doe made easy, yet powerful, with design-expert® software ... · a. culture-medium temperature b....

Copyright © 2018 Stat-Ease, Inc. Do not copy or redistribute in any form. 1

1

Making the most of this learning opportunityStat-Ease worldwide webinars attract many attendees, so, to prevent audio disruptions, all must be muted by the presenter. Also, to avoid interruptions and keep the presentation to about one hour, please hold all questions until afterwards and address them to [email protected].

– Mark

P.S. Find slides posted now at www.statease.com/webinar.html and, barring technical issues, a recording put up afterwards.

By Mark J. Anderson, PE, CQEStat-Ease, Inc., Minneapolis, MN

[email protected]

DOE Made Easy, Yet Powerful, with Design-Expert® Software

DOE Made Easy with DX

0.00 3.60 7.21 10.81 14.42 18.02 21.63

0

10

20

30

50

70

80

90

95

99

Half-Normal Plot

|Standardized Effect|

Half-N

orm

al %

Pro

bability

A-Temperature

C-ConcentrationD-Stir Rate

AC

AD

0.00

1.40

2.80

4.20

5.59

6.99

8.39

9.79

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Pareto Chart

Rank

t-V

alu

e o

f |E

ffe

ct|

Bonferroni Limit 3.82734

t-Value Limit 2.22814

A-Temperature

AC

AD

D-Stir Rate

C-Concentration

The WIIFM* for this Webinar*(What’s in it for me)

This demo of Design-Expert features factorials—the core tool for design of experiments (DOE), plus a few peeks at the peak technology for DOE: response surface methods (RSM).

See how you can quickly converge on the ‘sweet spot’—the most desirable combination of process parameters and product attributes.

Some WIFFMs for you to take home will be:

❖ Appreciation for multifactor testing.

❖ A tried-and-true strategy of experimentation.

❖ Inspiration to use stat tools that can greatly accelerate your research.

DOE Made Easy with DX 2

Please press the raise hand now if you are with me.

mailto:[email protected]

http://www.statease.com/webinar.html



3

Multi-Factorial (VS OFAT)(life from accelerated test)

"To make knowledge work productive will be the great management task of this century."

-- Peter Drucker

A- A+B-

B+

C-

C+

17

19

26

16

25

21

85

128Relative efficiency = 16/8

2 to 1!Start point for One Factor at a Time (OFAT)

Multifactor TestingMany experimenters cannot believe it

What did you do today?

No way! You tested more than one factor at a time?


I learned how to juggle, I tamed a wild horse and I did a DOE.

http://www.internet-wizard.net/


5

Strategy of Experimentation


6

Screening/Characterization

Purpose: Quickly sift through a large number

of potential factors to discard the trivial

many. Then follow-up with an experiment

that focuses on the vital few.

Tool: Two-level factorial designs:

1. Fractional for resolving main effects in

minimal runs.

2. Full (or less fractional) to resolve two-

factor interactions.


7

Screening/CharacterizationCase Study

A research scientist at a biotechnology firm reports* the use of DOE

in pilot studies on a pharmaceutical made via recombinant E. coli

fermentation. He screened six factors:

A. Culture-medium temperature

B. Fermentation-medium pH

C. Dissolved oxygen concentration (DO)

D. Induction optical density (OD) – a measure of cell density

E. Feed rate of proprietary complex solution

F. Feed rate of inducer solution

The primary response was harvest OD – a measure of cell-growth.

*(Arun Tholudur, et al, “Using Design of Experiments to Assess Escherichia Coli Fermentation Robustness,” BioProcess – posted at http://www.statease.com/pubs/fermentationrobustness.pdf.)


8

Full and Fractional Two-Level Designs

Factors

Run

s

A ‘happy medium’ for screening 6 factors: adequate resolution (IV) with

decent power (n=8 high vs low).


http://www.statease.com/pubs/fermentationrobustness.pdf


9

Experimental design template (1/4th fraction of 26 via std design with E=ABC and F=ABD)

Std A:

Temp.

B:

pH

C:

DO

D:

In. OD

E:

Soln 1

F:

Soln 2

1 Lo (–) Lo (–) Lo (–) Lo (–) Lo (–) Lo (–)

2 Hi (+) Lo (–) Lo (–) Lo (–) Hi (+) Hi (+)

3 Lo (–) Hi (+) Lo (–) Lo (–) Hi (+) Hi (+)

4 Hi (+) Hi (+) Lo (–) Lo (–) Lo (–) Lo (–)

5 Lo (–) Lo (–) Hi (+) Lo (–) Hi (+) Lo (–)

6 Hi (+) Lo (–) Hi (+) Lo (–) Lo (–) Hi (+)

7 Lo (–) Hi (+) Hi (+) Lo (–) Lo (–) Hi (+)

8 Hi (+) Hi (+) Hi (+) Lo (–) Hi (+) Lo (–)

9 Lo (–) Lo (–) Lo (–) Hi (+) Lo (–) Hi (+)

10 Hi (+) Lo (–) Lo (–) Hi (+) Hi (+) Lo (–)

11 Lo (–) Hi (+) Lo (–) Hi (+) Hi (+) Lo (–)

12 Hi (+) Hi (+) Lo (–) Hi (+) Lo (–) Hi (+)

13 Lo (–) Lo (–) Hi (+) Hi (+) Hi (+) Hi (+)

14 Hi (+) Lo (–) Hi (+) Hi (+) Lo (–) Lo (–)

15 Lo (–) Hi (+) Hi (+) Hi (+) Lo (–) Lo (–)

16 Hi (+) Hi (+) Hi (+) Hi (+) Hi (+) Hi (+)


Effects: Half-Normal Plot

Half-N

orm

al %

Pro

babili

ty

|Effect|

0 8 16 24 32

015

10

Hi

30

50

70

80

90

95

99

A

D

AD


How Half-Normal Plots* Help Experimenters*One of Design-Expert’s greatest strengths!

11DOE Made Easy with DX

Half-normal (aka “Daniel”) plots provide many advantages for factorial analysis:

❖ Easy to produce❖ Gives a quick visual of important effects❖ Effects can be compared to a reference—the near-zero

contrasts that emanate from the origin❖ Adapts to split-plot experiments [for hard-to-change factors]

“The greatest single advantage of the Daniel plot is its ability to stimulate discussion by encapsulating, in a single display, such a variety

of information.”

* (David Steinberg, “Discussion: On Daniel Plots”, Journal of Quality Technology, V47, N2,April 2015, p110.)

12

Effects: Ranked

Rank

T-v

alu

e o

f |E

ffect|

0.00

2.38

4.76

7.14

9.51

T Value Limit 2.16037

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

D

AD

A


Design-Expert makes it far easier to interpret this ordered bar chart of effects by it’s innovative enhancements:➢ Limit lines ➢ Color coding*➢ Interactivity*

* Also on half-normal


13

Interaction Effect

Lo Mid Hi

A: Temperature

Ha

rve

st O

D

48

66

84

101

119

D- Induction OD =Lo

Arun Tholudur's Fermentation


14

Minimum-Run Designs (up to 50 factors)Considerable Savings Over Standard Fractions

Factors Std Res V MR5* Factors Std Res IV MR4**

6 32 22 9 32 18

7 64 30 10 32 20

8 64 38 11 32 22

9 128 46 12 32 24

10 128 56 13 32 26

11 128 68 14 32 28

12 256 80 15 32 24

13 256 92 16 32 26

14 256 106 17 64 28

* Oehlert & Whitcomb, “Small, Efficient, Equireplicated Resolution V Fractions of 2k designs …”, Fall Technical Conference, 2002: www.statease.com/pubs/small5.pdf

** Anderson & Whitcomb, “Screening Process Factors In the Presence of Interactions,” Annual QualityCongress, American Society of Quality, Toronto, 2004: www.statease.com/pubs/aqc2004.pdf

Characterization Screening


http://www.statease.com/pubs/small5.pdf

http://www.statease.com/pubs/aqc2004.pdf


15

RSM

Strategy of Experimentation


16

Response Surface Methods (RSM)When to Apply It

1. Fractional factorials for screening such as MR4

2. High-resolution fractional, such as MR5, or full factorial for characterization of interactions (add center points at this stage to test for curvature)

3. Response surface methods (RSM) for optimization.

Contour maps (2D) and 3D surfaces guide you to the peak.



17

RSM: When to Apply It

Region of Operability

Region of InterestUse factorial design to get close to the peak. Then RSM to climb it.


18

RSM vs OFAT



19

RSM: Process Flowchart

Process

Vital Few Factors (x’s)

Measured Response(s) (y(s))

Subject Matter Knowledge(Plus Factorial Screening)

Polynomial Model

Fitting*

Response Surface

“All models are wrong, but some are useful.” - George BoxDOE Made Easy with DX

20

Carbon Monoxide EmissionsCase Study – RSM Design & Analysis

Via a response surface method (RSM) design aimed at minimizing carbon monoxide (CO) emission, Ford Motor UK varied 5 factors :

A. Engine loadB. Engine speedC. Spark advanceD. Air-to-fuel ratioE. Exhaust gas recycle

They used two test engines and set this up in two blocks.

Emission

Rebuild, analyze and optimize



21

Mixture DesignRSM Tailored to Formulation Space

Considerations:

➢ Factors are ingredients of a mixture.

➢ The response is a function of proportions, not amounts.

❖Given these two conditions, fixing the total (an equality

constraint) facilitates modeling of the response as a

function of component proportions.


22

A (1)

A (0)

B (0)

C (1)

C (0)

260

460

660

860

1060

E

lasticity

B (1)

Mixture Design:Rocket Science

Mixture model (special cubic):

Elasticity=

+ 351 * A

+ 446 * B

+ 653 * C

- 6 * AB

+1008 * AC

+1597 * BC

+6141 * ABCFile: RocketDOE Made Easy with DX


23

Mixture Case Study

Formulators* measured the effects of three mixture components on

an aqueous dispersion of polymeric nanospheres:

A. Poloxamer 188 NF (“P 188”)

B. Polyoxyethylene 40 monostearate NF (“Pe 40 Ms”)

C. Polyoxyethylene sorbitan fatty acid ester NF (“P.S.F.A.E.”)

They also studied the film-forming properties of this pharmaceutical

preparation. The measured responses were:

1. Particle size, nanometers (< 250 nm desired)

2. Glass transition, degrees C (< 17.5 deg desired)

Is there a sweet spot? See next slide!*(Frisbee, S.E., McGinity, J.W., “Influence of Nonionic Surfactants on the Physical and Chemical Properties of a

Biodegradable Pseudolatex,” European Journal of Pharmaceutics and Biopharmaceutics, V 40, No. 6 (Dec 1994).)


Framing the Sweet Spot for QbD

➢ Graphical optimization now frames the functional “design space” where all modeled responses for a unit operationfall within confidence intervals: Ideal tool for quality by design (QbD), for example, to see the real sweet spot in this pharmaceutical recipe.

24

A: P 188 1.000

B: Pe 40 Ms

1.000

C: P.S.F.A.E.

1.000

0.000 0.000

0.000

Overlay Plot

Particle size: 260.000

Particle size CI High: 260.000

Glass transition: 20.000

Glass transition CI High: 20.000

Glass transition CI High: 20.000

2

File: CEP-Mix



Plus free DOE appetizersonline!

252525

Stat-Ease Training: Sharpen Up via Computer-Intensive Workshops

Shari Kraber,Workshop Manager & Master [email protected]

Experiment DesignMade Easy

Response Surface Methods for Process Optimization

Mixture and CombinedDesigns for

Optimal Formulations

Designed Experiments forPharma, Life Sciences,

Medical Devices,Assay Optimization,

Oilfield Services,Food Science

Modern DOE for Process Optimization


26

References*

2nd edition 2016

* Taylor & Francis/CRC/Productivity Press

New York, NY.

3rd edition 2015 1st edition 2018


mailto:*[email protected]


The WIIFM* for this Webinar*(What’s in it for me)

This demo of Design-Expert features factorials—the core tool for design of experiments (DOE), plus a few peeks at the peak technology for DOE: response surface methods (RSM).

See how you can quickly converge on the ‘sweet spot’—the most desirable combination of process parameters and product attributes.

Some WIFFMs for you to take home will be:

❖ Appreciation for multifactor testing.

❖ A tried-and-true strategy of experimentation.

❖ Inspiration to use stat tools that can greatly accelerate your research.

PS (per bonus material): Formulators will do best by using mixture design.


Now you know.

Statistics Made Easy®


Best of luck for your experimenting!

Thanks for listening!

-- Mark

[email protected]

“Statistics used as a catalyst to engineering creation will always result in the fastest and most economical progress.”

- George Box


doe made easy, yet powerful, with design-expert® software ... · a. culture-medium temperature b....

Documents