|
General Question
and Answers on DOE
Q1: What kind of problems can
be solved by Design of Experiment (DOE) technique?
A1: Any problem or
product/process design issues that are suspected to have influence
from two or more factors (input variables) will benefit from DOE.
Q2: Why should I apply DOE
when there are other simpler problem solving techniques available?
A2: Use simpler
technique that you know to solve problems when you can. generally,
if your problem is of technical nature (production process or
complex product design) with more than two factors involved, common
disciplines of problem solving will not be able to determine a
statistically valid solution.
Q3: What are the
advantages of using Taguchi version of DOE?
A3:The Taguchi version of
DOE is relatively simpler to apply for people with limited
background in statistical science. Below are some of the advantages.
- Relatively smaller size
of the experiment
- Easier to learn and
apply
- Two factor interactions
are easily handled in selected orthogonal arrays
- Mixed level designs are
accomplished by simple modification of the orthogonal arrays
- Noise factor effects
are nicely incorporated in design with OUTER ARRAY designs
- Signal-to-Noise ratio
analysis of multiple sample results allows better analysis of
variability and robustness
- LOSS FUNCTION allows
translation of improvement in terms of dollars with statistical
validity.
Q4: If we were interested
in benefiting from the Taguchi DOE technique to solve product or
process design issues, how should we go about it?
A4: To learn and apply the
technique, you have several options.
- Retain experienced
practitioners to help you with current problems that affect your
bottom line.
- Dedicate time and
effort to learn the skill by attending seminar/workshop with focus
on application knowledge
- Build in-house
capabilities by whatever means available (Books , software,
training, etc.)
Q5: What level of
training or skill building do we need to start applying the
technique to solve problems?
A5: Start with
simpler experiments (3 -11 factors). Secure help/guidance from
experience practitioners for first few experiments.
Q5: What is a cost
effective way, train our own people or secure consulting support?
A5: Depending on your
need, select the option. Definitely, starting with a few
applications and have success stories to back your training effort
is highly recommended.
Q6: If we go for
consulting support, what should we expect the consultant to do?
A6: Consultants can help
you with all THREE PHASES of the application: (1) Experiment
PLANNING, (2) Experiment design, and (3) Analysis of results. Time
permitting, you may also receive an overview of the technique and
detailed understanding of the activities involved in each step.
Q7: Our process is quite
complex, is it possible for consultant to know our process and give
us some design recommendation?
A7:A consulting
knowledgeable in Taguchi/DOE need not be an expert in your product
or process under study. Nor should the consultant participate in
major decisions about the project. He or she simply should
facilitative and help you walk through the structure of discussions
recommended (
http://nutek-us.com/wp-exptplanning.html ) during the experiment
planning session.
Q8: If we secure
consulting support, what kind of time involvement the team members
have to offer?
A8: Involvement of the
team member will range between 1/2 day to 1 1/2 day. The rest of the
work on the project most likely be done by the responsible parties.
Q9: What is time
requirement to complete a typical project?
A9: The project
completion time is mainly dictated by the nature of the project and
other logistics for conducting the experiments.
Beyond the experiment
planning (2 days), the design can be laid out within 2 - 5 days.
This allows the experiments (say 4 - 16) conducted. The time for
running the experiments will depend on the nature of the project and
number of samples tested in each trial condition (cannot be
estimated). The final part the consultant provide is analysis
service which may take another week.
Q10: Is there a way
to estimate the range of size of experiment that would be
involved in our projects?
A10: Most projects
will involve 8 to 12 separate experimental conditions.
|
application specialist, 