Data-Driven and Model-Based Methods for Fault

Detection and Diagnosis 1st Edition - eBook PDF

download

https://ebooksecure.com/download/data-driven-and-model-based-
methods-for-fault-detection-and-diagnosis-ebook-pdf/

Download more ebook from https://ebooksecure.com

We believe these products will be a great fit for you. Click
the link to download now, or visit ebooksecure.com
to discover even more!

(eBook PDF) Bayesian Networks in Fault Diagnosis:

Practice and Application

http://ebooksecure.com/product/ebook-pdf-bayesian-networks-in-
fault-diagnosis-practice-and-application/

Optimal State Estimation for Process Monitoring, Fault

Diagnosis and Control 1st Edition Ch. Venkateswarlu
M.Tech Phd - eBook PDF

https://ebooksecure.com/download/optimal-state-estimation-for-
process-monitoring-fault-diagnosis-and-control-ebook-pdf/

Data-Driven Traffic Engineering: Understanding of

Traffic and Applications Based on Three-Phase Traffic
Theory 1st Edition - eBook PDF

https://ebooksecure.com/download/data-driven-traffic-engineering-
understanding-of-traffic-and-applications-based-on-three-phase-
traffic-theory-ebook-pdf/

Data Driven: Harnessing Data and AI to Reinvent

Customer Engagement 1st Edition Tom Chavez - eBook PDF

https://ebooksecure.com/download/data-driven-harnessing-data-and-
ai-to-reinvent-customer-engagement-ebook-pdf/
(eBook PDF) Model Driven Development for Embedded
Software: Application to Communications for Drone Swarm

http://ebooksecure.com/product/ebook-pdf-model-driven-
development-for-embedded-software-application-to-communications-
for-drone-swarm/

Quantitative systems pharmacology : models and model-

based systems with applications 1st Edition - eBook PDF

https://ebooksecure.com/download/quantitative-systems-
pharmacology-models-and-model-based-systems-with-applications-
ebook-pdf/

Highway Safety Analytics and Modeling: Techniques and

Methods for Analyzing Crash Data 1st edition - eBook
PDF

https://ebooksecure.com/download/highway-safety-analytics-and-
modeling-techniques-and-methods-for-analyzing-crash-data-ebook-
pdf/

Data-driven Solutions to Transportation Problems 1st

Edition- eBook PDF

https://ebooksecure.com/download/data-driven-solutions-to-
transportation-problems-ebook-pdf/

Urgent Care Dermatology: Symptom-Based Diagnosis 1st

Edition - eBook PDF

https://ebooksecure.com/download/urgent-care-dermatology-symptom-
based-diagnosis-ebook-pdf/
DATA-DRIVEN
AND
MODEL-BASED
METHODS FOR
FAULT DETECTION
AND DIAGNOSIS
DATA-DRIVEN
AND
MODEL-BASED
METHODS FOR
FAULT DETECTION
AND DIAGNOSIS

MAJDI MANSOURI
MOHAMED-FAOUZI HARKAT
HAZEM N. NOUNOU
MOHAMED N. NOUNOU
Elsevier
Radarweg 29, PO Box 211, 1000 AE Amsterdam, Netherlands
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom
50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States
Copyright © 2020 Elsevier Inc. All rights reserved.
No part of this publication may be reproduced or transmitted in any form or by any means,
electronic or mechanical, including photocopying, recording, or any information storage and
retrieval system, without permission in writing from the publisher. Details on how to seek
permission, further information about the Publisher’s permissions policies and our arrangements
with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency,
can be found at our website: www.elsevier.com/permissions.
This book and the individual contributions contained in it are protected under copyright by the
Publisher (other than as may be noted herein).
Notices
Knowledge and best practice in this field are constantly changing. As new research and experience
broaden our understanding, changes in research methods, professional practices, or medical
treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in
evaluating and using any information, methods, compounds, or experiments described herein. In
using such information or methods they should be mindful of their own safety and the safety of
others, including parties for whom they have a professional responsibility.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors,
assume any liability for any injury and/or damage to persons or property as a matter of products
liability, negligence or otherwise, or from any use or operation of any methods, products,
instructions, or ideas contained in the material herein.

Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress

British Library Cataloguing-in-Publication Data

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

ISBN: 978-0-12-819164-4

For information on all Elsevier publications

visit our website at https://www.elsevier.com/books-and-journals

Publisher: Susan Dennis

Acquisitions Editor: Anita Koch
Editorial Project Manager: Sara Valentino
Production Project Manager: Vignesh Tamil
Designer: Miles Hitchen
Typeset by VTeX
Contents

List of figures ix
List of tables xvii
About the authors xix
Acknowledgments xxi
List of acronyms xxiii
Nomenclature xxv

1. Introduction 1
References 6

2. PCA and PLS-based generalized likelihood ratio for fault detection 11

2.1. PCA and PLS-based generalized likelihood ratio for fault detection 12
2.1.1. Introduction 12
2.1.2. Principal component analysis (PCA) 13
2.1.3. Fault detection using PCA method 16
2.1.4. Statistical hypothesis testing 18
2.1.5. Fault detection using a PCA-based GLRT 21
2.1.6. PCA-based GLRT and applications 23
2.1.7. Conclusion 36
2.2. PLS-based generalized likelihood ratio for fault detection 37
2.2.1. Introduction 37
2.2.2. Partial Least Square (PLS) method 38
2.2.3. PLS-based GLRT for fault detection 40
2.2.4. PLS-based GLRT fault detection and applications 41
2.2.5. Conclusions 43
References 45

3. Kernel PCA- and Kernel PLS-based generalized likelihood ratio

tests for fault detection 49
3.1. Kernel PCA-based generalized likelihood ratio test for fault detection 49
3.1.1. Introduction 49
3.1.2. Kernel Principal Component Analysis (KPCA) description 50
3.1.3. Fault detection using KPCA method 53
3.1.4. Enhanced monitoring using kernel GLRT chart 54
3.1.5. Kernel GLRT fault detection chart with applications 57
3.1.6. Conclusion 62
3.2. Kernel PLS-based generalized likelihood ratio test for fault detection 63
3.2.1. Introduction 63
3.2.2. Kernel Partial Least Squares (KPLS) method 65

v
vi Contents

3.2.3. KPLS-based GLRT and application to fault detection in CSTR

process 69
3.2.4. Conclusion 73
References 75

4. Linear and nonlinear multiscale latent variable-based

generalized likelihood ratio for fault detection 79
4.1. Linear multiscale latent variable-based generalized likelihood ratio for
fault detection 79
4.1.1. Introduction 79
4.1.2. Multiscale PCA-based GLRT for fault detection 80
4.1.3. Multiscale PLS-based GLRT for fault detection 98
4.1.4. Conclusions 112
4.2. Multiscale nonlinear latent variable-based generalized likelihood ratio
test for fault detection 113
4.2.1. Introduction 113
4.2.2. Multiscale kernel PCA-based GLRT for fault detection 116
4.2.3. Multiscale kernel PLS-based GLRT for fault detection 122
4.2.4. Conclusion 132
References 132

5. Linear and nonlinear interval latent variable approaches for fault

detection 135
5.1. Interval latent variable approaches for fault detection 136
5.1.1. Introduction 136
5.1.2. Interval PCA-based GLRT for fault detection 137
5.1.3. Interval PLS-based GLRT for fault detection 170
5.1.4. Conclusion 185
5.2. Interval nonlinear latent variable approaches for fault detection 188
5.2.1. Introduction 188
5.2.2. Interval kernel PCA-based GLRT for fault detection 190
5.2.3. Interval kernel PLS-based GLRT for fault detection 204
5.2.4. Conclusion 214
References 215

6. Model-based approaches for fault detection 221

6.1. Introduction 221
6.2. State estimation 225
6.2.1. State estimation problem formulation 225
6.2.2. State estimation techniques 226
6.3. Fault detection-based state estimation approaches 230
6.3.1. Fault detection using multiscale EWMA chart 230
6.3.2. Application to wastewater treatment plant 233
 Contents vii

6.4. Fault detection-based state estimation approach 240

6.4.1. Fault detection using optimized weighted SS-DEWMA chart 240
6.4.2. Optimized WSS-DEWMA and application to fault detection 245
6.5. Conclusions 252
References 255

7. Conclusions and perspectives 259

7.1. Conclusions 259
7.2. Perspectives and research proposals 260
7.2.1. Project 1: water distribution networks: modeling, sensor
placement, leak and quality monitoring 260
7.2.2. Project 2: enhanced operation of wastewater treatment plants 266
7.2.3. Project 3: enhanced monitoring of photovoltaic systems 270
7.2.4. Project 4: enhanced data validation of an air quality monitoring
networks 273

Appendix 279
References 286
Index 289
List of figures

Figure 2-1 Geometric Interpretation of PCA. 17

Figure 2-2 Principle of PCA projection. 17
Figure 2-3 Outliers detection using PCA. 17
Figure 2-4 A schematic diagram of the PCA-based GLRT fault detection
method. 22
Figure 2-5 (A) Quarter-hourly ozone measurements and (B) ACF of ozone
measurements. 25
Figure 2-6 Variance captured by each principal component. 26
Figure 2-7 PC2 versus PC1. 26
Figure 2-8 PCA model prediction of ozone concentrations for the first
four network stations. 27
Figure 2-9 PCA model prediction of ozone concentrations for the last
three network stations. 28
Figure 2-10 Monitoring a simple sensor fault using PCA-based Q. 30
Figure 2-11 Monitoring a simple sensor fault using PCA-based GLRT. 30
Figure 2-12 Monitoring a multiple faults using PCA-based Q. 31
Figure 2-13 Monitoring a multiple faults using PCA-based GLRT. 31
Figure 2-14 The variance captured by each principal component. 32
Figure 2-15 Histograms showing the normality of the residuals. 33
Figure 2-16 Monitoring a fault in CA using PCA-based Q. 34
Figure 2-17 Monitoring a fault in CA using PCA-based GLRT. 34
Figure 2-18 Monitoring a fault in T using PCA-based Q. 35
Figure 2-19 Monitoring a fault in T using PCA-based GLRT. 35
Figure 2-20 Monitoring of multiple faults in CA and T using PCA-based Q. 36
Figure 2-21 Monitoring of multiple faults in CA and T using PCA-based
GLRT. 36
Figure 2-22 Correct detection rate vs false alarm rate for the PCA-based
GLRT fault detection method and the conventional PCA
method. 37
Figure 2-23 The time evolution of the generated data X . 42
Figure 2-24 Monitoring fault in temperature using PLS-based Q chart. 42
Figure 2-25 Monitoring fault in temperature using PLS-based GLRT chart. 43
Figure 2-26 Monitoring TEP IDV 2 fault using PLS-based Q chart. 43
Figure 2-27 Monitoring TEP IDV 2 fault using PLS-based GLRT chart. 44
Figure 3-1 Schematic illustration of KGLRT algorithm. 58
Figure 3-2 Evolutions of SPE, KGLRT, and EWMA-KGLRT in faulty case. 59
Figure 3-3 Evolutions of SPE, KGLRT, and EWMA-KGLRT in faulty case. 60

ix
x List of figures

Figure 3-4 The time evolution of the SPE (A), KGLRT (B) and
EWMA − KGLRT (C) statistics on a semi-logarithmic scale
in the presence of a single fault in T . 61
Figure 3-5 The time evolution of the SPE (A), KGLRT (B) and
EWMA − KGLRT (C) statistics on a semi-logarithmic scale
in the presence of simultaneous faults in CA and T . 62
Figure 3-6 KPLS diagram for nonlinear regression. 67
Figure 3-7 The time evolution of PLS-based Q statistic on a
semilogarithmic scale in the presence of a fault in CA . 69
Figure 3-8 The time evolution of KPLS-based Q statistic on a
semilogarithmic scale in the presence of a fault in CA . 70
Figure 3-9 The time evolution of PLS-based GLRT statistic on a
semilogarithmic scale in the presence of a fault in CA . 70
Figure 3-10 The time evolution of KPLS-based GLRT statistic on a
semilogarithmic scale in the presence of a fault in CA . 71
Figure 3-11 The time evolution of PLS-based Q statistic on a
semilogarithmic scale in the presence of a fault in T . 71
Figure 3-12 The time evolution of KPLS-based Q statistic on a
semilogarithmic scale in the presence of a fault in T . 71
Figure 3-13 The time evolution of PLS-based GLRT statistic on a
semilogarithmic scale in the presence of a fault in T . 72
Figure 3-14 The time evolution of KPLS-based GLRT statistic on a
semilogarithmic scale in the presence of a fault in T . 72
Figure 3-15 The time evolution of PLS-based Q statistic on a
semilogarithmic scale in the presence of simultaneous faults
in CA and T . 73
Figure 3-16 The time evolution of KPLS-based Q statistic on a
semilogarithmic scale in the presence of simultaneous faults
in CA and T . 73
Figure 3-17 The time evolution of PLS-based GLRT statistic on a
semilogarithmic scale in the presence of simultaneous faults
in CA and T . 74
Figure 3-18 The time evolution of KPLS-based GLRT statistic on a
semilogarithmic scale in the presence of simultaneous faults
in CA and T . 74
Figure 4-1 A schematic diagram of data representation at multiple
scales [1]. 81
Figure 4-2 Schematic illustration of MSPCA model. 83
Figure 4-3 Schematic illustration of proposed MSPCA-based MW-GLRT
algorithm. 86
 List of figures xi

Figure 4-4 The effect of the window length on the performance of

the moving-window GLRT (the fault size equals twice the
residuals standard deviation). (A) Missed detection rate, (B)
false alarm rate and (C) average run length (ARL) vs. window
length. 88
Figure 4-5 Monitoring a fault of magnitude unity using PCA-based T 2
chart. 89
Figure 4-6 Monitoring a fault of magnitude unity using MSPCA-based
T 2 chart. 90
Figure 4-7 Monitoring a fault of magnitude unity using PCA-based Q
chart. 90
Figure 4-8 Monitoring a fault of magnitude unity using MSPCA-based Q
chart. 90
Figure 4-9 Monitoring a fault of magnitude unity using MSPCA-based
GLRT chart. 91
Figure 4-10 Monitoring a fault of magnitude unity using MSPCA-based
MW-GLRT chart (WL = 4). 91
Figure 4-11 Monitoring a fault of magnitude unity using MSPCA-based
MW-GLRT chart (WL = 8). 91
Figure 4-12 Monitoring a fault of magnitude 1σ using PCA-based T 2 chart. 92
Figure 4-13 Monitoring a fault of magnitude 1σ using MSPCA-based T 2
chart. 92
Figure 4-14 Monitoring a fault of magnitude 1σ using PCA-based Q chart. 92
Figure 4-15 Monitoring a fault of magnitude 1σ using MSPCA-based Q
chart. 93
Figure 4-16 Monitoring a fault of magnitude 1σ using MSPCA-based
GLRT chart. 93
Figure 4-17 Monitoring a fault of magnitude 1σ using MSPCA-based
MW-GLRT chart (WL = 4). 93
Figure 4-18 Monitoring a fault of magnitude 1σ using MSPCA-based
MW-GLRT chart (WL = 8). 94
Figure 4-19 Monitoring TEP fault 12 using PCA-based T 2 chart. 101
Figure 4-20 Monitoring TEP fault 12 using MSPCA-based T 2 chart. 102
Figure 4-21 Monitoring TEP fault 12 using PCA-based Q chart. 102
Figure 4-22 Monitoring TEP fault 12 using MSPCA-based Q chart. 102
Figure 4-23 Monitoring TEP fault 12 using MSPCAbased GLRT chart. 103
Figure 4-24 Monitoring TEP fault 12 using MSPCA-based MW-GLRT chart
(WL = 4). 103
Figure 4-25 Monitoring TEP fault 12 using MSPCA-based MW-GLRT chart
(WL = 8). 103
xii List of figures

Figure 4-26 Monitoring a fault of magnitude unity using the

MSPCA-based MW-GLRT (WL = 2) chart. 104
Figure 4-27 Monitoring a fault of magnitude unity using the
MSPCA-based MW-GLRT (WL = 4) chart. 104
Figure 4-28 Monitoring a fault of magnitude unity using the
MSPCA-based MW-GLRT (WL = 8) chart. 104
Figure 4-29 Monitoring a fault of magnitude unity using the
MSPCA-based EWMA-GLRT chart. 105
Figure 4-30 Monitoring a fault of magnitude 1σ using the MSPCA-based
MW-GLRT (WL = 2) chart. 105
Figure 4-31 Monitoring a fault of magnitude 1σ using the MSPCA-based
MW-GLRT (WL = 4) chart. 105
Figure 4-32 Monitoring a fault of magnitude 1σ using the MSPCA-based
MW-GLRT (WL = 8) chart. 106
Figure 4-33 Monitoring a fault of magnitude 1σ using the MSPCA-based
EWMA-GLRT chart. 106
Figure 4-34 Representation of MSPLS fault detection model. 107
Figure 4-35 Monitoring multiple faults in temperature using PLS-based Q
chart. 109
Figure 4-36 Monitoring multiple faults in temperature using MSPLS-based
Q chart. 110
Figure 4-37 Monitoring multiple faults in temperature using PLS-based
GLRT chart. 110
Figure 4-38 Monitoring multiple faults in temperature using MSPLS-based
GLRT chart. 111
Figure 4-39 Monitoring TEP IDV 2 fault using PLS-based Q chart. 113
Figure 4-40 Monitoring TEP IDV 2 fault using MSPLS-based Q chart. 114
Figure 4-41 Monitoring TEP IDV 2 fault using PLS-based GLRT chart. 114
Figure 4-42 Monitoring TEP IDV 2 fault using MSPLS-based GLRT chart. 115
Figure 4-43 Representation of MSKPCA fault detection model. 116
Figure 4-44 Evolutions of SPE, KGLRT, and MS-KGLRT in faulty case. 118
Figure 4-45 Evolutions of SPE, KGLRT, and MS-KGLRT in faulty case. 118
Figure 4-46 Evolutions of SPE, KGLRT, and MS-KGLRT in faulty case. 119
Figure 4-47 Monitoring faults (A) IDV1 , (B) IDV2 , (C) IDV3 and (D) IDV4
using SPE, KGLRT, and MS-KGLRT charts. 121
Figure 4-48 Monitoring faults (A) IDV5 , (B) IDV6 , (C) IDV7 and (D) IDV8
using SPE, KGLRT, and MS-KGLRT charts. 122
Figure 4-49 Monitoring faults (A) IDV9 , (B) IDV10 , (C) IDV11 and (D) IDV12
using SPE, KGLRT, and MS-KGLRT charts. 123
Figure 4-50 Monitoring faults (A) IDV13 , (B) IDV14 , (C) IDV15 and (D) IDV16
using SPE, KGLRT, and MS-KGLRT charts. 124
 List of figures xiii

Figure 4-51 Monitoring faults (A) IDV17 , (B) IDV18 , (C) IDV19 , (D) IDV20 and
(E) IDV21 using SPE, KGLRT, and MS-KGLRT charts. 125
Figure 4-52 Representation of MSKPLS fault detection model. 126
Figure 4-53 The time evolution of the generated data X . 128
Figure 4-54 Time evolution of detection using PLS-based GLRT method. 128
Figure 4-55 Time evolution of detection using KPLS-based GLRT method. 129
Figure 4-56 Time evolution of detection using MSPLS-based GLRT method. 129
Figure 4-57 Time evolution of detection using MSKPLS-based GLRT
method. 130
Figure 4-58 Time evolution of detection using KPLS-based GLRT method. 131
Figure 4-59 Time evolution of detection using MSKPLS-based GLRT
method. 131
Figure 5-1 Time evolution of simulated interval-valued data. 150
Figure 5-2 Measurements and estimations of interval-valued data using
CIPCA model. 151
Figure 5-3 Time evolution of univariate interval GLR. 151
Figure 5-4 Time evolution of univariate interval weighted GLR. 152
Figure 5-5 Time evolution of multivariate GLR. 152
Figure 5-6 Time evolution of multivariate interval weighted GLR. 152
Figure 5-7 Distillation column interval-valued measurements. 153
Figure 5-8 Evolution of the VIRE with respect to the number of principal
components . 154
Figure 5-9 Measurements and estimations. 154
Figure 5-10 Time evolution of univariate interval GLR. 155
Figure 5-11 Time evolution of univariate interval weighted GLR. 155
Figure 5-12 Time evolution of multivariate interval GLR. 155
Figure 5-13 Time evolution of multivariate interval weighted GLR. 156
Figure 5-14 Ozone concentrations for single-valued and interval-valued
data. 157
Figure 5-15 Evolution of VIRE with respect to the number of principal
components. 157
Figure 5-16 Measurements and estimations of O3 station 1. 158
Figure 5-17 Measurements and estimations of O3 station 3. 158
Figure 5-18 Time evolution of univariate interval GLR with a fault on x7 . 159
Figure 5-19 Time evolution of univariate interval weighted GLR with a
fault on x7 . 159
Figure 5-20 Time evolution of multivariate interval GLR with a fault on x7 . 159
Figure 5-21 Time evolution of multivariate interval weighted GLR with a
fault on x7 . 160
Figure 5-22 Time evolution of simulated data. 163
xiv List of figures

Figure 5-23 The time evolution of the MRPCA-based SPE statistic in the
presence of faults in x2 and x3 . 165
Figure 5-24 The time evolution of the MRPCA-based EWMA statistic in the
presence of faults in x2 and x3 . 165
Figure 5-25 The time evolution of the MRPCA-based GLRT statistic in the
presence of faults in x2 and x3 . 166
Figure 5-26 The time evolution of the MRPCA-based EWMA statistic in the
presence of faults in x2 and x3 . 166
Figure 5-27 Measurements and estimations of O3 station 1. 167
Figure 5-28 Measurements and estimations of O3 station 3. 167
Figure 5-29 The time evolution of the MRPCA-based SPE statistic in the
presence of faults in O3 . 168
Figure 5-30 The time evolution of the MRPCA-based EWMA statistic in the
presence of faults in x2 and x3 . 168
Figure 5-31 The time evolution of the MRPCA-based GLRT statistic in the
presence of faults in O3 . 169
Figure 5-32 The time evolution of the MRPCA-based EWMA statistic in the
presence of faults in O3 . 169
Figure 5-33 Time evolution of interval-valued simulated variables. 177
Figure 5-34 Scatter plots of predicted and observed training data y1 . 177
Figure 5-35 Scatter plots of predicted and observed training data y2 . 178
Figure 5-36 Evolution of Q1,x in both fault-free and faulty cases. 178
Figure 5-37 Evolution of Q1,y in both fault-free and faulty cases. 179
Figure 5-38 Evolution of Q2,x in both fault-free and faulty cases. 179
Figure 5-39 Evolution of Q2,y in both fault-free and faulty cases. 180
Figure 5-40 Evolution of Q3,x in both fault-free and faulty cases. 180
Figure 5-41 Evolution of Q3,y in both fault-free and faulty cases. 181
Figure 5-42 Evolution of Q4,x in both fault-free and faulty cases. 181
Figure 5-43 Evolution of Q4,y in both fault-free and faulty cases. 182
Figure 5-44 Evolution of I Gx in both fault-free and faulty cases. 182
Figure 5-45 Evolution of I Gy in both fault-free and faulty cases. 183
Figure 5-46 Distillation column interval-valued measurements. 184
Figure 5-47 Evolution of Q1,x and Q1,y with a fault on variable x2 . 184
Figure 5-48 Evolution of Q2,x and Q2,y with a fault on variable x2 . 185
Figure 5-49 Evolution of Q3,x and Q3,y with a fault on variable x2 . 185
Figure 5-50 Evolution of Q4,x and Q4,y with a fault on variable x2 . 186
Figure 5-51 Evolution of IGLRx and IGLRy with a fault on variable x2 . 186
Figure 5-52 Evolution of Q4,x and Q4,y with a fault on variable y2 . 187
Figure 5-53 Evolution of IGLRx and IGLRy with a fault on variable y2 . 188
Figure 5-54 3-D scatter plot of the generated interval-valued data. 199
 List of figures xv

Figure 5-55 Time evolution of indices T 2 and SPE in KPCA model with a
fault on variable x1 . 200
Figure 5-56 Time evolution of indices T 2 and SPE in IKPCACR model
with a fault on variable x2 . 200
Figure 5-57 Time evolution of x3 . 201
Figure 5-58 Time evolution of indices T 2 and SPE in KPCA model with
IDV-1 fault. 202
Figure 5-59 Time evolution of indices T 2 and SPE in IKPCA model with
IDV-1 fault. 203
Figure 5-60 Time evolution of indices T 2 and SPE in IKPCA model with
IDV-1 fault. 203
Figure 5-61 Time evolution of indices IGLRTUL in IKPCA model with IDV-1
fault. 204
Figure 5-62 Time evolution of indices IGLRTCR in IKPCA model with IDV-1
fault. 204
Figure 5-63 Time evolution of univariate GLRT index based on IKPLSUL
model. 211
Figure 5-64 Time evolution of multivariate GLRT index based on IKPLSUL
model. 211
Figure 5-65 Time evolution of univariate GLRT index based IKPLSCR model. 213
Figure 5-66 Time evolution of multivariate GLRT index based on IKPLSCR
model. 213
Figure 6-1 Plots of samples of normal and faulty signals. 222
Figure 6-2 General flow-chart of the multiobjective optimization process. 232
Figure 6-3 Multiscale EWMA strategy. 233
Figure 6-4 PF-based MS-EWMA fault detection strategy. 234
Figure 6-5 State estimation of the variables (A) XDCO , (B) SO and (C)
XBH using UKF and PF. 236
Figure 6-6 State estimation of the variables (A) SNH , (B) SNO and (C)
XBA using UKF and PF. 237
Figure 6-7 Monitoring a bias fault in SO using (A) Shewhart, (B) EWMA,
and (C) MS-EWMA methods. 239
Figure 6-8 Monitoring a drift fault in SO using (A) Shewhart, (B) EWMA,
and (C) MS-EWMA methods. 241
Figure 6-9 Monitoring a fault in X using (A) EWMA, (B) SS-DEWMA, and
(C) OWSS-DEWMA charts. 247
Figure 6-10 Qualitative model of the CSEC (simplified). 248
Figure 6-11 Estimation of state variables using various state estimation
techniques. 249
Figure 6-12 Monitoring a multiple faults in cadaverine Cadav using (A)
EWMA, (B) SS-DEWMA and (C) OWSS-DEWMA charts. 251
xvi List of figures

Figure 6-13 Monitoring a fault in Cadav using (A) EWMA, (B) SS-DEWMA
and (C) OWSS-DEWMA charts. 253
Figure 6-14 Monitoring a multiple faults in Cadav and Lys using (A)
EWMA, (B) SS-DEWMA and (C) OWSS-DEWMA charts. 254
Figure 1 Tennessee Eastman process. 279
Figure 2 Basic distillation column controlled with LV-configuration. 283
Figure 3 Ozone concentrations for the first three stations. 285
Figure 4 Ozone concentrations for the first station, single-valued and
interval-valued representations. 285
List of tables

Table 2-1 Summary of MDRs (%), FARs (%) and ARL1 . 42

Table 2-2 Missed detection rates (MDR %), False alarm rates (FAR %),
and ARL1 values for TEP data. 44
Table 3-1 FAR %, MDR %, and ARL1 for the presented fault detection
charts (simulated example). 60
Table 3-2 FAR % and MDR % for the presented fault detection charts
(CSTR process) for a single fault. 63
Table 3-3 FAR % and MDR % for the presented fault detection charts
(CSTR process) for multiple faults. 63
Table 4-1 Summary of missed detection (%), false alarms (%), and
ARL1 for simulated data using PCA and MSPCA models. 95
Table 4-2 Summary of missed detection rates (%) for TEP data using
PCA and MSPCA. 96
Table 4-3 Summary of missed detection rates (%) for TEP data using
MSPCA-based GLRT and MW-GLRT. 97
Table 4-4 Summary of false alarm rates (%) for TEP data using PCA and
MSPCA. 98
Table 4-5 Summary of false alarm rates (%) for TEP data using
MSPCA-based GLRT and MW-GLRT. 99
Table 4-6 Summary of ARL1 for TEP data using PCA and MSPCA. 100
Table 4-7 Summary of ARL1 for TEP data using MSPCA-based GLRT and
MW-GLRT. 101
Table 4-8 Summary of missed detection (%), false alarms (%), and ARL1
for simulated data using MSPCA. 106
Table 4-9 Summary of MDR (%), FAR (%), and ARL1 . 110
Table 4-10 Missed detection rates (%) for TEP data. 111
Table 4-11 False alarm rates (%) for TEP data. 112
Table 4-12 ARL1 values for TEP data. 113
Table 4-13 FAR %, MDR %, and ARL1 for the presented fault detection
charts (simulated example). 118
Table 4-14 FAR %, MDR %, and ARL1 for the presented fault detection
charts (CSTR process). 119
Table 4-15 Summary of MDR, FAR, and ARL1 values for TEP data. 120
Table 4-16 Summary of missed detection rate (%), false alarm rate (%),
and ARL1 . 130
Table 4-17 Summary of missed detection rate (%), false alarm rate (%),
and ARL1 (CSTR process). 132

xvii
xviii List of tables

Table 5-1 VIRE of different interval-valued for CIPCA model. 150

Table 5-2 FAR, MDR, and ARL1 . 153
Table 5-3 FAR, MDR rates, and ARL1 for the distillation column. 156
Table 5-4 FAR, MDR, and ARL1 for air quality data. 160
Table 5-5 MSE using CPCA, CIPCA, and MRPCA models. 164
Table 5-6 Summary of missed detection (MDR %), false alarms (FAR%)
and ARL1 . 166
Table 5-7 Summary of missed detection (MDR %), false alarms (FAR %),
and ARL1 . 169
Table 5-8 FAR % and MDR % for the presented fault detection charts
(simulated example). 183
Table 5-9 FAR % and MDR % for the presented fault detection charts
(scenario 1 of distillation column). 187
Table 5-10 FAR % and MDR % for the presented fault detection charts
(scenario 2 of distillation column). 188
Table 5-11 Summary of MDR and FAR values for the simulation example. 201
Table 5-12 Missed detection rate (MDR %) and False Alarm Rate (FAR %)
values for TEP data sets. 205
Table 5-13 Selected monitored variables in the TE process. 210
Table 5-14 Selected output measured variables in the TE process. 211
Table 5-15 Missed detection rate (MDR %) and false alarm rate (FAR %)
values for TEP data sets using IKPCA-based GLRT approach. 214
Table 5-16 Missed detection ratio (MDR %)-based Q and GLRT for the 21
faults of the TEP. 215
Table 6-1 Comparison of the MSE for the UKF and PF techniques. 238
Table 6-2 Summary of MDR (%) and FAR (%) for bias fault. 240
Table 6-3 Summary of MDR (%) and FAR (%) for drift fault. 242
Table 6-4 Summary of MDR (%) and FAR (%) for different values of s. 242
Table 6-5 Summary of MDR (%) and FAR (%) for different values of a. 242
Table 6-6 MDR (%) and FAR (%) evaluation. 248
Table 6-7 CSEC parameters. 249
Table 6-8 RMSE of estimated states using EKF, UKF, and PF methods. 250
Table 6-9 MDR (%) and FAR (%) evaluation. 252
Table 6-10 MDR (%) and FAR (%) evaluation. 252
Table 6-11 MDR (%) and FAR (%) evaluation. 255
Table 1 Manipulated variables. 280
Table 2 Measured variables. 281
Table 3 Process faults of TEP. 282
Table 4 Distillation column process variables. 284
About the authors

Majdi Mansouri Dr. Majdi Mansouri received the engineering degree in

Electrical Engineering in 2006 from the Higher School of Communication
of Tunisia (SUPCOM), Tunisia. He received his master degree of Electrical
Engineering from the School of Electronics, Informatics and Radiocom-
munications in Bordeaux (ENSEIRB), France, in 2008. He received his
PhD degree of Electrical Engineering from the University of Technology
of Troyes (UTT), France, in 2011. In December 2019, he received the
degree of HDR (Accreditation To Supervise Research) of Applied Mathe-
matics and Statistics for Electrical Engineering from University of Orleans
in France. He joined the Electrical Engineering Program at Texas A&M
University at Qatar, in 2011, where he is currently an Associate Research
Scientist.
He has over ten years of research and practical experience in systems
engineering and signal processing. His work focuses on the utilization
of applied mathematics and statistics concepts to develop statistical data
and model driven techniques and algorithms for modeling, estimation,
fault detection, fault classification, monitoring and diagnosis, which aim
to improve process operations and enhance the data validation. Dr. Majdi
Mansouri is the author of more than 150 refereed journal and conference
publications and book chapters, and has worked on several projects as lead
principal investigator (LPI) and principal investigator (PI). Dr. Mansouri is
a member of IEEE.

Mohamed-Faouzi Harkat Dr. Mohamed-Faouzi Harkat received his

Eng. degree in Automatic control from Annaba University, Algeria in 1996,
his Ph.D. degree from Institut National Polytechnique de Lorraine (INPL),
France, in 2003. He is now Professor in the Department of Electronics at
Annaba University, Algeria. His research interests include fault diagnosis,
process modeling and monitoring, multivariate statistical approaches, and
neural networks. Dr. Harkat is the author of more than 100 refereed jour-
nal and conference publications and book chapters.

Hazem Numan Nounou Dr. Hazem N. Nounou (SM’08) is a professor

in the Electrical and Computer Engineering Program and the Assistant
Dean for Academic and Student Services at Texas A&M University at
xix
xx About the authors

Qatar. In 2015–2017, he was the holder of Itochu Professorship. He re-

ceived the B.S. degree (Magna Cum Laude) from Texas A&M University,
College Station, in 1995, and the M.S. and Ph.D. degrees from Ohio State
University, Columbus, in 1997 and 2000, respectively, all in electrical en-
gineering. In 2001, he was a Development Engineer for PDF Solutions, a
consulting firm for the semiconductor industry, in San Jose, CA. Then, in
2001, he joined the Department of Electrical Engineering at King Fahd
University of Petroleum and Minerals in Dhahran, Saudi Arabia, as an
Assistant Professor. In 2002, he moved to the Department of Electrical
Engineering, United Arab Emirates University, Al-Ain, UAE. In 2007, he
joined the Electrical and Computer Engineering Program at Texas A&M
University at Qatar, Doha, Qatar, where he is currently a professor. He
published more than 200 refereed journal and conference papers and book
chapters. He served as Associate Editor in technical committees of sev-
eral international journals and conferences. His research interests include
data-based control, monitoring and fault detection, intelligent and adaptive
control, control of time-delay systems, system biology, and system identifi-
cation and estimation. Dr. Nounou is a senior member of IEEE.

Mohamed Numan Nounou Dr. Mohamed Nounou is a professor of

Chemical Engineering at Texas A&M University-Qatar. He received the
B.S. degree (Magna Cum Laude) from Texas A&M University, College
Station, in 1995, and the M.S. and Ph.D. degrees from the Ohio State
University, Columbus, in 1997 and 2000, respectively, all in chemical en-
gineering. From 2000 to 2002 he was with PDF Solutions, a consulting
company for the semiconductor industry, in San Jose, CA. In 2002, he
joined the Department of Chemical and Petroleum Engineering at the
United Arab Emirates University. In 2006, he joined the Chemical Engi-
neering Program at Texas A&M University at Qatar, where he is currently
a professor. He has received research funding over $5M and published more
than 190 refereed journal and conference papers and book chapters. He also
served as Associate Editor in technical committees of several international
journals and conferences. His research interests include process modeling,
monitoring, estimation, system biology, and intelligent control. He is a se-
nior member of the American Institute of Chemical Engineers (AIChE)
and a senior member of the Institute of Electrical and Electronics Engi-
neers (IEEE).
Acknowledgments

This work was made possible by NPRP grant NPRP9-330-2-140 from

the Qatar National Research Fund (a member of Qatar Foundation). The
statements made herein are solely the responsibility of the authors.

xxi
List of acronyms

GLRT Generalized likelihood ratio test

KGLRT Kernel generalized likelihood ratio test
EWMA Exponential weighted moving average
CUSUM Cumulative sum
SPE Squared prediction error
LVR Latent variable regression
MSE Mean square error
PCA Principal component analysis
PLS Partial least square
MSPLS Multiscale partial least square
MSPCA Multiscale principal component analysis
IPCA Interval principal component analysis
KPCA Kernel principal component analysis
KPLS Kernel partial least square
MSPLS Multiscale kernel partial least square
MSKPCA Multiscale kernel principal component analysis
IKPCA Interval kernel principal component analysis
CPCA Centers principal component analysis
CPLS Centers Partial least square
MRPCA Midpoints-radii principal component analysis
MRPLS Midpoints-radii partial least square
CIPCA Complete information principal component analysis
CIPLS Complete information partial least square
SNR Signal-to-noise ratio
CSTR Continuous stirred tank reactor
AQMN Air quality monitoring network
TEP Tennessee Eastman process
FAR False alarm rate
MDR Missed detection rate
ARL1 Average run length
WL Window length
MW Moving window

xxiii
Nomenclature

Latin letters
X ∈ RN ×m Input data matrix
x ∈ Rm Input vector
y ∈ Rp Output vector
t ∈ R Latent variable
P∈R Eigenvector matrix
p Eigenvector
I Identity matrix
E Residual matrix
e Residual vector
m Number of inputs
n Number of samples
 Number of retained principal components
T2 Hotelling statistic
Q Squared prediction error statistic
T Generalized likelihood ratio test statistic
w Window length

Greek letters
 Eigenvalue matrix
 Model error
λ Eigenvalue
σ Standard deviation
φ(·) Nonlinear mapping
ψ Wavelet basis function
φ Orthonormal scaling function

xxv
CHAPTER 1

Introduction
Process monitoring is essential for proper and safe operation of various
industrial processes (like chemical and environmental processes), and it has
recently become even more important than ever before.
Proper operation of complex chemical processes, such as those in the oil
and gas industries, requires careful monitoring of certain key process vari-
ables to enhance the productivity of these processes and, more importantly,
to avoid disasters in the cases of failure [1]. Many serious accidents have
occurred in the past few decades in various chemical and petrochemical
plants all over the world. These accidents include the Union Carbide ac-
cident [2,3], the Piper Alpha accident [4,5], and the Al-Ahmedi (Kuwait)
accident [6]. The Union Carbide accident occurred in Bhopal, India, in
1984, where a major toxic gas leak resulted in over 3000 fatalities and in-
jured 400,000 others in the surrounding neighborhoods [2,3]. The 1988
accident in Piper Alpha (an oil production plant operated by Occidental
Chemical in the North Sea) involved an explosion killing 167 men, leaving
only 61 survivors [4,5]. The accident in Mina Al-Ahmedi in 2000, on the
other hand, was due to a failure in a condensate line in a refinery plant caus-
ing the death of 5 people and injuring 50 others [6]. These accidents show
that tight monitoring of chemical and petrochemical processes is essential
for safe and profitable operation of these plants.
Also, monitoring the atmospheric air pollution levels is extremely im-
portant for the safety of humans and the marine life, especially in areas with
large fuel productions or consumptions and large climate fluctuations [7].
For example, the heat wave in France in the summer of 2003 was linked to
an exceptional ozone pollution that affected the entire European commu-
nity [8]. The consequences of this heat wave demonstrated the importance
of having reliable warning systems to detect unexpected pollution levels
and any unforeseeable events [8]. Proper monitoring of air pollutants pro-
vides useful information that can help people take the needed precautions
to avoid undesirable consequences. During the past few decades, a lot of
effort has been made to improve air quality.
Fault detection is often used for process monitoring. Possible faults can
be due to malfunctioning sensor/s (called sensor faults) or to abnormal
changes in the process. Sensor faults are usually quantified by sudden (or
Data-Driven and Model-Based Methods for Fault Detection and Diagnosis Copyright © 2020 Elsevier Inc.
https://doi.org/10.1016/B978-0-12-819164-4.00010-8 All rights reserved. 1
2 Data-Driven and Model-Based Methods for Fault Detection and Diagnosis

quick) changes in a small number of process variables. Process faults, on

the other hand, are abnormal changes caused by deviations in the process
itself. These faults are usually quantified by slow drifts across several vari-
ables. The need for monitoring techniques that can accurately and quickly
detect abnormal situations (sensor or process faults) has greatly attracted the
attention of researchers and engineers. Over the past few decades, several
monitoring techniques have been developed [6,9–11]. Generally, fault de-
tection techniques can be classified into two main categories: data-based or
model-free techniques and model-based techniques.
Model-based monitoring methods rely on comparing the process mea-
surements with knowledge obtained from a mathematical process model,
which is usually derived using some fundamental understanding of the
process under fault-free conditions. The residuals, which are the differ-
ences between the measurements and the model predictions, can be used
as an indicator about the existence or absence of faults [12,13]. When the
monitored process is under normal operating conditions (no faults exist),
the residuals are zero or close to zero in cases of modeling uncertainties and
measurement noise. However, when a fault occurs, the residuals deviate sig-
nificantly from zero indicating the presence of a new condition that is sig-
nificantly distinguishable from the normal faultless working mode [12,13].
The model-based monitoring approaches include the observer-based meth-
ods [14,15], parity space approaches [16–19], and interval approaches [20].
Of course, the effectiveness of these model-based monitoring methods de-
pends on the accuracy of the models used.
The effective performance of various practical systems requires proper
process operations, such as modeling and monitoring. In these operations, it
is often assumed that the state variables of the process model are measurable
and that the model parameters are available. In many cases, however, obtain-
ing such measurements or determining the model parameters can be costly,
difficult, or sometimes impossible. To deal with this problem, state and/or
parameter estimators are often utilized. Several estimation techniques, such
as the extended Kalman filter, unscented Kalman filter, and more recently
the sequential Monte Carlo method have been developed and utilized in
many applications. The classical Kalman filter (KF) was developed in the
1960s [21] and has been widely applied in various engineering and science
areas, including communications, control, machine learning, neuroscience,
and many others. In the case where the model describing the system is
assumed to be linear and Gaussian, the KF provides an optimal solution
[22–25]. KF has also been formulated in the context of Takagi–Sugeno
Discovering Diverse Content Through
Random Scribd Documents
luck that adheres to you!’ I could have loved that woman. It was so
happy, so neat. That bit of wool drew Sulla and the Dictatorship to
her. You, what a blunderer you are. You have offended Domitian,
who may some day be greater than was Sulla, when you had it in
your power by a word, a look, a dimpled smile, to win him, and with
him the purple.”

“Mother, I do not covet it. You forget—I am promised to Lucius Ælius

Lamia.”

“Oh! Lamia! He could be bought off with a proconsulship.”

“I do not desire to be separated from him. I love him, and have

loved him since we were children together.”

“Well, you have done for your chances. If I surmise aright, the
young man entertains a great grudge against you.”

At that moment Eboracus came in.

“Madam,” said he, “the Illyrian legions have entered the city, under
Primus, and there is fighting in the streets. The people on the
housetops cheer on this side or that, as though they were at a show
of gladiators.”

“Well—those things happen. We shall know for certain which shall be

uppermost, and if fate favors Vitellius—Then, daughter, I shall not
scruple to give the young man up.”

The condition of the capital was frightful. Vitellius had called in levies
from the country to support him, and the prætorian soldiers stood
firm. But many men of direction were with the partisans of
Vespasian, who advanced steadily over the bodies of the troops
opposing them. Fifty thousand persons lost their lives in these
eventful days of the Saturnalia.
The legions under Primus succeeded in recapturing the Capitol,
which was still smoking, and pushed forward into the Forum.

Meanwhile, Vitellius, in the Palatine palace, a prey to irresolution,

had filled himself with wine, and then fled along with his cook and
pastrycook to his wife’s house on the Aventine. Then deceived by a
false report that his troops were successful, he returned to the
Palatine, and found it deserted, but a roar of voices rose from the
Forum below, and from the Capitol the cries of the legionaries were
wafted towards him along with the smoke.

He hastened to collect all the gold he could lay his hands on, stuffed
it into his cincture, assumed an old ragged suit, and then again
attempted to escape; but now he found every avenue blocked. Filled
with terror he crawled into the dog-kennel where the hounds,
resenting the intrusion, fell on him and bit his neck and hands and
legs. But now Vespasian’s soldiery invaded the palace, and a tribune,
Julius Placidius, dis covering the bloated, bleeding wretch, drew him
out by the foot, and he came forth thus, his hands full of dirty straw,
and strands adhering to his hair and garments. A howling rabble at
once surrounded him, leaping, jeering, throwing mud and stones; a
few soldiers succeeded in surrounding him. His hands were bound
behind his back, and a rope passed about his neck. Thus he was
dragged through the streets an object of insult to the people. Some
struck him in the face, some plucked out his hair. In the Forum the
rabble were breaking his statues and dragging them about. One
ruffian thrust a pike under the unfortunate prince’s chin and bade
him hold up his head. Then said Vitellius:—

“Thou, who thus addressest me—a tribune thou art, remember I

was once thy commander!”

Thereupon a German soldier, desirous of shortening his misery,

struck him down with a blow of his sword, and in so doing cut off
the ear of the tribune who had insulted the fallen Emperor.
At once the body of the prince, from whom the life was not sped,
was dragged to the Gemonian stair, a flight of steps down which the
corpses of malefactors were flung, and there he was despatched
with daggers.

Longa Duilia had been kept well informed as to all that took place.

No sooner was she assured that Vitellius was dead, than she rushed
into the apartment given up to Domitian.

“Salve, Cæsar! As the Gods love me, I am the first to so salute you,
son of the Augustus! Oh, I am so happy! And it might have been
otherwise, but you they never would have reached save over my
body.”
 CHAPTER XX.
CHANGED TACTICS.

The anarchy which had lasted from the 11th June, 68, when Nero
perished, came to an end on the 20th December, in the ensuing
year. In that terrible year of 69, three emperors had died violent
deaths, and Rome had been in a condition of disorder on each
occasion, and intermittent violence had lasted all the time. Men now
drew a long breath, they were disposed to blot out the memory of
those eighteen months of misery and national humiliation, as though
it had not been, and to reckon the strong Vespasian as prince next
after Nero. Indeed, on the morrow of the death of Vitellius, when
the Senate assembled and decreed the honors of the former princes,
they recited those of the first Cæsars, but ignored the three last who
had perished within a twelvemonth, as though they had never been,
and were to be forgotten as an evil dream.

That same day also, Domitian received the title of Cæsar, and was
made Prince of the Youths, and Præfect of Rome in the place of his
uncle, who had been murdered.

That day, also, Mucianus arrived with the Syrian legions, and with
plenitude of authority from Vespasian to act in his name.

To Duilia’s vast delight Domitian did not forget his obligation to her,
but paid frequent visits to her house, and it was a matter of pride to
her to have his attendant lictors standing outside her door, as in
former days.

When he came, she made a point of summoning her daughter, and

requiring her to be present during the interview. But she could not
make her speak or compel her to graciousness of manner towards
the visitor.

The young prince’s eyes watched the girl with question in them, but
he addressed all his conversation to the mother.

Longa Duilia did her utmost to disguise her child’s incivility,

attributed it to shyness, and used all her blandishments to make a
visit to her house agreeable to Domitian.

At length, the irksomeness caused by Domitia’s irresponsive manner

seemed to satisfy the mother that she did more harm than good in
enforcing her attendance, and she ceased to require the girl to
appear.

Some months passed, and Domitia had not given a thought to

Glyceria, and her offer to revisit the sick woman, when, all at once,
in a fit of weariness with all things that surrounded her, and a sense
of incapacity to find enjoyment anywhere, she started from her
languor to bid Eboracus go forth, buy honey-cakes and toys, and
accompany her on a visit to the Suburra.

As she was on her way, Domitian came by with his lictors and other
attendants. Since his elevation from poverty and insignificance to
ease and importance, he had acquired a swagger that made his
manner more offensive than before in his phase of cubbishness.

He at once addressed her, for though veiled he recognized her.

“May I attend you? I have at the moment nothing of importance to

occupy me.”

“I am bound for the Suburra.”

“For the Suburra! What can take you into the slums of Rome?”

“I am going to see the wife of Paris, the tragic actor.”

“Oh! the wife of the actor, Paris,” with a sneer.

“I said so—the wife of Paris the actor,” she withdrew her veil and
looked him straight in the eyes. He winced.

“And pray—is she a visiting acquaintance of the family?”

“She is our freedwoman. Paris was freed by my father likewise. Are

you content? I may add that she has met with an accident and is
crippled and confined to her bed.”

“Oh!” with a vulgar laugh, “and you are infected with the Christian
malady, and go among the sick and starving.”

“I know naught of this Christian malady. What is it?”

“We have had the contagion touch us. There is my cousin Clemens,
and his wife Domitilla, both taken badly with it. He is a poor, mean-
spirited fool. He has been offered excellent situations, with money to
be made in them, in bushels, but he refuses—will not swear by the
genius of my father, will not offer sacrifice to the Gods. Such thin
gruel minds I cannot away with. Were I Augustus, such as would not
serve the Commonwealth should be sent to kick their heels in a
desert island. These Christians are the enemies of the human race.”

“What, because they visit the sick and relieve the poor?”

“The sick are smitten by the Gods and should be left to die. The
poor are encumbrances and should be left to rot away. But a man of
rank and of family—”

“Flavius Clemens! of what family?”

Domitian bit his lip. The Flavians were of no ancestry; money-

lenders, tax-collectors, jobbers in various ways, with no connections
save through the mother of Vespasian, and that middle-class only.
“I say that a man who will not serve his country should be pitched
out of it.”

“About that I have no opinion.”

“Clemens was cast to the lions by Nero, but some witchcraft

charmed them, and they would not touch him.”

Domitia said nothing to this. She was desirous of being rid of her
self-imposed escort.

“You must wish me success,” said the young prince. “I am off to

Germany. There has been revolt there, and I go to subdue it.”

“By all means carry with you a pair of shears.”

“What mean you?”

“To obtain a crop of golden hair from the German women,

wherewith to grace your triumph.”

Domitian knitted his brows.

“You have a sharp tongue.”

“I need one. It is a woman’s sole defence.”

“Come, if a cousin, as your mother asserts,—though by the Gods! I

know not where the kinship comes in,—wish me well. Such words as
yours are of ill-omen.”

“I wish confusion and destruction to the worst enemies of Rome,”

answered Domitia.

“That suffices. I will offer the spoils to you.”

“Thank you, I do not yet wear wigs.”

He turned away with an expression of irritation.

“You are either silent, or stick pins into me,” he muttered.

Domitia continued her course, but as she entered the “Island” in

which was the home of Paris, she observed the young Cæsar still in
the street, at a corner watching her.

Much annoyed, and with her temper ruffled by this meeting, she
ascended the steps to the first story and at once turned towards the
apartments of Paris and Glyceria, but had to thread her way among
poor people, women weaving and spinning, and children romping
and running races.

She was welcomed with pleasure, Glyceria would have raised herself,
had she been able; as it was, she could show her respect only by a
salutation with the hand, and her pleasure by a smile and a word.

The chamber was fragrant with violets.

Domitia looked round and saw a small marble table on which stood a
statuette of a shepherd with panpipes, and a lamb across his
shoulders. Violets in a basin stood before the figure.

“Ah! Hermes,” said Domitia, and plucking a little bunch of the purple
flowers from her bosom she laid it in the bowl with the rest.

“Nay, dear Lady, not Hermes,” said Glyceria, “though indeed it was
sculptured to represent him—but to me that figure has another
meaning. And I hold your offering of the violets as made to Him who
to me is the Good Shepherd.”4

“Whom mean you? Atys?”

“Not Atys.”
Domitia was not particularly interested in the matter. She presumed
that some foreign cult was followed by Glyceria, and foreign cults at
this time swarmed in Rome.

“Do you believe me, Glyceria,” said Domitia, “as I came hither, the
Cæsar Domitian accompanied me, and said that I must be a
Christian to care for the sick and suffering. What are these
Christians?”

“I am one,” answered the paralyzed woman.

“What! and Paris?”

“Nay, he hovers between two opinions. His business holds him and
he will not give that up, he thinks that, were he to do so, he and I
might starve. But with the mind I think he is one.”

“And what are these Christians?”

“Those who believe in Christ.”

“And he?—is that his image?” pointing to the Good Shepherd.

“Oh Lady! it is only so much His image as the words Good Shepherd
written in characters are such, they call up a notion and so does that
figure. But in our worship we have no images, no sacrifices.”

“What is Christianity?”

“That is long to answer, but I may say in two words what it is to

me.”

“Say on.”

“The Daylight of the soul.”

“How mean you?”

“I once was in darkness. I knew not why I was set in the world,
whither I was going, what I ought to worship, what were my duties,
where was right and what was wrong. I had no light, no road, no
law. Now I have all.”

“So every votary of every new religion says. Where is your

guarantee that you are not in delusion?”

“Madam, when the sun rises and there is day, you do not suppose
the light, the splendor, the confidence inspired by it is a delusion.
You know that you see, and see that you may walk, and act with
purpose and direction. The soul has eyes as well as the body. These
eyes behold the light and cannot doubt it, by internal conscience
that distinguishes between the truth and falsehood. By that internal
conscience I am assured that the light is as real as that seen by eyes
of flesh.”

“I cannot understand you,” said Domitia. “Now for other matters—I

have made Eboracus bring you some dainties for yourself and
presents for the children who are so kind to you. Where is your
husband?”

“He is rehearsing. Better times have arrived, and he is now

occupied.”

“And you see less of him.”

“Yes—but we must live. When away from me, I know that in heart
he is with me.”

“You are sure of that?”

“Yes.”

“What, by the conscience that establishes between truth and

falsehood?”
“Nay—by trust. We must trust some one and some thing. We trust
God, we trust His Revelation, we trust in the goodness there is in
mankind.”

“There is evil rather than good.”

“There is good—but that is oft astray because of the darkness, and

does not know its course.”

Domitia did not remain long in the Insula. She bade farewell to the
wife of the actor and promised to revisit her. The presence of
Glyceria refreshed, soothed, sweetened the mind of the girl that was
heated, ruffled and soured by contact with so much there was in
pagan life that jarred against her noble instincts, by the
uncongeniality of her mother, and by the disgust she felt at
association with Domitian.

When she arrived at the palace, she heard that her mother had been
inquiring after her, and she at once went to her apartments.

Duilia asked where she had been, but did not listen for an answer, or
pay attention to what was said, when the reply came.

“What is this I hear?” said Duilia, in a tone of irritation. “Lucilla tells

me you have been chatting with Domitian, and in the street too——”

“I had no wish to speak with him. He came after me.”

“Oh! he went after you, did he? And pray what had he to say?”

“He is going to Germany to conclude a campaign already fought out

and come back and triumph for another man’s victories.”

“You did not say so to him?”

“Not in so many words.”

“My dear, it is true. He is going, and whether he be successful or
not, will return wearing the title Germanicus. I shall have a little
supper.”

“For whom?”

“For whom, do you ask? For him to be sure, to wish him good
success on the expedition.”

“You will allow me not to be present.”

“As you will, perverse girl. My dear,” in a confidential tone, “if kittens
can’t catch rats, cats can.”
 CHAPTER XXI.
THE VIRGIN’S WREATH.

“My dear,” said Longa Duilia to her daughter, “with wit such as you
have, that might be drawn through a needle’s eye, it is positively
necessary to have you married as quickly as possible. I can no
longer bear the responsibility of one so full of waywardness and
humors as yourself.”

“That, mother, is as Lamia chooses. You know that I can marry only
him.”

“And I do not ask you to take another. I will get it settled forthwith.
I’ll see his father by adoption and have the settlements looked to.
You are a good match. I presume you are aware of that, and this
explains certain poutings and bad temper. Well—reserve them for
Lamia, and don’t vex me. I wash my hands of you, when that you
are married. A camel carries his own hump, but a man his wife’s
humors.”

Domitia was sufficiently acquainted with her mother’s elasticity of

spirit and fertility of invention to be satisfied that she had a motive
for pressing on her marriage, and what that motive was seemed
obvious. But it was one that distressed her greatly.

“My dearest mother,” she said timidly, “I hope—I mean, since you
are so good as not to urge me further to break my engagement with
Lamia, that you have not set your mind—I mean your heart——”

“My excellent child,” answered Longa Duilia cutting her daughter

short, “make no scruple of blurting out what is on your tongue. You
allude to Domitian. Well! If you had common sense, you would know
that to get on in life, one must fit one’s heart with the legs of a
grasshopper, so as to be able to skip from an inconvenient, into any
suitable position. When a dish of ortolans is set on table, none but a
fool will dismiss it untasted to be devoured by the servants in the
kitchen!”

“But, mother, he is quite young.”

“By the favor of the Gods, Domitia, youths always fall in love with
women somewhat older than themselves. The Gods ordered it for
their good. If they, I mean the young men—would only follow their—
I mean the Gods’—direction, there would be fewer unhappy
marriages. For my part, I can’t see anything attractive in half-baked
girls.”

But the thoughts of her own future, and approaching happiness took
up the whole of Domitia’s brain, and left no space for consideration
of her mother’s schemes, and their chances of success.

The young prince was away. It was, as had been feared, too late for
him to reap laurels in Germany, the revolt had been quelled by
Cerealis, but as there was a ferment working in Gaul, it was deemed
advisable that Domitian should go thither and overcome the
dissatisfied instead of crossing the Alps. He had accordingly changed
his route, and had appeared in Lyons.

The marriage between Domitia and Lamia could not take place so
speedily as Duilia desired. She was wishful to have it over before the
return to Rome of Domitian, so that she might be left a freer hand,
and her daughter put out of the way who, she thought, exercised a
peculiar fascination over the young prince; but she was unable to
decide in her own mind whether what drew his eyes towards
Domitia was dislike or love; possibly it was a commingling of
resentment at her treatment of him, and admiration for her
loveliness.
But hindrances arose. Lamia was absent on his estates in Sicily,
where there had been disturbances among the slaves, and till
matters were settled there, he could not return.

Then came the month of May in which no marriages might be

performed owing to the hauntings of the Lemures, or ghosts of bad
men, and such as had not received burial. These, seen in the forms
of walking skeletons or bugbears, rioted in that sweetest month of
the whole year. Then they obtained opportunities among the
incautious to slip into their bodies, and possess them with madness,
or to take up their abodes in dwelling-houses and disturb the living
occupants by phantom appearances and mysterious sounds.

On three days in the month of May special means were adopted to

propitiate or scare away these spectres. On the 9th, 11th, and 13th,
at midnight, the master of a house, or, in the event of his death or
absence, his widow or wife, walked barefoot before the door to a
flowing fountain, where the hands were thrice washed, and then the
propitiator of the ghosts returned home, and threw black beans over
the shoulder, saying: “These I give to you, and with these beans I
ransom myself and mine.”

It was supposed that the ghost scrambled for the beans, and so
enabled the owner of the house to reach the door before them.
There stood the servants beating brazen vessels, pots and pans,
shouting, “Out with you! Out with you, ye ghosts!”

At the beginning of June was the cleansing of the Temple of Vesta,

and till that was completed, on the 15th, marriages were forbidden.

Consequently the wedding could not take place much before

midsummer, and to this Longa Duilia had to submit.

Domitia was content and happy. She had not been so happy since
her father’s death. Indeed till now she had not been able to shake
off the pain she had felt at his loss. For to her, that father was the
model of noble manhood, high-minded, full of integrity, strong yet
gentle. She had often marvelled at the manner in which he had dealt
with her mother, whom she indeed loved but who somewhat rasped
her. With his wife he had ever been firm yet forbearing. He allowed
her to form her little schemes, but always managed to thwart them
when foolish or mischievous, without her perceiving who had put a
spoke in the wheel.

Lucius Ælius Lamia she looked upon as formed in her father’s school,
upon his model. He was modest, honorable, true; a good man to
whom she could give her whole heart with full assurance that he
would treasure the gift, and that she could trust him to be as true to
her as she would be true to him.

Since her father’s death, Domitia had felt more than previously the
incompatibility of her mind with that of her mother. They had no
thoughts, no wishes, no feelings in common. Domitia was a dreamer,
speculative, ever with eager mind seeking the things beyond what
was known, whereas Duilia had not a thought, a care that were not
material. The lady Duilia cared not a rush about philosophy or the
theory of emanations. It was to her a matter of complete
indifference whether the established paganism was true or false. For
she had no apprehension of the importance of Truth. And she had
no wish that could not be gratified by money or the acquisition of
position.

Now also the haunting horror of those waking dreams that she had
seen in the Temple of Isis passed from the heart of the young girl,
like the vapors that roll away and disclose the blue heavens and the
glorious sun. She had been drifting purposeless; now she saw that
she was about to enter on a condition of life in which she would
have an object, and would find complete happiness in the pursuit of
that object,—in the fulfilment of her duties as housewife to a loved
husband, in whom she would find strength, sympathy and love.

And now also, for the first time since the death of Corbulo, she sang
as she went about the house, or worked at her bridal dress.
Lamia, on his return from Sicily was surprised to note the change in
her appearance. She had been as a beautiful flower bowed by rain
and pinched with cold, and now, as in renewed sunshine, she
bloomed with expanded petals. Light danced in her blue eyes, and a
delicate rose suffused her smooth cheeks. She had stepped back
into the childhood out of which she had passed on that terrible day
at Cenchræa.

And as he looked at her, her eyes sparkling with love and tears of
joy, he thought he had never seen one sweeter and to whom he
could so wholly devote himself as to his dear Domitia.

Then arrived the eve of the marriage.

The young girl was in the garden, stooping, picking the flowers of
which her virginal crown was to be woven, and singing as she
plucked.

Then she came with her lap full of herbs and blossoms to her
mother, who said:—

“That is right. None may gather the flowers but the bride. By the
way, have you heard? Domitian is back from Gaul. I was rejoiced at
the news, and have despatched an invitation to him to attend the
wedding.”

“Oh, mother! it is a bad omen.”

At the mention of the name, the vision of the red face, seen at Gabii
between her own and that of Lamia, started up before her, and she
let drop the lap of flowers, and they fell at her feet.

“By the Gods! what a silly thing thou art! Quick, gather up the herbs
and then go fetch thy dolls and toys of childhood, they must all this
evening be offered on the altar of the household gods.”

“I have them not, mother.”

“Not your dolls!”

“Not one.”

“But what have you done with them? I know they were all brought
from Antioch.”

“Mother, they have been given away.”

“Given away! to whom?”

“To Glyceria, the sister of Euphrosyne.”

“But what can have induced you to do this?”

“She is paralyzed, and served by little children in the story of the

Insula where she lives. I considered that it would amuse her to dress
the dolls afresh, and perhaps mend broken limbs, and after that she
will distribute them among the little willing children that help her in
her infirmity.”

“As the Gods love me!” exclaimed Duilia, “Whoever heard before of
such madness. Hellebore would not cure it. Verily the more you labor
at a hole the greater the hollow. You are a fool, and your folly grows
daily greater. You must present your toys of childhood to the Lares,
they expect it—it is the custom, it is right.”

“But I have none left.”

“Mother Ops! what is to be done? Run, Eboracus,—run and buy me

half a dozen dolls—dressed if possible. Domitia, you are determined
to bring ill-luck on yourself. There is nothing else to be done but for
you to spend an hour in playing with the dolls, and then you can
present them at the altar, and the Gods will be none the wiser.
Between me and you and the pillars of the peristyle, they are bigger
fools than us mortals, and easier gulled.”
Domitia stooped to collect the fallen flowers.

“What is that?” asked her mother—“Oh! right enough, natrix,5 that

drives away ghosts and nightmare. And that of course is in the
virginal wreath, myosotis (Forget-me-not) it dries tears. An Egyptian
slave I had—he fell ill, so I exposed him on the isle between the two
Bridges—he told me that if one ate the root in the month of Thoth—
that is August, one escaped sore eyes for a twelvemonth. That is
right also, the scarlet anemone, it betokens the flame of love—and
that evergreen its continuance. The centaury—that is the herb of
union, it will close a wound so as not to show even a scar—and in
marriage, no better symbol than that. What have you here? The
lysimachia, that gives harmony and agreement of mind. They say
that a plant of it fastened to the pole of a chariot will make the
wildest and most impatient horses pull together. And the herb of the
Twelve Gods! quite right, always remember the gods, they come in
useful. The vervain—of course, it will give you all you will. But, ye
Gods of Olympus! What have you done to pluck cypress! My dear
Domitia, are you mad? Thyme, mint, if you will—but cypress! the
tree of the infernal gods, and—as the Gods love me! let me look at
your hands! They are red—what have you plucked—plucked till your
hands are dyed—the androsœmum! Oh! Domitia! ill-fated child—
look, look at your hands, the juice has stained them, they are dipped
in blood.”
“ILL-FATED CHILD, LOOK AT YOUR HANDS.” Page 176.
 CHAPTER XXII.
QUONIAM TU CAIUS, EGO CAIA!

At the earliest rays of dawn the auguries were taken, not as of old
by the flight of birds, but by inspection of the liver and heart of a
sheep, that was slaughtered for the purpose by the Aruspices, and
this done they came to the palace of Duilia, bearing the skin of the
sheep, to announce that the portents were favorable, in fact, were
of extraordinarily good promise.

“That is as I hoped,” said Longa Duilia, “and that will counteract and
bring to naught the disastrous tokens of the wreath. Why, by Venus’s
girdle, the girl has not been able to get her hands white yet. The
stain of that nefast herb is on them still. But—ah! here she comes in
her flame-colored veil. By the Body of Bacchus! after all it means no
ill, for do not her hands agree in hue with her head-gear?”6

Domitia had laid aside her maidenly dress, the toga prætextata
woven with horizontal stripes, for the dress of a married woman, the
toga recta, with vertical stripes. About her waist was a woollen girdle
fastened in a peculiar manner, with the so-called knot of Hercules,
that was regarded as a charm against the evil eye, and was also
employed in binding up wounds and fractured bones. The girl’s
dress, as well as a net of red silk threads in which her hair had been
tied up on the previous day, had been offered on the altars of the
ancestral deities worshipped in the house.

Her hair had been divided that morning, not by a comb, but by the
head of a lance, into six tresses that were plaited with colored
ribbons. And about her head, beneath the veil, was the virgin’s
wreath woven out of the flowers she had herself picked—but the ill-
omened cypress and the blood distilling androsœmum had been
omitted.

And now with pipes and cymbals came the bridegroom attended by
all his friends, to fetch the bride home. The house door was
decorated with laurels, and incense smoked on the domestic altars,
in the vestibule, and in the atrium. The boxes that contained the
ancestral wax masks were open, and each face was wreathed about
with flowers. Green lines connecting the boxes united all to one
trunk forming a family tree. The household gods were not ignored,
lamps burned before them, flowers adorned their heads, and cakes
and wine were placed on shelves below them.

Slaves ran to and fro, and ran against each other. Ten witnesses,
kinsmen of the bride and bridegroom, assembled to take cognizance
of the marriage contract. Two seats were introduced into the hall,
and the legs bound together, and over both was spread the skin of
the sheep slaughtered that morning for the auspices.

Then bride and bridegroom were seated on these stools, the

marriage contract was read aloud, and they received the salutations
of their friends. The pronuba, a married female relative united their
hands, and that accomplished, the bridegroom rose, and attended
by the friends and kinsfolk of both parties, departed for the Temple
of Jupiter, where the flamen Dialis offered sacrifice to the gods of
marriage, to Jupiter, Juno, Tellus, and the old Latin half-forgotten
deities of Picumnus and Pilumnus.

Whilst the sacred sacrifice was being performed, in the house of the
bride all was being made ready for the wedding or meal after
midday.

The bride was now esteemed to have passed out of the family of her
father into that of her husband, his gods would be her gods, his
house her house, his name hers. In signification of this the formula
was used by her, “Since thou art Caius, I am Caia.” At a remote
period it would have been “Since thou art Lucius I am Lucia,” and
she would have lost her name of Domitia. But this was no longer
customary, only the liturgical form of surrender was employed.

It was past noon when the procession returned, swelled by more

friends and by all well-wishers, and as it entered the house, with a
shiver Domitia observed the glowing face and water-blue eyes of the
young prince, attended by his lictors. She caught his glance, but he
dropped his eyes the moment they encountered hers, and she saw
his cheeks pucker, as though with laughter. But she had no time to
give thought to him; she was required to acknowledge the
felicitations of the visitors, and to entreat them to partake of the
hospitality of the hour, and to offer a pinch of incense and a libation
to her happiness.

The supper was lengthy—many partook and came in relays, so that

the entire afternoon was consumed by it. To the relief of Domitia,
the prince Domitian had withdrawn. As each left the table he saluted
the bride with the exclamation, Feliciter.

For this long and tedious ceremonial feast, she was allowed to rest
on a couch, next to her husband, at the table, in the place of honor.

The meal lasted till evening, and then there ensued a movement.

The household goods of the bride, her spindle and distaff, her chest
containing robes, were brought forth, and placed on biers to be
conveyed to the new house.

Then Domitia rose, with tears in her eyes, and went to the several
chambers she had occupied, to say farewell to the kitchen, to salute
the hearth, to the shelf that served as chapel, to bid farewell to the
ancestral gods, to the wax forefathers in the hall, then to kiss her
mother, finally to turn, kneel and embrace the doorposts of the
paternal dwelling, and kiss the threshold from which she parted.
Without, the procession waited. She was gently disengaged from her
mother’s arms, and to the cries of Talasse! amidst a shower of
walnuts thrown among the boys by the bridegroom, the procession
started.

Domitia was attended by three lads, one went before carrying a

torch, the other two walked, one on each side, carrying spindle and
distaff. The torch, according to rule, was of whitethorn wood, and on
arrival at the house of the bridegroom would be scrambled for and
ripped to pieces by the guests, as every shred was esteemed to
carry good luck.

Now rose a burst of song, the so-called Fescennian lays, some old
and some new, accompanied by the flutes of musicians and the
clash of castanets and cymbals of dancing girls.

The procession descended the hill to the Forum, crowds lining the
way and shouting Feliciter!

At a corner there was a little clearing, for there lay a pallet, and on it
a sick woman, who had been brought from her dwelling to see the
sight. She extended and waved her hand, holding something as
Domitia approached, and the bride through her tears noticed her,
halted, went towards her, and said:—

“Glyceria! you here to wish me happiness!”

“And to give thee, dear lady, a little present.”

She extended to her a small amulet, that Domitia accepted

gratefully, and stooping kissed the paralyzed woman on the brow.

An unheard-of thing! unparalleled! A thing she would not have done,

had she been in full control over herself—a thing she would not have
done, had not her heart brimmed with love for all, at that moment.
She, a noble lady, belonging to one of the greatest houses in Rome,
kissed a poor actor’s wife, an enfranchised slave—and that before all
eyes.

About Glyceria was a dense throng of men and women and children,
the occupants of the “Island” in which she lived. It was they, who,
pitying her sufferings, desirous that she should see the procession,
had opened a space before her, and held it open, that none might
impede a full view of the marriage train.

And this throng of rude artisans, shoemakers, cordwainers, leather-

sellers, hawkers and their wives and children saw this act of Domitia.
For a moment they were silent, and then they broke into a roar of
“Feliciter! feliciter! the Gods be with thee, dear lady! The Gods
protect thee! The Gods shower blessings on thee!”

But Domitia might not tarry; confused, half ashamed of what she
had done, half carried off her feet by the thrill of joy that went from
the crowd to her, she advanced.

The train descended by the lake of Nero, now occupied by the

Colosseum, then ascended the Celian Hill to the house of Lamia.

On reaching his door, the procession spread out, and gave space for
the bride to advance.

Modestly, trembling with love, timidity, hope in her heart, she

anointed the doorposts with oil and then passed woollen strings
round them.

This accomplished, two young men started forward, caught her up,
made a seat for her of their hands, and bore her over the threshold,
which she might not touch with her feet, lest by accident or
nervousness she should stumble, and so her entry into the new
house be ill-omened. On being admitted into the habitation of her
husband, it was her duty to go to the hearth and make up the fire,
then to the fountain and draw water; next to worship the household
gods.
The house was pretty. It had been fresh painted, and was bright
with color, and sweet with flowers, for every pillar was wreathed and
each door garlanded. Numerous lamps illumined the chambers, and
in the atrium were reflected in the water tank. The air was vibrating
with music, as choirs sang Fescennian songs, and timbrels tinkled
and pipes twittered.

Domitia was received by the wife of L. Ælius Lamia, who had

adopted Domitia’s husband. He was a quiet man, who had no
ambition, had taken no offices, and had passed his time in taming
birds. He was the son of a better known man, who had been a friend
of Horace.

The old woman, gentle in manner, took Domitia by the hand and led
her into the tablinum, where was old Lamia, a cripple through gout,
and he kissed the girl, patted her hands and spoke an affectionate
welcome.

“Claudia and I,” said he, “were childless and so we adopted Lucius.
He has been a good son to us, and this is a happy day to all three,—
to him who has secured the sweetest flower of Rome, and to Claudia
and me who obtain so good a daughter. But, ah! we are old and
have our humors, I, with my gout, am liable to be peevish. You must
bear with our infirmities. You will have a worthy husband, one cut
out of the old rock of which were the ancient Romans, and not of
the Tiberine mud of which the present generation are moulded.”

“Come now,” said the old woman, “the guests are about to depart,
bid them farewell.”

Then she led the young girl back into the atrium.

There stood the Chaldæan, dark, stern, ominous.

Domitia in exuberant joy smiled at him, and said:

“Elymas! You see my happiness. Isis has for once been in error—we,
my Lamia and I, are united, and there have been no hands thrust
forth to part us.”

“My lady,” said the astrologer, “the day is not yet over.”

“And the auguries were all propitious.”

“The promise of the augurs may not jump with thy desire,” he
replied.

She had no time for more words, as her hand was caught by L.
Ælius Lamia, who drew her aside into the lararium or chapel.

“My dearest,” he said, “this is a day of trial to thee—but we shall be

left undisturbed shortly. The guests depart and the riot will cease.”

She looked at him, with eyes that brimmed with tears, and a sob
relieved her heart, as she cast herself on his breast and said:—

“Quoniam tu Caius, ego Caia.”

Welcome to Our Bookstore - The Ultimate Destination for Book Lovers
Are you passionate about testbank and eager to explore new worlds of
knowledge? At our website, we offer a vast collection of books that
cater to every interest and age group. From classic literature to
specialized publications, self-help books, and children’s stories, we
have it all! Each book is a gateway to new adventures, helping you
expand your knowledge and nourish your soul
Experience Convenient and Enjoyable Book Shopping Our website is more
than just an online bookstore—it’s a bridge connecting readers to the
timeless values of culture and wisdom. With a sleek and user-friendly
interface and a smart search system, you can find your favorite books
quickly and easily. Enjoy special promotions, fast home delivery, and
a seamless shopping experience that saves you time and enhances your
love for reading.
Let us accompany you on the journey of exploring knowledge and
personal growth!

ebooksecure.com