Full Factorial DoE Vs MetaCause
Accuracy of MetaCause data analysis software is verified on a benchmark full factorial design of experiment analysis presented by the National Institute of Standards and Technology in their online statistical handbook.
The detailed Statistical analysis procedures and corresponding results are available on http://www.itl.nist.gov/div898/handbook/pri/section4/pri471.htm
We look to demonstrate how MetaCause performs on this benchmark problem.
Aim: The purpose of this DoE study was to understand the effect of grinding parameters on 3 types of silicon nitrides measured as the ceramic strength
|
|
- Response Variable - Mean (over 15 repetitions) of Ceramic Strength
- Factors -
Process Parameters |
Lower Level |
Upper Level |
Table Speed |
0.025 m/s |
0.125 m/s |
Down Feed Rate |
0.05 mm |
0.125 mm |
Wheel Grit |
140/170 |
80/100 |
Direction |
Longitudinal |
Transverse |
Batch |
1 |
2 |
Full factorial design matrix and corresponding measured ceramic strength is given below.
- Click for MetaCause Data File -
The Analysis
Whether its Design of Experiments Data, Taguchi Data or Production Data, MetaCause analyses the data and learns from every observation. It learns which parameters are important in influencing the response and how.
The Results
The results generated by MetaCause are shown below:
- Click for Main Effects Report -
- Click for Interactions Report-
 |
As we can see from the results, they are not only accurate but also easy to read, interpret and understand. This is in contrast with the complex statistical procedures described on http://www.itl.nist.gov/div898/handbook/pri/section4/pri471.htm
We believe that conducting design of experiments for improving process should be third or fourth if not the last option. The challenge we face today is NOT about reducing defects from 20% to 5%. Our processes are already stable and we monitor lot of data.
Our challenge is to be able to fine tune the process by simultaneously adjusting number of parameters at the same time. We want to discover interactions among operators, machines and process variables and maintain optimal operations despite varying condition.
MetaCause's patented technology is designed to meet these challenges and help you in improving your process under practical conditions.
|
MetaCause identifies interactions when it sees evidence that the joint effect is stronger than the individual contribution. Calculation procedures for Main Effects and Interactions are very different in statistics and most of the time are unreliable. No interactions are discovered by MetaCause for this study. |
- Back -
.
|
