Dynamic mechanical analysis and chemometrics for polymer identification

The temperature profile of 252 polymers from eight different chemical groups has been studied by dynamic mechanical analysis, with the aim of producing a method for clustering polymers with the same origins into the same group. The cot δ profile is employed after interpolation, and principal components are used to visualise groupings. Hierarchical cluster analysis is then performed using both the Euclidean distance and the correlation coefficient on (a) the interpolated cot δ profiles and (b) the standardised scores of the first six principal components. Several methods for linkage are compared; the best approach involving single linkage and Euclidean distance on the cot δ data correctly assigns 245 of the samples into correct groups by visual inspection of dendrograms.