Have you always wondered why single-factor optimization is so difficult?
Or are you facing a new investigation/optimization and want to try another more promising method this time? As a DoE trainer, I am always asked why all the DoE effort when you can clearly optimize one factor after the other. ... Is this really the case? ... I have created a small overview, why “DoE” is possibly more elegant and sustainable than “OFAT” .... But, decide for yourself ...
1. By info@stefan-moser.com
DO you have an "optimization
problem" with some factors?
Which way is more effective?
OFAT DoE
„One Factor at a time“ „Design of Experiments“
On the following slides, some key
points are compared regarding the
further optimization methods!
2. By info@stefan-moser.com
The general focus of the
methodes is …
OFAT DoE
Start immediately and optimize
one factor at a time....
Identify and prioritize factors
and set variation width
3. By info@stefan-moser.com
… but, what happens between the
experiments...
OFAT DoE
Frequent, iterative
Discussions on whether and
how to proceed
The tests are all carried out in
one sequence if possible
4. By info@stefan-moser.com
What is not on the radar and still
affects the outcome?
OFAT DoE
Additional environmental
factors are less frequently
recorded as the concentration
is usually directed to the one
variable or factor .
All factors are varied
simultaneously according to a
systematic pattern, disturbances
can be recorded and evaluated
according to their trend
x1
x2
x1
x3
x2
5. By info@stefan-moser.com
How well things can be planned?
OFAT DoE
The experiments correspond
to a "feeling out" and can
therefore be poorly planned
systematically
Structured experimental designs
allow for better and more
forward-looking planning
Funktion
Budget
Ressourcen /&
Mitarbeiter
Zeitplan
6. By info@stefan-moser.com
OFAT DoE
Persistent but slow knowledge
building
Rapid increase in knowledge,
with significantly fewer attempts
What are the learning curves?
Wissen
Versuche
Wissen
Versuche
7. By info@stefan-moser.com
Basic orientation of the methods
OFAT DoE
Search for the best setting
and the best result
The aim is to describe the
entire experimental space and
to derive a cause-effect
model.
x1
x2
8. By info@stefan-moser.com
20% Yield
30
40
50
60 70
20% Yield
30
40
50
60 70
Limits of interpretation
OFAT DoE
Search for the best setting,
the best result, Process maps
creation is only possible to a
very limited extent
Aim to describe the entire design
space and generating process
maps is possible.
9. By info@stefan-moser.com
Description of interactions
OFAT DoE
Mostly not possible due to
lack of design in the
arrangement of the
experiments.
Possible by structured
arrangement of the
experiments
y1
x1
X2 ↑
X2 ↑
X2 ↓
X2 ↓
x1
x2
10. By info@stefan-moser.com
Describe nonlinear effects.
OFAT DoE
Only partially possible due
to arbitrary arrangement
of the experiments
Possible by structured
arrangement of the experiments
y1
x1
y1
x1
11. By info@stefan-moser.com
-1 1
Derivation of factor-effects
OFAT DoE
Mostly only partially possible
due to arbitrary depending
arrangement of the
experiments
Possible by structured
arrangement of the
experiments
y1
x1
y1
x1
𝑚 =
𝑌
𝑋 ≜ 2
2
12. By info@stefan-moser.com
x1 x2 x3 x4 x5 x6
Comparability of effects
OFAT DoE
Only very conditionally
possible due to arbitrary
arrangement of the
experiments
Possible by standardizing and
scaling the factor variation
y1
x1
13. By info@stefan-moser.com
Investigation of many
factors (10+)
OFAT DoE
Not possible!
No structured approach
Possible by using a design to
create the Experimental plan
2n +1 2n … 22 21 20
14. By info@stefan-moser.com
Investigation of many factors and
responses.
OFAT DoE
Not possible!
No structured approach.
Possible by using the design as
the basis of the
Cause-and-effect model
factors (X)
Input
Transfer function
Disturbances
Output
Cause - Effect - Model
System
oder
Prozess
Responses(y)
15. By info@stefan-moser.com
Follow-up studies of non-linear or
interaction effects
OFAT DoE
Mostly not possible due to lack
of design in the arrangement
of the experiments
Possible, by structured
complementary arrangement of
the experiments
x1
x2
x1
x3
x2
Missing most of the
time!
16. By info@stefan-moser.com
Buidling a Cause-effect model
OFAT DoE
Not possible!
No structured approach
Possible, by using a design for the
construction of the experimental
plan with subsequent cause-effect
model creation.
factors (X)
Input
Transfer function
Disturbances
Output
Cause - Effect - Model
System
oder
Prozess
Responses(y)
17. By info@stefan-moser.com
Evaluating multiple response
settings
OFAT DoE
Conditional (manually)
possible by comparing
heat maps, contour plots
Automated possible, through
multi-variable optimization e.g.
with the Simplex Algorithm
18. By info@stefan-moser.com
Limits in using the different
approaches for predicting
OFAT DoE
Conditionally possible, by
using heat maps, contour
plots
Possible through multi-linear
regression, good designs, contour
plots, and Venn diagrams.
X1
X2
X1
X2
Maximum 90%
Maximum 94%
19. By info@stefan-moser.com
Derive optimal adjustment / set
points and its ranges
OFAT DoE
Conditionally (manually)
possible by using
heatmaps, contour plots
Automated possible, through
multi-variable optimization e.g.
with the Simplex Algorithm
X1
X2
X1
X2
Max Yield
20. By info@stefan-moser.com
Determining of factor tolerances
OFAT DoE
Not possible, further tests
are required!
Automated possible by multi-size
optimization e.g. with Monte
Carlo simulation
21. By info@stefan-moser.com
Defining the space for Robustness
and the Design Space
OFAT DoE
Not possible, further
testing is required!
Automated possible by
considering individual model
errors as well as measurement
and adjustment accuracy
Venn Chart Probability
diagram
22. By info@stefan-moser.com
Find a robust work point
including its tolerances
OFAT DoE
Not possible, further tests
are required!
Automated possible, comparison
of operating points,
determination of a safe (robust)
operating point
Probability Diagramm
23. By info@stefan-moser.com
Preparation / summery to make
confident decisions....
Is there still room for
improvement?
What are the limits?
Is it safe / Robust?
Cost benefit ... is it
worth going on?
Can the project be
finally completed?
Do I have to expect
surprises?
Will I reach the goals?
OFAT DoE
24. By info@stefan-moser.com
✓ Provides a clear structure to think through (problem
formulation), to plan, and to organize
✓ An organized approach to creating experimental designs
and evaluating their results in a traceable manner.
✓ As much information as possible is obtained with as few
experiments as possible to create a cause - effect
models.
✓ In addition to the effect of the factors, further useful
information is obtained, such as the influence of
interaction and nonlinear behavior and noise.
✓ Valuable support for decision making based on process
maps, that can also address system boundaries and
contradictions within the light of variability.
✓ Tolerances are calculated holistically as wide as possible
and as narrow as possible
Summary: benefits of DoE?