Response surface methodology

process and product optimization using

designed experiments Fourth Edition
Anderson-Cook
Visit to download the full and correct content document:
https://textbookfull.com/product/response-surface-methodology-process-and-product-
optimization-using-designed-experiments-fourth-edition-anderson-cook/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

RSM Simplified Optimizing Processes Using Response

Surface Methods for Design of Experiments Second
Edition Mark J. Anderson

https://textbookfull.com/product/rsm-simplified-optimizing-
processes-using-response-surface-methods-for-design-of-
experiments-second-edition-mark-j-anderson/

Python Programming in Context, Fourth Edition Julie

Anderson And Jon Anderson

https://textbookfull.com/product/python-programming-in-context-
fourth-edition-julie-anderson-and-jon-anderson/

Pharmaceutical Drug Product Development and Process

Optimization-Effective Use of Quality by Design 1st
Edition Sarwar Beg (Editor)

https://textbookfull.com/product/pharmaceutical-drug-product-
development-and-process-optimization-effective-use-of-quality-by-
design-1st-edition-sarwar-beg-editor/

Stochastic Process Optimization using Aspen Plus® First

Edition Juan Gabriel Segovia-Hernández

https://textbookfull.com/product/stochastic-process-optimization-
using-aspen-plus-first-edition-juan-gabriel-segovia-hernandez/
Handbook of surface and colloid chemistry Fourth
Edition Birdi

https://textbookfull.com/product/handbook-of-surface-and-colloid-
chemistry-fourth-edition-birdi/

Youth Work Process Product and Practice 2nd Edition Ord

https://textbookfull.com/product/youth-work-process-product-and-
practice-2nd-edition-ord/

Fracture Mechanics: Fundamentals and Applications,

Fourth Edition Ted L. Anderson

https://textbookfull.com/product/fracture-mechanics-fundamentals-
and-applications-fourth-edition-ted-l-anderson/

Business Communication: Process & Product Mary Ellen

Guffey

https://textbookfull.com/product/business-communication-process-
product-mary-ellen-guffey/

Pocket Rough Guide London Fourth Edition Samantha Cook

https://textbookfull.com/product/pocket-rough-guide-london-
fourth-edition-samantha-cook/
RESPONSE SURFACE
METHODOLOGY
WILEY SERIES IN PROBABILITY AND STATISTICS

Established by WALTER A. SHEWHART and SAMUEL S. WILKS

Editors: David J. Balding, Noel A. C. Cressie, Garrett M. Fitzmaurice,

Geof H. Givens, Harvey Goldstein, Geert Molenberghs, David W. Scott,
Adrian F. M. Smith, Ruey S. Tsay, Sanford Weisberg
Editors Emeriti: J. Stuart Hunter, Iain M. Johnstone, Joseph B. Kadane,
Jozef L. Teugels

A complete list of the titles in this series appears at the end of this volume.
RESPONSE SURFACE
METHODOLOGY
Process and Product
Optimization Using
Designed Experiments

Fourth Edition

RAYMOND H. MYERS
Virginia Polytechnic University, Department of Statistics, Blacksburg, VA

DOUGLAS C. MONTGOMERY
Arizona State University, Department of Industrial Engineering, Tempe, AZ

CHRISTINE M. ANDERSON-COOK
Los Alamos National Laboratory, Los Alamos, NM
Copyright © 2016 by John Wiley & Sons, Inc. All rights reserved

Published by John Wiley & Sons, Inc., Hoboken, New Jersey

Published simultaneously in Canada

No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any
means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under
Sections 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the
Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center,
Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 750-8400, fax (978) 750-4470, or on the web at
www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions
Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748-6011, fax (201)
748-6008, or online at http://www.wiley.com/go/permission.

Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in
preparing this book, they make no representations or warranties with respect to the accuracy or completeness of
the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a
particular purpose. No warranty may be created or extended by sales representatives or written sales materials.
The advice and strategies contained herein may not be suitable for your situation. You should consult with a
professional where appropriate. Neither the publisher nor author shall be liable for any loss of profit or any other
commercial damages, including but not limited to special, incidental, consequential, or other damages.

For general information on our other products and services or for technical support, please contact our Customer
Care Department within the United States at (800) 762-2974, outside the United States at (317) 572-3993 or fax
(317) 572-4002.

Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be
available in electronic format. For more information about Wiley products, visit our web site at www.wiley.com.

Library of Congress Cataloging-in-Publication Data:

Myers, Raymond H.
Response surface methodology : process and product optimization using designed experiments.
– 4th ed. / Raymond H. Myers, Douglas C. Montgomery, Christine M. Anderson-Cook.
p. cm. – (Wiley series in probability and statistics)
Includes bibliographical references and index.
ISBN 978-1-118-91601-8 (cloth)
1. Experimental design. 2. Response surfaces (Statistics). I. Montgomery, Douglas C.
II. Anderson-Cook, Christine M. III. Title.
QA279.M94 2016
519.507–dc22 2016019012

Printed in the United States of America

10 9 8 7 6 5 4 3 2 1
CONTENTS

Preface xiii

1 Introduction 1
1.1 Response Surface Methodology, 1
1.1.1 Approximating Response Functions, 2
1.1.2 The Sequential Nature of RSM, 7
1.1.3 Objectives and Typical Applications of RSM, 9
1.1.4 RSM and the Philosophy of Quality Improvement, 11
1.2 Product Design and Formulation (Mixture Problems), 11
1.3 Robust Design and Process Robustness Studies, 12
1.4 Useful References on RSM, 12

2 Building Empirical Models 13

2.1 Linear Regression Models, 13
2.2 Estimation of the Parameters in Linear
Regression Models, 14
2.3 Properties of the Least Squares Estimators
and Estimation of 𝜎 2 , 22
2.4 Hypothesis Testing in Multiple Regression, 24
2.4.1 Test for Significance of Regression, 24
2.4.2 Tests on Individual Regression Coefficients and Groups of
Coefficients, 27
2.5 Confidence Intervals in Multiple Regression, 31
2.5.1 Confidence Intervals on the Individual Regression Coefficients β, 32
2.5.2 A Joint Confidence Region on the Regression Coefficients β, 32
2.5.3 Confidence Interval on the Mean Response, 33
2.6 Prediction of New Response Observations, 35

v
vi CONTENTS

2.7 Model Adequacy Checking, 36

2.7.1 Residual Analysis, 36
2.7.2 Scaling Residuals, 38
2.7.3 Influence Diagnostics, 42
2.7.4 Testing for Lack of Fit, 43
2.8 Fitting a Second-Order Model, 47
2.9 Qualitative Regressor Variables, 55
2.10 Transformation of the Response Variable, 61
Exercises, 66

3 Two-Level Factorial Designs 81

3.1 Introduction, 81
3.2 The 22 Design, 82
3.3 The 23 Design, 94
3.4 The General 2k Design, 103
3.5 A Single Replicate of the 2k Design, 108
3.6 2k Designs are Optimal Designs, 125
3.7 The Addition of Center Points to the 2k Design, 130
3.8 Blocking in the 2k Factorial Design, 135
3.8.1 Blocking in the Replicated Design, 135
3.8.2 Confounding in the 2k Design, 137
3.9 Split-Plot Designs, 141
Exercises, 146

4 Two-Level Fractional Factorial Designs 161

4.1 Introduction, 161
4.2 The One-Half Fraction of the 2k Design, 162
4.3 The One-Quarter Fraction of the 2k Design, 174
4.4 The General 2k−p Fractional Factorial Design, 184
4.5 Resolution III Designs, 188
4.6 Resolution IV and V Designs, 197
4.7 Alias Structures in Fractional Factorial and Other Designs, 198
4.8 Nonregular Fractional Factorial Designs, 200
4.8.1 Nonregular Fractional Factorial Designs for 6, 7, and 8 Factors in
16 Runs, 203
4.8.2 Nonregular Fractional Factorial Designs for 9 Through 14
Factors in 16 Runs, 209
4.8.3 Analysis of Nonregular Fractional Factorial Designs, 213
4.9 Fractional Factorial Split-Plot Designs, 216
4.10 Summary, 219
Exercises, 220

5 Process Improvement with Steepest Ascent 233

5.1 Determining the Path of Steepest Ascent, 234
5.1.1 Development of the Procedure, 234
5.1.2 Practical Application of the Method of Steepest Ascent, 237
 CONTENTS vii

5.2 Consideration of Interaction and Curvature, 241

5.2.1 What About a Second Phase?, 244
5.2.2 What Happens Following Steepest Ascent?, 244
5.3 Effect of Scale (Choosing Range of Factors), 245
5.4 Confidence Region for Direction of Steepest Ascent, 247
5.5 Steepest Ascent Subject to a Linear Constraint, 250
5.6 Steepest Ascent in a Split-Plot Experiment, 254
Exercises, 262

6 The Analysis of Second-Order Response Surfaces 273

6.1 Second-Order Response Surface, 273
6.2 Second-Order Approximating Function, 274
6.2.1 The Nature of the Second-Order Function and Second-Order
Surface, 274
6.2.2 Illustration of Second-Order Response Surfaces, 276
6.3 A Formal Analytical Approach to the Second-Order Model, 277
6.3.1 Location of the Stationary Point, 278
6.3.2 Nature of the Stationary Point (Canonical Analysis), 278
6.3.3 Ridge Systems, 282
6.3.4 Role of Contour Plots, 286
6.4 Ridge Analysis of the Response Surface, 289
6.4.1 Benefits of Ridge Analysis, 290
6.4.2 Mathematical Development of Ridge Analysis, 291
6.5 Sampling Properties of Response Surface Results, 296
6.5.1 Standard Error of Predicted Response, 296
6.5.2 Confidence Region on the Location of the Stationary Point, 299
6.5.3 Use and Computation of the Confidence Region on the Location
of the Stationary Point, 300
6.5.4 Confidence Intervals on Eigenvalues in Canonical Analysis, 304
6.6 Further Comments Concerning Response Surface Analysis, 307
Exercises, 307

7 Multiple Response Optimization 325

7.1 Balancing Multiple Objectives, 325
7.2 Strategies for Multiple Response Optimization, 338
7.2.1 Overlaying Contour Plots, 339
7.2.2 Constrained Optimization, 340
7.2.3 Desirability Functions, 341
7.2.4 Pareto Front Optimization, 343
7.2.5 Other Options for Optimization, 349
7.3 A Sequential Process for Optimization—DMRCS, 350
7.4 Incorporating Uncertainty of Response Predictions into Optimization, 352
Exercises, 357
viii CONTENTS

8 Design of Experiments for Fitting Response Surfaces—I 369

8.1 Desirable Properties of Response Surface Designs, 369
8.2 Operability Region, Region of Interest, and Metrics for Desirable
Properties, 371
8.2.1 Metrics for Desirable Properties, 372
8.2.2 Model Inadequacy and Model Bias, 373
8.3 Design of Experiments for First-Order Models and First-Order Models
with Interactions, 375
8.3.1 The First-Order Orthogonal Design, 376
8.3.2 Orthogonal Designs for Models Containing Interaction, 378
8.3.3 Other First-Order Orthogonal Designs—The Simplex Design, 381
8.3.4 Definitive Screening Designs, 385
8.3.5 Another Variance Property—Prediction Variance, 389
8.4 Designs for Fitting Second-Order Models, 393
8.4.1 The Class of Central Composite Designs, 393
8.4.2 Design Moments and Property of Rotatability, 399
8.4.3 Rotatability and the CCD, 403
8.4.4 More on Prediction Variance—Scaled, Unscaled, and Estimated, 406
8.4.5 The Face-Centered Cube in Cuboidal Regions, 408
8.4.6 Choosing between Spherical and Cuboidal Regions, 411
8.4.7 The Box–Behnken Design, 413
8.4.8 Definitive Screening Designs for Fitting Second-Order Models, 417
8.4.9 Orthogonal Blocking in Second-Order Designs, 422
Exercises, 434

9 Experimental Designs for Fitting Response Surfaces—II 451

9.1 Designs that Require a Relatively Small Run Size, 452
9.1.1 The Hoke Designs, 452
9.1.2 Koshal Design, 454
9.1.3 Hybrid Designs, 455
9.1.4 The Small Composite Design, 458
9.1.5 Some Saturated or Near-Saturated Cuboidal Designs, 462
9.1.6 Equiradial Designs, 463
9.2 General Criteria for Constructing, Evaluating, and Comparing Designed
Experiments, 465
9.2.1 Practical Design Optimality, 467
9.2.2 Use of Design Efficiencies for Comparison of Standard
Second-Order Designs, 474
9.2.3 Graphical Procedure for Evaluating the Prediction Capability of
an RSM Design, 477
9.3 Computer-Generated Designs in RSM, 488
9.3.1 Important Relationship Between Prediction Variance and Design
Augmentation for D-Optimality, 491
9.3.2 Algorithms for Computer-Generated Designs, 494
9.3.3 Comparison of D-, G-, and I-Optimal Designs, 497
 CONTENTS ix

9.3.4 Illustrations Involving Computer-Generated Design, 499

9.3.5 Computer-Generated Designs Involving Qualitative Variables, 508
9.4 Multiple Objective Computer-Generated Designs for RSM, 517
9.4.1 Pareto Front Optimization for Selecting a Design, 518
9.4.2 Pareto Aggregating Point Exchange Algorithm, 519
9.4.3 Using DMRCS for Design Optimization, 520
9.5 Some Final Comments Concerning Design Optimality and
Computer-Generated Design, 525
Exercises, 527

10 Advanced Topics in Response Surface Methodology 543

10.1 Effects of Model Bias on the Fitted Model and Design, 543
10.2 A Design Criterion Involving Bias and Variance, 547
10.2.1 The Case of a First-Order Fitted Model and Cuboidal Region, 550
10.2.2 Minimum Bias Designs for a Spherical Region of Interest, 556
10.2.3 Simultaneous Consideration of Bias and Variance, 558
10.2.4 How Important Is Bias?, 558
10.3 Errors in Control of Design Levels, 560
10.4 Experiments with Computer Models, 563
10.4.1 Design for Computer Experiments, 567
10.4.2 Analysis for Computer Experiments, 570
10.4.3 Combining Information from Physical and Computer
Experiments, 574
10.5 Minimum Bias Estimation of Response Surface Models, 575
10.6 Neural Networks, 579
10.7 Split-Plot Designs for Second-Order Models, 581
10.8 RSM for Non-Normal Responses—Generalized Linear Models, 591
10.8.1 Model Framework: The Link Function, 592
10.8.2 The Canonical Link Function, 593
10.8.3 Estimation of Model Coefficients, 593
10.8.4 Properties of Model Coefficients, 595
10.8.5 Model Deviance, 595
10.8.6 Overdispersion, 597
10.8.7 Examples, 598
10.8.8 Diagnostic Plots and Other Aspects of the GLM, 605
Exercises, 609

11 Robust Parameter Design and Process Robustness Studies 619

11.1 Introduction, 619
11.2 What is Parameter Design?, 619
11.2.1 Examples of Noise Variables, 620
11.2.2 An Example of Robust Product Design, 621
11.3 The Taguchi Approach, 622
11.3.1 Crossed Array Designs and Signal-to-Noise Ratios, 622
11.3.2 Analysis Methods, 625
11.3.3 Further Comments, 630
x CONTENTS

11.4 The Response Surface Approach, 631

11.4.1 The Role of the Control × Noise Interaction, 631
11.4.2 A Model Containing Both Control and Noise Variables, 635
11.4.3 Generalization of Mean and Variance Modeling, 638
11.4.4 Analysis Procedures Associated with the Two Response Surfaces, 642
11.4.5 Estimation of the Process Variance, 651
11.4.6 Direct Variance Modeling, 655
11.4.7 Use of Generalized Linear Models, 657
11.5 Experimental Designs For RPD and Process
Robustness Studies, 661
11.5.1 Combined Array Designs, 661
11.5.2 Second-Order Designs, 663
11.5.3 Other Aspects of Design, 665
11.6 Dispersion Effects in Highly Fractionated Designs, 672
11.6.1 The Use of Residuals, 673
11.6.2 Further Diagnostic Information from Residuals, 674
11.6.3 Further Comments Concerning Variance Modeling, 680
Exercises, 684

12 Experiments with Mixtures 693

12.1 Introduction, 693
12.2 Simplex Designs and Canonical Mixture Polynomials, 696
12.2.1 Simplex Lattice Designs, 696
12.2.2 The Simplex-Centroid Design and Its Associated Polynomial, 704
12.2.3 Augmentation of Simplex Designs with Axial Runs, 707
12.3 Response Trace Plots, 716
12.4 Reparameterizing Canonical Mixture Models to Contain A Constant
Term (𝛽 0 ), 716
Exercises, 720

13 Other Mixture Design and Analysis Techniques 731

13.1 Constraints on the Component Proportions, 731
13.1.1 Lower-Bound Constraints on the Component Proportions, 732
13.1.2 Upper-Bound Constraints on the Component Proportions, 743
13.1.3 Active Upper- and Lower-Bound Constraints, 747
13.1.4 Multicomponent Constraints, 758
13.2 Mixture Experiments Using Ratios of Components, 759
13.3 Process Variables in Mixture Experiments, 763
13.3.1 Mixture-Process Model and Design Basics, 763
13.3.2 Split-Plot Designs for Mixture-Process Experiments, 767
13.3.3 Robust Parameter Designs for Mixture-Process Experiments, 778
13.4 Screening Mixture Components, 783
Exercises, 785
 CONTENTS xi

Appendix 1 Moment Matrix of a Rotatable Design 797

Appendix 2 Rotatability of a Second-Order Equiradial Design 803

References 807

Index 821
PREFACE

This book deals with the exploration and optimization of response surfaces. This is a
problem faced by experimenters in many technical fields, where, in general, the response
variable of interest is y and there is a set of predictor variables x1 , x2 , … , xk . For example,
y might be the viscosity of a polymer and x1 , x2 , and x3 might be the process variables
reaction time, reactor temperature, and the catalyst feed rate in the process. In some systems
the nature of the relationship between y and the x’s might be known “exactly,” based on
the underlying engineering, chemical, or physical principles. Then we could write a model
of the form y = g(x1 , x2 , … , xk ) + 𝜀, where g is a known (often nonlinear) function and 𝜀
represents the “error” in the system. This type of relationship is often called a mechanistic
model. We consider the more common situation where the underlying mechanism is not
fully understood, and the experimenter must approximate the unknown function g with an
appropriate empirical model y = f(x1 , x2 , … , xk ) + 𝜀. Usually the function f is a low-order
polynomial, typically either a first-order or second-order polynomial. This empirical model
is called a response surface model.
Identifying and fitting an appropriate response surface model from experimental data
requires some knowledge of statistical experimental design fundamentals, regression mod-
eling techniques, and elementary optimization methods. This book integrates all three of
these topics into what has been popularly called response surface methodology (RSM). We
assume that the reader has some previous exposure to statistical methods and matrix alge-
bra. Formal coursework in basic principles of experimental design and regression analysis
would be helpful, but are not essential, because the important elements of these topics are
presented early in the text. We have used this book in a graduate-level course on RSM for
statisticians, engineers, and chemical/physical scientists. We have also used it in industrial
short courses and seminars for individuals with a wide variety of technical backgrounds.
This fourth edition is a substantial revision of the book. We have rewritten many sections
to incorporate new topics and material, ideas, and examples and to more fully explain some
topics that were only briefly mentioned in previous editions and added a new chapter. We
have also continued to integrate the computer tightly into the presentation, replying on JMP

xiii
xiv PREFACE

and Design-Expert for much of the computing, but also continuing to employ SAS for a
few applications.
Chapters 1 through 4 contain the preliminary material essential to studying RSM. Chapter
1 is an introduction to the general field of RSM, describing typical applications such as (a)
finding the levels of process variables that optimize a response of interest or (b) discovering
what levels of these process variables result in a product satisfying certain requirements or
specifications on responses such as yield, molecular weight, purity, or viscosity. Chapter
2 is a summary of regression methods useful in response surface work, focusing on the
basic ideas of least squares model fitting, diagnostic checking, and inference for the linear
regression model. Chapters 3 and 4 describe two-level factorial and fractional factorial
designs. These designs are essential for factor screening or identifying the correct set of
process variables to use in the RSM study. They are also basic building blocks for many
of the response surface designs discussed later in the text. New topics in these chapters
include additional information on analysis of unreplicated design and the use of nonregular
fractional factorial designs in factor screening experiments.
Chapter 5 presents the method of steepest ascent, a simple but powerful optimization
procedure used at the early stages of RSM to move the process from a region of rel-
atively poor performance to one of greater potential. Chapter 6 introduces the analysis
and optimization of a second-order response surface model. Both graphical and numerical
techniques are presented. A new chapter, Chapter 7, describes techniques for the simul-
taneous optimization of several responses, a common problem in the application of RSM
when experimenters want to extract information about several different characteristics of
the process or product.
Chapters 8 and 9 present detailed information on the choice of experimental designs
for fitting response surface models. Chapter 8 considers suitable metrics for evaluating
the quality of estimation and prediction of a design and focuses on standard designs,
including the central composite and Box–Behnken designs, and the important topic of
blocking a response surface design. A new topic in this chapter is the important new class
of designs called Definitive Screening Designs. Chapter 9 covers small response surface
designs, design optimality criteria, and methods for evaluation of the prediction properties of
response surface models constructed from various designs. We focus on variance dispersion
graphs and fraction of design space plots, which are very important ways to summarize
prediction properties. The section on computer generated designs has been expanded to
describe some of the common algorithms used to generate designs as well as how to
incorporate qualitative factors and multiple design criteria.
Chapter 10 contains more advanced RSM topics, including the use of mean square error
as a design criterion, the effect of errors in controllable variables, split-plot type designs
in a response surface setting, and the use of generalized linear models in the analysis
of response surface experiments. The section on design and analysis of experiments for
computer models has been updated to reflect new developments in this growing area.
Chapter 11 describes how the problem of robust parameter design originally proposed by
Taguchi can be efficiently solved in the RSM framework. We show how RSM not only
makes the original problem posed by Taguchi easier to solve, but also provides much more
information to the analyst about process or system performance. This chapter also contains
much information on robust parameter design and process robustness studies. Chapters 12
and 13 present techniques for designing and analyzing experiments that involve mixtures.
A mixture experiment is a special type of response surface experiment in which the design
factors are the components or ingredients of a mixture, and the response depends on the
 PREFACE xv

proportions of the ingredients that are present. Some new design techniques are introduced
in these chapters.
We have provided expanded end-of-chapter problems and an updated reference section.
The previous three editions of the text were written to emphasize methods that are useful
in industry and that we have found useful in our own consulting experience. We have
continued that applied focus in this new edition, though much new material has been added.
We develop enough of the underlying theory to allow the reader to gain an understanding
of the assumptions and conditions necessary to successfully apply RSM.
We are grateful to many individuals that have contributed meaningfully to this book. In
particular, Dr. Bradley Jones, Mr. Pat Whitcomb, Dr. Geoff Vining, Dr. Soren Bisgaard,
Dr. Connie Borror, Dr. Rachel Silvistrini, Dr. Scott Kowalski, Dr. Dennis Lin, Dr. George
Runger, Dr. Enrique Del Castillo, Dr. Lu Lu and Dr. Jessica Chapman made many useful
suggestions. Dr. Matt Carlyle and Dr. Enrique Del Castillo also provided some figures that
were most helpful. We also thank the many classes of graduate students that have studied
from the book, as well as the instructors that have used the book. They have made many
helpful comments and suggestions to improve the clarity of the presentation. We have tried
to incorporate many of their suggestions. We also thank John Wiley & Sons for permission
to use and adapt copyrighted material.

Blacksburg, Virginia Raymond H. Myers

Tempe, Arizona Douglas C. Montgomery
Los Alamos, New Mexico Christine M. Anderson-Cook
1
INTRODUCTION

1.1 RESPONSE SURFACE METHODOLOGY

Response surface methodology (RSM) is a collection of statistical and mathematical tech-

niques useful for developing, improving, and optimizing processes. It also has important
applications in the design, development, and formulation of new products, as well as in the
improvement of existing product designs.
The most extensive applications of RSM are in the industrial world, particularly in
situations where several input variables potentially influence performance measures or
quality characteristics of the product or process. These performance measures or quality
characteristics are called the response. They are typically measured on a continuous scale,
although attribute responses, ranks, and sensory responses are not unusual. Most real-
world applications of RSM will involve more than one response. The input variables are
sometimes called independent variables, and they are subject to the control of the engineer
or scientist, at least for purposes of a test or an experiment.
Figure 1.1 shows graphically the relationship between the response variable yield (y) in a
chemical process and the two process variables (or independent variables) reaction time (𝜉 1 )
and reaction temperature (𝜉 2 ). Note that for each value of 𝜉 1 and 𝜉 2 there is a corresponding
value of yield y and that we may view these values of the response yield as a surface lying
above the time–temperature plane, as in Fig. 1.1a. It is this graphical perspective of the
problem environment that has led to the term response surface methodology. It is also
convenient to view the response surface in the two-dimensional time–temperature plane,
as in Fig. 1.1b. In this presentation we are looking down at the time–temperature plane and
connecting all points that have the same yield to produce contour lines of constant response.
This type of display is called a contour plot.

Response Surface Methodology: Process and Product Optimization Using Designed Experiments, Fourth Edition.
by Raymond H. Myers, Douglas C. Montgomery, and Christine M. Anderson-Cook.
Copyright © 2016 John Wiley & Sons, Inc. Published 2016 by John Wiley & Sons, Inc.

1
2 INTRODUCTION

Figure 1.1 (a) A theoretical response surface showing the relationship between yield of a chemical
process and the process variables reaction time (𝜉 1 ) and reaction temperature (𝜉 2 ). (b) A contour plot
of the theoretical response surface.

Clearly, if we could easily construct the graphical displays in Fig. 1.1, optimization of
this process would be very straightforward. By inspection of the plot, we note that yield is
maximized in the vicinity of time 𝜉 1 = 4 hr and temperature 𝜉 2 = 525◦ C. Unfortunately,
in most practical situations, the true response function in Fig. 1.1 is unknown. The field
of response surface methodology consists of the experimental strategies for exploring
the space of the process or independent variables (here the variables 𝜉 1 and 𝜉 2 ), empirical
statistical modeling to develop an appropriate approximating relationship between the yield
and the process variables, and optimization methods for finding the levels or values of the
process variables 𝜉 1 and 𝜉 2 that produce desirable values of the responses (in this case that
maximize yield).

1.1.1 Approximating Response Functions

In general, suppose that the scientist or engineer (whom we will refer to as the experi-
menter) is concerned with a product, process, or system involving a response y that depends
on the controllable input variables 𝜉 1 , 𝜉 2 , … , 𝜉 k . These input variables are also sometimes
called factors, independent variables, or process variables. The actual relationship can be
written

y = f (𝜉1 , 𝜉2 , … , 𝜉k ) + 𝜀 (1.1)

where the form of the true response function f is unknown and perhaps very complicated, and
𝜀 is a term that represents other sources of variability not accounted for in f. Thus 𝜀 includes
effects such as measurement error on the response, other sources of variation that are
inherent in the process or system (background noise, or common/special cause variation
in the language of statistical process control), the effect of other (possibly unknown)
variables, and so on. We will treat 𝜀 as a statistical error, often assuming it to have a
normal distribution with mean zero and variance 𝜎 2 . If the mean of 𝜀 is zero, then

E(y) ≡ 𝜂 = E[f (𝜉1 , 𝜉2 , … , 𝜉k )] + E(𝜀)

(1.2)
= f (𝜉1 , 𝜉2 , … , 𝜉k )
 1.1 RESPONSE SURFACE METHODOLOGY 3

Figure 1.2 (a) Response surface for the first-order model 𝜂 = 50 + 8x1 + 3x2 . (b) Contour plot for
the first-order model.

The variables 𝜉 1 , 𝜉 2 , … , 𝜉 k in Equation 1.2 are usually called the natural variables,
because they are expressed in the natural units of measurement, such as degrees Celsius
(◦ C), pounds per square inch (psi), or grams per liter for concentration. In much RSM work
it is convenient to transform the natural variables to coded variables x1 , x2 , … , xk , which
are usually defined to be dimensionless with mean zero and the same spread or standard
deviation. In terms of the coded variables, the true response function (1.2) is now written as

𝜂 = f (x1 , x2 , … , xk ) (1.3)

Because the form of the true response function f is unknown, we must approximate it.
In fact, successful use of RSM is critically dependent upon the experimenter’s ability to
develop a suitable approximation for f. Usually, a low-order polynomial in some relatively
small region of the independent variable space is appropriate. In many cases, either a first-
order or a second-order model is used. For the case of two independent variables, the
first-order model in terms of the coded variables is

𝜂 = 𝛽0 + 𝛽1 x1 + 𝛽2 x2 (1.4)

Figure 1.2 shows the three-dimensional response surface and the two-dimensional con-
tour plot for a particular case of the first-order model, namely,

𝜂 = 50 + 8x1 + 3x2

In three dimensions, the response surface for y is a plane lying above the x1 , x2 space. The
contour plot shows that the first-order model can be represented as parallel straight lines of
constant response in the x1 , x2 plane.
The first-order model is likely to be appropriate when the experimenter is interested in
approximating the true response surface over a relatively small region of the independent
variable space in a location where there is little curvature in f. For example, consider a small
region around the point A in Fig. 1.1b; the first-order model would likely be appropriate
here.
4 INTRODUCTION

Figure 1.3 (a) Response surface for the first-order model with interaction 𝜂 = 50 + 8x1 + 3x2 −
4x1 x2 . (b) Contour plot for the first-order model with interaction.

The form of the first-order model in Equation 1.4 is sometimes called a main effects
model, because it includes only the main effects of the two variables x1 and x2 . If there is
an interaction between these variables, it can be added to the model easily as follows:

𝜂 = 𝛽0 + 𝛽1 x1 + 𝛽2 x2 + 𝛽12 x1 x2 (1.5)

This is the first-order model with interaction. Figure 1.3 shows the three-dimensional
response surface and the contour plot for the special case

𝜂 = 50 + 8x1 + 3x2 − 4x1 x2

Notice that adding the interaction term −4x1 x2 introduces curvature into the response
function. This leads to different rates of change of the response as x1 is changed for
different fixed values of x2 . Similarly, the rate of change in y across x2 varies for different
fixed values of x1 .
Often the curvature in the true response surface is strong enough that the first-order
model (even with the interaction term included) is inadequate. A second-order model
will likely be required in these situations. For the case of two variables, the second-order
model is

𝜂 = 𝛽0 + 𝛽1 x1 + 𝛽2 x2 + 𝛽11 x12 + 𝛽22 x22 + 𝛽12 x1 x2 (1.6)

This model would likely be useful as an approximation to the true response surface in a
relatively small region around the point B in Fig. 1.1b, where there is substantial curvature
in the true response function f.
Figure 1.4 presents the response surface and contour plot for the special case of the
second-order model

𝜂 = 50 + 8x1 + 3x2 − 7x11

2 2
− 3x22 − 4x1 x2

Notice the mound-shaped response surface and elliptical contours generated by this model.
Such a response surface could arise in approximating a response such as yield, where we
would expect to be operating near a maximum point on the surface.
 1.1 RESPONSE SURFACE METHODOLOGY 5

Figure 1.4 (a) Response surface for the second-order model 𝜂 = 50 + 8x1 + 3x2 − 7x1 2 − 3x2 2 −
4x1 x2 . (b) Contour plot for the second-order model.

The second-order model is widely used in response surface methodology for several
reasons. Among these are the following:

1. The second-order model is very flexible. It can take on a wide variety of functional
forms, so it will often work well as an approximation to the true response surface.
Figure 1.5 shows several different response surfaces and contour plots that can be
generated by a second-order model.

Figure 1.5 Some examples of types of surfaces defined by the second-order model in two variables
x1 and x2 . (Adapted with permission from Empirical Model Building and Response Surfaces, G. E.
P. Box and N. R. Draper, John Wiley & Sons, New York, 1987.)
6 INTRODUCTION

2. It is easy to estimate the parameters (the 𝛽’s) in the second-order model. The method
of least squares, which is presented in Chapter 2, can be used for this purpose.
3. There is considerable practical experience indicating that second-order models work
well in solving real response surface problems.

In general, the first-order model is

𝜂 = 𝛽0 + 𝛽1 x1 + 𝛽2 x2 + ⋯ + 𝛽k xk (1.7)

and the second-order model is

∑
k ∑
k ∑∑
k
𝜂 = 𝛽0 + 𝛽j xj + 𝛽jj xj2 + 𝛽ij xi xj (1.8)
j=1 j=1 i<j=2

In some situations, approximating polynomials of order higher than two are used. The
general motivation for a polynomial approximation for the true response function f is
based on the Taylor series expansion around the point x10 , x20 , … , xk0 . For example, the
first-order model is developed from the first-order Taylor series expansion

𝜕f ||
f ≅ f (x10 , x20 , … , xk0 ) + 𝜕x1 ||
(x1 − x10 )
x=x0
(1.9)
𝜕f | 𝜕f ||
+ 𝜕x || (x2 − x20 ) + ⋯ + 𝜕xk ||
(xk − xk0 )
2 |x=x x=x0
0

where x refers to the vector of independent variables and x0 is the vector of inde-
pendent variables at the specific point x10 , x20 , … , xk0 . In Equation 1.9 we have only
included the first-order terms in the expansion, so if we let 𝛽 0 = f(x10 , x20 , … , xk0 ),
𝛽1 = (𝜕f ∕𝜕x1 )||x=x , … , 𝛽k = (𝜕f ∕𝜕xk )||x=x , we have the first-order approximating model
0 0
in Equation 1.7. If we were to include second-order terms in Equation 1.9, this would lead
to the second-order approximating model in Equation 1.8.
Finally, note that there is a close connection between RSM and linear regression
analysis. For example, consider the model

y = 𝛽0 + 𝛽1 x1 + 𝛽2 x2 + ⋯ + 𝛽k xk + 𝜀

The 𝛽’s are a set of unknown parameters. To estimate the values of these parameters,
we must collect data on the system we are studying. Regression analysis is a branch of
statistical model building that uses these data to estimate the 𝛽’s. Because, in general,
polynomial models are linear functions of the unknown 𝛽’s, we refer to the technique
as linear regression analysis. We will also see that it is very important to plan the data
collection phase of a response surface study carefully. In fact, special types of experimental
designs, called response surface designs, are valuable in this regard. A substantial part of
this book is devoted to response surface designs. Note that analyses and designs need to be
carefully matched. If we are planning to analyze data from our planned experiment using
a first order model, then the design that we select should be well suited for this analysis.
Similarly, if we anticipate curvature similar to what can be modeled with a second-order
model, then a different design should be selected.
 1.1 RESPONSE SURFACE METHODOLOGY 7

Good response surface designs have been constructed to perform well based on a par-
ticular assumed model, but also have been structured so that they are able to evaluate the
assumptions of the model being analyzed to determine if the experimenter’s initial impres-
sions of the system under study match the true underlying relationship which produced the
data to be analyzed. Hence the experimenter should think carefully about the goals of a
particular experiment and what the anticipated analysis will involve before selecting the
design for data collection.

1.1.2 The Sequential Nature of RSM

Most applications of RSM are sequential in nature. That is, at first some ideas are generated
concerning which factors or variables are likely to be important in the response surface
study. This usually leads to an experiment designed to investigate these factors with a view
toward verifying the role of the factors in influencing the response and eliminating the
unimportant ones. This type of experiment is usually called a screening experiment. Often
at the outset of a response surface study there is a rather long list of variables that could
be important in explaining the response. The objective of factor screening is to reduce this
list of candidate variables to a relative few so that subsequent experiments will be more
efficient and require fewer runs or tests. We refer to a screening experiment as phase zero of
a response surface study. Since interest in a screening experiment lies in understanding the
gross behavior of the system and how factors are related to the response, a first-order model
is commonly selected. The class of response surface designs which are used for screening
experiments are well suited for gaining understanding about the main effects from different
independent variables and comparing their relative contributions to changes in the response
values. Since this represents an early stage in the planned sequence of experiments, the
goal is to determine which of the factors are more influential on the response while using as
small a fraction of the total experimental budget as possible. You should never undertake a
response surface analysis until a screening experiment has been performed to identify the
important factors.
Once the important independent variables are identified, phase one of the response
surface study begins. In this phase, the experimenter’s objective is to determine where the
collected data lie relative to an ideal response. Often, there are two possible outcomes with
the current levels or settings of the independent variables resulting in a value of the response
that is near the optimum (such as the point B in Fig. 1.1b), or the process is operating in
some other region that is (possibly) remote from the optimum (such as the point A in
Fig. 1.1b). If the current settings or levels of the independent variables are not consistent
with optimum performance, then the experimenter must determine a set of adjustments to
the process variables that will move the process toward the optimum. This phase of response
surface methodology makes considerable use of the first-order model and an optimization
technique called the method of steepest ascent. These techniques will be discussed and
illustrated in Chapter 5.
Phase two of a response surface study begins when the process is near the optimum.
At this point the experimenter usually wants a model that will accurately approximate
the true response function within a relatively small region around the optimum. Because
the true response surface usually exhibits curvature near the optimum (refer to Fig. 1.1),
a second-order model (or very occasionally some higher-order polynomial) will be used.
Once an appropriate approximating model has been obtained, this model may be analyzed
8 INTRODUCTION

to determine the optimum conditions for the process. Chapter 6 will present techniques for
the analysis of the second-order model and the determination of optimum conditions.
Response surface designs for modeling the response near the optimum are again selected
to match the anticipated analysis. Often, the plan is to characterize the relationship between
the response and the key independent variables using the second-order model of the form
in Equation 1.8. Designs are constructed to be able to estimate the response for input
factor combinations around the expected optimum, where curvature in the relationship is
common. Since this stage of experimentation is focused on determining a best set of input
values for which the process to operate, a generous portion of the experimental budget is
generally reserved for this portion of the process.
A final stage of experimentation, which generally does not require sophisticated response
surface designs or a large portion of the experimental budget, is a confirmatory experi-
ment. This data collection is generally simple and small, but is designed to confirm that
the identified optimum that was obtained in phase two can be achieved by setting the
independent variables at the designated settings.
This sequential experimental process is usually performed within some region of the
independent variable space called the operability region. For the chemical process illus-
trated in Fig. 1.1, the operability region is 0 hr < 𝜉 1 ≤ 7 hr and 100◦ C ≤ 𝜉 2 ≤ 800◦ C.
Suppose we are currently operating at the levels 𝜉 1 = 2.5 hr and 𝜉 2 = 500◦ C, shown as
point A in Fig. 1.6. Now it is unlikely that we would want to explore the entire region of
operability with a single experiment. Instead, we usually define a smaller region of interest
or region of experimentation around the point A within the larger region of operability.
Typically, this region of experimentation is either a cuboidal region, as shown around the
point A in Fig. 1.6, or a spherical region, as shown around point B. The choice of response
surface design matches the specified region of experimentation.

Figure 1.6 The region of operability and the region of experimentation.

 1.1 RESPONSE SURFACE METHODOLOGY 9

The sequential nature of response surface methodology allows the experimenter to learn
about the process or system under study as the investigation proceeds. This ensures that
over the course of the RSM application the experimenter will learn the answers to questions
such as (1) the location of the region of the optimum, (2) the type of approximating function
required, (3) the proper choice of experimental designs, (4) how much replication is neces-
sary, and (5) whether or not transformations on the responses or any of the process variables
are required. Because the nature of a response surface study has multiple stages with dif-
ferent goals, there are several improtant aspects that need to be managed throughout the
process. First, many studies have budget constraints that will dicate how much and what data
can be collected. It is important to plan for all of the stages of the study and to allow for ade-
quate resources to be available to effectively answer the important questions in each phase.
Second, since the knowledge gained in early phases of the study help to determine
what subsequent experiments will study, it is important to plan how the different phases
will connect to each other, and what information can be leveraged from early phases.
Thirdly, the selection of a model for the analysis of the data from each phase is based on
current understanding of the underlying process. It is important to think of the sequence of
experiments as a mechanism for not having to make too many assumptions at any stage,
Running a large complicated experiment that has many untested assumptions can lead to
costly errors and wasting of resources. Hence, a series of smaller experiments can verify
some assumptions early in the sequence and can allow the experimenter to proceed in later
stages with greater confidence.
Lastly, we again mention the connection between the choice of experiment and the
planned analysis. Before jumping in to collect data, the goals of each phase should be
clearly defined, and the nature of the response surface design selected should reflect the
goals and the planned analysis. Since there are often surprises when collecting and analyzing
data, it is helpful to consider what could go wrong with the experiment and to have a plan
for how to deal with some of these surprises. A substantial portion of this book—Chapters
3, 4, 8, and 9—is devoted to designed experiments useful in RSM.

1.1.3 Objectives and Typical Applications of RSM

Response surface methodology is useful in the solution of many types of industrial prob-
lems. Generally, these problems fall into three categories:

1. Mapping a Response Surface over a Particular Region of Interest. Consider the

chemical process in Fig. 1.1b. Normally, this process would operate at a particular
setting of reaction time and reaction temperature. However, some changes to these
normal operating levels might occasionally be necessary, perhaps to produce a prod-
uct that meets other specific customer requirements. If the true unknown response
function has been approximated over a region around the current operating condi-
tions with a suitable fitted response surface (say a second-order surface), then the
process engineer can predict in advance the changes in yield that will result from any
readjustments to the input variables, namely, time and temperature.
2. Optimization of the Response. In the industrial world, a very important problem is
determining the conditions that optimize the process. In the chemical process of
Fig. 1.1b, this implies determining the levels of time and temperature that result in
maximum yield. An RSM study that began near point A in Fig. 1.1b would eventually
10 INTRODUCTION

lead the experimenter to the region near point B. A second-order model could then be
used to approximate the yield response in a narrow region around point B, and from
examination of this approximating response surface the optimum levels or condition
for time and temperature could be chosen.
3. Selection of Operating Conditions to Achieve Specifications or Customer Require-
ments. In most response surface problems there are several responses that must be
simultaneously considered. For example, in the chemical process of Fig. 1.1, suppose
that in addition to yield, there are two other responses: cost and concentration. We
would like to maintain yield above 70%, while simultaneously keeping the cost below
$34/pound; however, the customer has imposed specifications for concentration such
that this important physical property must be 65±3 g/liter.

One way that we could solve this problem is to obtain response surfaces for all three
responses—yield, cost, and concentration—and then superimpose the contours for these
responses in the time–temperature plane, as illustrated in Fig. 1.7. In this figure we have
shown the contours for yield = 70%, cost = $34/pound, concentration = 62 g/liter, and con-
centration = 68 g/liter. The unshaded region in this figure represents the region containing
operating conditions that simultaneously satisfy all requirements on the process.
In practice, complex process optimization problems such as this can often be solved
by superimposing appropriate response surface contours. However, it is not unusual to
encounter problems with more than two process variables and more complex response
requirements to satisfy. In such problems, other optimization methods that are more effective
than overlaying contour plots will be necessary, and can often not only identify a region
which satisfies the minimal customer requirements, but also find an optimal combination
of input variables to achieve ideal performance. We will discuss methodology for solving
these types of problems in Chapter 7.

Figure 1.7 The unshaded region showing the conditions for which yield ≥70%, cost ≤$34/pound,
and 62 g/liter ≤ concentration ≤ 68 g/liter.
 1.2 PRODUCT DESIGN AND FORMULATION (MIXTURE PROBLEMS) 11

1.1.4 RSM and the Philosophy of Quality Improvement

During the last few decades, industrial organizations in the United States and Europe
have become keenly interested in quality and process improvement. Statistical methods,
including statistical process control (SPC) and design of experiments, play a key role in
this activity. Quality improvement is most effective when it occurs early in the product and
process development cycle. It is very difficult, expensive, and inefficient to manufacture
a poorly designed product. Industries such as semiconductors and electronics, aerospace,
automotive, biotechnology and pharmaceuticals, medical devices, chemical, and process
industries are all examples where experimental design methodology has resulted in shorter
design and development time for new products, as well as products that are easier to
manufacture, have higher reliability, have enhanced field performance, and meet or exceed
customer requirements.
RSM is an important branch of experimental design in this regard. RSM is a critical
technology in developing new processes, optimizing their performance, and improving the
design and/or formulation of new products. It is often an important concurrent engineering
tool, in that product design, process development, quality, manufacturing engineering,
and operations personnel often work together in a team environment to apply RSM. The
objectives of quality improvement, including reduction of variability and improved product
and process performance, can often be accomplished directly using RSM.

1.2 PRODUCT DESIGN AND FORMULATION (MIXTURE PROBLEMS)

Many product design and development activities involve formulation problems, in which
two or more ingredients are mixed together. For example, suppose we are developing a
new household cleaning product. This product is formulated by mixing several chemical
surfactants together. The product engineer or scientist would like to find an appropriate
blend of the ingredients so that the grease-cutting capability of the cleaner is good, and so
that it generates an appropriate level of foam when in use. In this situation the response
variables—namely, grease-cutting ability and amount of foam—depend on the percentages
or proportions of the individual chemical surfactants (the ingredients) that are present in
the product formulation.
There are many industrial problems where the response variables of interest in the
product are a function of the proportions of the different ingredients used in its formulation.
This is a special type of response surface problem called a mixture problem.
While we traditionally think of mixture problems in the product design or formula-
tion environment, they occur in many other settings. Consider plasma etching of silicon
wafers, a common manufacturing process in the semiconductor industry. Etching is usually
accomplished by introducing a blend of gases inside a chamber containing the wafers. The
measured responses include the etch rate, the uniformity of the etch, and the selectivity
(a measure of the relative etch rates of the different materials on the wafer). All of these
responses are a function of the proportions of the different ingredients blended together in
the etching chamber.
There are special response surface design techniques and model-building methods for
mixture problems. These techniques are discussed in Chapters 12 and 13.
Another random document with
no related content on Scribd:
in intaglio; others were beaten up with the hammer or chisel so as to
stand in gentle salience above their bed. But, speaking generally, no
attempt was made to model the nude figures of men or women in the
round. No suspicion of the wealth of suggestion latent especially in
the latter, seems to have dawned upon the Assyrian mind. If we
except a few terra-cotta statuettes, the artist who in some way gave
proof of so much resource, of so much skill and ingenuity, seems
never to have felt the charm of female beauty. The beauty of woman
is the light of nature, the perennial joy of the eye; to exclude it from
the ideal world created by the plastic arts is to condemn that world to
a perpetual twilight, to cast over it a veil of chill monotony and
sadness.
 Fig. 261.—Egyptian mirror, actual
size. Louvre. Drawn by Saint-Elme
Gautier.
In the arts of all those peoples who received the teachings of
Egypt and Chaldæa, whether at first hand, like the Phœnicians, or at
second, like the Greeks, the two distinct influences can always be
traced. Mesopotamia may be recognized in certain ornamental
motives, such as the “knop and flower,” the rosettes and palmettes,
as well as in its taste for the symmetry given by coupled figures; still
more clearly is it betrayed in motives into which lions and the whole
tribe of fantastic animals are introduced, struggling with and
devouring each other, and occasionally brought to the ground by
some individual dressed in a long gaberdine and crowned with a
tiara.
On the other hand, it is to Egypt that our thoughts are turned
when the human body meets our eyes in its unveiled nobility, with all
the variety of attitude and outline its forms imply. The peoples of
Western Asia learnt much in the school of the Chaldæan artist, but
the teaching given by the Egyptian sculptor was of a higher order,
and far better adapted to guide them in the way that leads to those
exquisite creations in which delicacy and certainty of hand are
happily allied with imaginative power. Sooner or later such teaching
must have aroused, in open and inquiring minds, a feeling for beauty
like that felt in her peculiar fashion by Egypt, a feeling to which
Greece, when once put in her right way, gave the fullest expression it
has ever received in marble and bronze.
In order to make good a comparison that no historian of art can
avoid, we have placed ourselves successively at two different points
of view, and from both we have arrived at the same result: as artists
the Egyptians take a higher rank than the Assyrians, than those
constructors who obstinately neglected the column even when they
built with stone, than those sculptors who avoided measuring
themselves with nature, and who shirked her difficulties by draping
their figures. But before thus bringing the two methods and the two
ways of looking at form into opposition, we ought perhaps to have
pointed out a difference in which this inequality is foreshadowed. In
all the monarchies of the East the great monuments were
anonymous, or, at least, if a name was given in the official texts it
was not that of the artist who conceived them, but of the king under
whom they were created. It is not till we arrive at Greece that we find
public opinion placing the work of art and its author so high that the
latter feels himself justified in signing his own creation. But although
this practice was not inaugurated in Egypt, numerous inscriptions
bear witness to the high rank held in Egyptian society by the artists
to whom the king confided the construction and decoration of his
buildings.[462] These men were not only well paid; they received
honours which they are careful to record, and their fame was spread
over the whole valley of the Nile. In the cuneiform texts we have so
far failed to discover the name of a single architect or sculptor, and it
does not appear that a reason for the omission is to be sought in the
peculiar conditions of Chaldæo-Assyrian epigraphy. Although
Babylon and Nineveh have not left us thousands of epitaphs like
those rescued from the sands of Egypt, we possess many private
contracts and agreements in which information similar to that
afforded in other countries by the sepulchral steles is to be found.
Neither there nor elsewhere do we find a trace of anything
corresponding to the conspicuous rank held under the Theban
princes of the eighteenth and nineteenth dynasties, by a Semnat, a
Bakhenkhonsou, or any other of the royal architects whose names
have been handed down to us in the texts.
It is unlikely that this difference will vanish when more texts have
been translated. The inequality in the position of the two artists is
readily explained by the unequal development of the two arts.
Egyptian architecture is learned and skilful after a fashion quite
distinct from that of Mesopotamia. It is not content, like the latter, to
spread itself out laterally and to heap up huge masses of earth, to be
afterwards clothed in thin robes of enamelled faïence, of painted and
sculptured alabaster. In spite of their rich decorations, palaces like
those of Nimroud and Khorsabad never quite threw off their
appearance of gigantic improvisations. Their plans once determined
—and Assyrian plans only varied within very narrow limits—the
method of roofing, flat or vaulted, fixed upon for each apartment, all
the rest was only a matter of foremen and their legions of half-skilled
workmen. At the very least we may say that the architect who
superintended the building of a Ninevite palace had a far easier task
than his rival of Thebes or Memphis. The arrangement of porticoes
and hypostyle halls demanded much thought and taste, and, if the
work when finished was at all to come up to the ideas of its creator,
the workmen who cut the graceful capitals and sturdy architraves
from the huge masses of granite, sandstone, or limestone, had to be
supervised with an unremitting care unknown and uncalled for in
Mesopotamia. The architects who raised the colonnades of Karnak
and the Ramesseum for Seti and his famous son, were the Ictinus
and Mnesicles of the East. We may become better acquainted than
we are now with the monumental history of Mesopotamia, but we
shall never find within her borders artists worthy to be placed on a
level with those Theban masters.
And if we compare the sculptors of Thebes and Nineveh, we shall
arrive at the same conclusion. On the one hand we find artists who,
whether they worked for the tomb or the temple, in the most
stubborn or the most kindly materials, chiselled images that either
delight us with their simple truth, or impress us with their noble
gravity and colossal size. A whole nation of statues issued from
those Egyptian studios through which we have conducted our
readers, many of them real masterpieces in their way. In
Mesopotamia, after early attempts that seemed full of promise, the
art of modelling statues was soon abandoned. In the glorious days of
Nineveh, all that was required of the sculptor was a talent, we might
say a knack, for cutting in the soft gypsum or limestone realistic
illustrations of the conquests and hunts of the reigning prince. He
had to turn out purely historical and anecdotic sculpture by the yard,
or rather by the mile; while in Egypt we see the whole nation, with its
kings and gods, revive to a second life in those forceful and sincere
portraits of which so many thousands have come down to our day.
In placing the distinctive features of the individual upon wood or
stone, the sculptor did something more than flatter the vanity of the
great; he prolonged their existence, he helped them to keep off the
assaults of death and to defy annihilation. From Pharaoh to the
humblest fellah, every one had to conciliate the man who possessed
such a quasi-magic power, and from whom such an all-important
service might have to be demanded. The common people bought
ready-made figures in a shop, on which they were content to cut
their names, but the kings and nobles commissioned their statues
from the best artists of the time, and some reflex from the respect
and admiration surrounding the sovereign must have fallen upon the
man to whom he confided the task of giving perpetuity to his royal
features, in those statues that during the whole of his reign would
stand on the thresholds and about the courts of the temple, and on
the painted walls of that happy abode to whose shadows he would
turn when full of years and eager for rest.
 If, before the advent of the Greeks, there were any people in the
ancient world in whom a passion for beauty was innate, they were
the people of Egypt. The taste of Chaldæa was narrower, less frank
and less unerring; she was unable, at least in the same degree, to
ally force with grace; her ideal had less nobility, and her hand less
freedom and variety. It is by merits of a different kind that she regains
the advantage lost in the arts. If her artists fell short of their rivals,
her savants seem to have been superior to those of Egypt. In their
easy-going and well-organized life, the Egyptians appear to have
allowed the inquiring side of their intellects to go to sleep. Morality
seems to have occupied them more than science; they made no
great efforts to think.
The Chaldæans were the reverse of all this. We have reason to
believe that they were the first to ask themselves the question upon
which all philosophy is founded, the question as to the true origin of
things. Their solution of the problem was embodied in the
cosmogonies handed down to us in fragments by the Greek writers,
and although their conceptions have only been received through
intermediaries by whom their meaning has often been altered and
falsified, we are still enabled to grasp their fundamental idea through
all the obscurities due to a double and sometimes triple translation,
and that idea was that the world was created by natural forces, by
the action of causes even now at work. The first dogma of the
Babylonian religion was the spontaneous generation of things from
the liquid element.[463]
The first vague presentiment and rough sketch, as it were, of
certain theories that have made a great noise in the world in our own
day, may be traced, it is asserted, in the cosmogonic writings of
ancient Chaldæa. Even the famous hypothesis of Darwin has been
searched for and found, if we may believe the searchers. In any case
it seems well established that the echo of these speculations
reached the Ionian sages who were the fathers of Greek philosophy.
Their traces are perceptible, some scholars declare, in the Theogony
of Hesiod. Possibly it is so; there are certainly some striking points of
resemblance; but where the influence of such ideas is really and
clearly evident is in those philosophic poems that succeeded each
other about the sixth century B.C., all under the same title:
concerning nature (περὶφύσεως).[464] These poems are now lost, but
judging from what we are told by men who read them in the original,
the explanation they gave of the creation of the world and of the first
appearance upon it of organized beings, differed only in its more
abstract character from that proposed many centuries before, and
under the form of a myth, by the priests of Chaldæa. If we may trust
certain indications, these bold and ingenious doctrines crossed over
from Ionia to the mainland of Greece, and reached the ears of such
writers as Aristophanes and Plato.
It does the greatest honour to Chaldæa that its bold speculations
should have thus contributed to awaken the lofty intellectual
ambitions and the scientific curiosity of Greece, and perhaps she
may have rendered the latter country a still more signal service in
teaching her those methods by whose use man draws himself clear
of barbarism and starts on the road to civilization; a single example
of this will be sufficient. It is more than forty years since Bœckh, and
Brandis after him, proved that all the measures of length, weight and
capacity used by the ancients, were correlated in the same fashion
and belonged to one scale. Whether we turn to Persia, to Phœnicia
or Palestine, to Athens or Rome, we are constantly met by the
sexagesimal system of the Babylonians. The measurements of time
and of the diurnal passage of the sun employed by all those peoples,
were founded on the same divisions and borrowed from the same
inventors. It is to the same people that we owe our week of seven
days, which, though not at first adopted by the western nations,
ended by imposing itself upon them.[465] As for astronomy, from a
period far away in the darkness of the past it seems to have been a
regular branch of learning in Chaldæa; the Greeks knew very little
about it before the conquests of Alexander; it was more than a
century after the capture of Babylon by the Macedonians that the
famous astrological tables were first utilized by Hipparchus.[466]
In the sequel we shall come upon further borrowings and
connections of this kind, whose interest and importance has never
been suspected by the historian until within the last few years. Take
the chief gods and demi-gods to whom the homage of the peoples of
Syria and Asia Minor was paid, and you will have no difficulty in
acknowledging that, although their names were often changed on
the way, Mesopotamia was the starting place of them all. By
highways of the sea as well as those on land, the peoples
established on the eastern shores of the Mediterranean entered into
relations with the tribes of another race who dwelt on the European
coasts of the same sea; they introduced them to their divinities and
taught them the rites by which those divinities were honoured and
the forms under which they were figured. Without abandoning the
gods they worshipped in common with their brother Aryans, the
Greeks adopted more than one of these Oriental deities. This is not
the place to consider the question in detail. We must put aside for
the present both the Cybele of Cappadocia and Phrygia and that
Ephesian Artemis, who, after being domiciled and naturalized in one
of the Hellenic capitals, so obstinately and so long preserved her
foreign characteristics; we must for the moment forget Aphrodite,
that goddess of a different fortune whose name is enough to call up
visions of not a few masterpieces of classic art and poetry. Does not
all that we know of this daughter of the sea, of her journeys, of the
first temples erected to her on the Grecian coasts and of the peculiar
character of her rites and attributes—does not all this justify us in
making her a lineal descendant of Zarpanitu, of Mylitta and Istar, of
all those goddesses of love and motherhood created by the
imagination and worshipped by the piety of the Semites of Chaldæa?
On the other hand the more we know of Egypt the less inclined are
we to think that any of the gods of her Pantheon were transported to
Greece and Italy, at least in the early days of antiquity.
Incomplete as they cannot help being, these remarks had to be
made. They will explain why in the scheme of our work we have
given similar places to Chaldæa and Egypt. The artist will always
have a predilection for the latter country, a preference he will find no
difficulty in justifying; but the historian cannot take quite the same
view. It is his special business to weigh the contributions of each
nation to the common patrimony of civilization, and he will
understand how it is that Chaldæa, in spite of the deficiencies of its
plastic art, worked more for others than Egypt and gave more of its
substance and life. Hidden among surrounding deserts the valley of
the Nile only opened upon the rest of the world by the ports on a
single short line of frontier. The basin of the Euphrates was much
more easily accessible. It had no frontier washed by the
Mediterranean, but it communicated with that sea by more numerous
routes than Egypt, and by routes whose diversity enhanced the
effect of the examples they were the means of conveying to the
outer world.
It is, to all appearance, to the ancient inhabitants of Mesopotamia
that humanity owes the cultivation of wheat, its chief alimentary
plant.[467] This precious cereal seems to have been a native of the
valleys of the Indus and Euphrates; nowhere else is it found in a wild
state. From those two regions it must have spread eastwards across
India to China, and westwards across Syria into Egypt and
afterwards on to the European continent. From the rich plains where
the Hebrew tradition set the cradle of the human race, the winds
carried many seeds besides those by which men’s bodies have so
long been nourished; the germs of all useful arts and of all mental
activities were borne on their breath like a fertilizing dust. Among
those distant ancestors of whom we are the direct heirs, those
ancestors who have left us that heritage of civilization which grows
with every year that passes, there are none, perhaps, to whom our
respect and our filial gratitude are more justly due than to the ancient
inhabitants of Chaldæa.
 ADDITIONS AND CORRECTIONS.

VOL I.

P. 155.—It was not only as mortar that bitumen was used. Mr.
Rassam tells us that he found at Abou-Abba (Sippara), in Chaldæa,
a chamber paved with asphalte much in the same fashion as a
modern street in London or Paris (Proceedings of the Society of
Biblical Archæology).
P. 200.—From a late communication to the Society of Biblical
Archæology we learn that Mr. Rassam found the Sippara tablet in
the corner of a room, under the floor; it was inclosed in an inscribed
earthenware box.
P. 242, line 12; for Shalmaneser III. read Shalmaneser II.
P. 266, line 8 from foot: for Plate X. read Plate IX.
P. 305.—Intercourse between the valley of the Nile and that of
the Tigris and Euphrates seems to have begun not sooner than the
eighteenth Egyptian dynasty. To this conclusion we are led both by
Egyptian texts and by the tablets in the library of Assurbanipal. Most
of the tablets are reprints—if we may say so—of texts dating
originally from Ur, and from the time of the ancient Chaldæan
monarchy. Now these texts seem to have been written by a people
who knew not Egypt; no mention of that country is to be found in
them. They contain a division of the world into four regions, in none
of which Egypt has a place (Sayce, The Early Relations of Egypt and
Babylonia, in Lepsius’s Zeitschrift, p. 150).
P. 349.—We may here draw attention to an object which may be
compared to that described by M. Clermont Ganneau, both for its
intrinsic character and its probable destination. It is a tablet in brown
limestone, portable, and surmounted by a ring or staple cut in the
material. On one face there is a bas-relief in which the goddess who
occupies the lower register in Péretié’s bronze again appears. She
has the head of a lioness, a snake dangles from each hand, the
arms are outstretched, and two animals, in which Layard recognises
a lioness and a sow, hang to her breasts. This goddess stands
before an animal which has a bull’s head in the engraving given by
Lajard. But its feet are those of a horse, and no doubt we should find
that the animal in question was a horse if we could examine the
original; but we do not know what has become of it. If, as there
seems reason to believe, this goddess is an infernal deity, it is easy
to understand why serpents were placed in her hands. These
reptiles are the symbols of resurrection; every year they quit their old
skins for new ones. The object in question is described in detail in
the Recherches sur le Culte de Vénus, p. 130, and figured in Plate
XVI, Fig. 1. Upon one of the larger faces of the tablet and upon its
edges there are inscriptions, magic formulæ according to M. Fr.
Lenormant.
This tablet was formerly in the cabinet of M. Rousseau, at one
time French consul at Bagdad. It was found in the ruins of Babylon.
Size, 24 inches high by 24 inches wide, and 3⅞ inches thick.
P. 384.—In speaking of the excavations made by Sir H.
Rawlinson at Borsippa, we forgot to mention his paper entitled On
the Birs Nimroud; or, The Great Temple of Borsippa (Journal of the
Royal Asiatic Society, vol. xviii. p. 1–32). Paragraphs 1 and 2 give an
account of the excavations, and we regret that we wrote of the
religious architecture of Chaldæa before having read them. Not that
they contain anything to cause us to change our conceptions of the
staged towers. The excavations seem to have been carried on with
great care, but they hardly gave results as complete as they might
have done had they been directed by a thoroughly-trained architect.
VOL. II.

P. 48.—Upon Arvil, the ancient Arbela, and the likelihood of great

discoveries in the mound which there rises 150 feet above the plain,
see a contribution from Sir H. Rawlinson to the Journal of the Royal
Asiatic Society, vol. i. new series, 1865, pp. 190–197. The mound is
at present crowned by a Turkish fort.
P. 176.—Herr Fritz Hommel, one of the few non-French
students of the remains from Tello, is no more inclined than we are to
allow that the igneous rocks from which they are cut were brought
from Egypt. He believes they were won from much nearer quarries,
viz., on the borders of the Arabian plateau (Die Vorsemitische
Kulturen in Egypten und Babylonien, pp. 211–223).
Pp. 188–190.—In enumerating the few monuments of Chaldæan
sculpture that we possess over and above those brought home by
M. de Sarzec, we forgot to mention a small Babylonian head in hard
alabaster, now in the Louvre (Fig. 262). Its workmanship resembles
that of the two heads from Tello (Plate VII.), and some of the small
heads from the same place. It is conspicuous for the same frank and
decided modelling, but it belongs to the period when long beards
were worn.
P. 202.—To the list of Chaldæan sculptures we should, perhaps,
add the rock-cut relief found by Sir H. Rawlinson in the district of
Zohab, about fifty leagues from the left bank of the Tigris, and to the
north-west of Bagdad, near the village of Sheikhan. This district
forms a part of the Persian province of Kirmanchah (Journal of the
Geographical Society, vol. ix. p. 31). The relief occurs, it seems, on
the high road between Babylon and Ecbatana, in the defile which is
now called Tak-i-Girrah, one of the passes leading up through Mount
Zagros to the plateau of Iran. There is a sketch of the relief from the
pencil of Sir H. Rawlinson in the Five Great Monarchies of his
brother (vol. iii. p. 7). The king stands with his foot on the body of a
conquered enemy. An individual, probably the royal general,
presents two kneeling captives, who are held by a cord attached to
rings put through their noses. More captives with ropes about their
necks are carved on the kind of plinth upon which the main group is
supported. The whole picture is about two feet wide and five high.
Near it there is an apparently unfinished inscription in Babylonish
cuneiform characters. The Chaldæan origin of the work is confirmed
by the flounces on the general’s robe. In the same neighbourhood
there are ruins which appear to date from a very early period.
 Fig. 262.—Fragment of a
Chaldæan statuette. Louvre.
P. 219.—We have here omitted to draw attention to one of the
differences between the art of the Sargonids and that of the
preceding dynasty. In the figures from Tello and in the bas-reliefs of
the time of Assurnazirpal the sculptor has left the eyeballs smooth
(Plate VII.; Vol. I. Fig. 15; Vol. II, Figs. 43, 64, 113). In the sculptures
of the time of Sargon and his successors, on the other hand, the
cornea is indicated in the figures both of men and animals, by a
clearly traced circle (Vol. I, Figs. 22 and 25; Vol. II. Fig. 118 and Plate
X.). It was, no doubt, the desire to give a more lifelike expression to
the physiognomy that led the artist thus to modify his proceeding.
There are a few figures in which the desire for imitative truth is
pushed even farther. In a bas-relief in the Louvre there is an eagle-
headed deity in which not only the cornea but the pupil also is
marked by a smaller circle within the first. See the De l’Expression
des Yeux dans la Statuaire of Doctor Debrou (Correspondant, April
10th, 1883). His special knowledge has enabled him to make more
than one remark upon the representation of the eye in ancient and
modern sculpture, to which writers upon art would do well to pay
attention.
P. 398.—On the subject of the female divinity whose worship was
so widely spread over the whole East and over the Mediterranean
coasts, the dissertation of Herr Gelzer, Zum Cultus der Assyrischen
Aphrodite (Lepsius’s Zeitschrift, 1875, p. 127) may be consulted with
profit.
 We received the admirable Guide to the Kuyundjik Gallery,
published by the authorities of the British Museum, too late to make
use of it for our work. It joins to an exhaustive account of the bas-
reliefs of Sennacherib and Assurbanipal a description of the smaller
objects contained in the glass cases of the same gallery. Many of
these objects date from a very early period, and many were found in
Chaldæa. Some of the more interesting texts are translated by Mr.
Pinches; of others he gives a summary. The body of the work is
preceded by an introduction giving such details of Assyrian history,
religion and manners as are required by the general student. When a
similar brochure is forthcoming for the Nimroud gallery—and the
energy of the English officials is a guarantee that we shall not have
to wait long for it—visitors to the museum will be in possession of all
that is necessary to enable them to profit to the fullest extent by its
superb collection.
 INDEX.

Abd-al-Medjid, Sultan, ii. 39.

Abdul-Hamid, ii. 76.
Abou-Abba (or Sippara), i. 200.
Abou-Sharein, i. 157, 190, 262; ii. 34.
Abraham, i. 15, 199.
Abydenus, i. 51;
spoken of by Eusebius, 57.
Abydos, on the Hellespont, ii. 220.
Accad, i. 14, 21, 59.
Acheron, i. 354.
Adar (Saturn?), i. 73.
Adrammelech, i. 103.
Agbatana, see Ecbatana.
Ahmes II., ii. 339.
Ahura-Mazda, i. 88.
Alabaster, found near Mossoul, i. 120;
its distribution, 121;
its constitution, 121;
its characteristics as a material for the sculptor, ii. 110.
Alabastron, ii. 301.
Alexander the Great, i. 54; ii. 382.
Alexander Polyhistor, i. 51.
Allat, i. 83, 345.
Alphabet, invention of, i. 22.
Altaï, i. 21.
Altars, their characteristic forms, i. 236;
with battlements, 255;
circular, 256;
sarcophagus-shaped, 256.
Amanus, ii. 340.
Amber, its absence from Mesopotamian remains, ii. 362.
Amen, i. 78, 79.
Ament, the Assyrian, i. 345.
Amiaud, M., i. 361.
Amoor, i. 19.
Amphora, ii. 300.
Amraphel, i. 36.
Amulets, ii. 251.
Anabasis quoted, i. 361.
Anaïtis, see Oannes.
Animals, grotesque and fantastic, in Assyrian Art, ii. 167–173;
on the seals, ii. 279.
Anthemius, i. 172.
Antioch, ii. 286.
Antiochus-Epiphanes, i. 33.
Aphrodite, ii. 398.
Apollonius of Tyana, i. 299.
Apsou (or Apason), i. 83.
Arade, ii. 265.
Aramaic, or Aramæan, came into common use with the second
Chaldæan Empire, i. 18.
Aram-Naharaim, i. 3.
Arbeles, Arvil (or Ervil), i. 6; ii. 48.
Arch, frequent use of, i. 132, 221;
invented in Chaldæa, 222;
at Mugheir, 222;
in the hanging gardens at Babylon, 223;
in Sargon’s gateways, 224;
in the sewers of the palaces, 227.
Archivolt, enamelled, at Khorsabad, i. 290.
Arioch, i. 36.
Aristophanes, ii. 397.
Aristotle, i. 71;
his Politics quoted in reference to the size of Babylon, ii. 56.
Arithmetic, Chaldæan, i. 68;
origin of the sexagesimal system, 68.
Armenia, annexed by Assyria, i. 7.
Arms, ii. 343.
Arrian, his Indian history, i. 57.
Artaxerxes Mnemon, i. 90.
Artists, their social position in Mesopotamia, ii. 394.
Aryans, said to compose part of the early Chaldæan population,
i. 18.
Aryballos, ii. 306.
Asia Minor, ii. 172.
Ass, the wild, in Assyrian sculpture, ii. 150.
Asshur, the builder of Nineveh, i. 14.
Assur, i. 83;
the Assyrian god par excellence, 84;
fell with Nineveh, 88.
Assurbanipal, i. 36, 44;
receives the homage of Gyges, king of Lydia, 44;
his cruelty, 47;
patron of literature and the arts, 47;
his library, 47;
numerous remains of sculpture dating from his reign, ii. 236.
Assurbilkala, ii. 101, 203, 289.
Assuredilani, i. 51.
Assurnazirpal, i. 42;
his statue in the round, ii. 126.
Assyria, its true boundaries, i. 7;
its successive capitals, 7;
its size, 7;
consolidation of its supremacy, 41;
first A. empire, 41;
second A. empire, 41;
expeditions into Armenia, Cappadocia, and Syria, 41;
strictly a military monarchy, 96;
its régime, 103;
Assyrian monarchy, solidity of the succession, 103;
characteristics of the Assyrian race, 105;
cruelty of the Assyrian kings, 105–7;
luxury of do., 105–7;
constitution of the Assyrian nation, 111;
comparative insignificance of civilian element, 112.
Assyro-Chaldæan language, the, i. 53.
Astarte, i. 345.
Astragali, i. 206.
Astrology, i. 65;
the forerunner of real astronomy, 67.
Athenæum quoted, i. 317.
Aturia, a variant of Assyria, i. 6.
Auben (or Auben-Ra), ii. 120.
Augustus, ii. 382.

Baal worshipped in Judah and Israel as well as Tyre and Sidon,

i. 16.
Baalazar, ii. 336.
Babel, i. 14, 53.
Bab-i-Houmaioun, ii. 72.
Babil, i. 130, 154;
its identity discussed, 384; ii. 35.
Babooshes, i. 238.
Babylon, age of its premiership, i. 38;
more tenacious of life than Nineveh, 54;
etymology of the name, 86;
natural elements of its prosperity, 92;
superiority of its situation over that of Nineveh, 93;
an “eternal city,” ii. 53;
its defences, 53;
incomplete nature of the explorations that have been carried
out on its site, 55;
its size discussed, 56–59;
the stone bridge, 57;
height of the walls, 63.
Bactriana, metals brought from, i. 125.
Bagdad, i. 40, 54.
Bahr-ul-nejef, ii. 176.
Bakhenkhonsou, ii. 394.
Balawat, gates of, i. 194;
steles figured on, 196;
standards figured on, 195;