Experience Seamless Full Ebook Downloads for Every Genre at textbookfull.

com

Reliability engineering. Probabilistic models and

maintenance methods Second Edition Nachlas

https://textbookfull.com/product/reliability-engineering-
probabilistic-models-and-maintenance-methods-second-edition-
nachlas-2/

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.

Reliability engineering. Probabilistic models and

maintenance methods Second Edition Nachlas

https://textbookfull.com/product/reliability-engineering-
probabilistic-models-and-maintenance-methods-second-edition-nachlas-2/

textboxfull.com

Stochastic Models in Reliability Engineering (Advanced

Research in Reliability and System Assurance Engineering)
1st Edition Lirong Cui (Editor)
https://textbookfull.com/product/stochastic-models-in-reliability-
engineering-advanced-research-in-reliability-and-system-assurance-
engineering-1st-edition-lirong-cui-editor/
textboxfull.com

Reasoning with Probabilistic and Deterministic Graphical

Models Exact Algorithms Second Edition Rina Dechter

https://textbookfull.com/product/reasoning-with-probabilistic-and-
deterministic-graphical-models-exact-algorithms-second-edition-rina-
dechter/
textboxfull.com

Reliability Based Aircraft Maintenance Optimization and

Applications A volume in Aerospace Engineering He Ren

https://textbookfull.com/product/reliability-based-aircraft-
maintenance-optimization-and-applications-a-volume-in-aerospace-
engineering-he-ren/
textboxfull.com
Graph Searching Games and Probabilistic Methods 1st
Edition Bonato

https://textbookfull.com/product/graph-searching-games-and-
probabilistic-methods-1st-edition-bonato/

textboxfull.com

Learning Probabilistic Graphical Models in R 1st Edition

David Bellot

https://textbookfull.com/product/learning-probabilistic-graphical-
models-in-r-1st-edition-david-bellot/

textboxfull.com

Probabilistic Methods in Geotechnical Engineering:

Proceedings of the conference, Canberra, 10-12 February
1993 First Edition Li
https://textbookfull.com/product/probabilistic-methods-in-
geotechnical-engineering-proceedings-of-the-conference-
canberra-10-12-february-1993-first-edition-li/
textboxfull.com

Advanced Maintenance Policies for Shock and Damage Models

1st Edition Xufeng Zhao

https://textbookfull.com/product/advanced-maintenance-policies-for-
shock-and-damage-models-1st-edition-xufeng-zhao/

textboxfull.com

Asset maintenance engineering methodologies First Edition

Farinha

https://textbookfull.com/product/asset-maintenance-engineering-
methodologies-first-edition-farinha/

textboxfull.com
RELIABILITY
ENGINEERING
Probabilistic Models and
Maintenance Methods
Second Edition
RELIABILITY
ENGINEERING
Probabilistic Models and
Maintenance Methods
Second Edition

JOEL A. NACHLAS

Boca Raton London New York

CRC Press is an imprint of the

Taylor & Francis Group, an informa business
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2017 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S. Government works

Printed on acid-free paper

Version Date: 20161019

International Standard Book Number-13: 978-1-4987-5247-3 (Pack - Book and Ebook)

This book contains information obtained from authentic and highly regarded sources. Reasonable
efforts have been made to publish reliable data and information, but the author and publisher cannot
assume responsibility for the validity of all materials or the consequences of their use. The authors and
publishers have attempted to trace the copyright holders of all material reproduced in this publication
and apologize to copyright holders if permission to publish in this form has not been obtained. If any
copyright material has not been acknowledged please write and let us know so we may rectify in any
future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information stor-
age or retrieval system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, please access www.copy-
right.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222
Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that pro-
vides licenses and registration for a variety of users. For organizations that have been granted a photo-
copy license by the CCC, a separate system of payment has been arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are
used only for identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
Dedicated to the memory of Betty Nachlas
Contents

Preface ................................................................................................................... xiii

Author .....................................................................................................................xv

1 Introduction .....................................................................................................1

2 System Structures ...........................................................................................5

2.1 Status Functions .................................................................................... 5
2.2 System Structures and Status Functions ...........................................7
2.2.1 Series Systems ..........................................................................7
2.2.2 Parallel System ......................................................................... 8
2.2.3 k-out-of-n Systems .................................................................. 10
2.2.4 Equivalent Structures ............................................................ 12
2.3 Modules of Systems ............................................................................ 17
2.4 Multistate Components and Systems ............................................... 18
Exercises .......................................................................................................... 19

3 Reliability of System Structures ............................................................... 23

3.1 Probability Elements........................................................................... 23
3.2 Reliability of System Structures ........................................................ 24
3.2.1 Series Systems ........................................................................ 24
3.2.2 Parallel Systems...................................................................... 25
3.2.3 k-out-of-n Systems .................................................................. 25
3.2.4 Equivalent Structures ............................................................ 26
3.3 Modules ................................................................................................ 31
3.4 Reliability Importance ........................................................................ 32
3.5 Reliability Allocation .......................................................................... 35
3.6 Conclusion............................................................................................ 36
Exercises .......................................................................................................... 37

4 Reliability over Time ................................................................................... 39

4.1 Reliability Measures ........................................................................... 39
4.2 Life Distributions ................................................................................44
4.2.1 Exponential Distribution ...................................................... 45
4.2.2 Weibull Distribution.............................................................. 46
4.2.3 Normal Distribution .............................................................. 49
4.2.4 Lognormal Distribution ........................................................ 51
4.2.5 Gamma Distribution ............................................................. 52
4.2.6 Other Distributions ............................................................... 52
4.3 System-Level Models ..........................................................................54
Exercises .......................................................................................................... 58

vii
viii Contents

5 Failure Processes ........................................................................................... 61

5.1 Mechanical Failure Models ............................................................... 62
5.1.1 Stress–Strength Interference ................................................ 62
5.1.2 Shock and Cumulative Damage ..........................................64
5.2 Electronic Failure Models .................................................................. 71
5.2.1 Arrhenius Model .................................................................... 71
5.2.2 Eyring Model .......................................................................... 72
5.2.3 Power Law Model .................................................................. 72
5.2.4 Defect Model........................................................................... 72
5.3 Other Failure Models.......................................................................... 73
5.3.1 Diffusion Process Model....................................................... 73
5.3.2 Proportional Hazards ........................................................... 78
5.3.3 Competing Risks ....................................................................80
Exercises ..........................................................................................................83

6 Age Acceleration ...........................................................................................85

6.1 Age Acceleration for Electronic Devices .......................................... 87
6.2 Age Acceleration for Mechanical Devices ....................................... 89
6.3 Step Stress Strategies .......................................................................... 92
6.4 Concluding Comment ........................................................................ 93
Exercises .......................................................................................................... 93

7 Nonparametric Statistical Methods .......................................................... 95

7.1 Data Set Notation and Censoring ..................................................... 96
7.2 Estimates Based on Order Statistics ................................................. 98
7.3 Estimates and Confidence Intervals ................................................. 99
7.4 Kaplan–Meier Estimates .................................................................. 102
7.4.1 Continuous Monitoring of Test Unit Status ..................... 102
7.4.2 Periodic Monitoring of Test Unit Status ........................... 105
7.5 Tolerance Bounds .............................................................................. 107
7.6 TTT Transforms ................................................................................. 109
7.6.1 Theoretical Construction .................................................... 109
7.6.2 Application to Complete Data Sets.................................... 113
7.6.3 Application to Censored Data Sets.................................... 118
7.7 Nelson Cumulative Hazard Estimation Method ......................... 122
Exercises ........................................................................................................ 124

8 Parametric Statistical Methods ................................................................ 129

8.1 Graphical Methods ........................................................................... 129
8.2 Method of Moments ......................................................................... 135
8.2.1 Estimation Expressions ....................................................... 136
8.2.2 Confidence Intervals for the Estimates ............................. 139
8.3 Method of Maximum Likelihood ................................................... 143
8.4 Maximum Likelihood Method with Data Censoring ................. 159
Contents ix

8.5 Special Topics..................................................................................... 161

8.5.1 Method of Moments with Censored Data ........................ 161
8.5.2 Data Analysis under Step Stress Testing .......................... 164
Exercises ........................................................................................................ 167

9 Repairable Systems I: Renewal and Instantaneous Repair ............... 173

9.1 Renewal Processes ............................................................................ 174
9.2 Classification of Distributions and Bounds on Renewal
Measures ............................................................................................ 181
9.3 Residual Life Distribution ............................................................... 186
9.4 Conclusion.......................................................................................... 189
Exercises ........................................................................................................ 190

10 Repairable Systems II: Nonrenewal and Instantaneous Repair ....... 193

10.1 Minimal Repair Models ................................................................... 194
10.2 Imperfect Repair Models ................................................................. 200
10.3 Equivalent Age Models .................................................................... 203
10.3.1 Kijima Models ...................................................................... 203
10.3.2 Quasi-Renewal Process ....................................................... 210
10.4 Conclusion.......................................................................................... 214
Exercises ........................................................................................................ 214

11 Availability Analysis ................................................................................. 217

11.1 Availability Measures....................................................................... 220
11.2 Example Computations ....................................................................223
11.2.1 Exponential Case ................................................................. 223
11.2.2 Numerical Case ....................................................................225
11.3 System-Level Availability ................................................................ 227
11.4 Nonrenewal Cases ............................................................................ 232
11.4.1 Availability under Imperfect Repair ................................. 233
11.4.2 Availability Analysis for the Quasi-Renewal Model ...... 235
11.5 Markov Models ................................................................................. 239
Exercises ........................................................................................................ 245

12 Preventive Maintenance............................................................................ 247

12.1 Replacement Policies......................................................................... 248
12.1.1 Elementary Models .............................................................. 248
12.1.2 Availability Model for Age Replacement ......................... 253
12.1.3 Availability Model for Block Replacement....................... 255
12.1.4 Availability Model for Opportunistic Age
Replacement ................................................................... 257
12.1.4.1 Failure Model ........................................................ 262
12.1.4.2 Opportunistic Failure Replacement Policy ...... 265
x Contents

12.1.4.3 Partial Opportunistic Age Replacement

Policy ................................................................ 268
12.1.4.4 Full Opportunistic Age Replacement Policy .... 271
12.1.4.5 Analysis of the Opportunistic Replacement
Models.................................................................... 271
12.2 Nonrenewal Models ......................................................................... 274
12.2.1 Imperfect PM Models .......................................................... 275
12.2.2 Models Based on the Quasi-Renewal Process ................. 277
12.2.3 Models Based on the Kijima Model .................................. 281
12.3 Conclusion.......................................................................................... 283
Exercises ........................................................................................................284

13 Predictive Maintenance............................................................................. 287

13.1 System Deterioration ........................................................................ 288
13.2 Inspection Scheduling ...................................................................... 289
13.3 More Complete Policy Analysis ...................................................... 290
13.4 Models and Analysis Based on Continuous Process
Monitoring ......................................................................................... 294
13.4.1 Observable Degradation Processes ................................... 294
13.4.2 Unobservable Degradation Processes ............................... 297
13.4.2.1 Time Series Methods ........................................... 298
13.4.2.2 Conditional Probability Methods ......................300
13.5 Conclusion..........................................................................................304
Exercises ........................................................................................................305

14 Special Topics .............................................................................................. 307

14.1 Statistical Analysis of Repairable System Data............................. 307
14.1.1 Data from a Single System .................................................. 307
14.1.2 Data from Multiple Identical Systems .............................. 310
14.2 Warranties .......................................................................................... 314
14.2.1 Full Replacement Warranties ............................................. 315
14.2.2 Pro Rata Warranties............................................................. 317
14.3 Reliability Growth ............................................................................ 319
14.4 Dependent Components .................................................................. 323
14.5 Bivariate Reliability........................................................................... 325
14.5.1 Collapsible Models............................................................... 326
14.5.2 Bivariate Models................................................................... 327
14.5.2.1 Stochastic Functions ............................................ 327
14.5.2.2 Correlation Models .............................................. 330
14.5.2.3 Probability Analysis............................................. 331
14.5.2.4 Failure and Renewal Models .............................. 335
Exercises ........................................................................................................ 341
Contents xi

Appendix A: Numerical Approximations ....................................................343

Appendix B: Numerical Evaluation of the Weibull Renewal
Functions ......................................................................................................... 347
Appendix C: Laplace Transform for the Key Renewal Theorem ............. 353
Appendix D: Probability Tables ..................................................................... 355
References ........................................................................................................... 359
Index ..................................................................................................................... 365
Preface

The motivation for the preparation of a second edition was my wish to expand
the treatment of several topics while maintaining an integrated introductory
resource for the study of reliability evaluation and maintenance planning.
The focus across all of the topics treated is the use of analytical methods to
support the design of dependable and efficient equipment and the planning
for the servicing of that equipment. The orientation of the topical develop-
ment is that probability models provide an effective vehicle for portraying
and evaluating the variability that is inherent in the performance and lon-
gevity of equipment.
The book is intended to support either an introductory graduate course
in reliability theory and preventive maintenance planning or a sequence of
courses that address these topics. A fairly comprehensive coverage of the
basic models and of various methods of analysis is provided. An under-
standing of the topics discussed should permit the reader to comprehend
the literature describing new and advanced models and methods.
Notwithstanding the emphasis upon initial study, the text should also
serve well as a resource for practicing engineers. Engineers who are involved
in the design process should find a coherent explanation of the reliability
and maintenance issues that will influence the success of the devices they
create. Similarly, engineers responsible for the analysis and verification of
product reliability or for the planning of maintenance support of fielded
equipment should find the material presented here to be relevant and easy
to access and use.
In preparing this second edition, the treatment of statistical methods for
evaluating reliability has been expanded substantially. Several methods for
constructing confidence intervals as part of the parametric estimation effort
are described and methods for treating data derived from operating repair-
able devices have also been added. In addition, the analysis of nonstation-
ary models of repairable equipment maintenance has been updated and
expanded. These expansions along with numerous other minor improve-
ments to the text should make this book an even more useful resource for
both students and practitioners.
The background required of the reader is a sound understanding of prob-
ability. This subsumes capability with calculus. More specifically, the reader
should have an understanding of distribution theory, Laplace transforms,
convolutions, stochastic processes, and Markov processes. It is also worth
mentioning that the use of the methods discussed in this book often involves
substantial computational effort, so facility with numerical methods and
access to efficient mathematical software is desirable.

xiii
xiv Preface

One caveat concerning the coverage here is that the treatment is strictly
limited to hardware. Reliability and maintenance models have been devel-
oped for applications to software, humans, and services systems. No criti-
cism of those efforts is intended but the focus here is simply hardware.
The organization of the text is reasonably straightforward. The elemen-
tary concepts of reliability theory are presented sequentially in Chapters 1
through 6. Following this, the commonly used statistical methods for eval-
uating component reliability are described in Chapters 7 and 8. Chapters
9 through 13 treat repairable systems and maintenance planning models.
Here again the presentation is sequential in that simple failure models pre-
cede those that include preventive actions and the renewal cases are treated
before the more realistic nonrenewal cases. In the final chapter, four inter-
esting special topics, including warranties, are discussed. It is worth noting
that four appendices that address aspects of numerical computation are pro-
vided. These should be quite useful to the reader.
Naturally, many people have contributed to the preparation of this text.
The principal factor in the completion of this book was the support and
encouragement of my wife Beverley. An important practical component of
my success was the support of Virginia Tech, especially during sabbaticals
when progress with writing is so much easier.
I acknowledge the significant computational capability provided to me by
the Mathematica software. Many of the analyses included in this text would
have been much more taxing or even impossible without the strength and
efficiency the Wolfram software provides.
I also wish to extend my thanks directly to three of my students, each of
whom contributed to my efforts. Edvin Beqari stimulated my increased inter-
est in and analysis of the diffusion models of degradation. He also directed
much of my analysis of that topic. Elliott Mitchell-Colgan helped to expand
the sets of exercises included at the end of the chapters. Paul D’Agostino
invested very many hours in verifying a majority of the complicated numeri-
cal analyses used for examples or for exercise solutions.
I express my profound gratitude to all of my graduate students who have
taught me so much about these topics over the years. May we all continue to
learn and grow and to enjoy the study of this important subject.
Author

Joel A. Nachlas received his BES from Johns Hopkins University in 1970, his
MS in 1972 and his PhD in 1976, both from the University of Pittsburgh. He
served on the faculty of the Grado Department of Industrial and Systems
Engineering at Virginia Tech for 41 years and retired in March 2016. His
research interests are in the applications of probability and statistics to prob-
lems in reliability and quality control. In addition to his normal teaching
activities during his time at Virginia Tech, he served as the coordinator for
the department’s graduate program in operations research and for their
dual master’s degree that is operated with École des Mines de Nantes in
France. From 1992 to 2011, he regularly taught reliability theory at the École
Polytechnique de l’Université Nice Sophia Antipolis. He is the coauthor of
more than 50 refereed articles, has served in numerous editorial and referee
capacities, and has lectured on reliability and maintenance topics through-
out North America and Europe.

xv
1
Introduction

Although we rarely think of it, reliability and maintenance are part of our
everyday lives. The equipment, manufactured products, and fabricated
infrastructure that contribute substantively to the quality of our lives have
finite longevity. Most of us recognize this fact, but we do not always fully
perceive the implications of finite system life for our efficiency and safety.
Many, but not all, of us also appreciate the fact that our automobiles require
regular service, but we do not generally think about the fact that roads and
bridges, smoke alarms, electricity generation and transmission devices, and
many other machines and facilities we use also require regular maintenance.
We are fortunate to live at a time in which advances in the understanding
of materials and energy have resulted in the creation of an enormous variety
of sophisticated products and systems, many of which (1) were inconceiv-
able 100 or 200 or even 20 years ago; (2) contribute regularly to our comfort,
health, happiness, efficiency, or success; (3) are relatively inexpensive; and
(4) require little or no special training on our part. Naturally, our reliance
on these devices and systems is continually increasing and we rarely think
about failure and the consequences of failure.
Occasionally, we observe a catastrophic failure. Fatigue failures of the fuse-
lage of aircraft [1], the loss of an engine by a commercial jet [1], the Three Mile
Island [1] and Chernobyl [1] nuclear reactor accidents, and the Challenger [2]
and Discovery [3] space shuttle accidents are all widely known examples of
catastrophic equipment failures. The relay circuit failure at the Ohio power
plant that precipitated the August 2003 power blackout in the northeastern
United States and in eastern Canada [4] is an example of a system failure
that directly affected millions of people. When these events occur, we are
reminded dramatically of the fallibility of the physical systems on which
we depend.
Nearly everyone has experienced less dramatic product failures such as
that of a home appliance, the wear out of a battery, and the failure of a light
bulb. Many of us have also experienced potentially dangerous examples of
product failures such as the blowout of an automobile tire.
Reliability engineering is the study of the longevity and failure of equip-
ment. Principles of science and mathematics are applied to the investiga-
tion of how devices age and fail. The intent is that a better understanding
of device failure will aid in identifying ways in which product designs can
be improved to increase life length and limit the adverse consequences of
failure. The key point here is that the focus is upon design. New product and

1
2 Reliability Engineering: Probabilistic Models and Maintenance Methods

system designs must be shown to be safe and reliable prior to their fabrica-
tion and use. A dramatic example of a design for which the reliability was
not properly evaluated is the well-known case of the Tacoma Narrows Bridge,
which collapsed into the Puget Sound in November 1940, a few months after
its completion [1].
A more recent example of a design fault with significant consequences is
the 2013 lithium-ion battery fire that occurred on a new Boeing 787 aircraft
while it was parked at the Boston airport [5]. Fortunately, the plane was
empty, so no one was injured, but the fire and two subsequent fires of the
same type resulted in all 787s being grounded until a modification to the
battery containment was made. The cost to the airlines using the planes was
estimated to be $1.1 million per day.
The study of the reliability of an equipment design also has important eco-
nomic implications for most products. As Blanchard [6] states, 90% of the life
cycle costs associated with the use of a product are fixed during the design
phase of a product’s life.
Similarly, an ability to anticipate failure can often imply the opportunity
to plan for an efficient repair of equipment when it fails or even better to per-
form preventive maintenance in order to reduce failure frequency.
There are many examples of products for which system reliability is far
better today than it was previously. One familiar example is the television
set, which historically experienced frequent failures and which, at present,
usually operates without failure beyond its age of obsolescence. Improved
television reliability is certainly due largely to advances in circuit technol-
ogy. However, the ability to evaluate the reliability of new material sys-
tems and new circuit designs has also contributed to the gains we have
experienced.
Perhaps the most well-recognized system for which preventive mainte-
nance is used to maintain product reliability is the commercial airplane.
Regular inspection, testing, repair, and even overhaul are part of the nor-
mal operating life of every commercial aircraft. Clearly, the reason for such
intense concern for the regular maintenance of aircraft is an appreciation of
the influence of maintenance on failure probabilities and thus on safety.
On a personal level, the products for which we are most frequently respon-
sible for maintenance are our automobiles. We are all aware of the inconve-
nience associated with an in-service failure of our cars and we are all aware
of the relatively modest level of effort required to obtain the reduced failure
probability that results from regular preventive maintenance.
It would be difficult to overstate the importance of maintenance and espe-
cially preventive maintenance. It is also difficult to overstate the extent to
which maintenance is undervalued or even disliked. Historically, repair and
especially preventive maintenance have often been viewed as inconvenient
overhead activities that are costly and unproductive. Very rarely have the
significant productivity benefits of preventive maintenance been recognized
and appreciated. Recently, there have been reports [7–9] that suggest that
Introduction 3

it is common experience for factory equipment to lose 10%–40% of pro-

ductive capacity to unscheduled repairs and that preventive maintenance
could drastically reduce these losses. In fact, the potential productivity gains
associated with the use of preventive maintenance strategies to reduce the
frequency of unplanned failures constitute an important competitive oppor-
tunity [9]. The key to exploiting this opportunity is careful planning based
on cost and reliability.
This book is devoted to the analytical portrayal and evaluation of equip-
ment reliability and maintenance. As with all engineering disciplines, the
language of description is mathematics. The text provides an exploration
of the mathematical models that are used to portray, estimate, and evalu-
ate device reliability and those that are used to describe, evaluate, and plan
equipment service activities. In both cases, the focus is on design. The models
of equipment reliability are the primary vehicle for recognizing deficiencies
or opportunities to improve equipment design. Similarly, using reliability
as a basis, the models that describe equipment performance as a function
of maintenance effort provide a means for selecting the most efficient and
effective equipment service strategies.
These examples of various failures share some common features and they
also have differences that are used here to delimit the extent of the analyses
and discussions. Common features are that (1) product failure is sufficiently
important that it warrants engineering effort to try to understand and con-
trol it and (2) product design is complicated, so the causes and consequences
of failure are not obvious.
There are also some important differences among the examples. Taking
an extreme case, the failure of a light bulb and the Three Mile Island reac-
tor accident provide a defining contrast. The Three Mile Island accident was
precipitated by the failure of a physical component of the equipment. The
progress and severity of the accident were also influenced by the response
by humans to the component failure and by established decision policies.
In contrast, the failure of a light bulb and its consequences are not usually
intertwined with human decisions and performance. The point here is that
there are very many modern products and systems for which operational
performance depends upon the combined effectiveness of several of (1) the
physical equipment, (2) human operators, (3) software, and (4) management
protocols.
It is both reasonable and prudent to attempt to include the evaluation of all
four of these factors in the study of system behavior. However, the focus of
this text is analytical and the discussions are limited to the behavior of the
physical equipment.
Several authors have defined analytical approaches to modeling the effects
of humans [10] and of software [11] on system reliability. The motivation for
doing this is the view that humans cause more system failures than does
equipment. This view seems quite correct. Nevertheless, implementation of
the existing mathematical models of human and software reliability requires
4 Reliability Engineering: Probabilistic Models and Maintenance Methods

the acceptance of the view that probability models appropriately represent

dispersion in human behavior. In the case of software, existing models
are based on the assumption that probability models effectively represent
hypothesized evolution in software performance over time. The appropri-
ateness of both of these points of view is subject to debate. It is considered
here that the human operators of a system do not comprise a homogeneous
population for which performance is appropriately modeled using a proba-
bility distribution. Similarly, software and operating protocols do not evolve
in a manner that one would model using probability functions. As the focus
of this text is the definition of representative probability models and their
analysis, the discussion is limited to the physical devices.
The space shuttle accidents serve to motivate our focus on the physical
behavior of equipment. The 1986 Challenger accident has been attributed to
the use of the vehicle in an environment that was more extreme than the one
for which it was designed. The 2002 Discovery accident is believed to have
been the result of progressive deterioration at the site of damage to its heat
shield. Thus, the physical design of the vehicles and the manner in which
they were operated were incompatible and it is the understanding of this
interface that we obtain from reliability analysis.
The text is organized in four general sections. The early chapters describe
in a stepwise manner the increasingly complete models of reliability and
failure. These initial discussions include the key result that our understand-
ing of design configurations usually implies that system reliability can be
studied at the component level. This is followed by an examination of statis-
tical methods for estimating reliability. A third section is comprised of five
chapters that treat increasingly more complicated and more realistic models
of equipment maintenance activities. Finally, several advanced topics are
treated in the final chapter.
It is hoped that this sequence of discussions will provide the reader with
a basis for further exploration of the topics treated. The development of new
methods and models for reliability and maintenance has expanded our
understanding significantly and is continuing. The importance of preventive
maintenance for safety and industrial productivity is receiving increased
attention. The literature that is comprised of reports of new ideas is expand-
ing rapidly. This book is intended to prepare the reader to understand and
use the new ideas as well as those that are included here.
As a starting point, note that it often happens that technical terms are
created using words that already have colloquial meanings that do not cor-
respond perfectly with their technical usage. This is true of the word reli-
ability. In the colloquial sense, the word reliable is used to describe people
who meet commitments. It is also used to describe equipment and other
inanimate objects that operate satisfactorily. The concept is clear but not par-
ticularly precise. In contrast, for the investigations we undertake in this text,
the word reliability has a precise technical definition. This definition is the
departure point for our study.
2
System Structures

The point of departure for the study of reliability and maintenance planning
is the elementary definition of the term reliability. As mentioned in Chapter
1, the technical definition of reliability is similar to the colloquial definition
but is more precise. Formally, the definition is as follows:

Definition 2.1

Reliability is the probability that a device properly performs its intended

function over time when operated within the environment for which it is
designed.

Observe that there are four specific attributes of this definition of reliability.
The four attributes are (1) probability, (2) proper performance, (3) qualifica-
tion with respect to environment, and (4) time. All four are important. Over
this and the next several chapters, we explore a series of algebraic models
that are used to represent equipment reliability. We develop the models suc-
cessively by sequentially including in the models each of the four attributes
identified in the previous definition. To start, consider the representation of
equipment performance to which we refer as function.

2.1 Status Functions

The question of what constitutes proper operation or proper function for a
particular type of equipment is usually specific to the equipment. Rather
than attempt to suggest a general definition for proper function, we assume
that the appropriate definition for a device of interest has been specified and
we represent the functional status of the device as

ì1 if the device functions properly

f=í
î0 if the device is failed

5
6 Reliability Engineering: Probabilistic Models and Maintenance Methods

Note that this representation is intentionally binary. We assume here that the
status of the equipment of interest is either satisfactory or failed. There are
many types of equipment for which one or more derated states are possible.
Discussion of this possibility is postponed until the end of this chapter.
We presume that most equipment is comprised of components and that
the status of the device is determined by the status of the components.
Accordingly, let n be the number of components that make up the device and
define the component status variables, xi, as

ì1 if component i is functioning
xi = í
î0 if component i is failed

so the set of n components that comprise a device is represented by the com-

ponent status vector

x = {x1 , x2 ,  , xn }

Next, we represent the dependence of the device status on the component

status as the function

f = f( x ) (2.1)

and the specific form for the function is determined by the way in which the
components interact to determine system function. In the discussions that
follow, ϕ(x) is referred to as a “system structure function” or as a “system sta-
tus function” or simply as a “structure.” In all cases, the intent is to reflect the
dependence of the system state upon the states of the components that com-
prise the system. A parenthetical point is that the terms “device” and “sys-
tem” are used here in a generic sense and may be interpreted as appropriate.
An observation concerning the component status vector is that it is defined
here as a vector of binary elements so that an n component system has 2n pos-
sible component status vectors. For example, a three-component system has
23 = 8 component status vectors. They are

{1, 1, 1} {1, 0, 1}

{1, 1, 0} {1, 0, 0}

{0, 1, 1} {0, 0, 1}

{0, 1, 0} {0, 0, 0}

Each component status vector yields a corresponding value for the system
status function, ϕ.
System Structures 7

From a purely mathematical point of view, there is no reason to limit the

definition of the system status function so forms that have no practical inter-
pretation can be constructed. In order to avoid any mathematically correct
but practically meaningless forms for the system status function, we limit
our attention to coherent systems.

Definition 2.2

A coherent system is one for which the system structure function is nonde-
creasing in each of its arguments.

This means that for each element of the component status vector, xi, there
exists a realization of the vector for which
f( x1 , … , xi -1 , 0, xi +1 , … , xn ) < f( x1 , … , xi -1 , 1, xi +1 , … , xn ) (2.2)

Throughout our study of reliability, we will limit our attention to algebraic

forms that comply with this restriction.
Generally, we expect that the physical relationships among the compo-
nents determine the algebraic form of the system status function, ϕ.

2.2 System Structures and Status Functions

Among reliability specialists, it is generally accepted that there are four
generic types of structural relationships between a device and its compo-
nents. These are (1) series, (2) parallel, (3) k out of n, and (4) all others. Consider
each of these forms in sequence.

2.2.1 Series Systems

The simplest and most commonly encountered configuration of components
is the series system. The formal definition of a series system is as follows:

Definition 2.3

A series system is one in which all components must function properly in

order for the system to function properly.

The conceptual analog to the series structure is a series-type electrical

circuit. However, unlike a series circuit, it is specifically not implied here that
the components must be physically connected in sequence. Instead, the point
of emphasis is the requirement that all components function. An example of
a series system in which the components are not physically connected is the
8 Reliability Engineering: Probabilistic Models and Maintenance Methods

1 2 3

FIGURE 2.1
Reliability block diagram for a series system.

set of legs of a three-legged stool. Another is the set of tires on an automobile.

In both examples, the components are not physically connected to each other
in a linear configuration. Nevertheless, all of the components must function
properly for the system to operate.
The concept of a series circuit is commonly used to define a graphical rep-
resentation of a series structure. For the three components, this is presented
in Figure 2.1.
In general, representations of system structures such as the one in Figure 2.1
are referred to as reliability block diagrams. They are often helpful in under-
standing the relationships between components.
For the series structure, the requirement that all components must func-
tion in order for the system to function implies that a logical algebraic form
for the system structure function is
f( x ) = min {xi } (2.3)
i

but an equivalent and more useful form is

n

f( x ) = Õx
i =1
i (2.4)

As examples, consider a three-component series system and the cases

x1 = x2 = 1, x3 = 0, and f( x ) = 0
x1 = 1, x2 = x3 = 0, and f( x ) = 0
x1 = x2 = x3 = 1 and f( x ) = 1

Only the functioning of all components yields system function.

2.2.2 Parallel System

The second type of structure is the parallel structure. The conceptual analog
is again the corresponding electrical circuit and the definition is as follows:

Definition 2.4

A parallel system is one in which the proper function of any one component
implies system function.
System Structures 9

It is again emphasized that no specific physical connections among the

components are implied by the definition or by the reliability block dia-
gram. The reliability block diagram for a three-component parallel system
is shown in Figure 2.2.
One example of a parallel system is the set of two engines on a two-
engine airplane. As long as at least one engine functions, flight is sustained.
However, this example implies that simply maintaining flight corresponds
to proper function. It is a worthwhile debate to discuss when it is and when
it is not an appropriate example of a parallel system.
Another example that is more appealing is the fact that the commu-
nications satellites presently in use have triple redundancy for each
communications channel. That is, three copies of each set of transmit-
ting components are installed in the satellite and arranged in parallel in
order to assure continued operation of the channel. In view of the fact
that this implies significant weight increases over the use of only single-
configuration transmitters, the satellite provides an example of the impor-
tance of reliability as well as one of a parallel structure.
In a similar manner to that for the series system, the structure function for
the parallel system may be defined as
f( x ) = max {xi } (2.5)
i

An alternate form that is more amenable to analytical manipulation can be

defined using a shorthand developed by Barlow and Proschan [12]:
n

f( x ) = x
i =1
i (2.6)

FIGURE 2.2
Reliability block diagram for a three-component parallel system.
10 Reliability Engineering: Probabilistic Models and Maintenance Methods

The inverted product symbol, Π, is called “ip” and is defined as

n n

i =1
xi = 1 - Õ (1 - x )
i =1
i (2.7)

Once mastered, this shorthand is very convenient. Example cases for the
three-component parallel system are

x1 = x2 = 1, x3 = 0, and f( x ) = 1
x1 = 1, x2 = x3 = 0, and f( x ) = 1
x1 = x2 = x3 = 0, and f( x ) = 0

Conceptually, a parallel system is failed only when all system components

are failed.
Before leaving the discussion of parallel structures, it is appropriate
to mention the fact that the parallel arrangement of components is often
referred to as “redundancy.” This is because the proper function of any
of the parallel components implies proper function of the structure.
Thus, the additional components are redundant until a component fails.
Frequently, parallel structures are included in product designs specifi-
cally because of the resulting redundancy. Often but not always, the par-
allel components are identical. At the same time, there are actually several
ways in which the redundancy may be implemented. A distinction is
made between redundancy obtained using a parallel structure in which
all components function simultaneously and that obtained using parallel
components of which one functions and the other(s) wait as standby units
until the failure of the functioning unit. Models that describe the reliabil-
ity of active redundancy and of standby redundancy are presented at the
end of Chapter 4.

2.2.3 k-out-of-n Systems

The third type of structure is the k-out-of-n structure. There is no obvious
conceptual analog for this structure. A formal definition of it is as follows:

Definition 2.5

A k-out-of-n system is one in which the proper function of any k of the n com-
ponents that comprise the system implies proper system function.
System Structures 11

The usual approach to constructing the reliability block diagram for the
k-out-of-n system is to show a parallel diagram and to provide an additional
indication that the system is k out of n.
An example of a k-out-of-n system is the rear axle of a large tractor trailer
on which the functioning of any three out of the four wheels is sufficient to
assure mobility. Another example is the fact that some (1 – k) electronic mem-
ory arrays are configured so that the operation of any 126 of the 128 memory
addresses corresponds to satisfactory operation.
The algebraic representation of the structure function for a k-out-of-n sys-
tem is not as compact as those for series and parallel systems. Given the defi-
nition of the relationship between component and system status, the most
compact algebraic form for the structure function is

ì n

ï1
f( x ) = í
if åx ³ k
i =1
i
(2.8)
ï
î0 otherwise

Example cases for a 3-out-4 system are

x1 = x2 = x3 = 1, x4 = 0, and f( x ) = 1
x1 = x2 = 1, x3 = x4 = 0, and f( x ) = 0
x1 = x2 = x3 = 0, x4 = 1, and f( x ) = 0

Note that a series system may be considered as an n-out-of-n system and a

parallel system may be viewed as a 1-out-of-n system. Thus, the k-out-of-n
form provides a generalization that is sometimes useful in analyzing system
performance in a generic context.
Before leaving the discussion of k-out-of-n systems, it is interesting to note
that a special case of this type of system is the consecutive k-out-of-n system of
which there are two types. The consecutive k-out-of-n:F system is comprised of
n-aligned independent components that fail when k or more consecutive com-
ponents fail. The k-out-of-n:G system functions when k or more consecutive
components function. These two realizations of the consecutive system have
been shown to be mirror images so it is sufficient to analyze either of them.
Examples of the k-out-of-n:F system have been provided by Chiang and
Niu [13]. One of those examples is an oil pipeline in which each of the equally
spaced pumps is capable of pushing the oil to a distance equal to k pumping
points so a failed pump does not fail the system as the next pump will main-
tain flow. If k consecutive pumps fail, the system fails.
As indicated earlier, the fourth class of component configurations is the set
of all others that are conceivable. This statement is not intended to be mis-
leading. Instead, it is intended to imply that we can establish an equivalence
12 Reliability Engineering: Probabilistic Models and Maintenance Methods

between any arbitrary component configuration and one based on series

and parallel structures. The process of constructing equivalent structures is
explained in the next section and is illustrated with a classic example.

2.2.4 Equivalent Structures

The selection of a component configuration is usually made by the device
designer in order to assure a specific functional capability. The configuration
selected may not match one of the classes discussed earlier. In such a case, there
are two ways by which we can obtain equivalent structural forms that may be
easier to analyze than the actual one. The two ways are to use either minimum
path or minimum cut analyses of the network representation of the system.
As a vehicle for illustrating the two methods, we use the Wheatstone
bridge. The reliability block diagram for the bridge is shown in Figure 2.3.
Notice that the bridge structure is not series, parallel, or k out of n. Thus,
the earlier algebraic representations cannot be used directly to provide a
statement of the system status function.
We can obtain a system status function for the bridge in several ways. One
obvious approach is to enumerate all of the component status vectors, to
determine the system status for each vector and to construct a table of system
status values. For the five-component bridge structure, this is readily done
and the result is shown in Table 2.1.
On the other hand, systems having a greater number of components can-
not be handled so easily.
The use of minimum paths will permit us to analyze the bridge structure
and other larger systems as well. Start with some definitions:

Definition 2.6

A path vector, x, is a component status vector for which the corresponding

system status function has a value of 1.

1 4

2 5

FIGURE 2.3
Reliability block diagram for a Wheatstone bridge.
Exploring the Variety of Random
Documents with Different Content
[700] Donkin’s evidence, Court Martial Proceedings, p. 448.

[701] Williamson’s evidence, Proceedings, p. 124.

[702] Williamson’s evidence, ibid., p. 125.

[703] Mr. Fortescue has, I think, misinterpreted this order, when he says that it
told Clinton to march to the same spot as the first (British Army, ix. p. 63), for
Constanti is not in the direction of the Gaya, but on the opposite flank, west of the
Francoli river.

[704] A queer misprint in this dispatch makes it say ‘the enemy will march.’

[705] Evidence of Captains Withers and Bathurst, R.N., Court Martial

Proceedings, pp. 86 and 95.

[706] Mackenzie in his evidence says his men began at 2 p.m. to get into the
boats.

[707] Evidence of Bentinck, ibid., p. 176. The cavalry went off at 3 p.m.

[708] The hours of this belated work are stated very differently by various naval
witnesses, some of whom say that they worked till 1 a.m., others till 4 a.m., others
till 7; one thinks that embarkations continued till well into the forenoon of the 13th
—say 11 o’clock. At any rate, the hour must have been long after daylight had
come—which was at 4.15, as is recorded by one witness.

[709] The total loss of Murray’s Army during the Tarragona operations was:
Killed. Wounded. Missing. Total.
British, Germans, Calabrese, Italian Levy 14 60 5
Sicilians — 15 — 102
Whittingham’s Spaniards 1 7 —
Bertoletti’s garrison lost 13 killed and 85 wounded = 98. The enemies did each
other little harm!

[710] Not apparently the whole division, for Mackenzie calls it ‘a small body of
infantry.’

[711] Suchet, Mémoires, ii. p. 315.

[712] Date stated by some as the 16th, but the earlier day seems correct. See
Mackenzie’s evidence, pp. 152-3: he was uncertain as to the date.
[713] Hallowell says on the evening of the 14th or the 15th, he forgets which. But
the latter date must be the true one.

[714] Hallowell’s speech, p. 554 of the Court Martial Proceedings.

[715] Pannetier’s rearguard followed on the 17th. See letter of the Alcalde of
Perello, Court Martial, p. 361.

[716] Hallowell’s speech, Court Martial Proceedings, p. 556.

[717] By Mr. Fortescue, History of the British Army, ix. 67.

[718] See above, p. 515.

[719] Supplementary Dispatches, viii. pp. 18-19, June 23.

[720] Ibid., p. 20.

[721] Supplementary Dispatches, viii. p. 22.

[722] Wellington to Murray, July 1, Dispatches, x. p. 487.

[723] Dispatches, x. p. 495, to Lord Bathurst, July 2.

[724] To the same, Dispatches, x. p. 496.

[725] To O’Donoju, Minister of War, Dispatches, x. pp. 492-3; to Castaños, x. p.

475; to Lord Bathurst, x. pp. 473-4.

[726] ‘We and the powers of Europe are interested in the success of the War in
the Peninsula. But the creatures who govern at Cadiz appear to feel no such
interest. All that they care about really is the praise of their foolish Constitution....
As long as Spain shall be governed by the Cortes, acting upon Republican
principles, we cannot hope for any permanent amelioration.’ Dispatches, x. p. 474,
Wellington to Bathurst, June 29.

[727] Wellington to Henry Wellesley, Dispatches, x. p. 491.

[728] Dispatches, x. pp. 523-4.

[729] Ibid., x. p. 521.

[730] Ibid., x. pp. 553-4. When this letter was written to Lord William Bentinck,
Wellington had received no London dispatch for twenty days, mainly owing to bad
weather in the Bay of Biscay.
[731] Ibid., x. pp. 613-14, to Lord Bathurst.

[732] Ibid., x. p. 478, to Bentinck.

[733] Ibid., x. pp. 477-9.

[734] Ibid., x. p. 531.

[735] Full details in O’Donnell’s report to Wellington of July 1, 1813. Dispatches,

p. 503. Toreno makes an odd mistake in calling the French commander de Ceva:
this was the name of the junior officer who drew up the capitulation.

[736] Not to be confounded with General Cassagne, who long commanded a

division in the Army of Andalusia.

[737] Late Hamilton’s division in 1810-11-12.

[738] Lord Dalhousie was left in command—a great slight to Picton—all the more
so after what had happened at Vittoria. See Supplementary Dispatches, viii. p.
249.

[739] Late Cadogan’s.

[740] Pringle arrived and took command of this brigade a fortnight later.
Meanwhile the senior battalion commander led it.

[741] Reille had only two chasseur regiments with him.

[742] See the indignant letters of French officials quoted in Vidal de la Blache, i.
pp. 69 and 165-7.

[743] For the Army of Portugal only two chasseur regiments were left: for the
Army of the North only one: for the Army of the Centre only the weak Nassau
squadrons. But 13 out of the 14 cavalry regiments of the Army of the South
remained behind.

[744] Chassé’s brigade lost 850 men of 1,700 present, and the Nassau regiment
in the German brigade had similar casualties, though the Baden regiment got off
more lightly.

[745] It should be noted that Daricau’s 6th Division was now led by Maransin—its
old commander having been badly wounded at Vittoria. Maransin’s late brigade,
still extra-divisional as at Vittoria, had been made over to Gruardet.
[746] Jourdan answered that it was a false movement, but that on June 29th he
could not possibly foresee that the King would change his mind as to the
destination of the Army of the South. See Vidal de la Blache, i. p. 103.

[747] See Vidal de la Blache, i. pp. 103-4.

[748] Under Cameron of the 92nd as senior colonel—Cadogan who fell at Vittoria
not having yet been replaced.

[749] Viz. Villatte’s and Maransin’s divisions, and Gruardet’s brigade of his own
army, and Braun’s brigade of the Army of the Centre.

[750] On May 1 the three 2nd Division brigades had shown 7,200 bayonets—they
had lost 900 men in action at Vittoria. If we allow for sick and stragglers and other
casual losses, they cannot possibly have had 6,000 men in line on July 5.

[751] All these absurd theories are to be found in Gazan’s reports to Jourdan of
July 4 and 5. See Vidal de la Blache, i. pp. 106-7.

[752] The troops of the Army of the North which Foy had collected from the
Biscay garrisons, the brigades of Deconchy, Rouget, and Berlier of which we have
heard so much in a previous chapter.

[753] Foy, Lamartinière, Maucune and Fririon (late Sarrut). There were behind
them the King’s Spaniards and the raw Bayonne reserve.

[754] Dispatches, x. p. 512. The total losses having been 124 on all three days,
Wellington’s ‘no loss’ means, of course, practically no loss.

[755] The clearest proof of Gazan’s resolute resolve not to stand, and of the
complete mendacity of his dispatches concerning his heavy fighting on the 4th-5th
and 7th, is that he returned the total of his losses at 35 killed and 309 wounded. As
he had six brigades, or 13,000 men at least, engaged, it is clear that there was no
serious fighting at all—a fact borne out by Hill’s corresponding return of 8 killed,
119 wounded, and 2 missing in the whole petty campaign.

[756] Cf. Lecestre, Lettres inédites, ii. p. 1037, where the Emperor says on July 3
that he cannot make out what is happening; and that Joseph and Jourdan are
incapables.

[757] See, e. g., Joseph to Clarke, p. 336 of vol. ix of his Correspondance.

[758] See vol. v, p. 97.

[759] ‘Les malheurs de l’Espagne sont d’autant plus grands qu’ils sont ridicules.’
Napoleon to Savary, Dresden, 20 July: Lecestre, Lettres inédites, ii.

[760] Even that he was withdrawing the British Army from Portugal. Lecestre, ii.
998, May 5.

[761] Though he did once make the observation that ‘on ne conduit pas des
campagnes à 500 lieues de distance,’ in a lucid interval.

[762] See v. pp. 194-6.

[763] Napoleon to Cambacérès, Lecestre, ii. 1055.

[764] See above, p. 88.

[765] See the very interesting pages of Vidal de la Blache, i. pp. 142-3.

[766] See Roederer’s account of the interview in Vidal de la Blache, i. pp. 132-3.
Napoleon had suggested him as the best person for the errand.

[767] Napoleon to Cambacérès, Lecestre, Lettres inédites, ii. 1055.

[768] ‘Joey Bottles’ is the English equivalent.

[769] See especially the caustic paragraphs in Lecestre, ii. 1045, 1047, 1055, to
Clarke and Cambacérès.

[770] So Jourdan’s Mémoires, p. viii. Vidal de la Blache, i. p. 140, says 5,000

francs only, which seems an impossibly small sum for Marshal’s half pay.

[771] Napoleon to Cambacérès, Lecestre, ii. 1045.

[772] Napoleon to Maret, No. 28 in Lettres de Napoléon non insérées dans la

Correspondance, Aug.-Sept.-Oct. 1813. Paris, 1907.

[773] Bathurst to Wellington, June 23, Supplementary Dispatches, viii. pp. 17-18.

[774] Liverpool to Wellington, Supplementary Dispatches, viii. pp. 64-5, dated

July 7.

[775] Wellington to Bathurst, July 12, from Hernani. Dispatches, x. p. 524.

[776] Wellington to Lord Liverpool, July 23, from Lesaca, ibid., x. p. 568. Cf. same
to same, x. p. 596.

[777] Dispatches, x. p. 570.

[778] See Wellington to Torrens, Dispatches, x. p. 616.

[779] See Wellington to Bathurst, Dispatches, x. p. 599, and other epistles on

same topic.

[780] Called the Mirador (’look-out’), Queen’s, and Principe batteries: there were
others facing sea-ward, which were of no account in this siege, as no attack from
the water-side took place.

[781] See Jones’s Sieges of the Peninsula, ii. p. 94.

[782] The governor surrendered the town on August 1, but retired into the castle
of La Mota, where he capitulated a few days later, just as Rey did in 1813.

[783] Jones’s Sieges of the Peninsula, ii. p. 14.

[784] Dickson’s diary, July 12, 1813, p. 960 of Colonel John Leslie’s edition of the
Dickson Papers.

[785] Jones, ii. p. 97.

[786] See above, p. 478.

[787] See Graham to Wellington, Supplementary Dispatches, viii. p. 62. The

K.G.L. brigade of the 1st Division was present for a few days.

[788] Wellington to Graham, Dispatches, x. p. 512.

[789] Wellington Dispatches, x. p. 525.

[790] Melville to Wellington, Supplementary Dispatches, viii. pp. 224-5.

[791] Wellington to Melville, Dispatches, xi. p. 115.

[792] See the interesting account of his cross-country ride on June 25-9 in his
Letters from the Peninsula, pp. 167-74.

[793] See Frazer, p. 195.

[794] In detail Jones gives them as twenty 24-pounders, six 18-pounders, four 68-
pound cannonades, six 8-inch howitzers, and four mortars.

[795] See Hartmann’s Life, pp. 153-4.

[796] Nos. 4, 5 in map.

[797] Nos. 1, 2, 3 in map.

[798] No. 6 in the map.

[799] Why does Belmas, who was very well informed, and used Jones’s book,
call the stormers ‘les Anglais’ and say that they lost 150 men? (Sièges, iv. p. 608).
He knew from Jones that they were Caçadores only (Jones, ii. p. 21), and that
their loss was under 70.

[800] Right column, to attack the cemetery and fortified houses—150 of 5th
Caçadores, 150 13th Portuguese Line, three companies 1/9th Foot, three
companies 3/1st Foot (Royal Scots) all under Hay, Brigadier of the 5th Division.
Left column: 200 of 5th Caçadores, 200 of 13th Portuguese Line, three companies
1/9th Foot—all under Bradford commanding Portuguese independent brigade.
Why did not Oswald use his own Portuguese brigade, but draw on Bradford?
Possibly because Spry’s brigade were discouraged by the failure of their Caçador
battalion on the 15th.

[801] Generally in British narratives called the Cask Redoubt, because wine
casks had been used to revet the shifting sand of which the soil was there
composed.

[802] Batteries 8 and 7 in the map.

[803] Batteries 13 and 14 in the map.

[804] Batteries 12 and 11 in the map.

[805] No. 6 in the map.

[806] There is a curious contradiction between Jones and Belmas as to the fate of
the Cask Redoubt. The latter says that the British took it—the former that the
garrison abandoned it, though not attacked.

[807] ‘From the looseness of the sand in which the battery was constructed, it
was found impossible to keep the soles of the embrasures sufficiently clear to use
the three short 24-pounders mounted on ship carriages—after a few rounds they
had to cease firing.’ Jones, ii. p. 28.
[808] Burgoyne’s Life and Correspondence, i. p. 267.

[809] Burgoyne, who took out the flag of truce, says that the French officer who
met him on the glacis used very angry words (ibid.).

[810] See Dickson Papers, ed. Col. Leslie, p. 970. The second breach is marked
as ‘Lesser Breach’ on the map.

[811] Burgoyne, whose diary of the siege is one of the primary authorities, says
that in his opinion the mine could have been much more useful than it was. ‘On the
discovery of the drain, I should have immediately have altered the whole plan of
attack. I would have made a “globe of compression” to blow in the counterscarp
and the crest of the glacis, and then at low water have threatened an attack on the
breaches, exploded the mine, and have made the real assault on the hornwork,
which not being threatened had few people in it, and would undoubtedly have
been carried easily.’ There was, he says, good cover in the hornwork, which would
have been easily connected with the parallel, and used as the base for attacking
the main front, with breaching batteries in its terre-plein and the crest of the glacis.
Burgoyne, i. p. 271. But this is wisdom after the event.

[812] Jones, ii. p. 36.

[813] For all these details see Belmas, iv. pp. 620-1.

[814] Burgoyne says (i. 369) that the engineers on the 24th settled that the mine
was no more than a signal ‘with a chance of alarming them’. On the 25th it would
seem that a little more attention, but not nearly enough, was given to this useful
subsidiary operation.

[815] Burgoyne says at 4.30.

[816] This is slurred over in the British narratives except Dickson’s Diary, p. 973.
Belmas gives some account of it, however, though he calls the assailants British
instead of Portuguese (iv. p. 623). They were some companies of the 8th
Caçadores.

[817] Most of this narrative is from Colin Campbell’s long and interesting letter to
Sir J. Cameron, printed on pp. 25-30 of his Life by General Shadwell.

[818] Gomm, p. 312.

[819] Frazer’s Letters from the Peninsula, p. 205.

[820] The 38th lost 53 men, the 9th 25, the Portuguese 138 in the side-attack.
Why need Belmas, who had Jones’s book before him, give the total of British
losses as 2,000? (Sièges, iv. 625).

[821] Though Jones says that he saw some wounded bayoneted.

[822] Printed in Maxwell’s Peninsular Sketches, vol. ii.

[823] Campbell’s letter quoted above in his Life, i. p. 30.

[824] ‘The men, panic stricken, turned and could never be rallied,’ writes Frazer
next day (p. 204). ‘One party, I believe of the 9th and 38th, went up to the breach
and then turned and ran away,’ says Larpent (p. 200). Neither saw the actual
assault in the dark.

[825] So at least he wrote to Castaños on the 24th: ‘j’espère que cette affaire est
finie.’ Dispatches, x. p. 564.

[826] See Frazer, p. 206, and Burgoyne, i. p. 269.

[827] See Wellington to Graham, night of the 25th, Dispatches, x. p. 566.

[828] Permission was given to leave four guns behind in the main breaching
batteries and two on Monte Olia, to keep up a semblance of continued attack.
Dispatches, x. p. 566.

[829] The British officer in command in the trenches, Major O’Halloran, was court
martialled, but acquitted. It was proved that he had given the correct orders to the
Portuguese captains of the companies on guard, who had not obeyed them. All the
prisoners except 30 were Portuguese.

[830] The history of this proclamation is curious. Clarke, or Napoleon himself,

considered it too full of insults of a person who was, after all, the Emperor’s
brother. So it had to be disavowed: Soult wrote to Paris that he had not authorized
it, and Clarke had the ingenuity to print in the French newspapers that it was an
invention of the English government, intended to disgust the Spanish partisans of
King Joseph, and to advertise the ill feeling that prevailed between the French
army and the Imperial family. See Vidal de la Blache, i. p. 138; as he remarks, the
style is all Soult’s, and there is not a trace of foreign diction in it. No Englishman or
Spaniard could have written it.

[831] Joseph to Napoleon, 1st February 1813.

[832] See notably the case of General Excelmans.

[833] See especially the proclamation of March 6, 1815.

[834] Mémoires of St. Chamans, p. 35.

[835] Maximilien Lamarque, ii. p. 182.

[836] Stanhope’s Conversations with Wellington, p. 20.

[837] The gendarmerie were those who had come from the ‘legions’, employed in
1811-12-13 as garrisons in Northern Spain. They were embodied in units, horse
and foot, and used as combatants (as at the combat of Venta del Pozo, for which
see p. 71).

[838] As Table XVI in the Appendix shows, Foy’s division received two of Sarrut’s
regiments: Cassagne’s (now Darmagnac’s) took all the French infantry of the old
Army of the Centre: Villatte’s (now Abbé’s) was given two of Abbé’s regiments of
the Army of the North: Conroux’s division absorbed Maransin’s independent
brigade: Barbot’s (now Vandermaesen’s) received two regiments of the Army of
the North: Daricau’s (now Maransin’s) got half Leval’s ‘scrapped’ division, Taupin
the other half of it: Maucune absorbed one of Vandermaesen’s old regiments,
Lamartinière one of Sarrut’s.

[839] 120th Line of Lamartinière.

[840] 2nd Léger of same, which suffered heavily at Vittoria while under Sarrut.

[841] 20,957 to be exact.

[842] Not only the Afrancesados but some of the Army of the North troops
withdrawn from the Biscay garrisons had a poor record, and had disgusted Foy in
his recent Tolosa fight. These were high-numbered battalions, recently made up
from the Bayonne conscript reserve.

[843] The best proof of the efficiency of the bulk of Villatte’s corps is that when
Vandermaesen’s and Maucune’s divisions were cut to pieces in the battles of the
Pyrenees, Soult made up a new brigade for each of them out of the Reserve.
Joseph’s French Guards fought splendidly at San Marcial. The Germans were very
steady veteran troops.

[844] Vidal de la Blache, i. p. 160.

[845] See above, p. 533. Jourdan to Joseph, July 5. The memorandum had been
made over to Soult. Cf. Clere, Campagne du Maréchal Soult, p. 46, and Vidal de la
Blache, i. p. 182.

[846] One asks oneself why Soult did not give Reille the Maya attack, saving him
two-thirds of his journey, and send D’Erlon to join Clausel at St. Jean-Pied-du-Port,
by a march much shorter than Reille was asked to make.

[847] It is said that persons acquainted with the country told Soult to send the
whole column round by Bayonne, on account of the artillery, but that he refused.
As a matter of fact, Lamartinière’s division and some of the guns did go that
détour, owing to the broken bridge.

[848] Wellington to Graham, July 22. Dispatches, x. p. 559.

[849] Ibid., p. 563, same to same.

[850] Supplementary Dispatches, viii. p. 113.

[851] Ibid., p. 114.

[852] Wellington to Graham, July 24, Dispatches, x. p. 563.

[853] Same morning, to Giron, ibid., p. 564.

[854] Wellington to Graham, 25th July, Dispatches, x. p. 566.

[855] Wellington to Graham, Dispatches, x. p. 570.

[856] See especially Lemonnier-Delafosse, pp. 211-12, and Wachholz of

Brunswick-Oels, p. 321.

[857] Vittoria and light companies of Doyle, La Union, and Legion Estremena.

[858] La Union and the Legion, minus their light companies.

[859] Doyle and 2nd of Jaen.

[860] Sometimes called the Puerto de Val Carlos.

[861] See the very interesting letter of Bainbrigge of the 20th, printed as an
Appendix to the regimental history of that corps, p. 390.

[862] Bainbrigge says that it was 7 a.m. before the regiment reached the Linduz,
but that it was an hour earlier is demonstrated by the fact that they heard firing at
Roncesvalles after arriving. Now Byng’s fight on the Leiçaratheca began at 6 a.m.
Therefore Ross was on the Linduz earlier.
[863] What became of this Spanish company? Captain Tovey of the 20th (see
history of that corps, p. 408) says that the French ‘made the Spanish picquet, who
were posted to give us intelligence, prisoners, without their firing a shot’. Another
account is that having seen Ross arrive, they quietly went off to rejoin their
brigade, without giving any notice.

[864] There is a curious and interesting account of all this in the Memoirs of
Lemonnier-Delafosse, aide-de-camp to Clausel, who was twice sent to stir up
Barbot, whose conduct he describes in scathing terms (pp. 212-14). Clausel says
that the 50th stormed the Leiçaratheca. That it stormed an abandoned position is
shown by the figure of its losses. What Clausel does not tell can be gathered from
Byng’s workmanlike dispatch to Cole, in Supplementary Dispatches, viii. pp. 128-9.

[865] Of the 27th and 130th Line.

[866] I confess that I doubt these figures. Martinien’s lists show the 27th Line with
seven officer-casualties, the 1st Line with two, the 25th Léger with three, the 130th
with two. Fourteen officer-casualties ought to mean more like 280 than 100
casualties of all ranks. In the whole Pyrenean campaign the French army lost 120
officers to 12,300 men—nearly 30 men to each officer. Clausel asks us to believe
that at Roncesvalles the proportion was one officer to twelve men! Yet, of course,
such disproportion is quite possible.

[867] While we have quite a number of good personal narratives of the fight on
the Linduz, I have found for the fight on the Leiçaratheca nothing but the official
reports of Clausel, Byng, and Morillo, save the memoirs of George L’Estrange of
the 31st and of Lemonnier-Delafosse, who is interesting but obviously inaccurate,
since he says that the French regiment which carried the hill was the 71st. Not
only was it the 50th, as Clausel specially mentions, but 71 was a blank number in
the French Army List.

[868] Four, not five, because the light company of the 20th was absent with the
other light companies far to the right: so the wing was only four companies strong,
or three deducting Tovey’s men. Wachholz forgets this.

[869] Wachholz, p. 322.

[870] Tovey fortunately wrote a narrative of this little affair, which may be found in
the history of the 20th, p. 408. He says: ‘The enemy’s light troops opened so
galling a fire that Major-General Ross called out for a company to go to the front.
Without waiting for orders I pushed out with mine, and in close order and double-
quick cleared away the skirmishers from a sort of plateau. They did not wait for us:
on reaching its opposite side we came so suddenly on the head of the enemy’s
infantry column, which had just gained a footing on the summit of the hill, that the
men of my company absolutely paused in astonishment, for we were face to face
with them. The French officer called to us to throw down our arms: I replied
“bayonet away,” and rushing on them we turned them back down the descent.
Such was the panic and confusion caused by the sudden onset, that our small
party (for such it was compared to the French column) had time to regain the
regiment, but my military readers may rest assured that it required to be done
double quick.’

[871] This ditch had been cut by the Spaniards in 1793 as an outer protection to
their redoubt on the Linduz.

[872] Wachholz, p. 324.

[873] The 6th Léger, 69th (2 battalions), 76th, and 36th show casualties, the rear
regiments (39th and 65th) none. Nor does Maucune’s division. Similarly on the
British side none of Anson’s or Stubbs’s battalions contribute to the list.

[874] As we have seen already, Clausel puts his loss at the Leiçaratheca at 160,
to Byng’s and Morillo’s 120. At the other end of the line Ross’s brigade had lost
216 men—139 of them in the 20th, 31 in the 7th, 42 in the 23rd, 4 in the Brunswick
company. [I know not where Napier got his strange statement that this company
lost 42 men: their captain, Wachholz, reports 2 killed and 2 wounded.] Foy’s six
front battalions had lost 10 officers and 361 men. The total Allied loss was about
350, there having been a few casualties among Campbell’s Portuguese and
among the Spaniards at Orbaiceta. The total French loss was not less than 530.
Both figures are very moderate. Cole estimated the French casualties at 2,000
men! Soult wrote that he had almost exterminated the 20th, whose total loss had
been 139.

[875] Cole to Wellington, Supplementary Dispatches, viii. p. 127.

[876] Wellington to Liverpool, Dispatches, x. p. 596.

[877] See vol. iv. pp. 389-90.

[878] Cole to Murray, Linzoain, July 26th. Wrongly dated July 27th in
Supplementary Dispatches, viii. p. 124.

[879] See diary of Dr. Henry, who was at Elizondo, and notes how all the senior
officers rode out eastward (p. 161).

[880] Bell, vol. i. p. 102; Cadell, p. 161.

[881] One from each battalion plus the odd company of the 5/60th attached to
each 2nd Division brigade.
[882] See Hope’s Military Memoirs, p. 319. Sceptical observers with telescopes
said that the objects seen were droves of bullocks.

[883] See Moyle Sherer (who commanded the picquet), p. 257.

[884] Major Thorne, assistant quartermaster-general. Moyle Sherer says that

Thorne owned that there was a small column on the move, but that he judged it to
be a battalion shifting its quarters, or a relief of outposts.

[885] Mr. Fortescue (History of the Army, ix. p. 258) thinks that the 34th got up in
time to join in their last struggle. But Bell of that regiment says ‘we laboured on, but
all too late—a forlorn hope—our comrades were all killed, wounded, or prisoners.
The enemy had full possession of the ground.’ Bell’s Rough Notes, i. p. 103.

[886] Bell’s Rough Notes, i. p. 103.

[887] Bell, i. p. 104.

[888] All this is most difficult to follow, our numerous sources contradicting each
other in matters of detail in the most puzzling fashion. For this part of the narrative
I have used, beside the dispatch of William Stewart, the books of Moyle Sherer of
the 34th, who commanded the Aretesque picquet and was taken prisoner—Sir
George Bell of the same regiment, Cadell of the 28th, Hope and Sergeant
Robertson of the 92nd, Patterson of the 50th, the two anonymous diarists ‘J. S.’
and the ‘Scottish Soldier’ of the 71st, besides D’Erlon’s and Darmagnac’s original
dispatches, lent me by Mr. Fortescue. I take it that each authority may be followed
for the doings of his own corps, but is of inferior weight for those of other units.
Patterson says that the 34th was at one time in close touch with the 50th, Cadell
that the 28th and 92nd worked together, while Hope says that the 28th was only
seen by the 92nd right wing after it had ended its terrible first entry into the fight.
Patterson says that he saw O’Callaghan of the 39th fighting along with the 50th in
the third episode of the combat, when, according to other sources, that regiment
had already retreated south toward the valley with the 34th. Stewart’s dispatch
only speaks of the 28th and 34th retiring in that direction, not the 39th. A confused
fight has left confused memories. I cannot be sure of all the details.

[889] The statement in Napier and succeeding writers that the wounded of the
right wing of the 92nd formed a bank behind which the French advance halted,
and stood to receive the fire of the left wing of that same corps, whose bullets hit
many of its comrades, comes from the narrative of Norton of the 34th (Napier, V.
appendix, p. 442), who was some way off. That the troops which came up were
the right wing 71st, and not the left wing 92nd, seems to me proved by the
narrative of Hope of the 92nd, who distinctly says that the right wing were relieved
by the 71st, and that the left wing were still holding the Maya position and under
Stewart, who had just arrived, along with the left wing of the 71st (Military
Memoirs, p. 210).

[890] He himself in his dispatch only says that it was after 1 p.m.

[891] Tulloh (commanding 2nd division batteries) to Dickson, in Dickson Papers,

p. 1022. Wellington’s censure of Stewart may be found in Dispatches, x. p. 588,
and his reply to the latter’s self-defence in xi. p. 107. The details are hard to follow:
Wellington says that Pringle ordered the guns to be taken off by the road to Maya
—that Stewart directed that they were to go back, and look to ‘the mountain road
to Elizondo’ as their proper line of retreat. When it became necessary for them to
retire at all costs, that road was already in the hands of the French. But I do not
know precisely what Wellington meant by the mountain-road to Elizondo. Does it
mean the track by which the 28th and 34th had retired?

[892] See Stewart’s Report to Hill, Berueta, July 26.

[893] Robertson, pp. 109-10.

[894] Stewart’s dispatch says that it was the 82nd who fought with stones.

[895] This was not the brigade to which the 82nd belonged, but the reserve
brigade of the 7th Division, short of one of its units, the 3rd Provisional.

[896] Cf. Dispatches, x. pp. 597-8.

[897] ‘Lettres de l’Empereur Napoléon non insérées dans la Correspondance,

publiées par X. Paris and Nancy, 1909,’ page 3. It is amusing to find out what
Napoleon III omitted of his uncle’s letters.

[898] ‘Lettres de l’Empereur Napoléon non insérées dans la Correspondance,’ p.

13.

[899] Wellington’s letter to Graham, giving the false report that D’Erlon had been
repulsed at Maya, is dated at 10 p.m. The letter to O’Donnell must be a little later,
as it repeats this error, but adds that a note has come in from Cole, saying that he
was heavily engaged at noon. Dispatches, x. pp. 566-7.

[900] The dispatch giving this information (Dispatches, x. p. 570) is wrongly dated
in the Wellington correspondence. It should be July 26th at 4 a.m. The hour of the
receipt of Hill’s and Stewart’s reports is not given.

[901] All in Supplementary Dispatches, viii. pp. 120-1.

[902] Hill to Murray, Supplementary Dispatches, viii. p. 121.

[903] The orders to Pack and Dalhousie may be found in Supplementary
Dispatches, xix. p. 258-9, dated from Almandoz—obviously before Cole’s dispatch
had come to hand.

[904] This letter in Supplementary Dispatches, viii. pp. 124-5, is there wrongly
dated July 27th (for 26th). Cole, of course, was no longer at Linzoain on the 27th.

[905] To Picton from Almandoz, Supplementary Dispatches, xiv. p. 259.

[906] Picton to Murray, 8.30 p.m., Supplementary Dispatches, xviii. pp. 121-2.

[907] See above, p. 622.

[908] See above, p. 591.

[909] These are his own words, in his Report of August 2.

[910] Foy to Reille, July 20.

[911] Reille to Soult, July 27.

[912] The tall hat is vouched for by George L’Estrange of the 31st, and Wachholz
from Ross’s brigade, the furled umbrella by Bainbrigge of the 20th, all eye-
witnesses, whose narratives are among the few detailed accounts of this retreat.

[913] Words overheard by Bainbrigge in his own company.

[914] See Quartermaster-General to Picton, enclosing letter for Cole, sent off
from Lesaca on July 23 (Supplementary Dispatches, viii. pp. 112-13), which must
have reached Picton at Olague on the 24th.

[915] This seems a more controvertible plea. Orders went out from Lesaca on the
23rd, and must have reached Picton not very late in the day on the 24th.
Supposing he had marched from Olague on the afternoon of the 24th, he would
have been at Zubiri (only 6 miles off) on that same night, or even at Viscarret. And
from Zubiri to Roncesvalles is not an excessive day’s march for the 25th,
especially when firing was to be heard at the front.

[916] The remaining four were in the Caçador battalion of Stubbs’s Portuguese
brigade.

[917] Unfortunately all French losses are given en bloc for the six days July 27 to
August 1, and the casualties of each day cannot be disentangled. The casualties
of Maya and Roncesvalles can be ascertained, but not those of the subsequent
days.
[918] Viz. about 6,000 of Cole’s division, 5,000 of his own, 1,700 of Byng’s
brigade, 2,500 of Campbell’s Portuguese at Eugui, only a few miles away, and
something under 4,000 of Morillo’s Spaniards.

[919] See Belmas, iv. p. 803.

[920] Napier says (v. p. 225), and all subsequent historians have followed him,
that Picton originally intended to place Cole on a line between Oricain and Arleta, i.
e. on the low back-slope of the ridge. This seems to me almost incredible, as this
ground is all running downhill, completely commanded by the much loftier crests
about the Col. Surely no one, according to the tactical ideas of 1813, would take
up a defensive position half-way down a slope whose summit is abandoned to the
enemy. I can find no authority save Napier (who was not in the battle) for this
curious statement. And I am justified, I think, in holding that the San Miguel hill was
the place where Picton intended to place Cole, by the narrative and sketch-map of
Wachholz of Ross’s brigade, who places the first position of the 4th Division on a
well-marked hill immediately to the right of Villaba, and close to the 3rd Division’s
ground at Huarte. This must mean San Miguel.

[921] R. Hill’s, Ponsonby’s, the Hussar brigade, and D’Urban’s Portuguese,

Fane’s brigade, which was observing on the side of Aragon, did not arrive this day.

[922] One of the 4th Line.

[923] Reille’s report of August 1st.

[924] Clausel in his report says that he arrived in time to see the 4th Division
cross the hill of Oricain.

[925] Perhaps Carlos de España’s division, arriving from the south.

[926] All this from the very interesting narrative of Clausel’s aide-de-camp
Lemonnier Delafosse (p. 220), who bore the first message to Soult, and was (like
his chief) much irritated by the Marshal’s caution and refusal to commit himself.
Clausel had got a completely erroneous notion of the enemy’s intentions—like Ney
at Bussaco.

[927] Quartermaster-General to Sir R. Hill, Supplementary Dispatches, viii. pp.

259-60.

[928] Wellington to Pack, Supplementary Dispatches, viii. p. 122, wrongly dated 1

o’clock—it should be 10 o’clock. Wellington was at Sorauren by 11.

[929] Final destination not given—clearly it might be down the high-road to

Pampeluna; but if Picton had retreated still further and raised the siege, it might be
to Lizaso, to join Hill and the rest.

[930] Wellington described his ride to Larpent, his Judge-Advocate General, a

week later, in the following terse language (Larpent, p. 242): ‘At one time it was
rather alarming, certainly, and a close run thing. When I came to the bridge of
Sorauren I saw French on the hills on one side, and it was clear that we could
make a stand on the other hill, in our position of the 28th, but I found that we could
not keep Sorauren, as it was exposed to their fire and not to ours. I was obliged to
write my orders accordingly at Sorauren, to be sent back instantly. For if they had
not been dispatched back directly, by the way I had come, I must have sent them
four leagues round, a quarter of an hour later. I stopped therefore to write
accordingly, people saying to me all the time, “The French are coming!” “The
French are coming!” I looked pretty sharp after them every now and then, till I had
completed my orders, and then set off. I saw them just near the one end of the
village as I went out of it at the other end. And then we took up our ground.’
Wellington then added, in a confidential moment, that there need have been no
fuss or trouble, if only Cole had kept sending the proper information on the 26th
and 27th. If only his intention of going right back to Pampeluna had been known
earlier, the 6th and 7th Divisions could have been up on the 27th, and Hill’s corps
too, which had been kept at Irurita and Berueta for 36 hours, because the situation
in the south was concealed by Cole’s reticence. ‘We should have stopped the
French much sooner.’

[931] French critics expressed surprise that Wellington did not tell Pack to fall on
Clausel’s flank and rear. But the 6th Division, attacking from Olague, would have
been out of touch with the rest of the army, and Wellington did not believe in
attacks by isolated corps uncombined with the main army, and unable to
communicate with it. See Dumas’ Campagne du Maréchal Soult, p. 163.

[932] Bainbrigge’s narrative in Smyth’s History of the XXth, p. 396.

[933] Ibid. Bainbrigge was standing close to both.

[934] Larpent, p. 243.

[935] Lemonnier-Delafosse, p. 219.

[936] Soult to Clarke, July 28.

[937] Supplementary Dispatches, viii. p. 123.

[938] Ibid., p. 124.

[939] See above, p. 661.

[940] Reports of Maucune and Lamartinière dated August 3rd and 4th.

[941] Clausel’s report of August 2.

[942] There is a most curious and difficult point in this history of the first phase of
the action. Clausel says, and he is of course a primary authority, that though
Conroux was already deeply engaged with the 6th Division, ‘was being fired on
from all sides, was suffering severe losses, and had already had one of his
brigadiers disabled’ [Schwitter], he told him that he must join in the attack
‘swerving to the left so as to mount the hill in the direction originally assigned to
him’, which was done and Conroux immediately repulsed. I cannot see how this
was physically possible. How could Conroux, if already disadvantageously
engaged with the 6th Division, and ‘fired at from all sides’, break off this fight and
attack any point of the hill of Oricain? If he had gone away in that direction, who
was there to hold Sorauren against Pack’s people, who were pressing in on it, and
(as Clausel says) only a musket-shot away from it? As far as I can make out,
Conroux must have been sufficiently employed in fending off Pack and maintaining
Sorauren, so as to cover the flank of the other divisions, for the next hour or two.
No other authority but Clausel gives any hint that Conroux got away from Pack and
joined in the general assault. And I am constrained to think that Clausel (strange
as it may seem) is making a misstatement—and that when Conroux is said to have
been ordered to attack the hill by swerving to the left, he can only have been
keeping off Pack. I note that Vidal de la Blache and Mr. Fortescue try to accept
Clausel’s story, but that General Beatson (With Wellington in the Pyrenees, pp.
170-2) ignores it.

[943] I include, in reckoning Picton’s force at Bussaco, his own division and the
three battalions of Leith’s first brigade which brought him help. In Cole’s Oricain
figures are reckoned the 4th Division, Byng’s brigade, Campbell’s Portuguese, and
two Spanish regiments.

[944] Lemonnier-Delafosse of the 31st Léger.

[945] This exceptional use of grenadiers in the skirmishing line, I get from an
observation of Bainbrigge of the 20th, who expresses his surprise that the troops
with whom he was engaged, though acting as tirailleurs, were not light infantry, but
men in tall bearskin caps like the Guard, ‘some of the finest-looking soldiers I ever
met’ (p. 400).

[946] The 10th Caçadores, Campbell’s light battalion, was a very weak unit of
only 250 bayonets.

[947] Clausel’s report of August 2.

[948] D’Haw of the 34th Léger.

[949] The fourth battalion of the brigade, the 1/40th was detached below on the
Spaniards’ Hill.

[950] Lemonnier-Delafosse, pp. 227-8.

[951] The Buffs lost only 2 men, the 1st Provisional (2/31st and 2/66th) only 5—so
can hardly have been engaged,—but the 1/57th had 63 casualties.

[952] The above narrative is reconstructed from Reille’s two reports (the divisional
report of Lamartinière, however, is useless) and from narratives of Stretton of the
40th in Maxwell’s Peninsular Sketches, and Mills in the history of the regiment by
Smythies.

[953] There is little about this affair in the British narratives. Diarists were rare in
the 6th Division. The only point of interest I found in them is the mention of mule-
guns used by the French.

[954] Larpent, p. 221. Cf. Napier, v. p. 226: ‘That will give time for the 6th Division
to arrive, and I shall beat him’—words true in thought but perhaps never spoken by
Wellington.

[955] Lapéne, p. 80.

[956] Soult to Clarke, report of the battle.

[957] See statistics in Appendix XXII.

[958] Narrative of Captain G. Wood of the 1/82nd, pp. 192-3.

[959] See Dickson Papers, Tulloh’s letter, p. 1022.

[960] Hill to Quartermaster-General, Supplementary Dispatches, viii. p. 142.

[961] Dalhousie to Quartermaster-General, ibid.

[962] 28th Léger.

[963] 22nd Chasseurs.

[964] Soult’s general orders of July 23rd.

[965] Soult to Clarke, from Zabaldica, evening of the 28th.

[966] See St. Chamans, quoted above, p. 590.

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