Design and Analysis of Engineering Experiments

Chapter 11 Based on Design & Analysis of Experiments 7E 2009 Montgomery

1

Design and Analysis of Engineering Experiments

Ali Ahmad, PhD

Chapter 11 Design & Analysis of Experiments 7E 2009 Montgomery

2

Response Surface

Methodology


3

• Text reference, Chapter 11• Primary focus of previous chapters is

factor screening– Two-level factorials, fractional factorials are

widely used• Objective of RSM is optimization• RSM dates from the 1950s; early

applications in chemical industry• Modern applications of RSM span many

industrial and business settings


4

Response Surface Methodology

• Collection of mathematical and statistical techniques useful for the modeling and analysis of problems in which a response of interest is influenced by several variables

• Objective is to optimize the response


5

Steps in RSM

1. Find a suitable approximation for y = f(x) using LS {maybe a low – order polynomial}

2. Move towards the region of the optimum 3. When curvature is found find a new

approximation for y = f(x) {generally a higher order polynomial} and perform the “Response Surface Analysis”


6

Response Surface Models

0 1 1 2 2 12 1 2y x x x x

0 1 1 2 2y x x

2 20 1 1 2 2 12 1 2 11 1 22 2y x x x x x x

• Screening

• Steepest ascent

• Optimization


7

RSM is a Sequential Procedure

• Factor screening• Finding the

region of the optimum

• Modeling & Optimization of the response


8

The Method of Steepest Ascent• Text, Section 11.2• A procedure for moving

sequentially from an initial “guess” towards to region of the optimum

• Based on the fitted first-order model

• Steepest ascent is a gradient procedure

0 1 1 2 2ˆ ˆ ˆy x x


9

Example 11.1: An Example of Steepest Ascent


10


11


12


13


14


15


16


17

• Points on the path of steepest ascent are proportional to the magnitudes of the model regression coefficients

• The direction depends on the sign of the regression coefficient

• Step-by-step procedure:


18

Second-Order Models in RSM

• These models are used widely in practice• The Taylor series analogy• Fitting the model is easy, some nice designs are available• Optimization is easy• There is a lot of empirical evidence that they work very well


19


20


21


22

Characterization of the Response Surface

• Find out where our stationary point is • Find what type of surface we have

– Graphical Analysis – Canonical Analysis

• Determine the sensitivity of the response variable to the optimum value– Canonical Analysis


23

Finding the Stationary Point

• After fitting a second order model take the partial derivatives with respect to the xi’s and set to zero– δy / δx1 = . . . = δy / δxk = 0

• Stationary point represents… – Maximum Point – Minimum Point – Saddle Point


24

Stationary Point


25

Canonical Analysis

• Used for sensitivity analysis and stationary point identification

• Based on the analysis of a transformed model called: canonical form of the model

• Canonical Model form: y = ys + λ1w1

2 + λ2w22 + . . . + λkwk

2


26


27

Eigenvalues• The nature of the response can be determined by the

signs and magnitudes of the eigenvalues – {e} all positive: a minimum is found– {e} all negative: a maximum is found – {e} mixed: a saddle point is found

• Eigenvalues can be used to determine the sensitivity of the response with respect to the design factors

• The response surface is steepest in the direction (canonical) corresponding to the largest absolute eigenvalue


28


29


30


31


32


33


34

Ridge Systems


35


36


37


38

Overlay Contour Plots


39

Mathematical Programming Formulation


40

Desirability Function Method


41

1/1 2( ... ) m

mD d d d


42


43


44


45

Addition of center points is usually a good idea


46


47

The Rotatable CCD 1/ 4F


48


49

The Box-Behnken Design


50

A Design on A Cube – The Face-Centered CCD


51

Note that the design isn’t rotatable but the prediction variance is very good in the center of the region of experimentation


52

Other Designs

• Equiradial designs (k = 2 only)• The small composite design (SCD)

– Not a great choice because of poor prediction variance properties

• Hybrid designs– Excellent prediction variance properties– Unusual factor levels

• Computer-generated designs


53


54


55


56

Blocking in a Second-Order Design


57


58

Computer-Generated (Optimal) Designs

• These designs are good choices whenever– The experimental region is irregular– The model isn’t a standard one– There are unusual sample size or blocking

requirements• These designs are constructed using a

computer algorithm and a specified “optimality criterion”

• Many “standard” designs are either optimal or very nearly optimal


59


60


61


62

Which Criterion Should I Use?

• For fitting a first-order model, D is a good choice– Focus on estimating parameters– Useful in screening

• For fitting a second-order model, I is a good choice– Focus on response prediction– Appropriate for optimization


63

The Adhesive Pull-Off Force Experiment – a “Standard” Design


64

A D-Optimal Design


65


66


67


68


69


70

Evolutionary Operation (EVOP)• An experimental deign based technique for

continuous monitoring and improvement of a process

• Small changes are continuously introduced in the important variables of a process and the effects evaluated

• The 2-level factorial is recommended• There are usually only 2 or 3 factors considered• EVOP has not been widely used in practice• The text has a complete example

Design and Analysis of Engineering Experiments

Documents