . . . . . . 1
1.1.2 Uncertainties in Geotechnical Model . . . . . . . . . . . . . . 3
1.1.3 Back Analysis Methods . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Bayesian Probabilistic Approach . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.1 Parametric Identification . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.2 Model Class Selection . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.3 Numerical Simulation for the Representation
of the Updated PDF . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Soil Water Retention Property of Unsaturated Soil . . . . . . . . . . 12
1.3.1 Soil Suction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.2 Soil–Water Characteristic Curve (SWCC) . . . . . . . . . . 13
1.3.3 Influencing Factors on SWCC . . . . . . . . . . . . . . . . . . . 15
1.3.4 Estimation Methods for SWCC . . . . . . . . . . . . . . . . . . 15
1.4 Creep Behavior of Soft Soil . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.4.1 Mechanism of Creep Behavior for Soft Soil . . . . . . . . . 18
1.4.2 Time-Dependent Model for Creep Analysis . . . . . . . . . 19
1.4.3 Bjerrum’s Time Line Conceptual Model . . . . . . . . . . . 19
1.4.4 Yin and Graham’s Elastic Viscoplastic
(EVP) Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.5 Critical State Behavior of Granular Soils . . . . . . . . . . . . . . . . . 24
1.5.1 Shear Behavior of Granular Soils . . . . . . . . . . . . . . . . 24
1.5.2 Critical State Line of Granular Soils . . . . . . . . . . . . . . 26
1.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2 Estimation of SWCC and Permeability for Granular Soils . . . . . . 39
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.2 Estimation of SWCCs with Different Initial Dry Densities . . . . . 41

xvii
xviii Contents

2.2.1 Fredlund and Xing [7] Equation for SWCC . . ....... 41

2.2.2 Effect of Initial Dry Density and Proposed
Estimation Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.3 Verification and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.4 Bayesian Approach and Confidence Interval of SWCC . . . . . . . 49
2.5 Estimation of Relative Permeability Function kr . . . . . . . . . . . . 51
2.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3 Modeling SWCC for Coarse-Grained and Fine-Grained Soil . . . . 59
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.2 Establishment of Relationships . . . . . . . . . . . . . . . . . . . . . . . . 61
3.2.1 Relationship of Volumetric Water Contents . . . . . . . . . 61
3.2.2 Relationship of Suctions . . . . . . . . . . . . . . . . . . . . . . . 63
3.3 Determination of the Adjustment Parameter b . . . . . . . . . . . . . 64
3.3.1 Established of Predictive Model . . . . . . . . . . . . . . . . . 65
3.3.2 Model Class Selection . . . . . . . . . . . . . . . . . . . . . . . . 67
3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.4.1 Optimal Model for b . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.4.2 Comparison of SWCCs Predicted by Different
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.4.3 Verification and Uncertainty Analysis . . . . . . . . . . . . . 78
3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4 Model Updating and Uncertainty Analysis for Creep of Clay . . . . 89
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.2 Review of Time-Dependent Models for Soft Soils . . . . . . . . . . 91
4.3 Model Updating with Bayesian Method . . . . . . . . . . . . . . . . . . 95
4.4 Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.4.1 Case Study 1: Intact Soft Soil Sample
of Vanttila Clay . . . . . . . . . . . . . . . . . . . . . . . . . .... 98
4.4.2 Case Study 2: Reconstituted Sample
of Hong Kong Marine Clay . . . . . . . . . . . . . . . . . . . . 103
4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
5 Effect of Loading Duration on Uncertainty in Creep
Analysis for Clay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.2 Data for Training and Testing . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.3.1 Results by TMCMC Method . . . . . . . . . . . . . . . . . . . . 114
Contents xix

5.3.2 Estimations by Updated Models . . . . . . . . . . . . . . . . . 119

5.3.3 Verification and Uncertainty Analysis . . . . . . . . . . . . . 127
5.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
6 Model Class Selection for Sand with Generalization Ability
Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
6.2 Representative Advanced Sand Models . . . . . . . . . . . . . . . . . . 134
6.3 Model Class Selection Approach and Generalization
Ability Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
6.3.1 Bayesian Model Class Selection . . . . . . . . . . . . . . . . . 137
6.3.2 Evaluation of Generalization Ability . . . . . . . . . . . . . . 138
6.3.3 General Procedure of Model Class Selection . . . . . . . . 139
6.3.4 Illustration Case of Bayesian Model Class Selection . . . 140
6.4 Selection of Sand Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
6.4.1 Model Class Selection Based on Different Test Sets . . . 140
6.4.2 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . 143
6.5 Validation by Other Sand . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
6.5.1 Model Class Selection Based on Tests
of Karlsruhe Sand . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
6.5.2 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . 155
6.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
6.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
7 Parametric Identification of Advanced Soil Models for Sand . . . . . 163
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
7.2 Enhanced DE-TMCMC-Based Bayesian Identification
with Parallel Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
7.2.1 Framework of Bayesian Parameter Identification . . . . . 165
7.2.2 Proposition of Enhanced DE-TMCMC . . . . . . . . . . . . 166
7.2.3 Parallel Computing DE-TMCMC . . . . . . . . . . . . . . . . 168
7.3 Performance of Parallel Computing DE-TMCMC . . . . . . . . . . . 170
7.3.1 Numerical Validation on Synthetic Laboratory
Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
7.3.2 Performance on Real Laboratory Tests . . . . . . . . . . . . 175
7.3.3 Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . 177
7.4 Application to In Situ Testing . . . . . . . . . . . . . . . . . . . . . . . . . 183
7.4.1 Parametric Identification Parameters from
Pressuremeter Test . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
7.4.2 Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . 186
xx Contents

7.5 Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189

7.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
8 Estimation of Pullout Shear Strength of Grouted Soil Nails . . . . . 195
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
8.2 Construction of Predictive Formula Candidates . . . . . . . . . . . . . 197
8.3 Bayesian Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
8.3.1 Bayesian Model Class Selection . . . . . . . . . . . . . . . . . 198
8.3.2 Multivariate Linear Model . . . . . . . . . . . . . . . . . . . . . 198
8.4 Analysis of Laboratory Test Data and Proposed Design
Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
8.4.1 Influence of Different Prior PDFs . . . . . . . . . . . . . . . . 202
8.4.2 Proposed Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
8.5 Comparison with the Effective Stress Method . . . . . . . . . . . . . . 208
8.6 Estimation of Field Pullout Test Data . . . . . . . . . . . . . . . . . . . . 210
8.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
9 Selection of Physical and Chemical Properties of Natural Fibers
for Predicting Soil Reinforcement . . . . . . . . . . . . . . . . . . . . . . . . . 215
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
9.2 Basic Index of Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
9.2.1 Chemical Components of Natural Fibers . . . . . . . . . . . 218
9.2.2 Physical Properties of Natural Fibers . . . . . . . . . . . . . . 219
9.2.3 Soil Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
9.3 Bayesian Nonparametric General Regression Method . . . . . . . . 221
9.4 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
9.4.1 Data Preparation and Model Class Selection . . . . . . . . 225
9.4.2 Validation of the Optimal Model . . . . . . . . . . . . . . . . . 229
9.4.3 Robustness of the BNGR Algorithm Using
the K-Fold Cross-Validation Method . . . . . . . . . . . . . . 232
9.4.4 Sensitivity Analysis of Different Influencing Factors . . . 233
9.4.5 Relationships Between the UCS and the Input
Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
9.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
10 An Efficient Probabilistic Back-Analysis Method for Braced
Excavations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
10.2 Method for Updating the Soil Parameters . . . . . . . . . . . . . . . . . 244
10.2.1 Framework for Bayesian Updating . . . . . . . . . . . . . . . 244
Contents xxi

10.2.2 Procedure for the Markov Chain Monte Carlo

Simulation Using the Metropolis–Hastings
Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
10.2.3 Response Surface Method . . . . . . . . . . . . . . . . . . . . . . 248
10.3 Measurement Errors and Model Errors . . . . . . . . . . . . . . . . . . . 249
10.3.1 Model Errors of the Wall Deflections in Braced
Excavations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
10.3.2 Measurement Errors of Inclinometer Instruments . . . . . 253
10.4 Illustrative Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
10.4.1 Excavation of a Metro Project in Hangzhou,
China and FEM Simulation . . . . . . . . . . . . . . . . . . . . . 256
10.4.2 Prior Knowledge of Soil Parameters . . . . . . . . . . . . . . 259
10.4.3 Construction of Response Surfaces . . . . . . . . . . . . . . . 259
10.4.4 Bayesian Updating Using Prior Information
from the Laboratory Test . . . . . . . . . . . . . . . . . . . . . . 260
10.4.5 Effect of Prior Distribution of Soil Parameters . . . . . . . 265
10.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267

Appendix A: Matlab Code for Linear Model Class Selection . . . . . . . . . 271

Appendix B: Matlab Code for Soft Soil Creep Models with Finite
Difference Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
Appendix C: Matlab Code for Transitional Markov Chain Monte
Carlo Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
Appendix D: Matlab Code for Identifying Parameters of a Given
Model or Model Class Selection for Given Experimental
Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
Appendix E: Matlab Code of DE-TMCMC Method . . . . . . . . . . . . . . . . 291
Appendix F: Fortran Code of SIMSAND Model in Format of UMAT
of ABAQUS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
Appendix G: Matlab Code for Using Bayesian Nonparametric
General Regression Method . . . . . . . . . . . . . . . . . . . . . . . . . 315
Appendix H: Matlab Code for Markov Chain Monte Carlo
Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
About the Authors

Wan-Huan Zhou is currently Head and Associate Professor of the Department of

Civil and Environmental Engineering and Interim Director of Centre for Regional
Oceans at the University of Macau. She received her B.Eng. degree in 2002 and
M.Sc. degree in 2005 from Zhejiang University in China and Ph.D. in Geotechnical
Engineering from The Hong Kong Polytechnic University in 2008. After graduation,
she worked as a Lecturer at The Hong Kong Polytechnic University and later joined
the University of Macau in September 2009. She has acted as Principal Investigator
on 15 research projects, including grants from the National Natural Science
Foundation (NSFC) and the Macao Science and Technology Development Fund
(FDCT). She has published more than 40 SCI articles as the first or correspondent
author in geotechnical journals, such as Journal of Geotechnical and
Geoenvironmental Engineering, ASCE, and Canadian Geotechnical Journal. She
won the 2011 best paper award of the International Association for Computer
Methods and Advances in Geomechanics. In 2020, she was awarded the NSFC
(National Natural Science Foundation of China)—the Excellent Young Scientist
Fund (Hong Kong and Macau).

Zhen-Yu Yin is Associate Professor of Geotechnical Engineering at The Hong

Kong Polytechnic University since 2018. He received his B.Sc. in Civil
Engineering from Zhejiang University in 1997, followed by a 5-year engineering
consultancy at the Zhejiang Jiahua Architecture Design Institute. Then, he obtained
his M.Sc. and Ph.D. in Geotechnical Engineering at Ecole Centrale de Nantes
(France) in 2003 and 2006, respectively. Then he has been working as Postdoctoral
Researcher at Helsinki University of Technology (Finland), University of
Strathclyde (Glasgow, UK), Ecole Centrale de Nantes, and University of
Massachusetts (Umass-Amherst, USA). In 2010, he jointed Shanghai Jiao Tong
University as Special Researcher and received “Professor of Exceptional Rank of
Shanghai Dong-Fang Scholar”. In 2013, he jointed Ecole Centrale de Nantes as
Associate Professor before moving to Hong Kong. He has published over 170
articles in peer-reviewed international journals with an H index of 36 according to

xxiii
xxiv About the Authors

Web of Science. Since 2012 he is an Associate Committee member of the granular

materials committee of American Society of Civil Engineers.

Ka-Veng Yuen is the Dean of Graduate School and Distinguished Professor of

Civil and Environmental Engineering (CEE) at the University of Macau. He
received his Ph.D. in civil engineering from the California Institute of Technology
(Caltech) after a study period of 2 years and 7 months. His major awards include
Young Investigator Award, International Chinese Association on Computational
Mechanics; QIAN Weichang programme scholar, K.C. Wong Education
Foundation; Phi-Tau-Phi Honorary member; and the Housner’s Fellowship. His
paper “Model selection using response measurements: Bayesian probabilistic
approach” (coauthored with J. L. Beck) is one of the top 10 most cited papers
among all papers published in the Journal of Engineering Mechanics (established
since 1983) of the American Society of Civil Engineers. Furthermore, his
single-author book “Bayesian methods for structural dynamics and civil engi-
neering” (published by the top publisher John Wiley & Sons in 2010) has attracted
great attention of Bayesian methods for structural dynamics. It has become a
classical reading in this research field. His H index is 33 according to Web of
Science.
Abbreviations

a Constant of fines content effect in silty sand

(SIMSAND+fr)
a Target inclination of yield surface related to volumetric
strain (ASCM)
Ad Constant of magnitude of stress-dilatancy (0.5*1.5)
b Constant controlling the amount of grain breakage
(SIMSAND+Br)
b Target inclination of yield surface related to deviatoric
plastic strain
Caei Intrinsic secondary compression index (remolded clay)
D Stiffness matrix of material
E Young’s modulus
e, e0 Void ratio and initial void ratio
E0 Referential Young’s modulus (dimensionless)
ec0 Initial critical state void ratio (SIMSAND); virgin initial
critical state void ratio before breakage
ecuf Fractal initial critical state void ratio due to breakage
ed General shear strain
Eh, Ev Horizontal and vertical Young’s modulus
ehc,c0 Initial critical state void ratio of pure fine soils (fc = 0%)
ehf,c0 Initial critical state void ratio of pure coarse soils (fc =
100%)
emax Maximum void ratio
Eu Undrained Young’s modulus
fth Threshold fines content from coarse to fine grain
skeleton (20*35%)
f Fines content
G Shear modulus
G0 Referencial shear modulus
Gvh Shear modulus

xxv
xxvi Abbreviations

I 1, I 2, I 3 The first, second, and third invariants of the stress

tensor
I1', I2', I3' The first, second, and third invariants of the strain
tensor
J 1, J 2, J 3 The first, second, and third invariants of the deviatoric
stress tensor
J1', J2', J3' The first, second, and third invariants of the deviatoric
strain tensor
K Bulk modulus
K0 The coefficient of earth pressure at rest
kp Plastic modulus-related constant in SIMSAND; plastic
modulus-related parameter in ASCM
Kw Bulk modulus of water
M Constraint modulus in elasticity; slope of critical state
line in p'-q plane
m Constant of fines content effect in sandy silt
Mc Slope of critical state line in triaxial compression in p'-
q plane
n Porosity of soil; elastic constant controlling nonlinear
stiffness
nd Phase transformation angle-related constant (1)
np Peak friction angle-related constant (1)
p' Mean effective stress
pat Atmosphere pressure
pb0 Initial bonding adhesive stress
pc0 Initial size of yield surface; initial size of yield surface
of grain breakage (SIMSAND+Br)
pexcess Excess pore pressure
psteady Steady pore pressure
q Deviatoric stress
Rd Ratio of mean diameter of sand to silt D50/d50
Ra Stress relaxation coefficient
sij Deviatoric stress tensor
ux, u y, u z Displacements
ak0 Initial inclination of yield surface
b Rate-dependency coefficient
v0 Initial bonding ratio
dij Kronecker symbol
e 1, e 2, e 3 Principle strains
ea, er Axial strain and radial strain
eij Strain tensor
em Mean strain
ev Volumetric strain
cxy ; cyx ; cyz ; czy ; czx ; cxz Engineering shear strains
Abbreviations xxvii

u Friction angle
j Swelling index of the isotropic compression test
(in e-lnp’ plane)
ji Intrinsic swelling index (of remolded soil, in e-lnp’
plane)
k Lame constant in elasticity; compression index
(in e-lnp’ plane); constant controlling the nonlinearity
of CSL in SIMSAND
k' Compression index under the plane of loge-logp′
ki Intrinsic compression index (of remolded soil, in e-lnp’
plane)
mu Undrained Poisson’s ratio
mvh' Horizontal Poisson’s ratio
mvv' Vertical Poisson’s ratio
h Lode angle
q Constant controlling the movement of CSL
ra, rr Axial stress and radial stress
rij Stress tensor
rm(p) Mean stress
rn, rh Vertical and horizontal stresses
rp0 Preconsolidation pressure
rw Pore water pressure
rx, ry, rz Normal stresses
r1, r2, r3 First, second, and third principle stresses
s Reference time (Oedometer test s = 24h) (ANICREEP)
sxy, syx, syz, szy, szx, sxz Shear stresses
t Poisson’s ratio
x Absolute rotation rate of the yield surface
xd Rotation rate of the yield surface related to the
deviatoric plastic strain
n Constant controlling the nonlinearity of CSL
(SIMSAND); absolute rate of bond degradation
nb Degradation rate of the inter-particle cohesive bonding
nd Constant controlling the deviatoric strain-related bond
degradation rate
w Dilatancy angle
Chapter 1
Problem of Uncertainties in Geotechnical
Engineering

1.1 Uncertainties in Geotechnical Engineering

Different types of uncertainties are involved in the geotechnical engineering, for

example, uncertainty of concept, uncertainty of material classification, soil parame-
ters, constitutive or empirical models, boundary conditions, other uncertainty caused
by measurement error, and so on. All these uncertainties affect the evaluation of the
soil or structure performance, and they are required to be considered and analyzed
for reliable evaluation results. According to the type of uncertainty, different anal-
ysis methods have been utilized to evaluate the uncertainty, for example, stochastic
theory, fuzzy mathematics, theory of gray systems, artificial neural networks, genetic
algorithms, and so on. In some cases, these methods are coupled to solve the uncertain
problems. Among these methods, the probabilistic method in stochastic theory has
been widely applied to many aspects of geotechnical engineering, for example, slope
analysis [1–4], liquefaction analysis [5, 6], reliability analysis of reinforced embank-
ment [7], seepage analysis [8], soil–water characteristic curve [9–11], reducing uncer-
tainties in empirical correlations [12], and so on. In the following sections, the sources
of uncertainty and their analysis methods for predicting the performance of soil or
structure are reviewed.

1.1.1 Uncertainties in Soil Properties

Soils are porous and discontinuous media and they are formed by the weathering of
rocks. Under the natural conditions, the state of soil is affected by various factors,
for example, stress level and stress history, water seepage, physical and chemical
changes. Even in the homogeneous soil, the soil property also varies with the site,
and it possesses the spatial variability. Due to these characteristics of soil, there are
inevitable uncertainties involved in the determined soil properties. The uncertainties
in measured soil properties are associated with the soil sample disturbance, test

© Tongji University Press 2021 1

W. Zhou et al., Practice of Bayesian Probability Theory in Geotechnical Engineering,
https://doi.org/10.1007/978-981-15-9105-1_1
2 1 Problem of Uncertainties in Geotechnical Engineering

conditions, human factors and statistical uncertainties. The uncertainties of estimated

soil properties by models are caused by the limited test data of soil samples, the
imperfection of adopted estimation formula and the uncertainty of other relevant soil
properties.
Tang [13] focused on the uncertainty characterization of soil properties in a
homogeneous soil layer, and stated three main sources of uncertainties: inherent
spatial variability, systematic uncertainties from the tests and estimation errors from
insufficient samples. He also proposed a simplified method to incorporate these
major sources of uncertainties in determining the statistic of the spatial average soil
properties.
Phoon and Kulhawy [14, 15] also summarized three primary sources for the
geotechnical uncertainty: inherent variability, measurement error and transforma-
tion uncertainty, and pointed out that the relative impact of these three sources to
the uncertainty in the designed soil property lies on the site conditions, degree of
equipment and procedural control, and precision of the correlation models. They
mentioned that it is necessary to evaluate these three primary uncertainties sepa-
rately and to develop a systematic method that can combine these uncertainties
consistently. In their studies, a second-moment probabilistic approach was adopted
to evaluate the uncertainties in the design soil properties by combining the appro-
priate uncertainties. The variability of designed soil properties can be evaluated by
the corresponding coefficient of variation (COV). There have been many studies on
the common soil properties, that is, effective stress friction angle, in situ horizontal
stress coefficient, undrained shear strength and Young’s modulus.
Christian et al. [16] presented how probabilistic descriptions of soil properties
can be achieved from laboratory and field data and utilized in the stability analysis.
The uncertainties of soil properties for the reliability analysis of slope stability were
considered to have two major sources: data scatter and systematic error. The former
contains inherent spatial variability and random test error. The latter contains system-
atic statistical error caused by sampling and bias in the measurement processes. They
also analyzed the relative contributions of the uncertainties of different parameters
to the reliability of embankment using the first-order second-moment approach.
Many other researchers have also identified and classified major components of
uncertainties associated with the estimated soil properties by different approaches.
Gui et al. [17] studied the effect of stochastic hydraulic conductivity on the slope
stability of an embankment dam by conducting the combination analysis of random
field simulation, seepage analysis and slope analysis. Baecher and Christian [18]
discussed the nature and philosophy of uncertainty in geological and geotechnical
engineering and addressed fundamentals and limits of probabilistic and statistical
methods. For the calculation of hydrodynamic pressures caused by rainfall, Zhang
et al. [19] discussed the errors from model representation and from variability of
model parameters separately. They found that different design assumptions and
models can obtain significant different factors of safety of the slope. Zhang et al.
[20] conducted numerous field pullout tests to evaluate four factors influencing the
pullout resistance of soil nails.
1.1 Uncertainties in Geotechnical Engineering 3

1.1.2 Uncertainties in Geotechnical Model

Besides the uncertainties in estimated soil properties, the influence of two types of
model uncertainty on the performance of soil or structure should also be noted, which
are associated with the test of geotechnical physical model and performance model
itself. The physical model tests, for example, centrifuge tests and full-scale field tests,
are usually conducted to study the performance of soil or structure. The performance
factors such as the pore water pressures and displacements can be measured directly,
which are influenced by the inherent variability of soil properties. Similar to the
soil property tests, the test results of physical model also suffer from the systematic
and statistical uncertainties. The measurements may be influenced by model scaling,
preparation of soil sample, instrumentation technology, interpretation of test results,
and so on [21].
In addition, the uncertainties of model input variables, which are measured or
estimated for soil parameters at a specific site, cause the variability of the model
outputs. The systematic errors associated with the performance model itself, for
example, uncertainties in constitutive behavior, imperfect model expression, initial
condition, boundary condition [19, 22], can also contribute to the uncertainties of
model outputs. In order to evaluate the performance of soil or underground structure,
the model uncertainties must be considered in the analysis.

1.1.3 Back Analysis Methods

In geotechnical engineering, many complicated mathematical models have been

proposed to study various properties. For instance, with respect to the stress–strain
relationship of soil, suitable constitutive models can describe the nonlinear, irre-
versible and time-dependent behavior of soil. In order to obtain good predictions
for the deformation behavior of soil, the parameters contained in the models often
play very important role. Among the model parameters, some parameters are related
to the soil properties, and they can be determined through the survey methods in
the laboratory or field site. It has been noted that the reliability and applicability of
measured soil parameters are significantly influenced by the test conditions which
have large difference with the engineering practice. In addition, the soil parameters
may change with time and the project process, such as construction technology and
construction parameters. If the influences of these factors on the soil parameters
are not considered, the design results or simulations with these soil parameters may
have large errors. For the other model parameters, they may have no physical mean-
ings and cannot be directly determined through the measurements. Fortunately, the
stress, displacement and strain in the engineering practice can be measured, and then
it became possible to determine the soil properties or model parameters using back
analysis method based on the measured data.
4 1 Problem of Uncertainties in Geotechnical Engineering

Since the late 1970s, back analysis has been paid much attention and also obtained
fast development. At first, the information, such as soil physical properties, constitu-
tive models, boundary conditions and measured data, have been processed as deter-
ministic quantities, that is, deterministic methods for back analysis. This method
is easy to be implemented without considering many influencing factors, and its
results basically meet for engineering applications. However, the soil engineering
practice is an uncertainty system, and there is uncertainty in many aspects, such as
the soil physical properties, the constitutive models and boundary conditions. The
measured properties, for example, displacement and stress, are all uncertain vari-
ables containing the measurement error. Therefore, the deterministic analysis is not
realistic to study the problems in geotechnical engineering.
At present, several methods for probabilistic back analysis have been conducted
in geotechnical engineering, for example, maximum likelihood method [23–25], and
Bayesian method [25, 28]. The Bayesian method is considered to be more effective
for updating the soil parameters and response predictions [22, 26–28], and it can
also evaluate the uncertainty of soil parameters by the obtained posterior proba-
bility distributions. Bayesian updating method has been successfully used to analyze
different geotechnical problems. Kay [29] evaluated the safety factor for single pile
in sand using this method, and Honjo et al. [24] studied the embankment on soft clay.
Yan et al. [30] analyzed the correlation between the soil compression index and their
influencing factors and proposed the predictive formulae for the compression index.
Chiu et al. [10, 11] presented a Bayesian approach for estimating the water retention
curve. Cao and Wang [31] and Wang et al. [32] developed Bayesian approaches to do
underground soil stratum identification, soil classification and soil properties based
on the cone penetration tests. Wang et al. [33] conducted the identification for the
soil strata in the London Clay Formation using the water content data. Cao and Wang
[34] presented a Bayesian model comparison approach to obtain the most plausible
correlation function for a specific site based on the project-specific test results and
previous site information. There are also many studies on the multistage excavation
analysis by utilizing the Bayesian approach [35–38].

1.2 Bayesian Probabilistic Approach

Given a specific model and the observations, the above-mentioned back analysis
methods can be utilized to evaluate the influences of the associated parameter uncer-
tainties on the model predictions. However, different constitutive models or empir-
ical models, which have different mathematical forms or complexity, are always
proposed to solve the same issue in geotechnical engineering. It is unknown how
these models perform in the engineering practice and which one has better ability
to fit with the observations and provide better predictions in a specific project. The
Bayesian approach as a useful tool can be adopted to investigate the predictive perfor-
mance of models. Bayesian approach can not only update the model parameters and
1.2 Bayesian Probabilistic Approach 5

characterize the uncertainties using their posterior PDFs but also find the most plau-
sible model for predicting the responses by Bayesian model class selection. In this
section, the Bayesian approach for parametric identification and model class selection
is briefly outlined. One may refer to the details from Yuen [39].

1.2.1 Parametric Identification

Consider a linear or nonlinear relationship (model class C) to describe the quantity

of concern y:

y = y(x; θm , C) (1.1)

where θm is the uncertain parameters contained in the model C; x is the measured

variable vector. Use Y to denote the measurement of y and it is assumed that the
difference can be adequately modeled as Gaussian random variable:

Y = y+ε (1.2)

where E is a Gaussian random variable with zero mean and variance σε2 referred
hereafter as prediction-error variance, and it is adopted to represent the measurement
 T
noise and modeling error. The uncertain parameter vector θ = θmT , σε2 includes the
model parameter vector θm and the prediction-error variance σε2 . Their uncertainties
can be represented by using the posterior PDFs, and the probability expression of
posterior PDF given the data D is written as:

p(θ|D, C) = c0 p(θ|C) p(D|θ, C) (1.3)

where c0 = 1/ p(D|C) is a normalizing constant; p(D|C) is the evidence of model

class C; p(θ|C) is the prior PDF of the uncertain parameters in θ, which is based
on the previous knowledge or user’s judgment; p(D|θ, C) is the likelihood function
expressing the level of data fitting. If the prediction errors in different measured data
are statistically independent, the likelihood function can be computed by:

N
p(D|θ, C ) = (2π σε2 )− 2 exp[−
N
Jg (θm ; D, C)] (1.4)
2σε2

where N is the number of measured data; Jg (θm ; D, C) is the goodness-of-fit

function, and is given by:

1 
N
Jg (θm ; D, C) = [Y (n) − y(n)]2 (1.5)
N n=1
6 1 Problem of Uncertainties in Geotechnical Engineering

The most probable model parameter vector θ∗m can be computed by maximizing
the posterior PDF in Eq. (1.3). If N is large enough or adopt the improper prior, this
will be equivalent to minimizing the goodness-of-fit function Jg (θm ; D, C):

θ∗m = arg min Jg (θm ; D, C) (1.6)

θm

If Jg (θm ; D, C) is implicit, numerical optimization is required to achieve the

optimal model parameters [39].
The most probable value of the prediction-error variance σε2 can also be obtained
by maximizing the posterior PDF, and its closed form solution is available:

σε2∗ = min Jg (θm ; D, C) = Jg (θ∗m ; D, C) (1.7)

θm

According to Beck and Katafygiotis [40], the posterior PDF p(θ |D, C) given
a large number of data D is approximately Gaussian distribution, and its mean
value is the optimal parameter vector θ∗m and its covariance matrix can be obtained
by computing the inverse of the Hessian matrix of the objective function J (θ) =
− ln[ p(θ|C) p(D|θ, C)] at θ∗m :

   ∂2 
H(l,l ) θ∗ = − ln[ p(θ|C) p(D|θ, C)] (1.8)
∂θl θl  θ=θ∗

The posterior PDF in Eq. (1.3) provides the probabilistic description for the model
parameters, but its topology may be very complicated in general. Especially for the
case with large number of uncertain parameters, the distribution is difficult to be
visualized. However, the numerical simulation based on the approximation method
is found to be effective for solving this difficulty, and three types of methods will be
presented in Sect. 1.2.3.

1.2.2 Model Class Selection

The probability/plausibility of a model class conditional on a database D can be

obtained by using the Bayes’ theorem as follows [41]:
   
  p D|C j · P C j
P C j |D = , j = 1, 2, . . . , Nc (1.9)
p(D)
1.2 Bayesian Probabilistic Approach 7

where N C is the number of model classes to be considered; p(D) =


Nc      
p D|C j P C j is a normalizing constant and P C j expresses the user’s judg-
j=1

Nc  
ment on the initial plausibility of the model classes with P C j = 1. In general,
 j=1
uniform prior plausibility
 can
 be assumed, that is, P C j = 1 Nc . The evidence of
the model class C j p D|C j can be obtained using the theorem of total probability:

     
p D|C j = p D|θ, C j p θ|C j dθ, j = 1, 2, . . . , Nc (1.10)


where the parameter vector θ belongs to the parameter space  j ⊂  N j and it is

defined in each model class (C j ). In general, Eq. (1.10) involves a high-dimensional
integral and it is computationally expensive for direct computation. In the globally
identifiable cases, the posterior PDF given large
 volume
 of data can be approximated
by a Gaussian distribution, so the evidence p D|C j can be approximated by using
Laplace’s method for asymptotic expansion [42]:
      Nj   − 1
p D|C j ≈ p D|θ∗ , C j p θ∗ |C j (2π ) 2 H j θ∗  2 , j = 1, 2, . . . , Nc (1.11)

where N j is the number of uncertain parameters

 for model class C j , θ∗ is the
optimal parameter vector that maximizes p θ|D, C j . The maximum likelihood
value p D|θ∗ , C j is larger for the model class that can better fit the data. Thus, the
model class with more uncertain parameters is more likely to have greater maximum
likelihood value. The remaining terms in Eq. (1.11) are called Ockham factor [43]:
  Nj   − 1
O j = p θ∗ |C j (2π ) 2 H j θ∗  2 (1.12)

The Ockham factor provides a measurement for the robustness of the model class
and its value decreases exponentially with the number of uncertain parameters in the
model class.  
On the other hand, the exact expression for p D|C j can be obtained based on
the Bayes’ theorem:
   
  p D|θ∗ , C j p θ∗ |C j
p D|C j =   (1.13)
p θ∗ |D, C j
   
where the ratio p θ∗ |C j p θ∗ |D, C j is the Ockham factor.
Given the same prior plausibility for all model class candidates, the model class
with the largest value in Eq. (1.11) is regarded to have a good tradeoff between the
data-fitting capability and the robustness to model error. Once again, the updated or
Another Random Scribd Document
with Unrelated Content
 CHAPTER IV.
A DOOR OPENED TO WICKEDNESS.

Soon after daybreak, upon the morning following the occurrence of

the incidents related in the preceding chapter, Lady Wynde’s gray
companion departed from Hawkhurst for Canterbury in a dog-cart
which, with its driver, the baronet’s wife had ordered to be always at
Artress’ disposal. She took the early train up to London, her business
a secret between her mistress and herself.
At the usual breakfast hour, eight o’clock, Lady Wynde descended to
the breakfast room. Sir Harold was already there, and greeted her
with his usual tender smile, although he looked somewhat careworn.
Their greetings were scarcely over, and the couple had taken their
places at the table, when the butler appeared, bringing in the
morning mail bag.
Sir Harold produced his key and unlocked it. There were a few
newspapers for himself, some packets of silk samples, and a letter
from Madame Elise, her dressmaker, for Lady Wynde. There were
two letters for the baronet, one quite unimportant, which he tossed
aside. The other bore the Indian post-mark.
“A letter from George,” said Sir Harold, his eyes brightening. “No, it’s
not from George. The address is not in his hand. Who can have
written to me in his stead?”
He tore open the letter hastily, his countenance falling.
His first glance was at the date; his second at the signature. An
exclamation broke from his lips as he read aloud the name
appended to the letter: “Cooper Graham, Regimental Surgeon.”
“What can this mean?” he exclaimed, in sudden agitation. “Can
George be ill? Octavia, read the letter to me. The words seem all
blurred.”
Lady Wynde took the letter, reading it aloud.
It was long, too long to transcribe here, and its import was terrible to
the baronet. It opened with the announcement that the writer was the
surgeon of Captain Wynde’s regiment, and that Captain Wynde was
a patient under his care. It went on to say that Captain Wynde was
the victim of a terrible and incurable disease under which he had
been suffering for months, and the surgeon had learned that the
poor young man had not written home to his friends the fact of his
peril. His disease was a cancer, which was preying upon his vitals.
Captain Wynde had been relieved of his regimental duties, and sent
up into the hill country, where he now was. The young man’s
thoughts by day and night were of his home—his one longing was to
see his father before he died. Surgeon Graham went on to say that
Captain Wynde could not possibly survive a sea journey; that he
could not bear the bracing sea air, nor the fatigues of the overland
route, and he would assuredly die on his way home. But, he added,
that in the cool and quiet seclusion of his upcountry bungalow, his
life could probably be prolonged for some three months.
Surgeon Graham concluded his startling letter with a further
reference to Captain Wynde’s anxiety to look once more on his
father’s face before he died. He said that the poor young man had
desired that the letter should not be written to Sir Harold, and that
the baronet should be informed of his son’s illness only in the letter
which should announce that son’s death.
This terrible news was a fearful shock to Sir Harold. His son George,
the heir of his name and estates, was dying in a far, foreign land,
with a frightful disease, with no relative nor friend about him to
smooth his pillow in his last agony, or to wipe the death-damp from
his brows. The father sobbed aloud in his agony.
“My boy! my poor boy!” he cried, in a broken voice. “My poor dying
boy!”
“It is very sad,” said Lady Wynde, wondering in her own heart if
George Wynde’s death could be made to benefit her pecuniarily.
“The surgeon seems a very kind-hearted person, and he says that
George has an excellent native nurse, George’s man-servant—”
Sir Harold interrupted his wife by a gesture of impatience.
“The man is a Hindoo,” he said. “What consolation can he offer
George in the hour of his death, when his eyes should rest on a
tender, loving face—when his dying hands should grasp the hands of
a friend? My poor brave boy! How could I ever consent to his going
out to India? All his bright, military genius, all his longings to
distinguish himself in the army, must end in an early Indian grave!
But he shall not die with not one of his kindred beside him. We must
go to him, Octavia. We shall reach him in time.”
Sir Harold seized upon his unopened Times, and glanced over the
advertisements.
“A steamer sails from Marseilles two days hence,” he announced.
“We must be off to-day, immediately, to catch it. I will have a bag
packed at once. Order your maid to pack your trunks, Octavia—”
He paused, not comprehending the surprised stare in her ladyship’s
bold black eyes.
“You seem to be laboring under a mistake Sir Harold,” said Lady
Wynde, coolly. “If you choose to go out to India, you can do so.
George is your son and heir, and I suppose it would really look better
if you were to go. But as to my hurrying by sea and land, by day and
night, to witness the death of a young man I never saw, the idea is
simply preposterous. My health could never endure the strain of
such a fatigue. You would have two graves to make instead of one.”
The lines in Sir Harold’s face contracted as in a sudden spasm.
“I—I was selfish to think of your going, Octavia,” he said sorrowfully.
“It is true that we should have to travel day and night to reach
Marseilles in time to catch the steamer. The passage of the Red Sea
would also be hard for you. But I was thinking of my poor brave boy
dying there among strangers, with no woman beside him. If—if you
could have gone to him, my wife, and let him feel that he was going
from one mother here to another mother there—”
“I should like to go, if only my health would permit,” sighed Lady
Wynde. “But why do you not take your daughter with you?”
The father shook his head.
“She is so young,” he said. “She is so fond of poor George. I cannot
cast so heavy a shadow over her future life as that visit to her
brother’s death-bed would be. No, Octavia, I will go alone.”
He arose and went out, leaving his breakfast untouched. Lady
Wynde sipped her coffee leisurely, and ate her breakfast with
untroubled appetite. Then she proceeded to her own private sitting-
room and took her place at one of the windows, watching the
whirling snow-flakes of the February storm.
Sir Harold found her here when he came in, dressed for his journey.
He had ordered a carriage, which was ready. His travelling bag was
packed, and had been taken below. He had come in to say good-bye
to his wife.
“What a great change a single hour has wrought in our lives!” he
said, as he came up to Lady Wynde and put his arms around her.
“Octavia, my darling, it wrings my heart to leave you. Write to me by
every post. I shall remain with my boy until all is over. Tell me all the
home news. You will have Neva home at Easter, and love her for my
sake! She will be our only child soon!”
He embraced his wife with passionate affection, and murmured
words of anguished farewell. He tore himself from her, but at the
door he turned back, and spoke to her with a solemnity she had
never seen in him before.
“Octavia,” he said, “at this moment a strange presentiment comes
over me—a sudden horror—a chill as of death! Perhaps I am to die
out there in India! If—if anything happens to me, Octavia, promise
me to be good to my Neva.”
“It is not necessary to promise,” said Lady Wynde, “but to please
you, I promise!”
Sir Harold’s keen blue eyes, full of anguish, rested in a long steady
gaze upon that false handsome face, and the solemnity of his
countenance increased.
“You will be Neva’s guardian, if I die,” he said, in a broken voice. “I
trust you absolutely. God do unto you, Octavia, as you do unto my
orphan child!”
How those words rang in the ears of Lady Wynde long afterward!
Sir Harold gave her a last embrace, and dashed down the stairs and
sprang into the carriage. Lady Wynde watched him with tearless
eyes as he drove down the avenue.
When he had disappeared from her sight, she said to herself:
“Of course I could have done nothing to put an end to Sir Harold’s
life this morning. I only hope he will die in India—to save me the
trouble of—of doing anything when he gets back!”
Sir Harold proceeded to Canterbury with all speed. On arriving, he
proceeded directly to his solicitor’s, had a new will drawn up,
constituting Lady Wynde his daughter’s personal guardian, and
making Neva his sole heiress in the event of her brother’s death,
Lady Wynde having been sufficiently provided for by her marriage
settlements. The will duly signed and witnessed, Sir Harold hastened
to the station, catching the train for Dover.
He crossed to Calais by the first boat, and went on to Marseilles, by
way of Paris, without stopping even to see his daughter. He was not
only in time to get passage by the Messageries Imperiales steamer,
but had an hour to spare. In this hour he wrote a long and very
tender letter to his daughter, telling her of her brother’s illness, and
hinting of the gloom that had settled down upon his own soul. He
begged her if anything happened to him on this journey, to love her
step-mother, and to obey her in all things, regarding Lady Wynde’s
utterances as if they came from Sir Harold.
He also wrote a note to his wife, and sent the two ashore to be
posted by one of the agents of the company, just as the vessel
weighed anchor for Suez.
In thirty-five days after leaving home he was in the Indian hill country,
and beside his dying son.
Lady Wynde went out very little after her husband’s departure. She
gave no more dinner parties, and behaved with such admirable
discretion that her neighbors were full of praises of her. Although
young, handsome and admired, presiding over one of the finest
places in the county, with no one to direct or thwart her movements,
the most censorious tongue could find nothing to condemn in her.
The only recreation she allowed herself were her weekly visits to
London, ostensibly to see Madame Elise, but as the ashen-eyed
Artress always accompanied her, they excited no comment even in
her own household.
Easter drew near, and Lady Wynde wrote to her step-daughter that it
would not be convenient to have her at Hawkhurst during the
holidays, and ordered her to remain at her school.
The spring months passed slowly. Lady Wynde wrote by every post
to her husband, and received letters as frequently. George’s
minutest symptoms were described to her by the anxious father, and
George himself, looking at his step-mother through his father’s eyes,
sent her loving and pathetic messages, to which she duly
responded.
Thus the time wore on until the midsummer.
About the middle of July, Lady Wynde received a black-bordered
letter from her husband stating that his son and heir was dead. He
had died at his up-country bungalow, after an illness which had been
protracted considerably beyond the anticipations of his surgeon. Sir
Harold wrote that he was exhausted by long nursing, and that he
should remain a fortnight longer at his son’s bungalow to recruit his
own health, and that he should then start for home.
“I wish he would come,” said Lady Wynde discontentedly, to her gray
companion. “I am tired of this dull existence. I am anxious to rid
myself of the trammels of my present marriage, and to be free to
marry again.”
“You can be free within a week after Sir Harold’s return,” said
Artress. “And he will be here in September.”
“I shall be free in September,” mused Lady Wynde, with sparkling
eyes. “A widow with four thousand a year! Ah, if only some good
demon would bring about that happy fact, leaving my hands
unstained with crime?”
It seemed as if her familiar demon had anticipated her prayer.
Some two weeks later, a second black-bordered letter was brought
to Lady Wynde. It was in an unfamiliar handwriting, and proved to be
from Surgeon Graham.
It announced the death of Sir Harold Wynde!
The surgeon stated that the baronet had made all arrangements for
returning to England, and that he had gone for a last ride among the
hills. He had taken a jungle path, but being well armed and attended
by a Hindoo servant, had anticipated no trouble. Some hours after
he had set out on his ride, about the time the surgeon looked for his
return, the Hindoo servant, covered with dust, rode up alone in a
very panic of terror. With difficulty he told his story. Sir Harold Wynde
had been attacked by a tiger that had leaped upon him from the
jungle, and before his terrified servant could come to his aid, he had
been dragged from his saddle, with the life-blood welling from his
torn throat and breast. The servant, appalled, had not dared to fire,
knowing that no human power could help Sir Harold in his extremity,
and the baronet had been killed before his eyes. The Hindoo had
then fled homeward to tell the awful story.
The surgeon added, that a party had been made up to visit the
scene of the tragedy. A pool of blood, fragments of Sir Harold’s
garments, the bones of his horse, and the foot-prints of a tiger, all
tended to the confirmation of the Hindoo’s story. A hunt was
organized for the tiger, and he was found near the same spot on the
following day and killed.
We have given a brief epitome of the letter that declared to Lady
Wynde that her prayer was answered, and that she was a widow.
She was sitting in the drawing-room at Hawkhurst when the letter
was brought in to her. She was still sitting there, the letter lying on
her lap, twice read, when her gray companion stole into the room.
“A letter from Sir Harold, Octavia?” said Artress, glancing at the
black-bordered missive.
“No, it is from that Surgeon Graham,” answered her ladyship, with an
exultant thrill in her low, soft voice. “You cannot guess the news,
Artress. Sir Harold is dead!”
“Dead?”
“Yes,” cried Lady Wynde, “and I am a widow. Is it not glorious? A
widow, well-jointured and free to marry again! Ha, ha! Tell the
household the sad news, Artress, and tell them all that I am too
overcome with grief to speak to them. Let the bell at the village be
set tolling. Send a notice of the death to the Times. I am a widow,
and the guardian of the heiress of Hawkhurst! You must write to my
step-daughter of her bereavement, and also drop a note to Craven.
A widow, and without crime. The heiress of Hawkhurst in my hands
to do with as I please! Your future is to be linked with mine, my
young Neva, and a fate your father never destined for you shall be
yours. I stand upon the pinnacle of success at last.”
 CHAPTER V.
SETTLING INTO HER PLACE.

The announcement of Sir Harold Wynde’s death in India, so soon

too after the death of his son and heir, produced a shock throughout
his native county of Kent, and even throughout England; for,
although the baronet had been no politician, he had been one of the
best known men in the kingdom, and there were many who had
known and esteemed him, who mourned deeply at his tragic fate.
The London papers, the Times, the Morning Post, and others, came
out with glowing eulogies of the grand-souled baronet whose life had
been so noble and beneficent. The local papers of Kent copied these
long obituaries, and added thereto accounts of the pedigree of the
Wynde family, and a description of the young heiress upon whom, by
the untimely deaths of both father and brother, the great family
estates and possessions, all excepting the bare title, now devolved.
The retainers of the family, the farmers and servants—those who
had known Sir Harold best—mourned for him, refusing to be
comforted. They would never know again a landlord so genial, nor a
master so kindly: and although they hoped for much from his
daughter, yet, as they mournfully said to each other, Miss Neva
would marry some day, and the chances were even that she would
give to Hawkhurst a harsh and tyrannical master.
The little village of Wyndham, near Hawkhurst, the very ideal of a
Kentish village, had been mostly owned by Sir Harold Wynde. To him
had belonged the row of shops, the old inn with its creaking sign,
and most of the neat houses that stood in gardens along the single
street. It was Sir Harold who had caused to be built the little new
stone church, with its slender spire, and in this church the mourning
villagers gathered to listen to the sermon that was preached in
commemoration of the baronet’s death.
Lady Wynde was not present to listen to this sermon. Her gray
companion, attired in deep mourning, with the entire household of
Hawkhurst, was there, and the young clergyman made a feeling
allusion to “the bereaved young widow, sitting alone in her darkened
chamber and weeping for her dead, refusing like Rachel of old, to be
comforted.” Many of the kindly women present shed tears at this
picture, but Artress smiled behind her double mourning vail. She
knew that Lady Wynde was lying upon a sofa in her luxurious sitting-
room at Hawkhurst, busy with a French novel, and she knew also
that not one tear had dimmed her ladyship’s black eyes since the
news had come of Sir Harold’s horrible fate.
Neighbors and friends thronged to Hawkhurst to offer their
condolences to the young widow. For the first week she was
reported inconsolable, and refused to see any one; but a box of the
most elegant and fashionable mourning having come down from
London, Lady Wynde began to receive her visitors. She affected to
be quite broken down by her bereavement, and for weeks did not go
out of doors. And when, finally, being urged to take care of her health
and to become resigned to her loss, she took morning drives, her
equipage looked like a funeral one, her carriage and horses being
alike black, and her own face being shrouded in double folds of
sombre crape.
Artress had written to Sir Harold’s daughter immediately upon the
arrival of the news of Sir Harold’s death, but the letter had been cold
and practical, and contained merely the terrible announcement,
without one line to soften its horror. About a week later, no letter
having been received from Neva, Lady Wynde wrote a very pathetic
letter, full of protestations of sympathy, and setting forth her own
mock sorrow as something genuinely heart-rending, and declaring
herself utterly prostrated in both body and mind. Her ladyship offered
her condolences to the bereaved daughter, assuring her that
henceforth they “must be all the world to each other,” and concluded
her letter by the false statement that it had been the late Sir Harold’s
wish that his daughter should remain at her Paris school a year
longer, and, as the wishes of the dead are sacred, Lady Wynde had
sacrificed her own personal feelings in the matter, and had
consented that Neva should remain another year “under the care of
her excellent French teachers.”
“That disposes of the girl for a year,” commented Lady Wynde, as
she sealed the missive. “I won’t have her here to spy upon me until
the year of mourning is over, and I am free to do just as I please.”
So the letter was dispatched, and the baronet’s daughter was
condemned to continue her school tasks, even though her heart
might be breaking. There was no leisure for her in which to weep for
the fate of her noble father; no one who had known him with whom
she might talk of him; and only in the long and lonely night times was
she free to weep for him, and then indeed her pillow was wetted with
her tears.
About three weeks after the receipt of the letter from India
announcing Sir Harold’s death, the baronet’s solicitor at Canterbury
received a note from the widow, requesting him to call at Hawkhurst
on the following day. He obeyed the summons, bringing with him a
copy of Sir Harold’s will, made, as will be remembered, upon the day
of the baronet’s departure from England. Lady Wynde, clad in the
deepest weeds of woe, and attended by Artress, also in mourning,
received the solicitor in the library, a grand apartment with vaulted
ceiling, and lofty walls lined with books in uniform Russia leather
bindings.
“I have sent for you, Mr. Atkins,” said Lady Wynde, when the
customary greetings had been exchanged, “to learn if poor Sir
Harold left a will. I had his desk searched, and no document of the
sort can be found. If he made no will, I am anxious to know how I am
to be affected by the omission.”
Mr. Atkins, a thin, small man, with a large, bald head, looked
surprised at the simple directness of this speech. He had expected
to find her ladyship overcome with grief, as report portrayed her; but
her eyes were as bright and tearless, her cheeks as red, her features
as composed, as if the business in hand were of the most trivial and
unimportant description. Atkins, who had appreciated Sir Harold’s
grand nature, felt an aversion to Lady Wynde from this moment.
“She didn’t care for him,” he mentally decided on the instant. “She’s
an arrant humbug, and poor Sir Harold’s love was wasted on her.
Upon my soul, I believe all she cared about him was for the title and
his money.”
Lady Wynde’s sharp eyes did not fail to perceive the unfavorable
impression she had made. She bit her lip fiercely, and her cheeks
flushed hotly. Her brows arched themselves superciliously, and Mr.
Atkins, marking her impatience, hastened to answer:
“Sir Harold left a will, my lady. It was drawn up at my office at
Canterbury upon the day on which he left England for India. You will
remember that he left Hawkhurst in the morning and drove to
Canterbury. He came direct to my office, and dictated and signed his
will. He then proceeded directly to the station and went by train to
Dover, and crossed to Calais. The will was left in my keeping and is,
there can be no question, the last will and testament of Sir Harold
Wynde.”
“I presume no one will care to question the will,” said Lady Wynde
coldly, “although Sir Harold was in a very excited frame of mind that
morning, on account of the news of his son’s illness, and the pain of
leaving his home and me. Nevertheless, I dare say he was quite
competent to dictate a will. I sent you the particulars of Sir Harold’s
death, with some of the letters detailing the sad event which I have
received from India. There being no possible doubt of his awful fate,
it is time to prove his will. I wish you to give me some idea of its
contents.”
The solicitor drew out a long leathern pocket-book and took from it a
neatly folded paper.
“I have here a copy of the will,” he said briefly. “Is it your ladyship’s
wish to have the will formally read, in the presence of witnesses?”
“No, that is unnecessary. Leave out the usual useless preamble and
tell me what disposition my husband made of his property—the
freehold farms, the money in bank, the consols, the bonds and
mortgages? All these he was free to leave to whom he pleased. I
desire to know to whom he did leave them.”
There was a greediness in the looks and tones of Lady Wynde that
chilled Atkins. In her anxiety to learn the contents of the will, her
ladyship half dropped her mask and displayed something of her true
character, and he was quick to read it.
“Sir Harold Wynde, in expectation of the death of his son and heir,”
replied Atkins, in his most formal tones, “bequeathed all the property
you have mentioned, all his real and personal property, to his
daughter, Miss Neva Wynde.”
“All to her?” muttered Lady Wynde. “All, you say?”
“All, my lady. Miss Wynde also inherits Hawkhurst and the entailed
property. She is one of the richest heiresses in England.”
“And—and my name is not mentioned?”
“Sir Harold declares that you are provided for by the terms of the
marriage settlement. You have Wynde Heights for your dower house
and four thousand pounds a year during your life, with no restrictions
in regard to a second marriage—a very liberal provision I consider
it.”
“And a very shabby one I consider it,” cried Lady Wynde, with a
black frown. “Sir Harold’s daughter seventy thousand pounds a year,
and I have a paltry four. It is a shame, a miserable, burning shame!”
“It is unjust, scandalous!” muttered Artress.
“Sir Harold thought the sum sufficient, and I must say I agree with
him,” declared Atkins. “Your ladyship was contented with the
provision at your marriage. If the allowance was unsatisfactory, why
did you not expostulate with Sir Harold at that time? Why wait until
he is dead to accuse him of injustice?”
“We will not argue the matter,” said Lady Wynde superciliously. “I
shall not contest the will. And now about my rich young step-
daughter. Who are her appointed guardians?”
There was a perceptible anxiety in her manner, which Atkins noticed
with some wonder. He referred to his copy of the will, which was
open in his hands.
“Sir Harold appointed yourself, my lady, the personal guardian of his
daughter,” he said slowly. “Miss Wynde is to reside at Hawkhurst
under your care until she becomes of age or marries. Upon the
occurrence of either of those events your ladyship is to retire to
Wynde Heights, or to whatsoever place you may prefer, leaving Miss
Wynde absolute mistress of Hawkhurst. Of course if Miss Wynde
desires you to remain after her marriage, or the attainment of her
majority, you are at liberty to do as you please. I think you
comprehend Sir Harold’s meaning. If it is not precisely clear, I will
read the will—”
“Do not!” interrupted Lady Wynde impatiently. “I abhor all that tedious
phraseology. I understand that I am Miss Wynde’s sole personal
guardian, that I am to direct her actions, introduce her into society,
and that she is to give me the simple, unhesitating obedience of a
daughter. Is this not so?”
“It is,” assented Atkins, rather hesitatingly. “Sir Harold expresses the
hope that his widow and his daughter will love each other; and that
your ladyship will give to his orphan child a mother’s tenderness and
affection.”
“Sir Harold knew that he could depend upon my kindness to his
child,” said Lady Wynde hypocritically. “I promised him before he
went away to be a mother to her, although I shall be but a young
mother, to be sure. I shall be very good to the poor girl, whom I love
already. I don’t know anything about law, Mr. Atkins, but is not some
other guardian also necessary—some one to see to the property,
you know?”
“There are three trustees appointed to look after the estate during
Miss Wynde’s minority,” answered Atkins. “Sir John Freise is one.
You know him well, my lady, and a more incorruptible, honest-souled
gentleman than he does not exist. He is a man of fine business
capacity, and Sir Harold could not have chosen better. I am also a
trustee, and I can answer for my own probity, and for my great
devotion to the interests of Miss Wynde.”
“And the third trustee—who is he?”
“The young Earl Towyn. He is the son of one of Sir Harold’s dearest
friends, as you probably know, and his youth admirably balances Sir
John’s age.”
Lady Wynde looked thoughtful. Her gray companion bent over her
work, embroidering a black monogram upon a black-bordered
handkerchief, and did not look up. Her ashen-hued lashes lay on her
ashen cheeks, and she looked dull, spiritless, a mere gray shadow,
as we have called her, but Atkins, studying her face, had an
uncomfortable impression that under all that coldness a fire was
burning.
“She’s more than she looks to be,” he thought keenly. “I wonder Sir
Harold tolerated her in his house. How singularly she resembles a
cat!”
Lady Wynde presently broke the silence.
“I understand the situation of affairs,” she said, “and I am obliged to
you for your prompt attendance upon my summons, Mr. Atkins. I
shall leave my money affairs in your hands. I desire my jointure to be
paid into the bank and placed to my credit, so that I may draw upon it
when I please. There is nothing more, I think.”
“I would like to make a few inquiries about Miss Wynde, if you
please, my lady,” said Atkins, with quiet firmness. “I understand that
she is not at home. Has she not been summoned from her school
since her father’s death?”
“She has not,” answered Lady Wynde haughtily.
“Pardon me, madam, but are you not about to summon her?”
“I am not. Miss Wynde will remain this year at school. Her studies
must be interrupted upon no account at this time.”
“Not even by her father’s death?” asked Atkins bitterly. “Sir Harold
mentioned to me his desire to have her at home—”
“Sir Harold Wynde is no longer master of Hawkhurst,” interposed
Lady Wynde, with increased superciliousness. “I believe, by the
terms of the will, that I am mistress here during Neva’s minority. Let
me tell you, Mr. Atkins, that I am my step-daughter’s sole personal
guardian, and that I will submit to no dictation whatever in my
treatment of the girl. If my husband had sufficient confidence in me
to make me his daughter’s guardian, the trustees whom he himself
appointed have no need nor right to comment upon my actions or
interfere in my plans. Permit me to assure you that I will brook no
interference, and if you try to sow dissension between Neva and me
you are proving unfaithful to Sir Harold—as well as oblivious of your
own interests.”
Mr. Atkins sighed, and murmured an apology. He soon after took his
leave, and drove away in the chaise in which he had come. His heart
was very heavy and his face overcast as he emerged from the
Hawkhurst grounds into the highway, and journeyed toward
Canterbury.
“It was a sorry day for Neva Wynde when her father died,” he
murmured, looking back at the grand old seat—“a sorry day! This
handsome black-eyed Lady Wynde, that everybody is praising for an
angel of love and devotion to her husband, is at heart a demon! She
means mischief, though I can’t see how. Poor Neva is booked for
trouble!”
Enough of honest Mr. Atkins’ sentiments had been apparent in his
countenance to prejudice Lady Wynde against him, and to warn her
that he comprehended something of her real character. As may be
supposed, therefore, she did not again summon him to Hawkhurst.
The days and weeks and months of Lady Wynde’s widowhood
passed on without event. She carried herself circumspectly in the
eyes of the world. No visitors were invited to Hawkhurst, and her
ladyship’s visits to London were few and far between. She seldom
went to Canterbury, and her drives about the neighborhood of
Hawkhurst were always of the most funereal description, with black
coach, black horses and black attire, and a slow gait. Her ladyship
was found every Sunday in the baronet’s great square pew in the
little Wyndham church, and as she always sat with the silken
curtains drawn, no one could know that she was not absorbed in the
church services. In short, during the year she had determined to
devote to mourning for her dead husband, the conduct of Lady
Wynde was such as to deepen her popularity throughout the county.
Sir John Freise enthusiastically declared her an angel, her neighbors
praised her, and only honest Mr. Atkins shook his head doubtfully
when her virtues were lauded, and dared to suggest that she might
not be all she seemed.
The year slowly wore away, and midsummer had come again. The
languor of Lady Wynde’s dull existence had begun to give place to a
strange restlessness. Her deep mourning had grown odious in her
sight, and was replaced by the lovely combinations of white and
black, the delicate lavenders and soft gray hues which are supposed
to indicate a mitigated grief. The hideous widow’s cap, not at all
becoming to her ladyship, was exchanged for lavender ribbons in her
hair, and jewels took the place of the orthodox mourning ornaments
of jet. In her “half mourning,” Lady Wynde appeared more than ever
a strikingly handsome woman.
“Artress,” she said one morning to her gray companion, as she
looked out of her sitting-room window upon the fair domain of
Hawkhurst, “this dreaded year is over at last. I have satisfied the
demands of society; I have hoodwinked the jealous and envious
eyes of neighbors, and am free at last. If I were to marry to-morrow,
no one could say that I had not treated the memory of Sir Harold
Wynde with respect. With the sacrifice of but little over two years of
my life, I have won a fine income, a splendid home during Neva’s
minority, and the guardianship of one of the greatest heiresses in
England. That office is worth three thousand a year to me while I
hold it. Surely I have played my part well.”
“You have indeed,” echoed Artress.
“Neva must come home soon, but my own business must be settled
before her advent on the scene. I shall write to Craven immediately. I
will have no further delay.”
She went to a small, beautifully inlaid writing desk, which stood in a
recessed window, and sitting down by it, wrote upon heavy velvet
paper the following words:
 “Craven: You may come to me at last. There is no further obstacle
between us.
“Octavia.”
This brief missive she inclosed in a square envelope, and stamped
with pale green wax and her favorite device.
The letter she addressed to The Hon. Craven Black, The Albany,
London, W.
She then touched her bell. To the servant who came at her summons
she gave the letter, ordering it to be posted at Wyndham village
without delay. When her messenger had gone, her ladyship gave a
sigh of consent, and murmured:
“I am about to reap the reward of all my schemes. Craven will be
here to-morrow!”
 CHAPTER VI.
HER LADYSHIP’S ACCOMPLICE.

The morrow to which Lady Wynde looked forward with feverish

expectation dawned at last, bright and clear, and deepened into a
sultry afternoon. The baronet’s widow spent hours at her toilet, and
the effect of her labors was satisfactory to her. She surveyed her
reflection in a full-length mirror in her dressing-room with a smile of
complacency. Her black hair was arranged in braids, curls, and finely
crimpled waves, after the fashion of the day, and in the midst of its
prodigal luxuriance, above her forehead, a jeweled spray flashed
and glittered. Her dress, made low in the neck and short in the
sleeves, to display her finely rounded shoulders and arms, was of
lustrous silk of lavender hue, and was draped with a black lace
overskirt. A necklace and bracelets incrusted with diamonds added
brilliancy to her appearance. Her liquid black eyes shone and
glittered; her cheeks were red as damask roses; she had never
looked half so handsome in the days when she had fascinated Sir
Harold Wynde and made him adore her.
She had dismissed her maid, and was giving a last touch to the short
curls that dropped over her forehead, while she talked with Artress,
when wheels were heard coming up the drive. The gray companion
flitted to a shuttered window and peeped out. A cab was
approaching the house, and a man’s head was protruded from the
window. His face was half averted, as he apparently studied the
exterior of the dwelling, but Artress knew him. She glided back to
Lady Wynde with the words:
“He has come!”
A sudden agitation seemed to convulse the soul of the baronet’s
widow. A sudden paleness swept over her face. She leaned heavily
upon the back of a chair, and stood there motionless until a servant
brought up a silver tray on which lay a large square card with the
inscription, “The Honorable Craven Black,” and announced that the
gentleman had been shown into the drawing-room. Then her
ladyship started abruptly, the color returning to her face in ruddy
waves.
“Come, Artress,” she said, “we will go below. Yet stay. You may delay
your coming for half an hour. Surely no one can find fault with me for
seeing him alone a little while. Since I became a widow for the
second time, I have felt as if I lived in a glass lantern with the eyes of
all Kent upon me. Yet there is no need of carrying my caution too
far.”
She gave a last glance at her reflection in the mirror, a last deft touch
to her attire, and then swept from the room down the stairs, and
slowly entered the drawing-room.
A gentleman within arose from his seat, and came forward with
outstretched hands and eager face. He was tall, handsome, fair-
haired, with light eyes full of sinister gleams, and his full, sensual lips
wore even now a cynical smile that appeared habitual to them.
He was the same man who had watched, from the pier head at
Brighton, the rescue of Octavia Hathaway from the sea by Sir Harold
Wynde—the same man who had witnessed the marriage of the
baronet and the widow from behind a clustered pillar in the church,
and whose sinister comments, as he emerged into Hanover Square,
we have chronicled.
His quick glance swept the form and face of Lady Wynde; a look of
admiration burned in his eyes. He held out his arms. With a joyous
cry, the handsome widow sprang forward, and was clasped in his
embrace.
“At last! At last!” she murmured.
“Yes, at last!” echoed Mr. Black, in tones of exultation. “Nothing
stands between us now, Octavia! We have lost nothing by waiting.
We have been guilty of no crime, and fate itself has played into our
hands. And you, Octavia, in the prime of your beauty, are more
magnificent than ever.”
He drew her to a sofa and clasped an arm around her waist. Her
head drooped to his shoulder. The flush of intense joy mantled her
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

textbookfull.com