<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />Author: Lou Valente
Why Good: Great data set for EMP MSA variability study. Data is from 1952 moisture analysis of nylon molding powders at W.L. Gore. Ready for JMP EMP Gauge R&R measurement system analysis exploration (unscripted).
Scripts:
None
Files:
WL Gore Nylon Moisture.jmp
Background:
An Experiment in Variation
W.L. Gore-Statistical Methods for Chemical Experimentation 1952
Moisture analyses of nylon molding powders sampled during drying cycles were found to be anomalous and unreasonable. Moisture content would frequently increase as the time or temperature of drying was increased. Duplicate moisture analyses usually checked within 0.01 to 0.02% moisture, while the anomalous increases or sudden drops in the moisture content were on the order of 0.1 to 0.15%. The analysis consisted in a measurement of weight loss when a 10g sample was heated to a high temperature for a short time.
Several hypotheses were advanced for this paradoxical behavior of nylon:
- ) Different batches of nylon produced variable amounts of an unidentified compound during the drying cycle which volatilized during the high temperature weight loss test.
- ) The 10g duplicate sample taken from the same sample were not representative of the very large batch from which they were taken.
- ) The variability of the weight loss test was much greater than indicated by the excellent checks between duplicate samples.
In spite of extensive experimental evidence demonstrating the good reproducibility of the analysis between duplicate tests, it was decided to evaluate the reliability of the analysis and sampling on a broader basis.
The four factors that decided to include in the experiment were:
- ) Analysts-six different analysts would be compared
- ) Time-each analyst would run the same sample on two different days
- ) Sampling-five different samples would be taken at different strata from the nylon in the dryer
- ) Duplication-each analyst would carry out all tests in duplicate.
(6 analysts)(2 different times)(5 samples)(2 duplicates) = 120 analyses
The checks between duplicates are much better than between different analysts for the same sample at the same time
‘There seems to be a consistent increase in moisture between the first and second time an analyst runs a particular sample
The overall reliability of a moisture determination appears to be poor by this method since the average results of different analysts at different times of different samples from the same lot of material differ by as much as 0.26% which is nearly as large a range as the average figure for the moisture of the batch.
The dried batch is reasonably homogeneous throughout since variations are relatively small between the five samples.
