In these pages we show the Robust Design techniques
that iCT-M supports. Although it is difficult to show examples of every conceivable field, the principles of experimental design and R&D are universal. Therefore, we show some generic applications. We believe, you will recognize areas in which you can apply these strategies.
Are these familiar to you?
- My yield is very low
- There is too much variability, I cannot repeat the process
- There are too many factors and too many experiments
- I cannot optimize all the factors together
- I do not know what to do or how to do
- The calculations are very difficult and I am not a statistician
How do we do it?
iCT-M believes that Robust Design is generic and can be applied in
diverse applications. Shown here is the generic 2-Step Optimization.
Our strategy is as follows:
Identify the four types of factors:
Factor that... |
Affects the mean |
Does not the affect mean |
Affects variability |
Great caution is needed. Most researchers are hit here from the word "Go" |
Use these factors to reduce variability |
Does not affect variability |
Use these factors to reduce bias |
Use these factors to your advantage |
and then
- Reduce variability
- Reduce bias
by the 2-Step optimization method. How? Use any statistically feasible method such as Orthogonal Arrays, Latin Square, Full Factorial, Plackett-Burman, Composite Designs
to identify which aspects the factor controls. If the factor:
-
Affects the variability without affecting the
mean, use that factor to reduce the variability.
-
Affects the mean without affecting the
variability, use that factor to reduce the bias.
-
Affects neither the variability nor the mean, use
that factor to suit.
-
Affects both the variability and the mean, use
that factor with some compromise on variability or mean.
|