|Saturday, February 17|
|PS3 Poster Session 3 and Continental Breakfast||
Sat, Feb 17, 8:00 AM - 9:15 AM
Multivariate Statistical Analysis in Plastic Foam Research (303595)
*Wenyu Su, The Dow Chemical Company
Keywords: Multivariate Statistical Analysis, Foam Research
Dow is an industrial leader in supplying products for building and construction applications. Dow R&D initiated an effort to improve plastic foam products in response to more stringent regulatory requirements on fire retardancy and energy efficiency of building products. We analyzed the foam properties data using multivariate statistical analysis methods including principal component analysis (PCA), variable clustering, clustering analysis, and partial least squares. PCA was used to identify a small set of indices to represent a large set of foam properties measurements. Variable clustering technique was applied to choose a representative variable from each group of highly correlated performance measurements. Clustering analysis proved to be a powerful method to investigate the similarity between samples. Partial least squares modeling effectively linked the highly correlated formulation components with performance measurements. In addition, star plots provided an effective way to graphically compare the samples with respect to the various measurements considered. The statistical analysis identified optimal samples for development of improved insulation foam products.