科研之友
微信
新浪微博
Facebook
分享链接
摘要
Experimental trials are usually needed to integrate information reported on data sheets in order to properly drive membrane choice. It results in data-sets where each membrane is characterised by several performance descriptors. Multivariate data-mining (i.e. chemometrics) effectiveness in analysing such data-sets has been demonstrated through the comparison of seven commercial membranes. Each membrane was represented as an object described by 15 features got from trials with different single-component test solutions. Tree cluster analysis based on Ward amalgamation method was employed for multivariate data mining. The algorithm progressively grouped the membranes in clusters, adopting the Euclidean distance in the 15-dimensional feature space as a measure of similarity. Thus, a graphical output consisting into a similarity tree representing the membrane taxonomy was obtained. A restricted number of membranes, selected as representatives of each identified cluster, underwent to further experiments devoted to a systematic study on boron removal at various pH values.
- 出版日期2016
