CO2_2016 - page 34

32
Chimica Oggi - Chemistry Today
- vol. 34(2) March/April 2016
of the site as a location for cost-effective, high-value
manufacturing. Additional DOE case studies are available in
Goos and Jones (8).
CONCLUSIONS
Providing you are able to experiment or actively intervene
in your process, you can similarly use DOE to create the
data you need to statistically model your process and
increase real understanding. DOE might help you or your
company to:
-
-
Increase the predictability of design, development and
manufacturing projects.
-
-
Increase understanding of your products and
processes.
-
-
Optimize products and processes more efficiently.
-
-
Deliver a competitive edge.
-
-
Increase the availability of money for further innovation
and improvement.
Mining of prior data helps increase the success of DOE
through efficient and effective variable selection.
REFERENCES
1. Fisher-Box J. (1978) R.A. Fisher: The Life of a Scientist (Wiley
Series in Probability and Mathematical Statistics), ISBN: 978-0-
471-09300-8
linear effects can also be modeled to optimize the process
without the need for additional experimentation.
Figure 10 contains the completed worksheet for the 17
run Definitive Screening Design in seven factors. Figure
11 shows the profiler and 3d plot of the statistical model
resulting from the DOE. Just three of the seven factors had
an important effect on mill time to 200 nm – Temp(Pot),
%Beads and %Pigment. The settings of these three factors
needed to minimize mill time to 200 nm are also indicated on
F igure 11 and are Temp(Pot) of 50
o
C, % Beads 68.8 of and
%Pigment of 13.5 with a predicted mill time of 2.6 hours (95%
confidence interval of 2.2 to 3.0 hours). Three confirmation
runs at these settings verified the time to mill to 200 nm was
now well below the upper limit of 5 hours, see Figure 12.
SUMMARY
Data mining of prior data identified the variables to input to
a definitive screening design, which provided an efficient
experimental plan. Statistically modeling the resulting
data and exploring the model using profiler graphs of JMP,
allowed the model to be translated into process insight
from which a process setting was determined to satisfy the
business needs in cycle time to deliver customer demand
and at the same time run the process at reduced energy
costs. The problem was solved quickly, saving €100,000s off
the development budget, and enhanced the credibility
Figure 10.
Completed worksheet.
Figure 11.
Profiler and 3d plot
of statistical model.
Figure 12.
Three confirmation runs.
2. Draper N., and Smith H. (1981) Applied
Regression Analysis, 2nd Edition, Wiley, ISBN
0-471-02995-5
3. Box G.E.P., Hunter J.S., Hunter G.H. (2005),
Statistics for Experimenters: Design,
Innovation, and Discovery, 2nd Edition, Wiley,
ISBN: 978-0-471-71813-0
4. Jones B., and Nachtsheim C., (2011) "A Class
of Three-Level Designs for Definitive Screening
in the Presence of Second-Order Effects"
Journal of Quality Technology, Vol. 43, No. 1
5.
6.
JMP Pro Statistical DiscoveryTM from SAS:
-
pro.html
7. Hassall P., (2009). Milling-media review:
Bead milling operating parameters. Paint &
Coatings Industry, 25(3), 54-56.
8. Goos P, and Jones B, (2011), Optimal Design
of Experiments: A Case Study Approach,
Wiley ISBN: 978-0-470-74461-1
1...,24,25,26,27,28,29,30,31,32,33 35,36,37,38,39,40,41,42,43,44,...68
Powered by FlippingBook